Psychometric Properties of the Geriatric Anxiety Scale in Community-Dwelling,

Clinical, and Medical Samples of Older Adults

by

Anne Elizabeth Mueller

B.A., Saint Louis University, 2008

M.A., University of Colorado Colorado Springs, 2010

A dissertation submitted to the Graduate Faculty of the

University of Colorado Colorado Springs

in partial fulfillment of the

requirements for the degree of

Doctor of Philosophy

Department of Psychology

2014
This dissertation for Doctor of Philosophy degree by

Anne Elizabeth Mueller

has been approved for the

Department of Psychology by

Daniel L. Segal, Chair

Leilani Feliciano

Brandon Gavett

Amy Silva-Smith

Brian Yochim

Date
 iii

Mueller, Anne Elizabeth (Ph.D., Psychology)

Psychometric Properties of the Geriatric Anxiety Scale in Community-Dwelling,

Clinical, and Medical Samples of Older Adults

Dissertation directed by Professor Daniel L. Segal

This study examined the psychometric properties of the Geriatric Anxiety Scale

(GAS; Segal, June, Payne, Coolidge, & Yochim, 2010) in three samples of older adults

using both classical test theory (CTT) and item response theory (IRT) techniques. Study

One was conducted in a community-dwelling sample of older adults. Factor analysis

revealed that a one- or two- factor solution best fit the data. The GAS also had excellent

reliability and adequate convergent validity with other measures, though it lacked

discriminant validity with a measure of depression. Study Two examined the

psychometric properties in a clinical sample of older adults, and the GAS performed

largely similar in this population. Study Three utilized a medical sample of older

individuals, and found that the GAS had moderate relationships to self-reported

subjective health status. Study Four used IRT to examine the item properties of the scale

in all three samples, and a short form (GAS-10) was also created. Two items were

flagged for differential item functioning (DIF), but the degree of DIF was negligible.

Women scored significantly higher than men on the GAS and subscales, and adults

younger than 80 scored significantly higher on the Cognitive subscale than adults 80 and

up. Results from the studies indicated that the GAS has strong psychometric properties.

Implications and future directions of study are discussed.

 TABLE OF CONTENTS

CHAPTER

I. INTRODUCTION ........................................................................................... 1

Item Response Theory .............................................................................. 13

Existing Measures of Anxiety: Strengths and Limitations ....................... 15

Geriatric Anxiety Scale: Overview and Preliminary Psychometric

Properties ..................................................................................... 18

Statement of Problem and Purpose of the Study ..................................... 22

II. STUDY ONE ................................................................................................. 27

Method ...................................................................................................... 27

Results ....................................................................................................... 34

III. STUDY TWO……………………………,,,, ................................................. 44

Method ...................................................................................................... 44

Results ....................................................................................................... 45

IV. STUDY THREE ............................................................................................. 52

Method ..................................................................................................... 52

Results ...................................................................................................... 54

V. STUDY FOUR ...............................................................................................60

Method ..................................................................................................... 60

Results ..................................................................................................... 63

VI. GENERAL DISCUSSION ............................................................................. 77

REFERENCES ................................................................................................................. 93

APPENDICES

A. Geriatric Anxiety Scale – Version 1.0 ........................................................... 103

B. Scoring Instruction ........................................................................................ 104

C. Geriatric Anxiety Scale-8 ............................................................................... 105

D. Geriatric Anxiety Scale – 10 Item Version (GAS-10) .................................. 106

 vi

TABLES

Table

1. Means, Standard Deviations, and Ranges for All Demographic Information and
All Measures… ......................................................................................... 28

2. Cronbach’s Alpha Coefficients for GAS Total Scale and Subscales ....................34

3. Principal Axis Factoring with Direct Oblimin Rotation in Study 1 .......................36

4. Correlations between Demographic Variables, GAS, PHQ-9, BHS, SISE, 3LS,

and SF-36 Physical Functioning Subscale in Study 1…........................... 37

5. Correlations Between GAS items and PHQ-9 in Study 1 (N = 275) .................... 40

6. Correlations among GAS Items and SF-36 Physical Functioning Subscale Scores
in Study 1 (N = 270)..................................................................................43

7. Correlations among Demographic Variables, GAS, Subscales, GDS, and GAF in

Study 2 (N = 136) ..................................................................................... 48

8. Correlations among GAS items and GDS in Study 2 (N = 99) ........................... 50

9. Correlations among Demographic Variables, GAS, MoCA, GAI, BAI, PHQ-9,

and SF-36 total Scale Scores in Study 3 (N = 38).....................................56

10. Correlations among GAS, Subscales, BAI, GAI, PHQ-9, and SF-36 Subscales in
Study 3 ...................................................................................................... 57

11. IRT Calibration for GAS Items .............................................................................66

12. Standard Score Distribution for GAS Total Scale Scores (N = 542) ................... 73

13. Score Distribution for GAS Subscale Scores .........................................................74

14. Score Distribution for GAS-10 (N = 556).............................................................75

 vii

FIGURES

Figure

1. Screeplot for principal axis factoring on 25 GAS items (Study 1). ................ 35

2. Screeplot for principal axis factoring on 25 GAS items (Study 2) ................ 46

3. Test information function for 24-item GAS… ............................................... 66

4. Test (left panel) and DIF-item (right panel) characteristic curves by age
(young-old versus old-old)....................................................................... 67

5. Test (left panel) and DIF-item (right panel) characteristic curves by sex...... 68

6. Test Information Function for GAS-10 ........................................................ 70

 CHAPTER 1

INTRODUCTION

Anxiety disorders are among the most ubiquitous and debilitating mental

disorders in older adults. In fact, anxiety disorders in older adults are common, with a

prevalence estimate ranging from 3.2 to 14.2% depending upon diagnostic criteria (i.e.,

DSM-III or DSM-IV) and age cutoff (i.e., 55 and up versus 65 and up; Wolitzky-Taylor,

Castriotta, Lenze, Stanley, & Craske, 2010). Generalized Anxiety Disorder (GAD), a

common anxiety disorder across the lifespan, includes the following criteria according to

the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR, American

Psychiatric Association, 2000): excessive anxiety and worry more days than not for at

least six months, difficulty controlling this worry, and at least three anxiety symptoms

(restlessness, irritability, muscle tension, sleep disturbances, difficulty concentrating,

being easily fatigued). These symptoms must also cause clinically significant impairment

in functioning (e.g., social, occupational) and/or distress. Sub-syndromal anxiety

symptoms in late life are even more widespread than anxiety disorders, with a prevalence

ranging from 15% to 52.3% in community samples (Bryant, Jackson, & Ames, 2008).

Indeed, a significant number of older adults are impacted by anxiety. As this

subpopulation will steadily increase in years to come, the number of people who

experience anxiety will subsequently increase as well. As will be discussed in this

review, anxiety is associated with a multitude of dire outcomes in late life.

 2

Accurate and timely assessment of anxiety is paramount, as this is a precursor for

appropriate treatment. However, there are several challenges unique to anxiety in older

adults which complicates anxiety assessment in this specific population. The Geriatric

Anxiety Scale (GAS; Segal, June, Payne, Coolidge, & Yochim, 2010) is a self-report

assessment tool designed specifically for use with older adults to address such challenges.

Initial studies have suggested that the GAS is a reliable and valid tool (Segal et al., 2010;

Yochim, Mueller, June, & Segal, 2011). The purpose of the current study was to further

investigate the psychometric properties of the Geriatric Anxiety Scale in three samples of

older adults, with the intention of increasing its utility for both clinicians and researchers

alike.

There is a unique set of challenges associated with the assessment of anxiety in

late life. Such challenges highlight the necessity to utilize screening tools that address

these issues and are well-validated for use with this particular population. Anxiety is

often co-morbid with physical health conditions, which complicates the manner in which

older adults perceive their symptoms, describe their symptoms, and seek treatment for

these symptoms. Research has indicated that approximately one-third of adults with a

somatic health problem also experiences anxiety and depression (Stordal, Bjelland, Dahl,

& Mykletun, 2003). Specifically, anxiety occurs in high rates alongside health conditions

such as arthritis (Brock et al., 2011; Murphy, Sacks, Brady, Hootman, & Chapman,

2012), chronic obstructive pulmonary disease (Cully et al., 2006), diabetes, and

gastrointestinal health concerns (Wetherell, Ayers, Nuevo, Stein, Ramsdell, & Patterson,

2010). Anxiety is also associated with cognitive impairment (Yochim, Mueller, & Segal,

2012), urinary incontinence, sleep problems, and detrimental health behaviors such as
 3

smoking, physical inactivity, poor diet, and alcohol abuse (DeLuca et al., 2005; Mehta et

al., 2003; Strine, Chapman, Kobau, & Balluz, 2005). The co-occurrence of anxiety with

medical problems is troublesome as it is associated with increased functional impairment,

more physician visits (Kroenke et al., 2007), and decreased health-related quality of life

(Porensky et al., 2009). This combination can result in increased healthcare costs and

ineffective treatment. For example, an older adult may attribute particular symptoms of

anxiety (e.g., feeling jumpy, fatigue) to his or her medical conditions and/or medications,

and seek out services from his or her physician instead of a mental health clinician. Not

surprisingly, co-morbid anxiety with chronic medical problems has been associated with

reports of heightened somatic symptoms, even after controlling for the severity of the

medical condition itself (Katon, Lin, & Kroenke, 2007). Thus, the presence of anxiety in

conjunction with medical illness often results in heightened utilization of medical

services (Porensky et al., 2009), whereas medical treatment alone may not be optimal nor

sufficient in providing total patient care. Another way this co-morbidity between health

conditions and anxiety may impact assessment is that heightened somatic symptoms may

result in inflated scores on anxiety measures which contain many somatic items.

Anxiety also increases the risk for individuals to develop significant health

concerns and disability throughout the lifespan. For instance, anxiety is associated with

the onset of cardiovascular medical conditions, such as the development of coronary

heart disease (Suls & Bunde, 2005) and acute coronary syndrome (Rozanski, Blumenthal,

& Kaplan, 1999). Thus, those with anxiety are at a higher risk for mortality in late life.

Van Hout et al. (2004) found that older men with anxiety disorders had an 87% higher

risk of mortality over seven years than older men without anxiety disorders, even after
 4

controlling for co-morbid depression, smoking, alcohol use, and body mass index.

Moreover, the presence of anxiety symptoms is associated with an increased risk of

disability in completing activities of daily living, despite the presence of emotional

support, psychotropic medication, and physical activity (Brenes et al., 2005). Whereas

some physical health conditions in late life may be unavoidable, excessive anxiety, in

contrast, is a treatable condition (i.e., Ayers, Sorrell, Thorp, & Wetherell, 2007). Given

these findings, anxiety is clearly a serious concern in late life.

Unfortunately, anxiety in medical settings is highly prevalent but largely

undetected. For example, in a study of 965 adults in primary care settings, 19.5% were

found to have at least one diagnosable anxiety disorder (Kroenke et al., 2007). In a study

of adults with arthritis and co-morbid anxiety and/or depression who regularly attended

medical appointments, only half of individuals had sought out any help for their mental

health symptoms (i.e., not just behavioral health services; Murphy et al., 2012).

Alarmingly, the detection rate of GAD by physicians is as low as 1.5% (Calleo et al.,

2009). Physicians often rely on patient self-report to diagnose anxiety, and the older

patient may ascribe their symptoms to other factors such as physical illness and/or

depression (Segal, Qualls, & Smyer, 2011). Additionally, some older patients may

erroneously perceive their anxiety symptoms as expectable or normal in the context of

later life, and thus not report them to their doctors. Taken together, these factors stress

the need for appropriate and brief assessment tools to be administered routinely in

medical settings to coordinate appropriate treatment and to maximize patient well-being.

Furthermore, the overlap between somatic symptoms of anxiety and symptoms of

underlying physical health conditions emphasizes the importance of assessing other

 5

aspects of anxiety (i.e., cognitive, affective) in addition to assessing somatic symptoms.

Failure to do so may result in an incomplete and/or inaccurate assessment, misdiagnosis,

and ineffective treatment.

Another major challenge to assessment is the co-morbidity of anxiety with other

mental health problems. Although anxiety exists independently of other mental health

complications, it is often co-morbid with other psychopathology. Most notably, anxiety

in older adults is highly co-morbid with depressive symptoms (Beekman et al., 2000).

This could be in part due to the overlapping symptom criteria of both conditions (i.e.,

sleep disturbances, fatigue, difficulty concentrating). In a study of over 3,000 older

adults, anxiety symptoms occurred in 43% of the people who reported depression (Mehta

et al., 2003). Furthermore, in a population-based sample of community-dwelling older

adults, anxiety disorders were noted in 23% of participants who also met the full

diagnostic criteria for Major Depressive Disorder (Cairney, Corna, Veldhuizen,

Herrmann, & Streiner, 2008). The co-morbidity of anxiety and depression is problematic

as it has been associated with more severe anxiety symptomatology (Hopko et al., 2000),

lower levels of well-being, greater functional impairment (Cairney et al., 2008), as well

as poorer responsiveness to both pharmacological and psychological therapies

(Andresscu et al., 2007). These findings demonstrate the gravity of having more than one

psychiatric condition, and emphasize the importance of early and accurate detection of

anxiety symptoms as well as depression in order to initiate treatment as soon as possible.

However, the co-morbidity of anxiety with depression increases the likelihood of

detection and psychopharmacological treatment in primary care settings (Calleo et al.,

 6

2009). This is likely due to the increased severity of symptoms, although both conditions

are largely unrecognized in medical settings.

Another issue unique to the assessment of anxiety in later life is the role of

cognitive impairment, a common condition among older adults. Individuals with

cognitive impairment often present with co-occurring symptoms of anxiety, which may

complicate the manner in which anxiety is experienced and communicated (Wolitzky-

Taylor et al., 2010). For example, individuals with more severe cognitive impairment

may experience anxiety but lack the cognitive skills needed to describe and seek help for

such symptoms, stressing the need for caregivers to be aware of such symptoms.

Additionally, some symptoms of anxiety overlap with cognitive impairment (i.e.,

difficulty concentrating, agitation), increasing the risk of misdiagnosis of either

condition. The relationship between anxiety and cognitive impairment may be

bidirectional, such that anxiety symptoms may exacerbate cognitive impairment, but

awareness of cognitive impairment can lead to anxiety as well.

There are also cultural considerations in late-life anxiety assessment. Though the

lifetime prevalence rates of all anxiety disorders does not significantly differ among

various ethnic groups (Jimenez, Alegria, Chen, Chan, & Laderman, 2010), there is vast

diversity both among and within various ethnic groups in regards to the manner in which

anxiety is expressed and treated. The prevalence of generalized anxiety disorder and

social phobia is lower among Afro-Caribbean older adults than non-Latino white older

adults. Additionally, the prevalence of generalized anxiety disorder is higher among

older Asian and Latino immigrants than US-born Asian and Latino immigrants (Jimenez

et al., 2010). These subgroups may have greater difficulty in accessing behavioral health
 7

care due to language barriers, socioeconomic status, and stigma. However, ethnic

identity has a protective effect for some minority groups. Williams, Chapman, Wong,

and Turkheimer (2012) reported that higher levels of ethnic identity was associated with

lower levels of anxiety and depression in African American adults, but this association

was not found in European American adults. This suggests that ethnic identity is an

important area of assessment to consider with minority clients. Kim and colleagues

(2011) reported that self-reported mental health and psychiatric diagnoses is not

consistent among minority groups. The researchers found that self-reported mental

health was more related to anxiety disorders in non-Hispanic white individuals than

Hispanic, African-American, and Asian individuals. Thus, though anxiety disorders are

more prevalent in non-Hispanic white populations, it could be that screening measures

which assess self-reported mental health are not sensitive to symptoms in individuals

from minority groups.

Compared to younger adults, older adults may be reticent to identify their

symptoms as anxiety and subsequently seek treatment specifically for those symptoms.

For example, in an epidemiological study of older adults, less than a quarter of older

adults with GAD elicited mental health services (Mackenzie, Reynolds, Chou, Pagura, &

Sareen, 2011), again underscoring the need for routine screening in other contexts (e.g.,

primary care). Furthermore, older adults are less accurate than younger adults at

identifying symptoms of anxiety (Wetherell et al., 2009). Older adults often attribute

somatic symptoms of anxiety (e.g., fatigue, sleep troubles, muscle tension) to physical

illnesses versus identifying them as symptoms of a mental health problem, leading them

to elicit help from their physician rather than a specialized mental health clinician,
 8

particularly in the presence of other medical conditions. Additionally, older adults may

interpret affective symptoms of anxiety (i.e., decreased mood, worry) as depression

(Segal et al., 2011), which could also impact the manner in which older adults seek

treatment.

In addition, the experience of anxiety in late life may differ from the experience

of anxiety in earlier stages of life. This presents concerns regarding the content validity

of screening measures intended for use with younger adults, as some items on such

measures may not be appropriate (Kogan, Edelstein, & McKee, 2000). For example,

older adults report more concerns regarding health than younger adults, and younger

adults tend to report higher levels of work-related concerns (Diefenbach, Stanley, &

Beck, 2001). Other common areas of worry for older adults include loss of independence

and functional status, looking incompetent in front of others, and forgetting information

in front of others (Brock et al., 2011; Ciliberti, Gould, Smith, Chorney, & Edelstein,

2011). Inadequate content validity of assessment tools increases the risk of

misdiagnosing anxiety in late life, as some older adults with anxiety may not endorse

items designed for use with younger adults. Thus, development of age and cohort

appropriate assessment measures is imperative.

Wolitzky-Taylor and colleagues (2010) also raised concerns regarding the

differential experience of anxiety in late life versus earlier life. In their review of the

current literature on late-life anxiety, they found that in general, the symptom

presentations of anxiety disorders are similar to that of younger adults; however, they

highlighted that the current knowledge of anxiety disorders in older adults is bound by

the diagnostic classification system itself. Older adults may experience unique symptoms
 9

and/or patterns of symptoms which are not currently included in the DSM-IV-TR and thus

not routinely assessed, resulting in an inaccurate estimate regarding how many older

adults experience anxiety. For example, the current diagnostic system lacks age-specific

examples of maladaptive avoidance behaviors (i.e., avoiding situations due to excessive

fear of falling, excessive checking of blood pressure; Mohlman et al., 2011). There is

also controversy regarding the use of the word “excessive” in assessment, as older adults

may not view their concerns as exceeding the norm for their age (Mohlman et al., 2011).

Using more sensitive language (i.e., “How nervous are you compared to other people you

know?”) may be warranted. Therefore, there is a need for assessment tools designed

specifically for older adults, as well as better guidelines for diagnosing anxiety disorders

in this unique population.

Research has identified variables which differentially impact the incidence of

anxiety in late life. Such variables are important to consider in anxiety assessment as

they may be useful in identifying individuals who are at-risk for experiencing this

condition. These variables also influence normative data which may increase the

accuracy of assessment tools in correctly identifying individuals who experience anxiety

symptoms. The following section will describe two major variables identified in the

literature which appear to influence the occurrence of late life anxiety and their impact on

assessment.

Though anxiety disorders are common among older adults, they are generally

reported as less prevalent in older adults than in younger adults (Flint et al., 2010; Jorm,

2000; Owens, Hadjistavropoulos, & Asmundsmon, 2000). This finding is similar to the

relationship between other mental health problems (i.e., depression) and age. It is also
 10

important to bear in mind that older adults are heterogeneous in respect to various age

cohorts. Research has identified differences in anxiety prevalence between the “young-

old” and “old-old.” In a cross-sectional, community-based, epidemiologic study, Gum,

King-Kallimanis, and Kohn (2009) discovered that adults 75 years of age and older were

less likely than those between the ages of 65 to 69 to be diagnosed with an anxiety

disorder. Data from the Berlin Aging Study also suggest that the prevalence of anxiety

disorders continues to decrease in the “very old” (Schaub & Linden, 2000). The

prevalence of anxiety disorders (using DSM-III-R criteria) in adults aged 70-84 was

4.3%, whereas the prevalence in adults 85 years old and older was 2.3%. In sum, anxiety

generally occurs less frequently in older adults than in younger adults, and occurs even

less frequently in those who are considered “very old.” However, a multitude of adults in

late life experience anxiety regardless of age disparities in prevalence.

There are multiple hypotheses as to why the prevalence of anxiety tends to

decrease with advancing age. Findings from cross-sectional studies may reflect

underlying cohort differences versus true age effects (Segal et al., 2011). Additionally,

individuals with anxiety are more likely to be functionally impaired (Kroenke et al.,

2007), in which they may reside in assisted living facilities and thus be excluded from

epidemiological research (Segal et al.) or die earlier than those without anxiety (van Hout

et al., 2004). Furthermore, as mentioned previously, the experience of anxiety in late life

may be qualitatively different than anxiety in earlier stages of life. As such, the

differences in prevalence rates across the lifespan may be partly due to the diagnostic

criteria used to identify anxiety. Therefore, the finding that anxiety decreases with age
 11

may be due to diagnostic criteria and assessment tools better suited for younger adults

rather than older adults (Flint et al., 2010) as well as other methodological factors.

Sex is another variable which has been identified as a risk factor for anxiety.

Specifically, women tend to report higher levels of anxiety than men, a finding that is

reported consistently in the literature. For instance, Gum et al. (2009) found that

community-dwelling individuals who were diagnosed with an anxiety disorder were more

likely to be female. Furthermore, female sex has been associated with a greater

likelihood of anxiety chronicity in older adults (De Beurs, Beekman, van Dyck, & van

Tilburg, 2000), such that anxiety tends to persist in older women compared to older men.

Similarly, Owens et al. (2000) found that men above the age of 60 reported the lowest

levels of anxiety in comparison to younger men and older women. Other researchers

have identified female sex as an independent risk factor for heightened anxiety symptoms

(e.g., Lowe & Reynolds, 2005; Potvin et al., 2011). Despite these findings, there is some

evidence to suggest that sex differences in anxiety dissipate in very late adulthood. In a

study of community-dwelling adults aged 82 to 87, Pachana, McLaughlin, Leung, Byrne,

and Dobson (2011) found no significant sex differences in both anxiety and depression

after controlling for cognitive status, health, and level of education. Moreover, Brock

and colleagues (2011) found that women above the age of 80 tended to report fewer

worries than the younger women in their sample. Thus, though some research has drawn

attention to sex differences in anxiety prevalence, these differences may become less

salient with increasing age.

A number of variables have been proposed to explain sex differences in anxiety.

For example, Leach, Christensen, Mackinnon, Windsor, and Butterworth (2008) found
 12

that women tended to have poorer physical health, be more physically inactive, and have

more interpersonal problems than men, and reported that these variables mediated the

relationships among sex, anxiety, and depression. Thus, sex differences in anxiety could

be attributed to other factors which are associated with female sex. An additional

explanation could be that women are more emotionally aware and thus willing to report

or seek help for their anxiety symptoms than men, resulting in artificial discrepancies in

anxiety levels.

As can be noted from the previous sections, certain demographic variables

differentially impact the incidence of anxiety in older adults. Thus, particular

demographic groups (i.e., younger women) can be expected to obtain higher scores on

measures such as the GAS. Such variations could be the result of true differences in

prevalence rates, but could also reflect measurement bias. Measurement bias occurs

when a particular group of individuals has an unequal chance of endorsing an item than

another group of individuals, despite being matched upon the variable of interest. For

example, men and women with the same level of anxiety should have the same likelihood

of endorsing a particular item on a measure in the same manner. Thus, if an assessment

tool is biased against a certain variable, variations in prevalence rates would reflect this

measurement error instead of actual group differences. If a measure has no detectable

bias but differences between groups remain, the differences are more likely to reflect

actual variations between groups. Clearly, measurement bias has serious implications for

the conclusions drawn from assessment tools, both clinically and in research. Several

researchers have found item biases in various measures of anxiety. For instance, Van

Dam, Earleywine, and Forsyth (2009) found that removing a single item from the
 13

Anxiety Sensitivity Index (“It scares me when I feel faint”) eliminated significant gender

difference in scores but did not alter the internal consistency of the measure. Leach,

Christensen, and Mackinnon (2008) found that the Goldberg Anxiety and Depression

scales were free from gender-biased items, though the factor structures of the measures

became more similar between men and women when certain items were removed (i.e.,

items regarding sleep). This has serious implications for scale development, such that

individuals could appear more or less anxious than they are based upon item properties,

placing them at risk for misdiagnosis. Thus, though age and sex appear to have an impact

upon the incidence of anxiety in late life, item bias must be taken into account in scale

construction.

Item Response Theory

IRT is a set of statistical models used to measure latent variables (e.g., anxiety),

and posits that responses on a given item are a function of both person and item

properties (Edwards, 2009). According to IRT, individuals who have a greater level of

the latent trait should have a higher probability of endorsing a particular item measuring

that trait. Analyses are often presented as item characteristic curves (ICCs), plots which

indicate the likelihood of endorsing an item (e.g., symptom) as the level of the underlying

trait (e.g., anxiety) changes. The underlying trait is represented as theta (θ). More steep

slopes in ICCs indicate that the item under scrutiny is better able to discriminate among

people with high or low levels of the latent trait (represented as the discrimination

parameter, a). The threshold parameter (also known as item severity or difficulty

parameter; b) indicates the trait level at which the likelihood of endorsing a given

response choice is 50%. A higher threshold parameter indicates that the individual must
 14

have higher levels of the latent trait to have a 50% likelihood of endorsing the response

choice. Each item has an information function, depicted in item information curves

(IICs), which provide data about how much information the item yields about the

threshold parameter. In IICs, a steeper slope indicates that the item provides more

information about the threshold but over a more restricted range across the latent trait. A

less steep slope indicates that the item provides less information over a more broad range.

Item information functions are combined to create a test information function (TIF).

Within the IRT framework, the reliability of a test increases with the inclusion of better

or more informative items. Additionally, standard error is determined by calculating the

square root of the inverse of information. According to IRT, both information and

standard error are believed to vary across all trait levels, such that a particular item or

sum of items may be more informative for an individual with a higher level of a trait

compared with an individual with a lower level of the trait. Thus, IRT analyses are

pertinent to scale development as items that do not provide reliable information about a

person’s standing on a latent trait can be identified and either re-written or removed from

the measure. IRT can also be used to analyze item bias or measurement invariance (also

known as differential item functioning [DIF]). If an item is biased against a certain group

characteristic (e.g., age, sex), then the ICCs for that item will differ, despite the groups

being matched on the latent trait.

IRT modeling techniques are an alternative to classical test theory (CTT), which

has historically been popular among psychologists in scale development (Embretson &

Reise, 2000). While CTT statistical techniques are performed on a combined set of

items, IRT states that properties of the item are linked directly to test behavior. In other
 15

words, IRT can provide information regarding the latent variable under scrutiny by

analyzing each individual item of the test, whereas CTT examines the properties of the

test as a whole. IRT is growing in popularity among psychologists, particularly in

regards to clinical assessment (Reise & Waller, 2009), and is considered a favorable

alternative or supplement to CTT.

Existing Measures of Anxiety – Strengths and Limitations

In a systematic review of the literature, Therrien and Hunsley (2011) reported that

the majority of anxiety assessment tools used in research with older adults lack sufficient

psychometric evidence to justify their use with this population. Although there are a

multitude of empirically-validated measures of anxiety designed for use with younger

adults, there are few assessment tools designed specifically for older adult populations.

However, some existing measures of anxiety, while not intended for use with older

adults, have empirical support for use with this population. The following section will

detail some measures which are commonly used with older adults, including strengths

and limitations of each measure.

Beck Anxiety Inventory (BAI). The BAI (Beck, Epstein, Brown, & Steer,

1988) is a popular self-report measure of anxiety intended for use with adults of all ages.

It contains a list of 21 symptoms which are rated dimensionally from 0 (not at all) to 3

(severe). Though the BAI is not intended specifically for older adults, Kabacoff, Segal,

Hersen, and Van Hasselt (1997) found that the BAI had high internal consistency,

convergent, divergent, and factorial validity in a sample of older adults in a psychiatric

outpatient setting. Similar findings have been found in samples of older adults in medical

samples (i.e., Wetherell & Areán, 1997).

 16

However, the BAI may not be ideal for use with older adults as it contains many

somatic symptoms of anxiety. In fact, the majority of the 21 items on the BAI pertain to

somatic symptoms. Older adults may experience such symptoms for reasons other than

anxiety (i.e., physical illness), but could endorse high levels of anxiety as defined by the

measure. Thus, someone could screen positive for anxiety but not actually experience

this condition. Furthermore, the BAI was designed to exclude DSM symptoms of

depression which overlap with anxiety (i.e., difficulty concentrating, sleep troubles,

fatigue). Though this may help the clinician to distinguish whether the respondent is

experiencing either anxiety or depression, these symptoms are key diagnostic features of

anxiety and should be included in any routine assessment of anxiety symptoms. Thus,

though the BAI has demonstrated adequate psychometric properties in older adult

populations, it contains limitations which may restrict its clinical utility with older adults.

Geriatric Anxiety Inventory (GAI). The GAI (Pachana, Byrne, Siddle, Koloski,

Harley, & Arnold, 2007) is a self-report measure designed specifically for use with older

adults. Its intent is not to diagnose anxiety disorders per se, but rather to measure anxiety

symptomatology in older people. This may be considered a strength of the measure

given the inherent problems in the DSM and the prevalence of sub-syndromal anxiety in

older individuals. The GAI contains 20 statements (i.e., “I find it hard to relax,” “I think

of myself as a worrier, “Little things bother me a lot”) in which the respondent answers

yes or no. Pachana et al. also noted that the GAI was designed to assess few somatic

symptoms to minimize the risk of medical problems inflating scores on the measure. The

GAI has demonstrated sound psychometric properties, including high internal

consistency as well as convergent validity with other measures of anxiety (Pachana et

 17

al.). It has utility in various settings and contexts in which older adults receive services,

such as home care (Diefenbach et al., 2009), and has been translated and validated in

several different languages. A short form of the GAI has also been created and validated

in a community-dwelling sample of Australian women (Byrne & Pachana, 2011).

Despite these strengths, the GAI also has several limitations which may restrict its

clinical utility. For instance, the majority of items on the scale include statements

pertaining to worry and other cognitive aspects of anxiety (i.e., anticipating the worst,

difficulty making decisions). This may be problematic as an individual may experience

anxiety but their symptoms may not be predominantly cognitive in nature. Additionally,

the GAI utilizes a simple dichotomous response format. Though this may be preferable

for older adults with cognitive impairment and other limitations, a dimensional rating

system provides more information regarding the severity of the symptoms experienced by

the client. Overall, the GAI has strong psychometric properties as well as other ideal

features, but it also contains several features which may restrict its utility in assessing

anxiety in certain groups of older adults.

Adult Manifest Anxiety Scale-Elderly Version (AMAS-E). The AMAS-E

(Reynolds, Richmond, & Lowe, 2003) is a 44-item self-report measure of anxiety

designed for use with senior populations. The AMAS-E intends to measure chronic,

manifest anxiety, which refers to how one generally thinks, feels and/or acts. It contains

three subscales: Fear of Aging, Physiological Anxiety, and Worry/Oversensitivity. The

AMAS-E also contains a Lie scale, which provides a validity estimate which

distinguishes the measure from other anxiety assessment tools. Research has provided

evidence for the construct validity of the measure, as well as the factor structure (Lowe &
 18

Reynolds, 2006). However, similar to the GAI, the AMAS-E utilizes a dichotomous

response format. Additionally, the scale intends to measure chronic anxiety, which may

not apply to seniors who experience anxiety as a result of an acute life circumstance (i.e.,

changes in health/functional status, loss of spouse). Little validation research has been

conducted on the AMAS-E outside of the laboratory of the measure’s developer.

Older Adult Social Evaluative Scale (OASES). The OASES (Gould,

Gerolimatos, Ciliberti, Edelstein, & Smith, 2012) is a self-report assessment tool

designed to measure social anxiety in older adults. The measure consists of 37 items in

which the respondent is asked to rate how uncomfortable he or she would feel in a given

situation on a scale ranging from 0 (not at all) to 3 (severely). The respondent is also

asked to rate how often he or she avoids the given situation on a scale ranging from 0

(never) to 3 (usually). Preliminary research indicates that the OASES has excellent

internal consistency, convergent validity, and divergent validity (Gould et al.). The main

limitation of the OASES is that it does not intend to provide a measure of general anxiety

in older adults and thus is limited in its utility for such a purpose.

Geriatric Anxiety Scale: Overview and Preliminary Psychometric Properties

The GAS (Segal et al., 2010) is a 30-item self-report measure of anxiety

symptoms designed for use with older adults. The GAS is presented in Appendix A and

the scoring instructions are presented in Appendix B. There are 25 items assessing

symptoms of anxiety and 5 items which assess common aspects of worry among older

adults. The items were selected from a larger pool of items based upon item endorsement

frequency in a sample of older adults (Segal et al.). There are several qualities which

distinguish the GAS from other measures of anxiety and were intended to address
 19

limitations of existing assessment tools. First, the GAS was designed to include the full

spectrum of anxiety disorder symptoms as listed by the DSM-IV-TR (APA, 2000),

including symptoms which overlap with depression (in contrast to the BAI). This is

unique to the GAS as other anxiety assessment tools did not originate from DSM-IV-TR

symptoms, and indicates the GAS could potentially correspond more accurately with

anxiety diagnoses from the DSM-IV-TR than measures that do not contain such

symptoms. Second, the GAS contains three conceptually-derived subscales which are

intended to holistically assess anxiety: Somatic, Affective, and Cognitive. This feature of

the GAS allows the clinician or researcher to determine what type of symptoms are

particular challenges for the individual. It may also help the clinician rule out other

conditions which could impact their symptom presentation versus true anxiety symptoms.

For example, if an individual scores highly on the Somatic subscale but does not endorse

Cognitive or Affective symptoms, the clinician may wish to inquire further about this

discrepancy and rule out physical health conditions which could mimic anxiety

symptoms. Another distinctive attribute of the GAS is the gradated rating scale. Each

item on the GAS is rated dimensionally, with potential responses ranging from 0 (not at

all) to 3 (all of the time). This allows the respondent to endorse the severity of his or her

symptoms, which may provide additional data to the clinician for further inquiry.

Two studies have been published regarding the psychometric properties of the

GAS. Segal et al. (2010) examined the validity and internal consistency of the measure

in both community-dwelling and clinical samples of adults over the age of 60. In the

community-dwelling sample, the GAS was administered along with the Geriatric

Depression Scale, State-Trait Anxiety Inventory, Beck Anxiety Inventory, and Adult
 20

Manifest Anxiety Scale-Elderly Version). Segal et al. found that the GAS total score had

excellent internal consistency (Cronbach’s α = .93). The internal consistency of the

subscales ranged from good to excellent (Somatic α = .80; Affective α = .82; Cognitive α

= .90). Additionally, the GAS total score and subscale scores correlated significantly

with other measures of anxiety, providing evidence of convergent validity. However, the

authors also found that the GAS significantly correlated with GDS total scores as well,

raising concerns regarding the ability of the GAS to differentiate anxiety from

depression. Within the clinical sample of older adults, Segal et al. found similar internal

consistency coefficients for the GAS total score (α = .93) and the subscales (Somatic α =

.80; Affective α = .82; Cognitive α = .85). Convergent validity was noted in that the

GAS total score and its subscales correlated significantly with each other, as would be

expected. The GAS also demonstrated divergent validity with the Global Assessment of

Functioning (GAF) scale scores from the DSM-IV-TR, such that there were significant

negative correlations noted with the GAS total score, subscales, and the GAF.

Furthermore, neither the GAS total score nor the subscales correlated significantly with

education, additional evidence of discriminant validity. The authors concluded that the

GAS demonstrated strong preliminary evidence for convergent and divergent validity as

well as reliability. The study did not examine the impact of age or sex on measurement

bias. Additionally, the authors noted that the sample sizes were not large enough to

conduct factor analyses to examine the underlying factor structure of the GAS, and thus

the subscales remained conceptually-designed instead of empirically-based. Another

limitation is that medical problems were not assessed, and thus the impact of medical

burden on GAS scores was not known.

 21

Yochim et al. (2011) further examined the psychometric properties of the GAS in

a community-dwelling sample of 117 older adults. The convergent, divergent, and

discriminant validity of the GAS was assessed in comparison to the Beck Anxiety

Inventory and the Geriatric Anxiety Inventory. They also examined the ability to which

the GAS could identify clinically significant anxiety as determined by the cut-scores of

the Beck Anxiety Inventory. Additionally, Yochim et al. examined the impact of medical

burden on GAS scores by utilizing a self-report measure of health conditions. As in

Segal et al. (2010), the GAS was found to have excellent internal consistency (α = .90),

and significantly correlated with other measures of anxiety (evidence of convergent

validity). A noteworthy finding was that the GAS correlated more strongly with the GAI

(r = .69, p < .01) and BAI (r = .61, p < .01) than the GAI and BAI did with each other (r

= .36, p < .01). The GAS correlated weakly with reading ability and processing speed,

suggesting discriminant validity. Similar to Segal et al., the GAS correlated strongly with

measures of depression. Furthermore, those who reported clinically significant anxiety as

determined by the BAI cutoff scores reported significantly more anxiety on the GAS than

those who did not report significant anxiety on the BAI. Furthermore, Yochim et al.

found that the GAS total score correlated with self-reported medical burden, as did the

BAI. Not surprisingly, the Somatic subscale correlated with medical burden most

strongly (r = .38, p < .01), though the Affective and Cognitive subscales correlated with

medical burden as well (r = .22, p < .05 and r = .28, p < .01, respectively). The GAS

total score was not significantly correlated with sex. The correlations between GAS

subscales and sex were not reported. One limitation of the study was that it did not

include a clinical sample and thus the analyses were limited to community-dwelling older
 22

adults. Factor analysis was not conducted with this sample, nor was measurement bias

assessed.

Statement of Problem and Purpose of Study

Anxiety is a significant concern for older adults, and there are a multitude of

challenges unique to assessing anxiety in this population. As such, failure to accurately

detect anxiety in older adults has serious consequences for senior populations. Given the

significant number of older adults who are impacted by either sub-syndromal or clinically

significant anxiety, the need for brief, psychometrically sound assessment tools for use in

this specific population is imperative. Current measures of anxiety have a number of

limitations which may restrict their applicability for use with older adults and increase the

risk of misdiagnosing anxiety in this population, though they are used frequently

(Therrien & Hunsley, 2011). The Geriatric Anxiety Scale was created to address such

limitations, and initial research indicates that it is a promising measure for use with older

people (Segal et al., 2010; Yochim et al., 2011). However, research on the psychometric

properties of the GAS is in its early stages.

The present study sought to further investigate the psychometric properties of the

GAS in three distinct samples of older adults: community-dwelling, clinical, and medical.

The purpose of this study was to identify the psychometric characteristics of the existing

scale and, if needed, suggest modifications to maximize the utility of the measure for use

both clinically and for research purposes. This study is the first to use item response

theory to examine the scale properties of an anxiety measure in an older population. The

present study had the following aims:

1) Assess the reliability and validity of the GAS and its subscales;
 23

2) Examine the underlying factor structure of the GAS (both exploratory and

confirmatory);

3) Examine the ability of the GAS to discriminate anxiety from depression;

4) Examine the extent to which subjective health well-being impacts scores on

the GAS;

5) Identify age, sex, and education differences in GAS scores and item

functioning;

6) Utilize IRT to examine the item properties of the GAS;

7) Determine the extent to which measurement bias impacts item endorsement

on the GAS in regards to age, sex, and education;

8) Utilize item response theory (IRT) to create a short form of the GAS with

adequate psychometric characteristics;

9) Establish descriptive labels for scores on the GAS (mild, moderate, severe) to

assist with score interpretation.

Study 1. The analyses in Study 1 were conducted in a large community-dwelling

sample of older adults. The reliability of the GAS and each subscale was also examined

(Aim 1), and it was hypothesized that the GAS total score and subscales would each yield

sufficient internal consistency (α = .70 or higher). The factor structure of the GAS was

examined to determine whether the items load onto their corresponding subscales (Aim

2). It was hypothesized that principal axis factoring will confirm a three-factor solution,

consistent with the three conceptually derived subscales (Somatic, Cognitive, and

Affective). Convergent validity of each subscale and the total scale score was assessed

(Aim 3), and it was predicted that the GAS total score and each subscale would have
 24

sufficient validity. An additional aim of Study 1 was to examine the ability of the GAS

to discriminate anxiety from depression (Aim 4), and it was expected that items which

discriminate anxiety from depression would be identified. The final aim of Study 1 was

to examine the relationship among the GAS, its subscales, and items with a measure of

subjective health status (Aim 5). It was expected that the GAS and its subscales would

be related to lower ratings of subjective health status, and that the somatic scale and its

items would have the strongest relationships with subjective health ratings.

Study 2. Similar analyses were conducted in Study 2, but this study utilized a

clinical sample of older adults. The reliability of the overall scale and each subscale was

examined (Aim 1). The factor structure of the GAS was examined using principal axis

factoring (Aim 2). It was hypothesized that this analysis would have similar results to

those in Study 1. It was hypothesized that the total scale and each subscale would have

adequate internal consistency (α = .70 or higher). The convergent validity of each

subscale and the total scale score was assessed (Aim 3). It was predicted that the GAS

and each subscale would have adequate convergent validity. This study also examined

the ability of the GAS to differentiate anxiety from depression (Aim 4) and it was

expected that items that sufficiently discriminate anxiety from depression would be

identified. The results from each of the aforementioned analyses were compared to the

results of Study 1 to determine if differences in psychometric properties exist between the

two samples. It was expected that the results will be largely similar as Study 2, although

it was expected that the clinical sample would have elevated scores on the GAS.
 25

Study 3. This study explored the psychometric properties of the GAS in a

medical sample of older adults. The psychometric properties of the GAS and its

subscales were examined in this sample, including reliability (Aim 1). Convergent

validity and divergent validity was also examined (Aim 2). It is hypothesized that the

GAS in this sample will yield sufficient psychometric properties in all analyses. The

relationship between medical burden and GAS scores was also examined (Aim 3). It was

expected that those who experience more functional impairment would endorse more

items on the GAS, especially somatic items.

Study 4. This study combined data from Studies 1, 2 and 3 and employed both

CTT and IRT techniques. Exploratory and confirmatory factor analysis was performed

on the data to determine the unidimensionality criteria for IRT (Aim 1). As the GAS is

intended to be a clinically useful measure of anxiety, it was expected that items should be

able to discriminate individuals with high and low levels of anxiety. It was also expected

the test information curve peak would be above the mean level of anxiety. DIF by age,

sex, and education was assessed (Aim 2). This aim was exploratory and no specific

hypotheses were generated. A short form was created by identifying and retaining the

items which provide the greatest information and have the highest discrimination

parameters, while maintaining the integrity of the subscales (Aim 3). It was expected that

the short form would have adequate reliability and validity and function similar to the full

version of the GAS. Age, sex, and education differences were assessed at the group level

(Aim 4), with the expectation that individuals who are younger, less educated, and female

would score higher on the GAS than individuals who are older, more educated, and male.
 26

An additional aim was to establish descriptive categories (mild, moderate, severe) for the

GAS, subscales, and short form (Aim 5).

 CHAPTER II

STUDY ONE

Method

In this study, two existing datasets from community-dwelling samples of older

adults were combined to create one larger dataset.

Sample 1. Data were collected from 123 older adults over the age of 60.

Participants were volunteers from the community who participated in a larger study of

cognitive functioning and mental health. All participants provided informed consent and

were financially compensated for their time.

Sample 2. Data were collected on 284 community-dwelling older adults

recruited from the El Paso county voter registry. Participants provided informed consent

prior to their participation in the study.

Combined dataset. Merging these datasets yielded a sample of 407 older adults.

Their ages ranged from 60 to 96 (M = 73.78, SD = 7.14), and 57.5% (n = 234) were

female. The majority of participants were European American (n = 361). Participants

were generally well educated (years of education M = 14.98, SD = 2.94). Other

demographic information is presented in Table 1.

 28

Table 1

Means, Standard Deviations, and Ranges for All Demographic Information and All
Measures

N Mean SD Possible Range

Range
Study 1
Age 407 73.78 7.14 - 60-96
Education (Years) 405 14.98 2.95 - 8-25
GAS Total Scale 384 9.18 7.88 0-75 0-48
GAS Cognitive 398 1.92 2.64 0-27 0-17
GAS Affective 395 2.40 2.75 0-24 0-17
GAS Somatic 398 5.07 3.63 0-24 0-20
BAI 121 5.06 5.60 0-63 0-29
GAI 95 1.91 3.17 0-20 0-20
GDS 123 5.69 6.04 0-30 0-29
BDI-II 122 7.75 7.51 0-63
BHS 271 30.74 8.79 20-80 20-70
SISE 279 4.28 .87 1-5 1-5
3LS 280 3.91 1.41 3-9 3-9
SF-36 271 73.59 23.58 0-100 5-100
PHQ-9 276 2.73 3.57 0-27 0-22

Study 2 Age 136 68.60 7.60 - 60-88

Education (Years) 119 14.61 2.84 - 6-28
GAS Total Scale 121 19.95 11.50 0-75 0-54
GAS Cognitive 128 5.91 4.03 0-24 0-17
GAS Affective 127 6.10 3.66 0-24 0-16
GAS Somatic 131 8.01 4.70 0-27 0-26
GDS 99 11.80 8.12 0-30 0-28
GAF 124 64.85 9.38 0-100 41-86

Study 3 Age 38 69.92 7.98 - 60-90

Education (Years) 38 13.95 2.10 - 9-18
GAS Total Scale 38 15.16 11.12 0-75 1-47
GAS Cognitive 38 3.70 4.12 0-27 0-18
GAS Affective 38 4.32 3.88 0-24 0-14
GAS Somatic 38 7.05 4.17 0-24 1-16
BAI 38 10.45 9.07 0-63 0-35
MoCA 38 25.47 2.25 0-30 21-30
GAI 38 4.43 5.70 0-20 0-20
PHQ-9 38 6.84 6.03 0-27 0-21
SF-36 Total Scale 38 56.19 21.05 0-100 3-86
 29

Table 1 Continued

Study 4 Age 581 72.32 7.64 - 60-96

Education (Years) 562 14.83 2.89 - 6-28

GAS Total Scale 541 11.99 10.11 0-75 0-54
GAS Cognitive 562 2.94 3.53 0-24 0-18
GAS Affective 559 3.37 3.43 0-24 0-17
GAS Somatic 565 5.87 4.13 0-27 0-26
GAS-10 563 4.72 4.64 0-30 0-24

Note. GAS = Geriatric Anxiety Scale, BAI = Beck Anxiety Inventory, GAI = Geriatric
Anxiety Inventory, GDS = Geriatric Depression Scale, BDI-II = Beck Depression
Inventory, Second Edition, BHS = Beck Hopelessness Scale, SISE = Single-Item Self-
Esteem Scale, 3LS = Three-Item Loneliness Scale, SF-36 = 36-Item Health Survey, GAF
= Global Assessment of Functioning, PHQ-9 = Patient Health Questionnaire, GAS-10 =
Geriatric Anxiety Scale – 10 Item Version.

Measures

Geriatric Anxiety Scale (GAS). The GAS (Segal et al., 2010) is a self-report

measure of anxiety symptoms designed for use with older adult populations. Participants

are asked to rate symptoms of anxiety or stress by indicating how often they have

experienced each symptom during the past week on a Likert-type scale that ranges from 0

(not at all) to 3 (all of the time). Possible scores range from 0 to 75, with higher scores

indicating the presence of more severe anxiety. The GAS was administered to all older

adults included in the combined community-dwelling sample.

Beck Anxiety Inventory. The BAI (Beck et al., 1988) is a self-report measure

of anxiety intended for use with adults of all ages. It contains a list of 21 symptoms

which are rated from 0 (not at all) to 3 (severe). Possible scores range from 0 to 63, with

higher scores indicating more severe anxiety. Though the BAI is not gero-specific, the

measure has adequate psychometric properties in older adult samples (i.e., Kabacoff et

al., 1997; Wetherell et al., 1997). The BAI was administered to Sample 1 only.
 30

Geriatric Anxiety Inventory. The GAI (Pachana et al., 2007) is a 20-item self-

report assessment tool. Participants are asked to respond yes or no to statements regarding

their experience with anxiety during the past week. The internal consistency of the GAI

is high, as is its convergent validity with other measures (Pachana et al.). Possible scores

range from 0 to 20 with higher scores indicating the presence of more severe anxiety. The

GAI was administered to Sample 1 only.

Geriatric Depression Scale (GDS). The GDS (Yesavage et al., 1983) is a

widely used self-report measure of depressive symptoms. It contains 30 items in which

participants are asked to respond yes or no to each question. Possible scores range from

0-30, with higher scores indicating the presence of more depressive symptoms. The GDS

is a reliable and valid measure of depression in older adults (Yesavage et al.), and has

adequate internal consistency, test-retest reliability, and concurrent validity with diverse

measures of depression in diverse populations (Marty, Pepin, June, & Segal, 2011). The

GDS was administered to Sample 1 only.

Beck Depression Inventory – Second Edition (BDI-II). The BDI-II (Beck et

al., 1996) is a self-report measure containing 21 items which correspond with the DSM

criteria for major depressive disorder. Participants are asked to respond on a 4-point

Likert-type scale, ranging from 0 to 3. Possible scores range from 0 to 63, and higher

scores indicate more severe levels of depression. The BDI-II was administered to Sample

1 only.

Beck Hopelessness Scale (BHS). The BHS is a 20-item self-report measure that

assesses pessimism and hopelessness (Beck, Weissman, Lester, & Trexler, 1974).

Participants respond to questions on a 5-point Likert scale, ranging from 1 (rarely or

 31

none of the time) to 4 (most or all of the time). Higher scores on the BHS indicate greater

hopelessness or frequency of negative expectancies for the self or for the future, with

possible scores ranging from 20 to 80. The BHS has been validated among depressed

older adult outpatients (e.g., Hill, Gallagher, Thompson, & Ishida, 1988) and utilized in

research studies with older adults (i.e., Serrano, Latorre, Gatz, & Montanes, 2004). This

measure was administered to Sample 2 only.

Single-Item Self-Esteem Scale (SISE). The SISE is a one-item scale assessing

self-esteem (Robins, Hendin, & Trzesniewski, 2001). Respondents rate how much they

agree with the statement, “I see myself as someone who has high self-esteem,” on a 5-

point scale ranging from 1 (strongly disagree) to 5 (strongly agree). This measure was

administered to Sample 2.

Three-Item Loneliness Scale (3LS). The 3LS is a 3-item self-report measure of

loneliness (Hughes, Waite, Hawkley, & Cacioppo, 2004). Respondents rate the frequency

of loneliness, ranging from 1 (hardly ever) to 3 (often). Higher scores indicate more

loneliness, with possible scores ranging from 3 to 9. The 3LS has been validated in

research studies with older adults (Hughes et al., 2004). This measure was administered

to Sample 2 only.

Patient Health Questionnaire (PHQ-9). The PHQ-9 is a self-report measure of

depressive symptoms, based on DSM diagnostic criteria for Major Depressive Disorder

(Kroenke, Spitzer, & Williams, 2001). Respondents indicate how often they experienced

each symptom over the previous two weeks on a 4-point scale ranging from 0 (not at all)

to 3 (nearly every day). Higher total scores indicate greater severity of depression, with

possible scores ranging from 0 to 27. The PHQ-9 has demonstrated good reliability and
 32

validity among the general population (Martin, Rief, Klaiberg, & Braehler, 2006). This

measure was administered to Sample 2 only.

RAND 36-Item Health Survey 1.0 (SF-36). The SF-36 is a self-report

questionnaire measuring self-perceived health and functional status (Ware & Sherbourne,

1992). It contains 36 items assessing eight domains of health: 1) limitations in physical

activities due to health problems; 2) limitations in social activities due to physical or

emotional problems; 3) limitations in role obligations due to physical health problems; 4)

pain; 5) mental health; 6) limitations in role activities due to emotional problems; 7)

vitality; and 8) general perceptions of health. Possible scores for each variable range

from 0 to 100, and higher scores indicate better health. The SF-36 is widely used in

epidemiological research, and has demonstrated adequate psychometric properties in

older adult samples (Mishra et al., 2011). Only the Physical Functioning subscale of this

measure was administered to Sample 2.

Procedure

Sample 1. The measures administered to this sample were included in a two

hour battery of cognitive tests and mental health questionnaires. Testing occurred in

either a research lab at the university, or in a testing room at a mental health clinic.

Sample 2. Participants in this sample were mailed a packet of questionnaires

which they were asked to complete and return to the principal investigator. Completion

of the packet took approximately 30 minutes.

 33

Statistical Analyses

All analyses were conducted using PASW 18.0 using an alpha level of .05. The

reliability of the overall scale as well as each subscale was assessed by calculating

Cronbach’s alphas for the GAS total score and each subscale (Aim 1).

For Aim 2, the factorial structure of the GAS was examined by conducting an

exploratory factor analysis (principal axis factoring, PAF) on the merged sample of 407

older adults. Before performing PAF, the suitability of the data for factor analysis was

assessed. The inter-item correlation matrix was examined to identify correlations .30 and

above. Bartlett’s test of sphericity was inspected, with a statistically significant value (p

< .05) required for the analyses to be considered appropriate for the data (Bartlett, 1954).

Additionally, the Kaiser-Meyer-Oklin (KMO) value was examined, with a value of at

least .60 ideal for the analyses (Tabachnick & Fidell, 2007).

Factors with an eigenvalue of at least 1.0 were retained for further analysis. The

scree plot was also examined to assist in determining the ideal number of factors within

the data, testing the hypothesis that the three subscales (Somatic, Affective, and

Cognitive) will yield three corresponding factors. Convergent validity (Aim 3) was

assessed by correlating each subscale with the total scale score and with each other. In

addition, correlations were calculated with the PHQ-9, BHS, SISE, and 3LS.

To determine whether the GAS measures a construct distinct from depression

(Aim 4), PCA was performed on the combined items of the GAS and the PHQ-9 (using

data from Sample 2 only). This strategy was also used by Wetherell and Areán (1997) in

the development of the Beck Anxiety Inventory to determine the discriminant validity of

the BAI. As previous research has indicated that the GAS is highly correlated with
 34

measures of depression (Segal et al., 2010; Yochim et al., 2011), it was expected that

many items will load onto the same factor as PHQ-9 items. Correlations were also

calculated among individual GAS items and the PHQ-9 total score. Items which

correlated less strongly with the PHQ-9 were included in a short-form version of the GAS

which should theoretically be less correlated with measures of depression.

Finally, the relationship of the GAS with self-rated physical functioning was

assessed (Aim 5). The GAS, its subscales, and individual items were correlated with the

SF-36.

Results

Aim 1: Reliability analysis. Cronbach’s alpha was calculated on the total scale

and subscales to assess the reliability of the scale scores (see Table 2). The reliability of

the overall scale was excellent (Cronbach’s α = .91). The reliability of the subscales were

good (Cognitive α = .84, Affective α = .83, Somatic α = .79).

Table 2

Cronbach’s Alpha Coefficients for GAS Total Scale and Subscales

Total Scale Cognitive Affective Somatic
(Items 1-25) Subscale Subscale Subscale
Community Sample (Study 1) .91 .84 .83 .79
N = 398
Clinical Sample (Study 2) .93 .85 .80 .82
N = 136
Medical Sample (Study 3) .94 .89 .87 .79
N = 38
Combined Sample (Study 4) .94 .88 .86 .81
N = 581

Aim 2: Exploratory factor analysis. To examine the factor structure of the

GAS, the 25 items of the measure were subjected to principal axis factoring (PAF). Prior

to performing PAF, the suitability of data for factor analysis was assessed. The sample
 35

size (N = 384) was sufficient for PAF. The Kaiser-Meyer-Oklin value was .91,

exceeding the recommended value of .60 (Tabachnick & Fidell, 2007), and Bartlett’s

Test of Sphericity (Bartlett, 1954) reached statistical significance, supporting the

factorability of the correlation matrix.

PAF revealed the presence of 6 factors with eigenvalues exceeding 1, explaining

34.64%, 6.8%, 5.5%, 4.9%, 4.9%, and 4.1% of the variance, respectively. Collectively,

the factors explained 60.96% of the variance. The screeplot (Figure 1) revealed a clear

break after the first factor, with a much smaller break after the second factor.

Figure 1. Screeplot for principal axis factoring on 25 GAS items (Study 1).

As the factors were significantly correlated with one another, a direct oblimin

rotation was used to assist in the interpretation of the data. The data were analyzed using

both a one and two factor solution. In the one-factor solution, all items significantly

loaded onto the single factor. The interpretation of the two factor solution suggests that

affective and cognitive items tended to load onto the first factor, and somatic items

loaded onto the second factor (see Table 3). There was medium positive correlation
 36

between the two factors (r = .59). The results of this analysis suggest that the GAS in

this community-dwelling sample may be interpreted as one-dimensional, but a two-factor

solution also is also appropriate for the data. Although the results do not support the

theoretically-derived three-factor hypothesis, the two-factor solution supports the

conceptual basis for the subscales.

Table 3

Principal Axis Factoring with Direct Oblimin Rotation in Study 1

Factor 1 Factor 2. Commun

“Cognitive and “Somatic” -ality
Affective”
Item 1 (My heart raced or beat strongly.) .22 .14
Item 2 (My breath was short.) .34 .19
Item 3 (I had an upset stomach.) .36 .16
Item 4 (I felt like things were not real or like I was outside of .64 .29
myself.)
Item 5 (I felt like I was losing control.) .82 .55
Item 6 (I was afraid of being judged by others.) .42 .27
Item 7 (I was afraid of being humiliated or embarrassed.) .41 .25
Item 8 (I had difficulty falling asleep.) .69 .46
Item 9 (I had difficulty staying asleep.) .71 .45
Item 10 (I was irritable.) .37 .36
Item 11 (I had outbursts of anger.) .32 .20
Item 12 (I had difficulty concentrating.) .48 .47
Item 13 (I was easily startled or upset.) .49 .35
Item 14 (I was less interested in doing something I typically enjoy.) .51 .39
Item 15 (I felt detached or isolated from others.) .61 .46
Item 16 (I felt like I was in a daze.) .65 .39
Item 17 (I had a hard time sitting still.) .33 .17
Item 18 (I worried too much.) .60 .48
Item 19 (I could not control my worry.) .69 .56
Item 20 (I felt restless, keyed up, or on edge.) .67 .53
Item 21 (I felt tired.) .61 .51
Item 22 (My muscles were tense.) .43 .42
Item 23 (I had back pain, neck pain, or muscle cramps.) .58 .34
Item 24 (I felt like I had no control over my life.) .75 .57
Item 25 (I felt like something terrible was going to happen to me.) .51 .26
Note. Loadings less than .30 were suppressed.

Aim 3: Convergent and divergent validity. Convergent validity indicates that

the measure of interest should correlate moderately or highly with the other theoretically

similar constructs. Divergent validity indicates that the measure of interest should not
 37

correlate with measures of theoretically dissimilar constructs. In the current study, the

GAS and its subscales were expected to correlate with each other, the PHQ-9, BHS,

SISE, and the 3LS. All correlations are presented in Table 4. As predicted, GAS total

scale was correlated with the Cognitive subscale (r = .88, p < .01, 77% variance shared),

Affective subscale (r = .89, p < .01, 79% variance shared), and Somatic subscale (r = .88,

p < .05, 77% variance shared). Additionally, the correlation between the Cognitive and

Affective subscales was stronger (r = .77, p < .01, 59% variance shared) than the

correlation between the Cognitive and Somatic subscale (r = .63, p < .01, 40% variance

shared) and between the Affective and Somatic subscale (r = .63, p < .01, 40% variance

shared). This supports the results from the PAF analysis, indicating that the affective and

cognitive items are more conceptually similar than they are with the somatic items.

Table 4

Correlations between Demographic Variables, GAS, PHQ-9, BHS, SISE, 3LS, and SF-36
Physical Functioning Subscale in Study 1
Sex Edu GAS GAS GAS GAS PHQ- BHS SISE 3LS SF-36
Cog Aff Som 9

Age -.16** -.02 .00 -.03 -.04 .04 .09, -.06, -.02, -.41*
Sex - .26** -.05 -.10* -.01 -.07 -.08 -.04 .12* -.12 .13*
Edu - - -.11* -.13** -.12* -.08 -.06 -.14* .10 -.09 .11
GAS - - - .88** .89** .88** .74** .59** -.47** .50** -.27**
Cog - - - - .77** .63** .70** .63** -.41** .49** -.24**
Aff - - - - - .63** .60** .53** -.45** .49** -.14*
Som - - - - - - .63** .45** -.38** .42** -.27**

Note. Edu = Education (Years), GAS = Geriatric Anxiety Scale, Cog = Geriatric Anxiety
Scale – Cognitive subscale, Aff = Geriatric Anxiety Scale – Affective Subscale, Som =
Geriatric Anxiety Scale – Somatic Subscale, PHQ-9 = Patient Health Questionnaire, BHS
= Beck Hopelessness Scale, SISE = Single-Item Self-Esteem Scale, 3LS = Three-Item
Loneliness Scale, SF-36 = 36-Item Health Survey Physical Functioning Subscale.
Gender coded 0 = female, 1 = male.
 38

The GAS total scale was correlated with the PHQ-9 (r = .74, p < .001, 55%

variance shared), as was the Cognitive subscale (r = .70, p < .001, 49% variance shared),

Affective (r =.60, p < .001, 36% variance shared), and Somatic (r = .63, p < .001, 40%

variance shared), indicating that individuals who reported more anxiety also reported

more depressive symptoms. The GAS was also correlated with the BHS (r = .59, p <

.001, 35% variance shared), as were the subscales (Cognitive, r = .63, p < .001, 40%

variance shared; Affective, r = .53, p < .001, 28% variance shared; Somatic, r = .45, p <

.001, 20% variance shared), indicating that those who reported more anxiety also

reported more feelings of hopelessness. The GAS total scale was correlated with the

3LS, such that those with more anxiety also reported more feelings of loneliness (GAS

total scale, r = .50, p < .001, 25% variance shared, Cognitive, r = .49, p < .001, 24%

variance shared, Affective, r = .49, p < .001, 24% variance shared; Somatic, r = .42, p <

.001, 18% variance shared). Furthermore, the GAS and its subscales was negatively

correlated with the SISE, indicating that participants with greater self-esteem reported

less anxiety (GAS total scale, r = -.47, p < .001, 22% variance shared; Cognitive, r = -.41,

p < .001, 17% variance shared; Affective = -.45, p < .001, 20% variance shared; Somatic,

r = -.38, p < .001, 14% variance shared). All correlations were in the expected directions

and were moderate to strong in magnitude (Cohen, 1992).

The GAS total score and its subscales was not significantly correlated with age

(total scale r = -.02, p = .72, Cognitive r = .00, p = .98, Affective r = -.03, p = .57,

Somatic r = -.04, p = .47). The GAS total score, Affective subscale and Somatic subscale

were all not significantly correlated with sex (total scale r = -.05, p = .32, Affective r = -

.01, p = .87, Somatic r = -.07, p = .19). In contrast, the Cognitive subscale was
 39

significantly but weakly correlated with sex (r = .10, p < .05, 1% variance shared), such

that women were more likely to report cognitive anxiety symptoms. The GAS total scale

was significantly and negatively correlated with education (r = -.11, p < .05, 1% variance

shared), as were the Cognitive (r = -.13, p < .01, 2% variance shared) and Affective (r = -

.12, p < .05, 1% variance shared) subscales. This indicates those with less education

were more likely to endorse anxiety symptoms. The Somatic subscale was not

significantly correlated with education (r = -.08, p = .11). The relationships among these

variables are evidence of divergent validity.

Aim 4: Relationship with depression. Next, correlations were computed

between each individual item on the GAS with the PHQ-9 to determine which GAS items

correlated with depression most strongly (see Table 5). As can be seen, all 25 items

significantly correlated with the PHQ-9 total score.

The following eight items had correlations equal to or stronger than .50: Item 5 (I

felt like I was losing control; r = .50, p < .01), Item 14 (I was less interested in doing

something I typically enjoy, r = .50, p < .01), Item 15 (I felt detached or isolated from

others; r = .59, p < .01), Item 16 (I felt like was in a daze, r = .55, p < .01), Item 19 (I

could not control my worry; r = .54, p < .01), Item 20 (I felt restless, keyed up, or on

edge, r = .53, p < .01), Item 21 (I felt tired; r = .56, p < .05), and Item 24 (I felt like I had

no control over my life, r = .68, p < .01). With the exception of Item 21, all items were

from the Affective or Cognitive subscales.

Next, a version of the GAS which correlates less strongly with depression was

created (GAS-8). The correlations between individual GAS items and the PHQ-9 were

calculated, and the eight items with the correlations below .40 were selected for this
 40

Table 5

Correlations Between GAS items and PHQ-9 in Study 1 (N = 275)

GAS Item Correlation

with PHQ-
9
Item 1 (My heart raced or beat strongly.) .22**
Item 2 (My breath was short.) .31**
Item 3 (I had an upset stomach.) .27**
Item 4 (I felt like things were not real or like I was outside of myself.) .38**
Item 5 (I felt like I was losing control.) .50**
Item 6 (I was afraid of being judged by others.) .26**
Item 7 (I was afraid of being humiliated or embarrassed.) .14*
Item 8 (I had difficulty falling asleep.) .46**
Item 9 (I had difficulty staying asleep.) .43**
Item 10 (I was irritable.) .36**
Item 11 (I had outbursts of anger.) .24**
Item 12 (I had difficulty concentrating.) .44**
Item 13 (I was easily startled or upset.) .40**
Item 14 (I was less interested in doing something I typically enjoy.) .50**
Item 15 (I felt detached or isolated from others.) .59**
Item 16 (I felt like I was in a daze.) .55**
Item 17 (I had a hard time sitting still.) .24**
Item 18 (I worried too much.) .48**
Item 19 (I could not control my worry.) .54**
Item 20 (I felt restless, keyed up, or on edge.) .53**
Item 21 (I felt tired.) .56*
Item 22 (My muscles were tense.) .49**
Item 23 (I had back pain, neck pain, or muscle cramps.) .42**
Item 24 (I felt like I had no control over my life.) .68**
Item 25 (I felt like something terrible was going to happen to me.) .42**
Note. PHQ-9 = Patient Health Questionnaire. *p < .05, **p < .01.

scale. This version of the GAS is presented in Appendix C. Correlations were then

calculated among this version of the GAS and the PHQ-9 (r = .46, p < .01), Geriatric

Depression Scale (r = .61, p < .01, N = 122), and Beck Depression Inventory-II (r = .54, p

< .01, N = 122). Although still significantly correlated, the magnitude of these

correlations was less strong than the correlations between the GAS total scale and
 41

measures of depression. The reliability of the GAS-8 was acceptable (Cronbach’s alpha =

.76).

A Principal Components Analysis (PCA) was performed on the combined GAS

and PHQ-9 items to examine the underlying component structure of the GAS items in

relation to the PHQ-9 items. Prior to conducting the analysis, the data were deemed

suitable as determined by the sample size (N = 258), the KMO test (.90), and Bartlett’s

Test of Sphericity (p = .00). As it was hypothesized that there would be two components

(depression and anxiety), two factors were forced on the data. Overall, there were 7

components with eigenvalues greater than 1.0, with the first component explaining the

greatest percentage of variance (34.52%). Inspection of the component matrix revealed

that all GAS and PHQ-9 items loaded onto the first component, and several loaded onto

both components. The second component contained items from both the GAS and PHQ-

9 that were more somatic in nature. Overall, the GAS and PHQ-9 items appear to be one-

dimensional in nature, perhaps because both measures assess affective distress.

Furthermore, several items on both measures are similar in nature (i.e., difficulty

concentrating, sleep, loss of interest in activities).

Aim 5: Relationship with physical functioning. To assess the degree to which

the GAS and its subscales were related to self-reported physical functioning, correlations

were conducted with the physical functioning subscale of the SF-36. The GAS total scale

was significantly negatively correlated with the SF-36 (r = -.27, p < .001, 7% variance

shared), indicating that those with more anxiety rated their physical functioning as lower.

The subscales were also negatively correlated with the SF-36: Cognitive (r = -.24, p <

.001, 6% variance shared), Affective (r = -.14, p <.05, 2% variance shared), Somatic (r =

 42

-.27, p < .001, 7% variance shared). Correlations among individual GAS items and the

SF-36 total score are presented in Table 6. As can be seen, there were several items

which correlated significantly with the SF-36, although the magnitude of these

correlations were weak. Item 27 (I was concerned about my health) was significantly

correlated with the SF-36 (r = -.37, p < .001, 14% variance shared), indicating those with

more concerns in this regard had poorer self-rated physical functioning.

Summary of results. Overall, the GAS appears to have excellent reliability in

this community-dwelling sample, and the subscales had good reliability. An exploratory

factor analysis revealed that a two-factor solution is appropriate for the data, although the

GAS can also be considered one-dimensional. The first factor contains mainly affective

and cognitive items, whereas the second factor contains somatic items. The GAS

correlated moderately with variables related to anxiety (hopelessness, loneliness, and

self-esteem), indicative of convergent validity. The GAS had strong relationships with a

measure of depression. All individual items on the GAS are correlated with a measure of

depression, with 8 of items exceeding correlations of .50. This indicates substantial

overlap between anxiety as measured by the GAS, and depression as measured by the

PHQ-9. Principal components analysis revealed that the GAS and PHQ-9 items were

one-dimensional in nature. The GAS was negatively correlated with a measure of

subjective physical functioning, indicating that those with more poor self-rated health

experienced more anxiety.

 43

Table 6

Correlations among GAS Items and SF-36 Physical Functioning Subscale Scores in
Study 1 (N = 270)

GAS Item Correlation

with SF-36
Physical
Functioning

Item 1 (My heart raced or beat strongly.) -.14*

Item 2 (My breath was short.) -.36**
Item 3 (I had an upset stomach.) .00
Item 4 (I felt like things were not real or like I was outside of myself.) -.13*
Item 5 (I felt like I was losing control.) -.15*
Item 6 (I was afraid of being judged by others.) -.06
Item 7 (I was afraid of being humiliated or embarrassed.) .03
Item 8 (I had difficulty falling asleep.) -.12
Item 9 (I had difficulty staying asleep.) -.11
Item 10 (I was irritable.) -.13*
Item 11 (I had outbursts of anger.) -.01
Item 12 (I had difficulty concentrating.) -.18**
Item 13 (I was easily startled or upset.) -.11
Item 14 (I was less interested in doing something I typically enjoy.) -.14*
Item 15 (I felt detached or isolated from others.) -.12*
Item 16 (I felt like I was in a daze.) -.27**
Item 17 (I had a hard time sitting still.) .00
Item 18 (I worried too much.) -.13*
Item 19 (I could not control my worry.) -.17**
Item 20 (I felt restless, keyed up, or on edge.) -.15*
Item 21 (I felt tired.) -.23**
Item 22 (My muscles were tense.) -.24**
Item 23 (I had back pain, neck pain, or muscle cramps.) -.29**
Item 24 (I felt like I had no control over my life.) -.19**
Item 25 (I felt like something terrible was going to happen to me.) -.20**
Content Item 1 (I was concerned about my finances.) -.16**
Content Item 2 (I was concerned about my health.) -.37**
Content Item 3 (I was concerned about my children.) -.19**
Content Item 4 (I was afraid of dying.) -.12*
Content Item 5 (I was afraid of becoming a burden to my family or -.26**
children).

Note. SF-36 = Short-Form Health Survey. *p < .05, **p < .01.
 CHAPTER III

` STUDY TWO

Method

Data were collected on 136 adults above the age of 60 receiving psychological

services at a local community outpatient mental health clinic. Seventy-nine percent of

the sample was female, and 88% identified themselves as European American. The

remainder of the sample reported their ethnicity as Hispanic (N = 10), African American

(N = 1), Native American (N = 1), or other (N = 2). Other demographic information

about participants in this sample is presented in Table 1. Individuals at this clinic

typically sought services for concerns relating to caregiving, depression, anxiety,

personality disorders, and various other psychosocial challenges.

Measures

Geriatric Anxiety Scale (GAS). See Study 1 for a description of this measure.

Geriatric Depression Scale (GDS). See Study 1 for a description of this

measure.

Global Assessment of Functioning (GAF). The GAF is from Axis V of the

Diagnostic and Statistical Manual of Mental Disorders. It is typically administered by

clinicians to indicate the level of functioning of the client. Possible scores range from 0-
 45

100, with higher scores indicating superior levels of functioning (i.e., minimal

symptoms) and lower scores indicating severe symptoms which interfere with daily life

(i.e., severe suicidality).

Procedure

Participants indicated at intake whether they consent to have their data be used for

research purposes. The GAS and GDS were administered at the time of intake in

addition to other paperwork required by the clinic. The intake clinician determined the

GAF score for each client as well as provisional diagnoses on Axis I and II.

Statistical Analyses

The analyses conducted in this sample were similar to the analyses conducted in

Study 1. The reliability of the overall scale as well as each subscale was assessed by

calculating Cronbach’s alphas for the GAS total score and each subscale (Aim 1). For

Aim 2, the factor structure of the GAS was examined by conducting principal axis

factoring (PAF) on the data. Convergent validity was assessed by correlating each of the

subscales with the total scale score and with each other, as well as correlating the total

score and each subscale with GDS and GAF scores (Aim 3). To further analyze the

relationship between the GAS and depression, individual GAS items were also correlated

with the GDS total score.

Results

Aim 1: Reliability analysis. Cronbach’s alpha was calculated on the total scale

and subscales to assess the reliability of the measure (see Table 2). The reliability of the

overall scale was excellent (Cronbach’s α = .93). The reliability of the subscales were
 46

good (Cognitive α = .85, Affective α = .80, and Somatic α = .82, respectively). These

results are similar to those from Study 1.

Aim 2: Exploratory factor analysis. To examine the factor structure of the

GAS in a clinical sample of older adults (N = 136), the 25 items of the measure were

subjected to principal axis factoring (PAF). Cases were considered pair-wise to address

missing data. The sample size is lower than what is recommended for factor analysis, so

the results should be interpreted with caution bearing this in mind. Inspection of the

correlation matrix revealed several correlations below .30. The Kaiser-Meyer-Oklin

value was .87 exceeding the recommended value of .60 (Tabachnick & Fidell, 2007), and

Bartlett’s Test of Sphericity (Bartlett, 1954) reached statistical significance, supporting

the factorability of the correlation matrix.

PAF revealed the presence of 6 factors with eigenvalues exceeding 1, explaining

36.75%, 6.92%, 6.21%, 5.36%, 4.68%, and 4.21% of the variance, respectively.

Collectively, the factors explained 64.14% of the variance. The screeplot (Figure 2)

revealed a clear break after the first factor.

Figure 2. Screeplot for principal axis

factoring on 25 GAS items (Study 2).
 47

A direct oblimin rotation was used to assist in the interpretation of the data. Upon

inspection of the factor loadings, it was apparent that a one-factor solution was the best fit

for the data. When a two- or three-factor solution was forced, the majority of the items

loaded onto the first factor, and items that loaded onto the second or third factors were

very small in magnitude.

Aim 3: Validity. To gauge the level of convergent validity, correlations were

computed between GAS total scores with each GAS subscale, and among the subscales

with each other. Correlations are presented in Table 7. As predicted, GAS total scale was

correlated with the Cognitive subscale (r = .91, p < .001, 83% variance shared), Affective

subscale (r = .91, p < .001, 83% variance shared), and Somatic subscale (r = .92, p <

.001, 85% variance shared). Additionally, the Cognitive and Affective subscales were

correlated with each other (r = .77, p < .001, 59% variance shared) as were the Cognitive

and Somatic subscales (r = .72, p < .001, 52% variance shared) and the Affective and

Somatic subscales (r = .75, p < .001, 56% variance shared). These results are similar to

those from Study 1, although the magnitude of the correlations between the subscales

appears to be higher.

Correlations were also calculated with the GAS, subscales, and GAF scores.

Higher GAF scores indicate better functioning. Correlations are presented in Table 8.

The GAS total scale had a negative correlation with the GAF score (r = -.34, p < .001,

12% variance shared), as did its subscales (Cognitive r = -.33, p < .001, 11% variance

shared; Affective r = -.33, p < .001, 11% variance shared; Somatic r = -.28, p < .01, 8%

variance shared). The magnitudes of these correlations were small-to-medium (Cohen,

1992), and indicate that those who reported more anxiety symptoms were rated by their
 48

Table 7

Correlations among Demographic Variables, GAS, Subscales, GDS, and GAF in Study 2
(N = 136)

Sex Edu GAS GAS GAS GAS GDS GAF

Cog Aff Som
Age -.07 -.33** -.18 -.21* -.14 -.12 -.09 -.02
Sex - -.05 .11 .06 .10 .11 -.06 -.10
Edu - - .09 .01 .11 .10 -.08 .17
GAS - - - .91** .91** .92** .67** -.34**
GAS - - - - .77** .72** .62** -.33**
Cog
GAS - - - - - .75** .66** -.33*
Aff
GAS - - - - - - .57** -.28**
Som
GDS - - - - - - - -.54**

Note. Edu = Education (Years), GAS = Geriatric Anxiety Scale, GAS Cog = Geriatric
Anxiety Scale – Cognitive subscale, GAS Aff = Geriatric Anxiety Scale – Affective
Subscale, GAS Som = Geriatric Anxiety Scale – Somatic Subscale, GDS = Geriatric
Depression Scale, GAF = Global Assessment of Functioning.

clinician as having overall lower well-being. This provides evidence of convergent

validity, such that the GAS is related to a concept similar to anxiety (psychological well-

being) in an inverse relationship, as expected.

The GAS was not significantly correlated with age (r = -.18, p = .06). The

Cognitive subscale was significantly negatively correlated with age (r = -.21, p < .05, 4%

variance shared), but the Affective and Somatic subscales were not (r = -.14, p = .12 and

r = -.12, p = .17, respectively). The GAS and its subscales were not significantly

correlated with sex or education.

 49

Aim 4: Relationship with depression. Correlations were computed between the

GAS, subscales, and GDS. The GAS total scale was significantly correlated with the

GDS (r = .67, p < .001, 45% variance shared), as were the subscales (Cognitive r = .62, p

< .001, 38% variance shared; Affective r = .66, p < .001, 44% variance shared; Somatic r

= .57, p < .001, 32% variance shared), such that those who reported higher anxiety

symptoms also reported more depressive symptoms. Though moderate-to-strong, the

magnitudes of these correlations were slightly lower than those reported in Study One.

However, it should be noted that Study 2 utilized a different measurement of depression

which may account for these discrepancies.

Next, correlations were calculated among the individual GAS items and the GDS

total scale score to determine which items were most related to depression. As can be

seen in Table 8, all 25 items had significant correlations with the GDS total scale score.

Four items had correlations with magnitudes greater than .50: Item 12 (r = .57, p < .001),

Item 14 (r = .63, p < .001), Item 15 (r = .65, p < .001), and Item 24 (r = .58, p < .001),

and these items were from the Cognitive and Affective subscales. In comparison to

Study 1, the magnitude of the correlations overall were smaller. A lower sample size in

Study 2 and a different measure of depression used may account for these discrepancies.

However, there was substantial overlap in both studies in regards to the GAS and

measures of depression.

Summary of results. PAF indicates that a one-factor solution is most appropriate

in a clinical sample, although the results from this analysis should be interpreted in light

of a lower sample size than in Study 1. As in the community-dwelling sample, the GAS

had excellent reliability, and the reliability of the subscales was good.
 50

Table 8

Correlations among GAS items and GDS in Study 2 (N = 99)

GAS Item Correlation with GDS

total scale
Item 1 (My heart raced or beat strongly.) .23*
Item 2 (My breath was short.) .33**
Item 3 (I had an upset stomach.) .33**
Item 4 (I felt like things were not real or like I was outside of myself.) .28**
Item 5 (I felt like I was losing control.) .38**
Item 6 (I was afraid of being judged by others.) .39**
Item 7 (I was afraid of being humiliated or embarrassed.) .41**
Item 8 (I had difficulty falling asleep.) .42**
Item 9 (I had difficulty staying asleep.) .39**
Item 10 (I was irritable.) .28**
Item 11 (I had outbursts of anger.) .21*
Item 12 (I had difficulty concentrating.) .57**
Item 13 (I was easily startled or upset.) .46**
Item 14 (I was less interested in doing something I typically enjoy.) .63**
Item 15 (I felt detached or isolated from others.) .66**
Item 16 (I felt like I was in a daze.) .44**
Item 17 (I had a hard time sitting still.) .30**
Item 18 (I worried too much.) .41**
Item 19 (I could not control my worry.) .45**
Item 20 (I felt restless, keyed up, or on edge.) .34**
Item 21 (I felt tired.) .41**
Item 22 (My muscles were tense.) .38**
Item 23 (I had back pain, neck pain, or muscle cramps.) .49**
Item 24 (I felt like I had no control over my life.) .58**
Item 25 (I felt like something terrible was going to happen to me.) .37**
Content Item 1 (I was concerned about my finances.) .55**
Content Item 2 (I was concerned about my health.) .48**
Content Item 3 (I was concerned about my children.) .21*
Content Item 4 (I was afraid of dying.) .25*
Content Item 5 (I was afraid of becoming a burden to my family or .42**
children).
Note. GDS = Geriatric Depression Scale. *p < .05, **p < .01.

The GAS and its subscales demonstrated adequate convergent validity in its relationships

with a measure of overall psychological well-being. The GAS, subscales, and the
 51

majority of items significantly correlated with depression as measured by the GDS, but to

a slightly lesser degree than in the community-dwelling sample.

 CHAPTER IV

STUDY THREE

Method

Data were collected from 38 older adults aged 60 or above with at least one

chronic physical health condition (M number of health conditions = 3.36, SD = 1.67,

range = 1-7). In this study, chronic physical health condition was defined as a physical

health condition lasting longer than three months and was active at the time of testing.

The most common self-reported health conditions included: high blood pressure (N = 17),

high cholesterol (N = 9), heart disease/history of heart attack (N = 8) diabetes (N = 7),

arthritis (N = 8), and hypothyroidism (N = 5). Less commonly reported conditions

included Parkinson’s disease (N = 1), multiple sclerosis (N = 2), and Lyme disease (N =

1). This sample was 82% female and 90% European American, with the remaining 10%

identifying themselves as either African American or Hispanic. Other demographic

information from this sample is presented in Table 1. Participants were recruited via

flyers placed in primary care clinics for seniors (N = 9) as well as the Colorado Springs

Senior Center (N = 11). Participants were also recruited via snowball sampling (N = 2)

and the Gerontology Center database (N = 9). Additionally, three participants were

recruited from an intensive outpatient day treatment center for medically frail older

adults, and three participants were recruited from the independent living section of an

assisted living facility.

 53

Measures

Montreal Cognitive Assessment (MoCA). The MoCA is a brief screen for

cognitive impairment (Nasreddine et al., 2005). It contains several items assessing

visuospatial/executive skills, attention, abstraction, delayed verbal memory, verbal

fluency, language, and orientation. Possible scores range from 0-30, with higher scores

reflecting better cognitive functioning. Scores below 26 indicate cognitive impairment.

The MoCA has good internal consistency, test-retest reliability, and has excellent

sensitivity and specificity in identifying mild cognitive impairment and Alzheimer’s

disease (Nasreddine et al.). In the current study, MoCA scores ranged from 21 to 30 (M

= 25.47, SD = 2.25), indicating that participants had, on average, slight cognitive

impairment.

Geriatric Anxiety Scale (GAS). See Study 1 for a description of this measure.

BAI. See Study 1 for a description of this measure.

GAI. See Study 1 for a description of this measure.

Patient Health Questionnaire (PHQ-9). See Study 1 for a description of this

measure.

RAND 36-Item Health Survey (SF-36). See Study 1 for a complete description

of this measure. All subscales were administered to this sample. The mean SF-36 total

scale score was 56.19 (SD = 21.05), with scores ranging from 3-86. This value is lower

than average for a general older adult sample (Mishra et al.).

Procedure

All participants provided informed consent prior to their participation in the study.

The testing was conducted at a university research center. The participants recruited
 54

from the day treatment center and the assisted living facility were tested in private, quiet

rooms in each of the respective sites. Participants were first administered the MoCA and

then completed the packet of mental health questionnaires. Basic demographic

information as well as information regarding health conditions was also collected.

Testing took about 30-45 minutes, and participants were compensated with $10 for their

time.

Statistical Analyses

The reliability of the GAS and subscales in this sample was assessed by

examining Cronbach’s alpha (Aim 1). Convergent validity was assessed by correlating

the GAS with the BAI, GAI, and PHQ-9 (Aim 2). Additional validity analyses were

conducted by correlating the GAS with the Mental Health subscale of the SF-36. All

analyses were conducted with the GAS total score and all subscales. Each subscale was

correlated with each other as well as the total score. To analyze the extent to which the

presence of medical conditions and functional impairment impact scores on the GAS

(Aim 3), correlations were conducted among the GAS, subscales, SF-36 total score, and

SF-36 subscale scores.

Results

Aim 1: Reliability analysis. The reliability of the total scale in this sample was

excellent (Cronbach’s alpha = .94). The reliability of the subscales was good (Cognitive

= .89, Affective = .87, Somatic = .79). Cronbach’s alpha coefficients are presented in

Table 2. These values are largely comparable to Studies 1 and 2.

Aim 2: Validity. To test the convergent validity of the GAS in this sample,

correlations were calculated among the GAS, subscales, BAI, and GAI. All correlations
 55

are presented in Table 9. The GAS total scale was correlated with the Cognitive (r = .93,

p < .001, 86% variance shared), Affective (r = .93, p < .001, 86% variance shared), and

Somatic (r = .87, p < .001, 76% variance shared). The Cognitive subscale was correlated

with the Affective and Somatic subscales (r = .87, p < .001, 76% variance shared, and r =

.71, p < .001, 50% variance shared, respectively). The Affective and Somatic subscales

were correlated with each other (r = .68, p < .001, 46% variance shared).

The GAS was significantly and positively correlated with the BAI (r = .73, p <

.001, 53% variance shared), as were the subscales: Cognitive (r = .73, p < .001, 53%

variance shared), Affective (r =.66, p < .001, 44% variance shared), Somatic (r = .61, p <

.001, 37% variance shared). The GAS was also significantly and positively correlated

with the GAI (r = .82, p < .001, 67% variance shared), as were the subscales: Cognitive

(r = .85, p < .001, 72% variance shared), Affective (r = .80, p < .001, 64% variance

shared), Somatic (r = .60, p < .001, 36% variance shared). These correlations were in the

expected directions, and provide evidence of convergent validity.

The GAS was not significantly correlated with age (r = -.18, p = .28), sex (r = -

.15, p = .39), or education (r = -.17, p = .33). The subscales were not significantly

correlated with age (Cognitive: r = -.18, p = .28; Affective: r = -.14, p = .40; Somatic: r =

-.15, p = .39), sex (Cognitive: r = -.11, p = .51; Affective: r = -.13, p = .44; Somatic: r = -

.20, p = .23), or education (Cognitive: r = -.18, p = .28; Affective: r = -.18, p = .27;

Somatic: r = -.09, p = .60).

The MoCA was not significantly correlated with the GAS (r = -.19, p = .27) or

subscales: Cognitive (r = -.24, p = .15), Affective (r = -.18, p = .30), Somatic (r = -.10, p

= .55). This provides evidence of divergent validity, as the GAS was not significantly
 56

Table 9

Correlations among Demographic Variables, GAS, MoCA, GAI, BAI, PHQ-9, and SF-36
total Scale Scores in Study 3 (N = 38)

Sex Ed GAS Cog Aff Som GAI MoC BAI PHQ-9 SF-36
u A
Age -.17 .05 -.18 -.18 -.14 -.15 -.13 -.15 .17 -.26 -.15
Sex - -.05 -.15 -.11 -.13 -.20 -.17 -.22 -.03 -.11 .42*
Edu - - -.17 -.18 -.18 -.09 -.19 -.03 -.23 -.26 .09
GAS - - - .94** .93** .87** .82** -.19 .73** .84** -.68**
Cog - - - - .87** .71** .85** -.24 .73** .80** -.60**
Aff - - - - - .68** .80** -.18 .66** .83** -.62**
Som - - - - - - .61** -.10 .61** .67** -.66**
GAI - - - - - - - -.26 .70** .83** -.57**
MoC - - - - - - - - -.41** -.24 .26
A
BAI - - - - - - - - .71** -.60**
PHQ- - - - - - - - - - -.73**
9
Note. Edu = Education (Years), GAS = Geriatric Anxiety Scale, Cog = Geriatric Anxiety
Scale – Cognitive subscale, Aff = Geriatric Anxiety Scale – Affective Subscale, Som =
Geriatric Anxiety Scale – Somatic Subscale, GAI = Geriatric Anxiety Inventory, MoCA
= Montreal Cognitive Assessment, BAI = Beck Anxiety Inventory, PHQ-9 = Patient
Health Questionnaire, SF-36 = Medical Outcomes Survey Short Form Total Scale. Sex
coded 0 = female, 1 = male.
*p < .05, **p < .01.

related to global cognitive functioning, a construct that is theoretically distinct from

anxiety.

To examine the relationship with depression, correlations were calculated

between the GAS and the PHQ-9. The GAS total scale was significantly correlated with

the PHQ-9 (r = .84, p < .001, 71% variance shared), as were its subscales: Cognitive (r =

.80, p < .001, 64% variance shared), Affective (r = .83, p < .001, 69% variance shared),

and Somatic (r = .66, p < .001, 44% variance shared). These correlations indicate that

those who reported more anxiety symptoms also reported more depressive symptoms. As
 57

in Studies 1 and 2, the GAS had substantial overlap with depression as measured by the

PHQ-9. The magnitude of the correlation between the PHQ-9 and the Somatic subscale

was lower than the correlations between the Cognitive and Affective subscales with the

PHQ-9.

Aim 3: Relationship with health and functional status. To assess the degree

to which GAS scores were related to subjective health status, correlations were calculated

among the GAS, SF-36 total score, and SF-36 subscales. On all SF-36 subscales, higher

scores reflect better subjective health. These correlations are presented in Table 10. For

purposes of comparison, correlations among the BAI, GAI, PHQ-9 and SF-36 scores are

also included in Table 10.

Table 10

Correlations among GAS, Subscales, BAI, GAI, PHQ-9, and SF-36 Subscales in Study 3

Physical Role Social Body Genera Role Vitality General

Functionin Limitation Functioning Pain l Limitations Health
g s Mental due to Perceptions
Health Emotional
Problems

GAS -.13 -.64** -.75** -.53** -.89** -.59** -.63** -.59**

Cognitive -.04 -.61** -.70** -.42** -.83** -.58** -.48* -.55**
Affective -.12 -.60** -.68** -.49* -.87** -.56** -.61** -.41*
Somatic -.20 -.55** -.69** -.55** -.73** -.44* -.64** -.66**
BAI -.19 -.57** -.60** -.34** -.67** -.53** -.49** -.51**
GAI -.07 -.56** -.69** -.37** -.80** -.58** -.45** -.48**
PHQ-9 -.17 -.71** -.74** -.46** -.83** -.70** -.74** -.59**

*p < .05, **p < .01.

 58

The GAS total scale was significantly correlated with the SF-36 total scale (r = -

.68, p < .001, 46% variance shared), as were the subscales: Cognitive (r = -.60, p < .001,

38% variance shared), Affective (r = -.62 p < .001, 40% variance shared), Somatic (r = -

.65, p < .001, 42% variance shared). These correlations indicate that those who endorsed

more symptoms of anxiety also rated their overall subjective health status as poorer. The

total SF-36 score includes items from the General Mental Health and Role Limitations

due to Emotional Problems subscales. These items were removed and the total SF-36

score was recalculated. The magnitude of these correlations decreased with the mental

health items removed: total score (r = -.57, p < .001, 32% variance shared), Cognitive (r

= -.48, p < .01, 23% variance shared), Affective (r = -.50, p < .01, 25% variance shared),

Somatic (r = -.59, p < .001, 35% variance shared).

The Physical Functioning subscale was not significantly correlated with the GAS

total scale (r = -.13, p = .47). The GAS was significantly correlated with the General

Health Perceptions subscale (r = -.59, p < .001, 35% variance shared), indicating those

who rated their general health as lower also reported more anxiety. The GAS was

significantly correlated with the Role Limitations (r = -.64, p < .001, 41% variance

shared) and Social Functioning (r = -.75, p < .001, 56% variance shared) subscales,

indicating those who perceived their health as more restricting reported more anxiety

symptoms. The GAS was significantly correlated with the Body Pain (r = - .53, p < .01,

28% variance shared) subscale, indicating those who reported more body pain also

reported more anxiety. The GAS was also significantly negatively correlated with the

Vitality (r = -.63, p < .001, 40% variance shared) subscale, indicating that those with less
 59

self-rated vitality (energy) also had more anxiety. Across all subscales, the correlations

among the GAS and SF-36 subscales were largely similar.

The GAS was significantly negatively correlated with the General Mental Health

(r = -.89, p < .001, 79% variance shared) and Role Limitations from Emotional Problems

(r = -.59, p < .001, 35% variance shared) subscales, indicating those with more anxiety

had poorer self-rated mental health and reported more restrictions in daily activities from

these problems. These relationships are further evidence of convergent validity.

Summary of results. The reliability of the GAS and its subscales ranged from

good to excellent in this medical sample of older adults. The GAS demonstrated

convergent validity in its relationships to other measures of anxiety and mental health,

and demonstrated significant overlap with a measure of depression. These findings are

similar to those of Studies 1 and 2. The GAS had significant overlap with a measure of

subjective health status, indicating that those with anxiety are likely to also rate their

health more poorly. The overlap between the total SF-36 score and the GAS decreased

when the items pertaining to mental health were removed from analysis.
 CHAPTER V

STUDY FOUR

Method

Study 4 combined the participants from the three aforementioned studies to yield

one large database of older adults (N = 581). Trait heterogeneity is recommended in

estimating polytomous IRT models (i.e., used for scales with more than two response

options; Embretson & Reise, 2000); combining the three samples thus provided an

adequate sample size as well as adequate trait heterogeneity for the proposed analyses.

Measures and Procedure

See Studies 1-3 for summary of measures and a description of data collection for

each respective sample.

Statistical Analyses

The data were prepared for IRT by collapsing response categories which were

infrequently endorsed (also known as “sparse cells”). There were seven items which

needed to be collapsed due to infrequent responses in the most severe response category

(“All the Time”). The graded response model (GRM; Samejima, 1969) was used for IRT

analyses as GAS items use more than two response categories.

Prior to running analyses, the assumptions for IRT were tested. There are two

assumptions: unidimensionality (i.e., there is only one underlying factor within the data)

and local independence (items should not be correlated when the shared variance of the
 61

latent trait is removed). Both assumptions were tested by conducting a confirmatory

factor analysis using Mplus.

All IRT analyses were conducted in R version 2.13.2 (R Core Team, 2012).

Information curves, item characteristic curves, threshold parameters, and discrimination

parameters were analyzed to examine the item properties of the measure and identify

which items were more or less useful in reliably measuring trait levels of anxiety.

Discrimination parameters reflect the level of the latent trait in which the item best

discriminates among individuals with higher or lower levels of the latent trait. Baker

(2001) suggests that discrimination parameter values ranging from .01 to .24 are

considered very low, .25 to .64 are low, .65 to 1.34 are moderate, 1.35 to 1.69 are high,

and more than 1.7 are very high. Discrimination parameters are directly related to item

information, such that items with higher discrimination parameters are more informative.

Within the graded response model, there are k – 1 threshold parameters (k

indicates the number of response categories). As the GAS has four possible response

categories (not at all, sometimes, most of the time, all the time), there are three threshold

parameters presented for each item in this study. The first threshold parameter (scaled as

a z-score, M = 0, SD = 1, lower values reflecting less anxiety) reflects how much anxiety

is required to have a 50/50 chance of endorsing the “not at all” or “sometimes” response

categories. The second threshold parameter reflects how much anxiety is needed to have

a 50/50 chance of endorsing the “sometimes” or “most of the time” categories, and the

third threshold parameter reflects how much anxiety is needed to have a 50/50 chance of

endorsing the “most of the time” or “all of the time” categories. Response options with

very low or negative threshold parameters would be considered less useful items in
 62

measuring anxiety, as very low levels of anxiety would be needed to endorse these

response options. In contrast, response options with extremely high (i.e., 4 standard

deviations above the mean) threshold parameters would also be less useful, as extreme or

atypical levels of anxiety would be needed to endorse these response options. It would be

reasonable to expect the first threshold parameter to be much lower in magnitude than the

second or third threshold parameters, as the “not at all” or “sometimes” response

categories would be endorsed more frequently by respondents.

The test information curve was also examined to determine at what level of

anxiety the GAS provides the most information. In IRT, higher information indicates

higher reliability and lower standard error of the estimate. The test information function

is the summation of the information of all items, and provides an estimate of what level

of anxiety the test as a whole provides the most information (i.e., is the most useful and

reliable). It was expected that the test information curve would peak for individuals

above the mean level of anxiety, as the GAS is intended to assess individuals with

clinically significant anxiety.

Analyses were also conducted to identify items which may demonstrate

differential item functioning in regards to age, sex, and education (Aim 2). Age was

dichotomized into young-old (60-79) and old-old (80 and up). Sex was dichotomized as

male or female. Education was dichotomized into two groups: up to 12 and 13 and

above.

A short form was created by identifying 10 items with the highest discrimination

parameters while retaining the structure of the subscales, and the reliability and validity

of the short form was tested (Aim 3). It was expected that the short form would have
 63

similar psychometric properties as the full version of the scale, especially in regards to

reliability and validity. The items were then examined to ensure there were not

redundant or similar items included in the short form, and the threshold parameters were

inspected to determine if the parameters were reasonable in magnitude (i.e., not

extremely low or high). The procedure for the selection of the short form items followed

Edelen and Reeve (2007), who used IRT to create a 10 item short form from a larger set

of items on a depression measure. The test information curve for the short form was then

inspected to ensure that the short form provided a reasonable level of information in

comparison to the full version.

T-tests and ANOVAs were conducted on the data to test the hypotheses that

younger, less educated, and female individuals will report the highest levels of anxiety

(Aim 4).

Aim 5 of Study 4 was to create standard scores and descriptive categories (i.e.,

mild, moderate, severe) for scoring the GAS to increase its clinical utility. Raw scores

were converted into t-scores and percentiles. We then used clinical judgment to

determine at what level of anxiety the cutoff scores would be most meaningful. The

following cutoffs were used: minimal, t-scores below 50; mild, 50-59; moderate, 60-65;

severe, 66 and above. A t-score and percentile distribution table was created for the GAS

and all subscales.

Results

Aim 1: Exploratory factor analysis, confirmatory factor analysis, and item

response theory. To test the IRT assumption of unidimensionality, both exploratory

and confirmatory factor analyses were conducted on the data. Principal axis factoring
 64

was conducted on the data. The data met the statistical assumptions for running PAF

(KMO = .94, Bartlett’s test p < .001). PAF yielded 4 factors with an eigenvalue over 1,

and cumulatively explained 56.58% of the variance. The first factor (eigenvalue = 10.22)

explained the most variance (40.86%), and all items loaded onto the first factor. The

ratio of the first to the second eigenvalue was 6.55, which exceeds the recommended ratio

of 3.0 for identifying a unidimensional construct (Morizot, Ainsworth, & Reise, 2007).

An examination of factor loadings revealed that all items loaded onto the first factor.

Confirmatory factor analysis (CFA) suggested that a one factor model was acceptable, x2

(275) = 1582.97, p = .00, CFI = .923. TLI = .916, RMSEA = .091 (90% CI: .086 - .095).

It should be noted that the chi square test is inflated due to the large sample size. The

CFI, TLI, and RMSEA values were close to acceptable. Three pairs of items had high

modification indices, indicating there may be additional variables other than anxiety

which the pairs of items have in common. These item pairs included: Item 6 (“I was

afraid of being judged by others”) and 7 (“I was afraid of being humiliated or

embarrassed”), 8 (“I had difficulty falling asleep”) and 9 “I had difficulty staying

asleep”), and 18 (“I worried too much”) and 19 (“I could not control my worry”).

Removal of any of these items would increase the fit of the one-factor model to the data.

Based upon the redundancy of items 8 and 9, it was decided that item 9 would be

removed from the analysis. Also, although the other item pairs are similar in content, they

measure separate symptoms of anxiety (i.e., excessive worry and difficulty controlling

worry are two separate symptom criterion of Generalized Anxiety Disorder).

Furthermore, the magnitude of the modification index between items 8 and 9 was the

strongest, suggesting the largest potential for improved fit. CFA was performed on the
 65

remaining 24 GAS items and revealed a better fit to the data, x2 (252) = 1051.17, p = .00,

RMSEA = .074 (90% C.I.: .069 - .079), CFI = .951, TLI = .947. The results from the

CFA were deemed sufficient to meet the statistical assumption of unidimensionality for

the IRT analyses. Item 9 was excluded from the IRT analyses.

Due to sparse cells (less than 5 responses in a given category), items 4, 5, 7, 11,

13, 16, and 25 were collapsed (i.e., responses in the “all of the time” category were

combined with responses in the “most of the time” category). The discrimination and

threshold parameters for all GAS items are listed in Table 11. IRT analyses yielded

discrimination parameters ranging from 1.070 to 3.024. These parameters reflect values

in the moderate to very strong range (Baker, 2001). Items from the Somatic subscale (M

discrimination parameter = 1.339, SD = .366) tended to have lower discrimination

parameters than items from the Cognitive (M = 1.938, SD = .442) and Affective subscales

(M = 2.361, SD = .358). Threshold parameters in the current study ranged from -.993 to

1.670 for the first parameter, 1.387 to 3.582 for the second parameter, and 2.456 to 5.094

for the third parameter. As discrimination parameters are directly related ot item

information, the inspection of item information curves revealed that the Somatic items

tended to provide less information than the Affective and Cognitive items. Next, the test

information curve (TIC) was examined. The test information curve and the standard error

of the estimate are presented in Figure 3. The GAS provides the greatest amount of

information for individuals with average or higher levels of anxiety, as indicated by the

maximum TIC and minimum standard error of the estimate (SEE). The TIC peak was at

approximately 2.5 standard deviations above the mean level of anxiety. This was as

expected, as the GAS aims to detect clinically significant anxiety.

 66

Table 11

IRT Calibration for GAS Items

GAS Item Discrimination Threshold Threshold Threshold

a b1 b2 b3
Item 1 (My heart raced or beat strongly.) 1.160 .689 3.558 4.526
Item 2 (My breath was short.) 1.070 .542 3.258 5.094
Item 3 (I had an upset stomach.) 1.086 .878 3.508 4.908
Item 4 (I felt like things were not real or like I was 1.937 1.670 3.248 -
outside of myself.)
Item 5 (I felt like I was losing control.) 2.473 0.983 2.440 -
Item 6 (I was afraid of being judged by others.) 1.516 .767 2.870 4.013
Item 7 (I was afraid of being humiliated or 1.396 1.031 3.251 -
embarrassed.)
Item 8 (I had difficulty falling asleep.) 1.067 -.062 2.108 3.392
Item 10 (I was irritable.) 2.026 -.057 2.458 3.368
Item 11 (I had outbursts of anger.) 1.466 .762 3.582 -
Item 12 (I had difficulty concentrating.) 1.987 .090 2.185 3.297
Item 13 (I was easily startled or upset.) 2.041 .643 2.425 -
Item 14 (I was less interested in doing something I 2.172 .438 2.151 2.985
typically enjoy.)
Item 15 (I felt detached or isolated from others.) 2.226 .765 1.978 2.896
Item 16 (I felt like I was in a daze.) 2.214 1.180 2.740 -
Item 17 (I had a hard time sitting still.) 1.360 .876 2.651 4.032
Item 18 (I worried too much.) 2.225 -.072 1.568 2.535
Item 19 (I could not control my worry.) 2.657 .569 1.802 2.789
Item 20 (I felt restless, keyed up, or on edge.) 2.657 .296 1.998 2.848
Item 21 (I felt tired.) 1.758 -.993 1.410 2.456
Item 22 (My muscles were tense.) 2.040 .040 1.839 2.805
Item 23 (I had back pain, neck pain, or muscle 1.171 -.874 1.387 2.679
cramps.)
Item 24 (I felt like I had no control over my life.) 3.024 .754 1.818 2.481
Item 25 (I felt like something terrible was going to 2.371 1.487 2.748 -
happen to me.)
Note. Items without b3 parameter estimates had been collapsed due to sparse cells.

Figure 3. Test information function for

24-item GAS. The dotted line
represents standard error estimate, and
the solid line represents the summative
test information.
 67

The TIC also indicated that the GAS provides less information above 3 standard

deviations above the mean level of anxiety, as well as one or more standard deviations

below the mean level of anxiety. This is reflected by low TIC and higher SEE. This

would indicate that the GAS does not provide useful or reliable information at markedly

low levels of anxiety, nor does it provide useful or reliable information at levels of

extremely severe anxiety.

Aim 2: Differential item functioning. Next, exploratory analyses were

conducted to detect differential item functioning (DIF) among the GAS items. Age was

dichotomized into young-old (60-79) versus old-old (80 and above). Item 3 (“I had an

upset stomach”) was flagged for age-related DIF. Test and item characteristic curves

(ICC) for this item and the GAS total scale are presented in Figure 4, and indicate that

younger adults were more likely to score higher on Item 3 than older individuals with

equivalent levels of anxiety. The plot on the left of this figure indicates that there were

no significant differences in GAS total scale scores by age, suggesting that total scale

scores were not affected by DIF on Item 3. An examination of McFadden’s pseudo R2

statistic indicated that the magnitude of DIF for this item was very small or negligible

(.0119; Zumbo, 1999).

Figure 4. Test (left panel) and DIF-item

(right panel) characteristic curves by age
(young-old versus old-old).
 68

Item 10 (“I was irritable”) was flagged for DIF in regards to sex. The test and

item characteristic curves for this item and the total scale are presented in Figure 5,

indicating that men were slightly more likely to score higher on this item than women

with equivalent levels of anxiety. The plot on the left of this figure indicates that there

were no significant differences in GAS total scale scores by sex, suggesting the amount

of DIF for this item was negligible. An examination of McFadden’s pseudo R2 statistic

also indicated that the magnitude of DIF for this item was negligible (.0118; Zumbo,

1999).

All Items DIF Items

1.0
40

0.8
30

0.6
TCC

TCC
20

0.4
10

0.2

Women (= 0) Women (= 0)
0.0
0

Men (= 1) Men (= 1)

-4 -2 0 2 4 -4 -2 0 2 4

theta theta

Figure 5. Test (left panel) and DIF-item (right panel) characteristic curves by sex.

No items were flagged for DIF when examining education dichotomized two

ways: up to 12 years versus more than 12 years, and up to 14 years versus over 14 years.
 69

Aim 3: Short form. To create the GAS short form (GAS-10), items with the

highest discrimination parameters and information curve peaks were retained from each

of the subscales. The GAS-10 is presented in Appendix D. Three items were selected

from both the Somatic (items 17, 21, 22) and Affective (10, 15, 20) subscales. Four

items were retained from the Cognitive subscale (16, 19, 24, 25) as items from this

subscale had the highest discrimination parameters. The items were examined to ensure

the short form would not contain redundant items. For instance, items 5 (“I felt like I was

losing control”) and 24 (“I felt like I had no control over my life”) both had high

discrimination parameters and information curve peaks, but both pertain to perceptions of

control. In this case, the latter was retained due to the higher discrimination parameter.

The threshold parameters of the 10 chosen items were also inspected, and revealed items

which could be endorsed by individuals with varying amounts of anxiety severity (i.e.,

the threshold parameters were not too high nor too low). Presented in Figure 6, the TIC

for the GAS-10 indicates the short form best assesses anxiety for individuals with

average up to 2.5 standard deviations above the mean level of anxiety. The GAS-10 does

not assess anxiety reliably for people below the mean level or above 3 standard

deviations above the mean of anxiety. This is comparable to the full version of the GAS.

The peak of the test information curve indicated a reasonable amount of information is

provided by the GAS-10 in comparison to the full GAS. The SEE indicated that the

GAS-10 did not lose precision as a result of reducing items.

The GAS-10 performed similar to the full version of the GAS. The GAS-10 had

excellent reliability (Cronbach’s alpha = .89), and was significantly and positively

correlated with the GAS total scale (r = .96, p < .001) and subscales (Cognitive: r = .92,
 70

Figure 6. Test Information Function for GAS-10. The dotted line represents standard
error estimate, and the solid line represents the summative test information.

p < .001, Affective: r = .89, p < .001, Somatic: r = .82, p < .001). The GAS-10 was

significantly and positively correlated with other measures of anxiety:

GAI (r = .80, p < .001), BAI (r = .65, p < .001), which provides evidence of convergent

validity. It was also significantly correlated with measures of depression: GDS (r = .72,

p < .001), BDI-II (r = .70, p < .001), PHQ-9 (r = .77, p < .001). Principal axis factoring

revealed all 10 items loaded onto a single factor that explained 52.03% of the variance.

Aim 4: Sex, gender, and education differences in scores. A series of

independent samples t-tests was conducted to compare GAS total scale and subscale

scores for men and women. Women (M = 13.12, SD = 10.58) scored significantly higher
 71

than men (M = 10.02, SD = 8.93) on the GAS total scale score, t(465.73) = 3.63, p < .001.

The magnitude of the difference in the means (mean difference = 3.10, 95% CI: 1.42 to

4.78) was small (eta squared = .02).

Women (M = 3.37, SD = 3.68) also scored higher than men (M = 2.17, SD = 3.13)

on the Cognitive subscale, t(476.66) = 4.09, p < .001. The magnitude of the difference in

means (mean difference = 1.20, 95% CI: .62 to 1.77) was small (eta squared = .03).

Similarly, on the affective subscale, women (M = 3.69, SD = 3.59) scored significantly

higher than men (M = 2.80, SD = 3.07), t(481.28) = 3.10, p < .01. The magnitude of the

difference in means (mean difference = .89, 95% CI: .33 to 1.45) was small (eta squared

= .02). Finally, there were sex differences on Somatic subscale scores as well, such that

women (M = 6.30, SD = 4.26) scored significantly higher than men (M = 5.10, SD =

3.77), t(563) = 3.34, p < .01. The magnitude of the difference in means (mean difference

= 1.20, 95% CI: .49 to 1.90) was small (eta squared = .02).

Next, a series of independent t-tests were conducted to examine potential age

differences in GAS scores among older adults only. Age was dichotomized: young-old

(60-79) and old-old (80 and above). In regards to GAS total score, there was no

significant difference between the young-old (M = 12.31, SD = 10.72) and old-old (M =

10.51, SD = 7.77; t(224.23) = -1.77, p = .08). In contrast, the young-old (M = 3.08, SD =

3.74) scored significantly higher than the old-old (M = 2.37, SD = 2.55) on the Cognitive

subscale, t(255.25) = -2.40, p < .05. The magnitude of the difference in means (mean

difference = -.71, 95% CI: -1.29 to -.13) was small (eta squared = .01).

Finally, there were no significant differences between the young-old (M = 3.49,

SD = 3.58) and the old-old (M = 2.91, SD = 2.75) on the Affective subscale, t(224.95) = -
 72

1.86, p = .06, although this approached statistical significance, and there were no

significant differences between the young-old (M = 5.98, SD = 4.23) and the old-old (M =

5.46, SD = 3.69) on the Somatic subscale, t(563) = -1.20, p = .23.

To examine the potential interaction effect between age and gender, a two-way,

between-groups analysis of variance (ANOVA) was conducted. The interaction between

age and sex was not statistically significant in regards to total GAS scores, F(1, 537) =

1.46, p = .23. Likewise, the interaction was not significant in regards to Cognitive

subscale scores, F(1,558) = 2.61, p = .64, Affective subscale scores, F(1, 555) = 1.03, p =

.31 . It was not significant for Somatic and Somatic subscale scores, F(1, 561) = 2.62, p

= .11.

There were no significant differences between education groups (up to 12 versus

13 and above) in regards to the total scale score (t(308.06) = -1.69, p = .09), Cognitive

subscale (t(296.40) = -1.77, p = .08), Affective subscale (t(308.29) = -1.59, p = .11), and

Somatic subscale (t(289.05) = -1.53, p = .13).

Aim 5: Norms and descriptive categories. Standard t-scores for the total scale,

subscales, and GAS-10 are presented in Tables 12 through 14. Descriptive categories

(minimal, mild, moderate, and severe) are also included.

Summary of results. Study 4 examined the GAS using both CTT and IRT

techniques. The GAS met the statistical assumptions for IRT. IRT analyses revealed that

the GAS provided the greatest level of information for individuals 2.5 standard deviations

above the mean level of anxiety, as expected. The items which had the highest

discrimination parameters tended to come from the Cognitive and Affective subscales,

and items from the Somatic subscale provided less information.

 73

Table 12

Standard Score Distribution for GAS Total Scale Scores (N = 542)

Raw T-Score Percentiles Descriptive

Category
1 39 13 Minimal
2 40 16 Minimal
3 41 18 Minimal
4 42 21 Minimal
5 43 25 Minimal
6 44 27 Minimal
7 45 32 Minimal
8 46 34 Minimal
9 47 37 Minimal
10 48 45 Minimal
11 49 47 Minimal
12 50 50 Mild
13 51 53 Mild
14 52 55 Mild
15 53 63 Mild
16 54 67 Mild
17 55 70 Mild
18 56 73 Mild
19 57 75 Mild
20 58 81 Mild
21 59 82 Mild
22 60 84 Moderate
23 61 87 Moderate
24 62 88 Moderate
25 63 91 Moderate
26 64 93 Moderate
27 65 94 Moderate
28 66 95 Severe
29 67 95 Severe
30 68 97 Severe
31 69 97 Severe
32 70 98 Severe
33 71 98 Severe
34 72 99 Severe
35 73 99 Severe
36 74 99 Severe
37 74 99 Severe
38 76 99 Severe
39 78 99 Severe
40 78 99 Severe
41 79 99 Severe
42 80 99 Severe
43 81 99 Severe
46 84 99 Severe
47 85 99 Severe
48 86 99 Severe
54 92 99 Severe
 74

Table 13

Score Distribution for GAS Subscale Scores

Somatic (N = 566)

Raw T- Percentile Descriptive

Score Category
1 38 13 Minimal
2 41 19 Minimal
3 43 25 Minimal
4 45 32 Minimal
5 48 45 Minimal
6 50 50 Mild
7 53 63 Mild
8 55 70 Mild
9 58 81 Mild
10 60 84 Moderate
11 62 88 Moderate
12 65 94 Moderate
13 67 95 Severe
14 70 98 Severe
15 72 99 Severe
16 75 99 Severe
17 77 99 Severe
18 79 99 Severe
19 82 99 Severe
20 84 99 Severe
27 101 99 Severe

Affective (N = 560)
Raw T- Percentile Descriptive
Score Category
1 43 25 Minimal
2 46 34 Minimal
3 49 47 Minimal
4 52 55 Mild
5 55 70 Mild
6 58 81 Mild
7 61 87 Moderate
8 63 91 Moderate
9 66 95 Severe
10 69 97 Severe
11 72 99 Severe
12 75 99 Severe
14 81 99 Severe
16 87 99 Severe
18 93 99 Severe
24 110 99 Severe
 75

Table 13 Continued

Cognitive (N = 563)
Raw T- Percentile Descriptive
Score Category
1 45 32 Minimal
2 47 37 Minimal
3 50 50 Mild
4 53 63 Mild
5 56 73 Mild
6 59 82 Mild
7 62 88 Moderate
8 64 93 Moderate
9 67 95 Severe
10 70 98 Severe
12 76 99 Severe
14 81 99 Severe
16 87 99 Severe
18 93 99 Severe
24 110 99 Severe

Table 14

Score Distribution for GAS-10 (N = 556)

Raw T- Percentile Descriptive

Score Category
1 42 21 Minimal
2 44 30 Minimal
3 46 34 Minimal
4 48 45 Minimal
5 51 53 Minimal
6 53 63 Minimal
7 55 70 Mild
8 57 75 Mild
9 59 82 Mild
10 61 90 Moderate
12 66 95 Severe
14 70 98 Severe
16 74 99 Severe
18 79 99 Severe
24 92 99 Severe
30 104 99 Severe

Differential item functioning analyses revealed two items flagged for DIF, but the degree

of DIF for both items was negligible and did not significantly impact total scale scores.
 76

A short form of the GAS was created (GAS-10), which had excellent psychometric

properties, though it correlated strongly and positively with measures of depression. In a

series of t-tests, women scored significantly higher on the GAS total scale score and all

subscales. Young-old adults (aged 60-79) scored significantly higher than old-old adults

(80 and above) on the Cognitive subscale, but there were no significant differences

detected on the total scale, Affective, or Somatic subscales. There were no significant

education differences in test scores. There was no significant age by sex interaction on

GAS scores.
 CHAPTER VI

GENERAL DISCUSSION

The purpose of this study was to examine the psychometric properties of the

Geriatric Anxiety Scale in three samples of older adults, using both classical test theory

and item response theory techniques. Across all three samples, the GAS had strong

psychometric characteristics, especially in regards to reliability, convergent and divergent

validity, factorability, and item properties. IRT analyses indicated that the GAS best

assesses anxiety for people with higher than average amounts of anxiety, as intended. A

main limitation of the measure is in its strong relationship with measures of depression,

reflecting poor discriminant validity with depression. Another limitation is in its

moderate relationship with subjective health status. This study adds to the current

literature on anxiety assessment in several novel ways. This is the first study to utilize

item response theory to examine a measure of anxiety designed specifically for use with

older individuals. Furthermore, this study utilized a large sample of older individuals

recruited from a number of diverse populations, and merged both classical test theory and

item response theory techniques. Overall, the results from the current study support the

use of the GAS in measuring anxiety in older individuals. As the number of older adults

who experience anxiety increases, evidence-based assessment tools become increasingly

necessary.
 78

The GAS had excellent reliability (internal consistency) across all four studies,

and the subscales and good reliability. This indicates that the GAS items tap into a similar

construct, but are not redundant. The Cronbach’s alpha coefficient of the total scale is

similar to those measures of anxiety used in older adult samples, including the Beck

Anxiety Inventory (Kabacoff et al., 1997), the Geriatric Anxiety Inventory (Pachana et

al., 2007), and the Adult Manifest Anxiety Scale-Elderly Version (Lowe & Reynolds,

2006). It is also similar to the reliability of the GAS reported in previous studies (Segal

et al., 2010; Yochim et al., 2011).

The GAS can be interpreted as a one- or two- factor scale. In the one-factor

model, all items on the GAS tap into the same underlying latent variable or construct

(general anxiety). The results from the confirmatory factor analysis in Study 4 also

support this one-factor model, and the model fit was improved with the removal of Item 9

(“I had difficulty staying asleep”). The strong correlations among the subscales and the

GAS total score also support the one-factor model. In the two-factor model, items from

the Affective and Cognitive subscales loaded onto the first factor, whereas somatic items

tended to load onto the second factor. In the two-factor model, it appears as if the

Affective and Cognitive subscales both tap into the same underlying construct:

Cognitive-Affective Distress. While a three-factor model of anxiety was hypothesized,

the findings from the current study support other conceptual models of anxiety in older

adults. For instance, the Beck Anxiety Inventory (Kabacoff et al., 1997) contains two

primary subscales: Somatic and Cognitive. The GAI has one single underlying factor

(Byrne & Pachana, 2010), which could be due to the scale’s deliberate exclusion of

somatic items. The subscales should be utilized with these empirical findings considered.
 79

The subscales have clinical or practical significance and utility, but clinicians should bear

in mind that the scale supports a one- or two-factor model of anxiety. Future research

should examine the possibility of the GAS having a bi-factor model structure.

Validity refers to both the accuracy and appropriateness of a scale in measuring a

theoretical construct of interest. Convergent validity reflects the degree to which the

scale relates with other theoretically similar variables. In the current study, the GAS

demonstrated strong convergent validity, such that the scale was significantly related to

measures of other theoretically similar constructs in the expected directions. Evidence

for convergent validity of the GAS was found in all three samples in this study. The

GAS and subscales were positively related to other measures of anxiety (BAI and GAI)

and a measure of general mental health, similar to previous studies (Segal et al., 2010;

Yochim et al., 2011) and as expected. The GAS and subscales were also significantly

related to other similar constructs, such as self-esteem, hopelessness, and loneliness.

Previous literature has indicated relationships among these variables (i.e., Barg, Huss-

Ashmore, Wittink, Murray, Bogner, & Gallo, 2005). The correlations between anxiety

and these three constructs were less strong than the relationships among the GAS and

other measures of anxiety, as would be expected.

The GAS correlated strongly with three measures of depression (PHQ-9, BDI-II,

and GDS) across samples. This finding is similar to previous studies of the GAS (Segal et

al., 2010; Yochim et al., 2011), which reported correlations ranging from .73 to .78 in

community-dwelling and clinical samples (using the GDS and BDI-II as measures of

depression). These correlations were also similar those between the BAI and GAI and

the PHQ-9 in Study 3, but a relatively low sample size may account for unreliable
 80

correlations in that sample. Other measures of geriatric anxiety, such as the OASES,

demonstrate moderate-to-strong relationships with depression (i.e., Gould et al., 2012).

Items from the Cognitive and Affective subscales tended to correlate more strongly with

these measures of depression than items from the Somatic subscale, similar to other

measures of anxiety (Wetherell & Areán, 1997). In Study 1, a principal components

analysis of the combined GAS and PHQ-9 items indicated that both scales tap into the

same underlying construct. Overall, the results of this study indicate the GAS lacks

discriminant validity in measuring anxiety as a construct distinct from depression. There

are several possible explanations for this finding. Perhaps most significantly, the

comorbidity of anxiety and depression is known to be common and ubiquitous among

older individuals (Beekman et al., 2000), perhaps reflecting a common underlying cause.

Because of these high rates of comorbidity, it is expected that the GAS will have some

degree of overlap with measures of depression. The GAS also contains several items

which assess symptoms of anxiety which overlap strongly with symptoms of depression

(i.e., items 8, 9, 12, 14, 15, 21), which reflects limitations of the current diagnostic

criteria for both conditions. Mixed anxiety-depression (MAD) has received some

attention in the empirical literature, with some arguing that anxiety and depression are not

discrete diagnostic categories. Some claim that MAD should receive its own diagnostic

category based upon its different symptom presentation and treatment outcomes,

especially in later life (Cassidy, Lauderdale, & Sheikh, 2005; Mohlman et al., 2011).

This topic warrants further research, especially as the diagnostic system continues to

evolve and the older adult population continues to increase steadily. It could be that the

GAS measures MAD better than it does “pure” anxiety as a discrete diagnosis. In Study
 81

1, a brief version of the GAS was created (GAS-8) which was less strongly correlated

with measures of depression (range of r = .46 to .61). The reliability of the GAS-8 was

acceptable (Cronbach’s alpha = .76), and was lower than the GAS, subscales, or GAS-10.

Although still moderately correlated, this version of the GAS can be used to help

distinguish anxiety symptoms from depressive symptoms should clinicians or researchers

be interested in doing so. However, some of these items also had lower discrimination

parameters than other items on the scale, indicating that the GAS-8 may provide less

information than the full version or GAS-10. This discrepant finding also highlights the

differences among CTT and IRT in scale development.

Aside from its relationship with depression, the GAS demonstrated solid evidence

of divergent validity in that it did not significantly or strongly correlate with unrelated

constructs. For example, the GAS did not significantly relate to scores on a measure of

general cognition (MoCA) in Study 3. It also did not significantly correlate with sex, and

correlated weakly with education in only Study 1 (r = -.11). The Cognitive subscale

correlated weakly with age (r = -.21, p < .05) in Study 2 only, and the total scale and

other subscales did not correlate with age. Previous studies on the GAS also found that

the measure had weak or statistically insignificant relationships with constructs

theoretically unrelated to anxiety, such as reading and processing speed (Yochim et al.,

2011) and education (Segal et al., 2010).

The GAS had moderate negative correlations with a measure of subjective health

status, indicating that people with anxiety also self-reported their subjective health as

poorer. These results are similar to Yochim et al. (2011), who reported that the GAS and

subscales had mild-to-moderate correlations with a weighted measure of medical burden

 82

(Charlson Comorbidity Index; CMI). This was found in both Study 1, which examined

the relationship between the Physical Functioning subscale of the SF-36, as well as Study

3, which found significant relationships among the SF-36 total scale score. Interestingly,

all SF-36 subscales correlated with the GAS in Study 3 aside from the Physical

Functioning subscale. Low sample size could account for this discrepant finding as well

as inflated correlations in Study 3. However, correlations detected with this sample size

reflect large effect sizes (Cohen, 1992). It was expected that somatic items would

correlate more strongly with the Physical Functioning score in Study 1, but only partial

support for this was found. Of the 17 items which correlated with the Physical

Functioning subscale, five were from the Somatic subscale, 8 were from the Cognitive

subscale, and 4 were from the Affective subscale. The magnitude of the correlations

between the Somatic items and the Physical Functioning subscale were slightly larger

than the Affective or Cognitive items. This suggests that aspects or symptoms of anxiety

other than somatic are related to physical health and functioning. This also suggests that

the relationship between subjective health well-being and anxiety cannot be explained by

somatic symptoms alone.

Other measures of anxiety have also demonstrated relationships with measures of

health. For example, the GAI was significantly related to self-reported number of health

conditions (r = .36) and self-rated health status (r = -.28) in a sample of community-

dwelling Australian women (Byrne et al., 2010), despite the GAI containing few somatic

items. The results of Studies 1 and 3 add support to previous literature highlighting the

relationship between anxiety symptoms and subjective health status. Anxiety occurs in

higher rates among those with chronic health conditions (Brock et al., 2011; Cully et al.,
 83

2006; Murphy, Sacks, Brady, Hootman, & Chapman, 2012; Wetherell, Ayers, Nuevo,

Stein, Ramsdell, & Patterson, 2010). It could also be that anxiety could predispose one to

exaggerate or catastrophize the extent of their health burden, leading to lower subjective

health ratings. Pessimism has been found to moderate the relationship between

physician-rated illness burden and anxiety in older adults (Hirsch, Walker, Chang, &

Lyness, 2012), suggesting that other psychosocial variables impact the complex

relationship between physical health and anxiety. The relationship between anxiety and

health status in older adults warrants further attention. For instance, future studies could

examine DIF among older individuals with medical burden to assess for item bias in this

population.

IRT analyses indicated that the GAS is most reliable in discriminating individuals

at the higher end of the anxiety continuum versus people with very low levels of anxiety.

As indicated by the peak of the test information curve and minimum standard error

estimate, the GAS provides the most information at 2.5 standard deviations above the

mean level of anxiety. This was predicted, as the GAS intends to detect clinically

significant levels of anxiety. The TIC and SEE also indicates that the GAS is less

reliable for individuals below the mean level of anxiety or three or more standard

deviations above the mean level of anxiety. Practically speaking, clinicians are less

interested in measuring anxiety in someone with below average levels of anxiety.

Individuals scoring three or more standard deviations above the mean of anxiety would

likely be experiencing extreme mental distress, which would be quite apparent in a

clinical setting. The GAS intends to be used as a clinical screening to help identify

individuals experiencing elevated levels of anxiety. Thus, the results of the IRT analysis
 84

support the use of the GAS as a clinically meaningful assessment tool. However, the

shape of the TIC indicates that the GAS may have limitations in regards to measuring

changes in anxiety scores, especially when one moves into or out of anxiety levels that

are not measured well by the test (i.e., >3 or <-1 SD). This is due to the fact that at these

levels of anxiety, measurement precision is low (SEE is high) at these levels of anxiety.

This indicates that it would take a very large change to know that the change is reliable

and not just due to measurement error. Clinicians and researchers should utilize the scale

bearing this limitation in mind, and future studies should examine this limitation further.

Baker (2001) suggests that discrimination parameter values ranging from .01 to

.24 are considered very low, .25 to .64 are low, .65 to 1.34 are moderate, 1.35 to 1.69 are

high, and more than 1.7 are very high. Thus, the discrimination parameter values in this

study (presented in Table 11) ranged from moderate to very high. Overall, the somatic

items (M discrimination parameter = 1.339, SD = .366) provided less information than

items from the Affective (M = 1.938, SD = .442) or Cognitive (M = 2.361, SD = .358)

subscales, although all items had discrimination parameters of at least 1.067. There were

no items which appeared problematic in this respect, and IRT analyses indicated that all

items had justification for remaining in the scale. However, the results indicate that the

Somatic items provide less information about anxiety than the Affective and Cognitive

items, likely because they are endorsed frequently by individuals with medical

conditions.

One item was flagged for DIF by age, and another item was flagged for DIF by

sex. No items were flagged for DIF by education. The young-old (60-79) were more

likely to score higher on Item 3 (“I had an upset stomach”) than the old-old group,
 85

despite controlling for level of anxiety. Men were more likely than women to score

higher on Item 10 (“I was irritable”) after controlling for level of anxiety. The degree of

DIF for both items was very small or negligible, indicating that the overall scale scores

were not impacted meaningfully by DIF. This indicates that the items flagged for DIF

are not biased in regards to age or sex, respectively. For these reasons, no modifications

to these items are proposed, although researchers should continue to look closely at these

two items in future studies.

This study also created a brief version of the GAS (GAS-10), which had excellent

psychometric properties in regards to reliability, convergent validity, and factor structure.

Similar to the regular version of the GAS, the GAS-10 was strongly correlated with

measures of depression. The items of this short form were selected upon their

discrimination parameters and information curves while retaining the structure of the

subscales, following an example set forth by Edelen and Reeve (2007). While retaining

only the items with the highest discrimination parameters or the highest information

curve peaks may have resulted in a short form with higher information, including items

from the other subscales ensured breadth of content coverage. Furthermore, as indicated

by the TIC presented in Figure 7, the GAS-10 performed similar to the full version of the

GAS with respect to the number of items retained. The GAS-10 best assessed anxiety for

individuals at the mean level of anxiety up to about 2.5 standard deviations above the

mean level of anxiety, as reflected by peak TIC and minimum standard error. The

detection of anxiety in primary care settings is poor (Calleo et al., 2009), highlighting the

dire necessity for behavioral health screening in medical settings. Short forms are also
 86

useful in epidemiological research and in clinical use with individuals with limited

cognitive capacity. The GAS-10 shows promise as a brief measure of anxiety.

To my knowledge, this is the first study to use item response theory to examine

the psychometric properties of an anxiety measure in older adults. IRT offers several

notable advantages over CTT (de Ayala, 2009). However, some argue that both theories

are best used when in conjunction with one another to attain the benefits of both models

(e.g., Edelen & Reeve, 2007). One main advantage is that IRT analyses are not test or

sample dependent as in CTT, resulting in more accurate parameter estimates. Because

IRT is not test or sample dependent, the item parameter estimates from the current study

are able to be generalized beyond the sample used in this study. Another advantage is

that IRT has stronger assumptions than CTT, meaning that the assumptions of

unidimensionality and local independence are more difficult to meet than the assumptions

in CTT. Stronger assumptions beget stronger findings; thus, the use of IRT is considered

more advanced than CTT techniques. Furthermore, a major limitation of CTT is the

assumption that the test functions the same for all levels of the latent trait. The IRT

results of this study indicate that the reliability and standard error of the GAS is different

for individuals among varying levels of anxiety. IRT methods can also be used to

improve scoring, such as using adaptive computer testing. One advantage of using this

scoring method is that is that each individual has his or her own unique SEE (and thus

unique confidence intervals around their estimated level of anxiety), which take into

account the specific patterns of responding across all items. Future studies should

examine the utility of IRT scoring in regards to the GAS.

 87

Women tended to score higher on the GAS total scale and subscales than men, as

hypothesized, although the effect sizes for these results were small. Sex differences in

anxiety prevalence have been reported previously (De Beurs et al., 2000; Gum et al.,

2009; Lowe & Reynolds, 2005; Owens et al., 2000; Potvin et al., 2011), with higher

prevalence rates occurring in women than men. Taken together with the results from the

DIF analyses, the results from the t-tests indicate true group differences in anxiety

symptoms rather than item bias. There are a number of reasons why women may score

higher on the GAS than men. Women tend to have more risk factors for anxiety, such as

poorer physical health, physical inactivity, and more interpersonal difficulties (Leach et

al., 2008). Women may also be more comfortable in expressing mental health symptoms

than men, leading to higher scores on self-report measures of mental health problems.

As far as age-differences in the full older adult sample, the young-old (60-79

years) scored significantly higher on the Cognitive subscale than the old-old (80 years old

and older). No significant differences between groups were found on the total scale,

Affective, or Somatic subscales, although the results approached significance. This was

not as hypothesized, and contradicts previous literature reporting differences among the

young-old and old-old in anxiety prevalence (e.g., Gum et al., 2009; Schaub & Linden,

2000). It could be that the Affective or Somatic subscales are not as sensitive to age

differences in anxiety symptoms among the young-old and old-old. It should be noted

that there are no agreed-upon standards for classifying the young-old and old-old in the

current literature, with researchers using various cutoffs. Older adults can also be split

into more than two groups (i.e., young-old, middle-old, old-old, very-old old), and some
 88

use 85 as a cutoff for the “very old-old” (i.e., Schaub & Linden, 2000). Categorizing age

in a different way could potentially have resulted in different results.

Furthermore, there was no education differences noted in GAS or subscale scores.

This contradicts previous literature stating that lower levels of education are a risk factor

for anxiety disorders (e.g., Gum et al., 2009). In the current study, there were few

individuals with less than 12 years of education. It could be also that these differences

are not notable for sub-syndromal levels of anxiety, and the current study lacked enough

individuals with anxiety disorders who also had less than 12 years of education to detect

significant differences. Categorizing education in a different way could have also

resulted in different results.

Both CTT and IRT techniques suggest that the GAS has strong psychometric

properties, and that the scale assesses anxiety at the level for which it was intended.

Clinicians and researchers should use caution in interpreting the scale if an individual

endorses more somatic items than affective or cognitive items, as these items appear to

provide less reliable information about anxiety than items from the other subscales. They

should also use caution in using GAS results to discriminate anxiety from depression, as

results from the study revealed the GAS lacks discriminant validity in this domain, a

problem not unique to the GAS. Overall, the suggestions for modifications to the scale

are minimal as the results of the current study support the use of the GAS in measuring

anxiety symptoms in late life. The t-scores, percentiles, and descriptive categories (mild,

moderate, severe) presented in Tables 12 through 14 aid in the interpretation of scores for

clinicians and researchers. It should be noted that the sample used for establishing these

norms include a clinical sample, which could have resulted in elevated levels of anxiety
 89

(versus a community-dwelling sample). Future studies should establish sample-specific

norms, should they significantly differ among various populations of older adutls.

Both sub-syndromal and clinically significant anxiety has been associated with

dire outcomes in older adults. However, some research has suggested that anxiety can be

beneficial in later life. For example, Price and Mohlman (2007) found that clinically

significant anxiety symptoms were associated with better inhibitory control, an executive

function. The researchers suggest worry is a cognitive avoidance strategy which requires

some degree of selective processing; over time, the cognitive skill of attentional control

becomes strengthened. Furthermore, mild levels of anxiety in older adults have been

associated with better performance on cognitive tasks, whereas severe anxiety was

associated with impaired performance on cognitive tasks (Bierman et al., 2005). Anxiety

in that study was measured using the Hospital Anxiety and Depression Scale-Anxiety

subscale. This curvilinear relationship is also known as the Yerkes-Dodson law (Yerkes

& Dodson, 1908), which suggests that some amounts of anxious arousal can be

beneficial, as anxiety has a motivational component. Too little anxiety may not be

arousing or motivating enough to achieve optimal performance, but too much anxiety

may act as a detriment to adaptive behavior. For instance, mild worry regarding one’s

health status may prompt one to visit his or her physician, leading to preventative

healthcare, while having no worry or debilitating levels of worry could prevent one from

visiting his or her physician altogether. However, recent research has also indicated that

sub-syndromal symptoms of anxiety are associated with lower memory and executive

functioning in community-dwelling older adults (Yochim, Mueller, & Segal, 2012).

Overall, the relationship among anxiety (especially sub-syndromal anxiety) and other
 90

health outcomes in later life appears complex, and individuals with mild anxiety

symptoms should be given special attention due to contradictory findings in the literature.

Evidence-based assessment is a necessary and important precursor to evidence-

based treatment, and there are a number of unique issues pertaining to anxiety assessment

in older adults. Thorough assessment of symptoms at the onset of treatment provides an

important baseline by which treatment progress can be monitored over time. Assessment

data also provides valuable diagnostic information to the clinician which can help guide

treatment. Research has indicated that self-report measures are more useful in identifying

sub-syndromal anxiety than other forms of assessment, such as diagnostic interviews

(Karsten, Nolen, Penninx, & Hartman, 2011). Diagnostic interviews may not be sensitive

to levels of anxiety that do not meet diagnostic criteria, but interfere with functioning

nonetheless. In regards to the treatment of anxiety, community-dwelling older adults

demonstrate a preference of psychotherapy delivered in a primary care, private practice,

or university setting versus psychopharmacotherapy or combined treatment (Mohlman,

2012). Taken together, these points emphasize the necessity for psychometrically sound

self-report assessment tools for use with older adults, and many measures currently in use

lack empirical support for older populations (Therrien & Hunsley, 2011). The results

from the current study reveal that the GAS is an appropriate measure to use with this

subpopulation.

This study is not without limitations. One limitation is the lack of ethnic and

educational diversity within the samples. The small number of ethnic minorities in this

sample prevented analyses being conducted due to limited statistical power (Cohen,

1992). Future research should examine the psychometric properties of the GAS in
 91

culturally diverse populations of older adults, including both ethnic and sexual minorities.

However, the use of IRT strengthens the generalizability of the results, such that the item

parameters are able to be generalized beyond the sample used in this study. An

additional limitation is in the variety of data collection approaches used, which has

implications the results from Study 4. For instance, some participants completed mail-in

surveys, whereas some were tested in person. This could have implications for self-

selection biases, and results should be interpreted in light of this limitation. Furthermore,

future studies should examine the divergent validity of the GAS with other measures of

mental health, examining its relationship with disorders such as schizophrenia, schizoid

personality disorder, and antisocial personality disorder. This would hep establish that

the GAS is measuring a construct separate from mental disorders to which it should not

be related. Another limitation is the low sample size in Study 3, which limited the

number and type of statistical analyses appropriate for the data. Additionally, this study

did not examine the psychometric properties in individuals with cognitive impairment.

As cognitive impairment and anxiety are often co-morbid in older adults (Wolitzky-

Taylor et al., 2010), the GAS should be validated in a sample of individuals with

cognitive impairment to determine its utility for use with such populations of older adults.

A collateral or caregiver version of the GAS could also be useful for use with these

populations. Moreover, the sensitivity and specificity of the GAS should be examined

using the diagnostic criteria for anxiety disorders to determine an appropriate cut-score

for clinically significant anxiety. This line of research should also be pursued with the

GAS-10. Furthermore, it should be noted that the DSM-5 will be released in May 2013,

which may reveal new diagnostic criteria for anxiety disorders. (Information on these
 92

potential changes was not available as of the writing of this manuscript.) As the GAS

was created based on the diagnostic criteria of anxiety disorders, it will be important to

review the new diagnostic criteria following this release. If the GAS is revised in the

future, these diagnostic revisions should be taken into account. Lastly, future studies

should continue to utilize IRT to examine the item properties of anxiety measures in older

individuals. It would be ideal to understand how the GAS items assess anxiety relative to

other measures. Based upon the results of this study, it would be expected that the BAI

would provide the greatest level of information for individuals on the lower end of the

anxiety spectrum given the number of somatic items on the measure. It would also be

expected that the GAI would provide the greatest level of information for individuals on

the higher end of the anxiety continuum, given the cognitive nature of GAI items.

Overall, this study supports the use of the GAS in assessing anxiety in older

adults. As the world population continues to age, the number of people affected by this

serious condition will exponentially increase as a result. As this occurs, empirically-

supported assessment tools will be increasingly important for use in settings in which

older adults receive healthcare. Concurrently, it is increasingly necessary to raise

healthcare providers’ awareness of the prevalence of this condition in older populations.

This study adds to the emerging literature on anxiety assessment in older adults by

merging classical test and item response theory approaches to assessing psychometrics.

Future studies should continue to examine the utility of the GAS in other populations,

and continue to merge methods for analyzing psychometric properties of assessment

measures.
 REFERENCES

American Psychiatric Association. (2000). Diagnostic and statistical manual of mental

disorders (4th ed. text revision). Washington, DC: Author.

Andresscu, C., Lenze, E. J., Dew, M. A., Begley, A. E., Mulsant, B. H., Dombrovski, A.
Y., et al. (2007). Effect of comorbid anxiety on treatment response and relapse
risk in late-life depression: controlled study. British Journal of Psychiatry, 190,
344-349. DOI:10.1192/bjp.bp.106.027169.

de Ayala, R. J. (2009). The theory and practice of item response theory. New York, NY:
The Guilford Press.

Ayers, C. R., Sorrell, J. T., Thorp, S. R., & Wetherell, J. L. (2007). Evidence-based
psychological treatments for late-life anxiety. Psychology and Aging, 22, 8-17.
DOI:10.1037/0882-7974.22.1.8

Baker, F. (2001). The basics of item response theory. College Park: ERIC Clearinghouse
on Assessment and Evaluation, University of Maryland.

Barg, F., Huss-Ashmore, R., Wittink, M. N., Murray, G. F., Bogner, H. R., & Gallo, J. J.
(2005). A mixed-methods approach to understanding loneliness and depression in
older adults. Journal of Gerontology B: Psychological Sciences, 61, 5329-5339.

Bartlett, M. S. (1954). A note on the multiplying factors for various chi square
approximations. Journal of the Royal Statistical Society, 16, 296-298.

Beck, A. T., Epstein, N., Brown, G., & Steer, R. A. (1988). An inventory for measuring
clinical anxiety: Psychometric properties. Journal of Consulting and Clinical
Psychology, 56, 893-897. DOI:10.1037/0022-006X.56.6.893

Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Beck Depression Inventory (2nd ed.).
San Antonio, TX: The Psychological Corporation.

Beck, A. T., Weissman, A., Lester, D., & Trexler, L. (1974). The measurement of
pessimism: The Hopelessness Scale. Journal of Consulting and Clinical
Psychology, 42, 861-865. DOI: 10.1037/h0037562
 94

Beekman, A. T. F., de Beurs, E., van Balkom, A. J. L. M., Deeg, D. J. H., van Dyck, R.,
& van Tilburg, W. (2000). Anxiety and depression in later life: Co-occurrence
and communality of risk factors. American Journal of Psychiatry, 157, 89-95.
Retrieved from http://ajp.psychiatryonline.org/

Bierman, E. J. M., Comijs, H. C., Jonker, C., & Beekman, A. T. F. (2005). Effects of
anxiety versus depression on cognition in later life. American Journal of Geriatric
Psychiatry, 13, 686-693. doi:10.1176/appi.ajgp.13.8.686
Brenes, G. A., Guralnik, J. M., Williamson, J. D., Fried, L. P., Simpson, C., Simonsick,
E. M., & Penninx, B. W. J. H. (2005). The influence of anxiety on the progression
of disability. Journal of the American Geriatrics Society, 53, 34-39.
DOI:10.1111/j.1532-5415.2005.53007.x

Brock, K., Clemson, L., Cant, R., Ke, L., Cumming, R. G., Kendig, H., & Mathews, M.
(2011). Worry in older community-residing adults. International Journal of Aging
and Human Development, 72, 289-301. DOI:10.2190/AG.72.4.a

Bryant, C., Jackson, H., & Ames, D. (2008). The prevalence of anxiety in older adults:
Methodological issues and a review of the literature. Journal of Affective
Disorders, 109, 233-250. DOI:10.1016/j.jad.2007.11.008

Byrne, G. J., & Pachana, N. A. (2011). Development and validation of a short form of
the Geriatric Anxiety Inventory – the GAI-SF. International Psychogeriatrics,
23, 125-131. DOI:10.1017/S1041610210001237

Byrne, G. J., Pachana, N. A., Goncalves, D. C., Arnold, E., King, R., & Khoo, S. K.
(2010). Psychometric properties and health correlates of the Geriatric Anxiety
Inventory in Australian community-residing older women. Aging & Mental
Health, 14, 247-254.

Cairney, J., Corna, L. M., Veldhuizen, S., Herrmann, N., & Streiner, D. L. (2008).
Comorbid depression and anxiety in later life: Patterns of association, subjective
well-being, and impairment. American Journal of Geriatric Psychiatry, 16, 201-
208. DOI:10.1097/01.JGP.0000300627.93523.c8

Calleo, J., Stanley, M. A., Greisinger, A., Wehmanen, O., Johnson, M., Novy, D. Wilson,
N. & Kunik, M. (2009). Generalized anxiety disorder in older medical patients:
Diagnostic recognition, mental health management, and service utilization.
Journal of Clinical Psychology in Medical Settings, 16, 178-185.
DOI:10.1007/s10880-008-9144-5

Cassidy, E. L., Lauderdale, S., & Sheikh, J. I. (2005). Mixed anxiety and depression in
older adults: Clinical characteristics and management. Journal of Geriatric
Psychiatry and Neurology, 18, 83-88. DOI: 10.1177/0891988705276060
 95

Ciliberti, C., Gould, C., Smith, M., Chorney, D., & Edelstein, B. (2011). A preliminary
investigation of developmentally sensitive items for the assessment of social
anxiety in late life. Journal of Anxiety Disorders, 25, 686-689. DOI:
10.1016/j,janxdis.2011.03.003

Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155-159.

Cully, J.A., Graham, D. P., Stanley, M. A., Ferguson, C. J., Sharafkhaneh, A., Souchek,
J., & Kunik, M. E. (2006). Quality of life in patients with chronic obstructive
pulmonary disease and comorbid anxiety or depression. Psychosomatics, 47, 312-
319. DOI:10.1176/appi.psy.47.4.312

De Beurs, E., Beekman, A. T. F., van Dyck, D. J. H. D., & van Tilburg, W. (2000).
Predictors of change in anxiety symptoms of older persons: Results from the
Longitudinal Aging Study Amsterdam. Psychological Medicine, 30, 515-527.

DeLuca, A. K., Lenze, E. J., Mulsant, B. H., Butters, M. A., Karp, J. F., Dew, M. A., et
al. (2005). Comorbid anxiety disorder in late life depression: Association with
memory decline over four years. International Journal of Geriatric Psychiatry, 20, 848-
854.

Diefenbach, G., Tolin, D. F., Meunier, S. A., & Gilliam, C. M. (2009). Assessment of
anxiety in older home care recipients. Gerontologist, 49, 141-153.
DOI:10.1093/geront/gnp019

Edelen, M. O., & Reeve, B. B. (2007). Applying item response theory (IRT) modeling to
questionnaire development, evaluation, and refinement. Quality of Life Research,
16, 5-18. DOI 10.1007/s11136-007-9198-0

Edwards, M. C. (2009). An introduction to item response theory using the Need for
Cognition Scale. Social and Personality Psychology Compass, 3, 507-529. DOI:
10.1111/j.1751-9004.2009.00194.x

Embretson, S.E., & Reise, S. P. (2000). Item response theory for psychologists. Mahwah,
NJ: Lawrence Erlbaum Associates, Inc.

Flint, A. J., Peasley-Miklus, C., Papademetriou, E., Meyers, B. S., Mulsant, B.,
Rothschild, A. J., & Whyte, E. M. (2010). Effect of age on the frequency of
anxiety disorders in major depression with psychotic features. American Journal
of Geriatric Psychiatry, 18, 404-412.

Gould, C. E., Gerolimatos, L. A., Ciliberti, C., Edelstein, B. A., & Smith, M.D. (2012).
Initial evaluation of the Older Adult Social-Evaluative Situations Questionnaire:
A measure of social anxiety in older adults. International Psychogeriatrics, 24,
2009-2018. doi:10.1017/S1041610212001275
 96

Gum, A. M., King-Kallimanis, B., & Kohn, R. (2009). Prevalence of mood, anxiety, and
substance-abuse disorders for older Americans in the National Comorbidity
Survey Replication. American Journal of Geriatric Psychiatry, 17, 769-781.
DOI:10.1097/JGP.0b013e3181ad4f5a

Hill, R. D., Gallagher, D., Thompson, L. W., & Ishida, T. (1988). Hopelessness as a
measure of suicidal intent in the depressed elderly. Psychology and Aging, 3,
230-232. DOI: 10.1037/0882-7974.3.3.230

Hirsch, J. K., Walker, K. L., Chang, E. C., Lyness, J. M. (2012). Illness burden and
symptoms of anxiety in older adults: Optimism and pessimism as moderators.
International Psychogeriatric Association, 24, 1-8.
doi:10.1017/S1041610212000762

Hopko, D. R., Bourland, S. L., Stanley, M. A., Beck, G., Novy, D. M., Averill, P. M., &
Swann, A. C. (2000). Generalized anxiety disorder in older adults: Examining the
relation between clinician severity ratings and patient self-report measures.
Depression and Anxiety, 12, 217-225. DOI: 10.1002/1520-6394

Hughes, M. E., Waite, L. J., Hawkley, L. C., & Cacioppo, J. T. (2004). A short scale for
measuring loneliness in large surveys: Results from two population-based studies.
Research on Aging, 26, 655-672. doi: 10.1177/0164027504268574

Jorm, A. F. (2000). Does old age reduce the risk of anxiety and depression? A review of
epidemiological studies across the adult life span. Psychological Medicine, 30,
11-22. DOI:10.1017/S0033291799001452

Jimenez, D. E., Alegria, M., Chen, C., Chan, D., & Laderman, M. (2010). Prevalence of
psychiatric illnesses in older ethnic minority adults. Journal of the American
Geriatrics Society, 58, 256-264. DOI: 10.1111/j.1532-5415.2009.02685.x

Kabacoff, R. I., Segal, D. L., Hersen, M., & Van Hasselt, V. B. (1997). Psychometric
properties and diagnostic utility of the Beck Anxiety Inventory and the State-Trait
Anxiety Inventory with older adult psychiatric outpatients. Journal of Anxiety
Disorders, 11, 33-47. DOI:10.1016/S0887-6185(96)00033-3

Karsten, J., Nolen, W. A., Penninx, B. W. J. H., & Hartman, C. A. (2011). Subthreshold
anxiety better defined by symptom self-report than by diagnostic interview.
Journal of Affective Disorders, 129, 236-243. DOI: 10.1016/j.jad.2010.09.006

Katon, W., Lin, E., & Kroenke, K. (2007). The association of depression and anxiety
with medical symptom burden in patients with chronic medical illness. General
Hospital Psychiatry, 29, 147-155. DOI:10.1016/j.genhosppsych.2006.11.005
 97

Kim, G., DeCoster, J., Chiriboga, D. A., Jang, Y., Allen, R. S., & Parmelee, P. (2011).
Associations between self-rated mental health and psychiatric disorders among
older adults: Do racial/ethnic differences exist? American Journal of Geriatric
Psychiatry, 19, 416-422. DOI: 10.1097/JGP.0b013e3181f61ede

Kogan, J. N., Edelstein, B. A., & McKee, D. R. (2000). Assessment of anxiety in older
adults: Current status. Journal of Anxiety Disorders, 14, 109-132.
DOI:10.1016/S0887-6185(99)00044-4

Kroenke, K., Spitzer, R. L., & Williams, J. B. W. (2001). The PHQ-9: Validity of a brief
depression severity measure. Journal of General Internal Medicine, 16, 606-613.
DOI: 10.1046/j.1525-1497.2001.016009606.x

Kroenke, K., Spitzer, R. L., Williams, J. B. W., Monahan, P. O., & Lowe, B. (2007).
Anxiety disorders in primary care: Prevalence, impairment, comorbidity, and
detection. Annals of Internal Medicine, 146, 317-325.

Leach, L. S., Christensen, H., & Mackinnon, A. J. (2008). Gender differences in the
endorsement of symptoms for depression and anxiety: Are gender-biased items
responsible? Journal of Nervous and Mental Disease, 196, 128-135.
DOI:10.1097/NMD.0b013e318162aa63

Leach, L. S., Christensen, H., Mackinnon, A. J., Windsor, T. D., & Butterworth, P.
(2008). Gender differences in depression and anxiety across the adult lifespan:
The role of psychosocial mediators. Social Psychiatry and Psychiatric
Epidemiology, 43, 983-998. DOI: 10.1007/s00127-008-0388-z

Lowe, P. A., & Reynolds, C. R. (2005). Do relationships exist between age, gender, and
education and self-reports of anxiety among older adults? Individual Differences
Research, 3, 239- 259.

Lowe, P. A., & Reynolds, C. R. (2006). Examination of the psychometric properties of

the Adult Manifest Anxiety Scale – Elderly Version scores. Educational and
Psychological Measurement, 66, 93-115. DOI:10.1177/0013164405278563

Mackenzie, C. S., Reynolds, K., Chou, K. – L., Pagura, J., & Sareen, J. (2011).
Prevalence and correlates of Generalized Anxiety Disorder in a national sample of
older adults. American Journal of Geriatric Psychiatry, 19, 305-315.
DOI:10.1097/JGP.0b013e318202bc62

Martin, A., Rief, W., Klaiberg, A., & Braehler, E. (2006). Validity of the Brief Patient
Health Questionnaire Mood Scale (PHQ-9) in the general population. General
Hospital Psychiatry, 28, 71-77. DOI:10.1016/j.genhosppsych.2005.07.003
 98

Marty, M. A., Pepin, R., June, A., & Segal, D. L. (2011). Geriatric Depression Scale. In
M. Abou Saleh, C. Katona, & A. Kumar (Eds.), Principles and practice of
geriatric psychiatry (3rd ed.). New York: Wiley.

Mehta, K. M., Simonsick, E. M., Penninx, B. W. J., Schulz, R., Rubin, S. M., Satterfield,
S., et al. (2003). Prevalence and correlates of anxiety symptoms in well-
functioning older adults: Findings from the health aging and body composition
study. Journal of the American Geriatrics Society, 51, 499-504.
DOI:10.1046/j.1532-5415.2003.51158.x

Mishra, G. D., Gale, C. R., Sayer, A. A., Cooper, C., Dennison, E. M. Whalley, L. J.,
Craig, L., Kuh, D., & Deary, I. J. (2011). How useful are the SF-36 subscales in
older people? Mokken scaling of data from the HALCyon programme. Quality of
Life Research, 20, 1005-1010. DOI:10.1007/s11136-010-9838-7

Mohlman, J., Bryant, C., Lenze, E. J., Stanley, M., Gum, A., Flint, A., … Craske, M. G.
(2011). Improving recognition of late life anxiety disorders in Diagnostic and
Statistical Manual of Mental Disorders, Fifth Edition: Observations and
recommendations of the Advisory Committee to the Lifespan Disorders Work
Group. International Journal of Geriatric Psychiatry, 27, 549-556. DOI:
10.1002/gps.2752

Mohlman, J. (2012). A community based survey of older adults’ preferences for

treatment of anxiety. Psychology of Aging, 27, 1182-190. DOI:
10.1037/a0023126

Morizot, J., Ainsworth, A. T., & Reise, S. (2007). Toward modern psychometrics:
Application of item response theory models in personality research. In R.W.
Robins, R. C. Fraley, & R. F. Krueger (eds.), Handbook of research methods in
personality psychology (pp. 407-423). New York, NY: Guilford Press.

Murphy, L. B., Sacks, J. J., Brady, T. J., Hootman, J. M., Chapman, D. P. (2012).
Anxiety is more common than depression among US adults with arthritis. Anxiety
Care & Research, 64, 968-976. DOI: 10.1002/acr.21685

Nasreddine, Z.S., Phillips, N. A., Bedirian, V., Charbonneau, S., Whitehead, V., Collin,
I., … Chertkow, H. (2005). The Montreal Cognitive Assessment, MoCA: A brief
screening tool for mild cognitive impairment. Journal of the American Geriatrics
Society, 53, 695-699.

Owens, K. M. B., Hadjistavropoulos, T., & Asmundson, G. J. G. (2000). Addressing the

need for appropriate norms when measuring anxiety in seniors. Aging & Mental
Health, 4, 309-314. DOI: 10.1080/13607860020010448.
 99

Pachana, N. A., Byrne, G. J., Siddle, H., Koloski, N., Harley, E., & Arnold, E. (2007).
Development and validation of the Geriatric Anxiety Inventory. International
Psychogeriatrics, 19, 103-114. DOI:10.1017/S1041610206003504

Pachana, N. A., McLaughlin, D., Leung, J., Byrne, G., & Dobson, A. (2011). Anxiety and
depression in adults in their eighties: Do gender differences remain? International
Psychogeriatrics, Available on CJO 2011 DOI:10.1017/S1041610211001372.

Porensky, E. K., Dew, M. A., Karp, J. F., Skidmore, E., Rollman, B. L., Shear, M. K., &
Lenze, E. (2009). The burden of late-life Generalized Anxiety Disorder: Effects
on disability, health-related quality of life, and healthcare utilization. American
Journal of Geriatric Psychiatry, 17, 473-482.
DOI:10.1097/JGP.0b013e31819b87b2

Potvin, O., Bergua, V., Meillon, C., Goff, M. L., Bouisson, J., Dartigues, J. F., &
Amieva, H. (2011). Norms and associated factors of the STAI-Y State anxiety
inventory in older adults: Results from the PAQUID study. International
Psychogeriatrics, 23, 869-879. DOI:10.1017/S1041610210002358

Price, R. B., & Mohlman, J. (2007). Inhibitory control and symptom severity in late-life
generalized anxiety disorder. Behaviour Research and Therapy, 45, 2628-2639.
doi:10.1016/j.brat.2007.06.007

R Core Team. (2012). R: A language and environment for statistical computing

[computer software]. Vienna, Austria: R Foundation for Statistical Computing.
Available from http://www.R-project.org.)

Reise, S. P., & Waller, N. G. (2009). Item response theory and clinical measurement.
Annual Review of Clinical Psychology, 5, 27-48.
DOI:10.1146/annrev.clinpsy.032408.153553

Reynolds, C. R., Richmond, B. O., & Lowe, P. A. (2003). The Adult Manifest Anxiety
Scale-Elderly Version (AMAS-E). Los Angeles: Western Psychological Services.

Robins, R. W., Hendin, H. M., Trzesniewski, K. H. (2001). Measuring global self-

esteem: Construct validation of a single-item measure and the Rosenberg Self-
Esteem Scale. Personality and Social Psychology Bulletin, 27, 151-161.
doi: 10.1177/0146167201272002

Rozanski, A., Blumenthal, J. A., & Kaplan, J. (1999). Impact of psychosocial factors on
the pathogenesis of cardiovascular disease and implications for therapy.
Circulation, 99, 2192-2217. DOI:10.1161/01.CIR.99.16.2192

Samejima, F. (1969). Estimation of latent ability using a response pattern of graded

scores. Psychometrika Monograph Supplement, 34, 100–114
 100

Schaub, R. T., & Linden, M. (2000). Anxiety and anxiety disorders in the old and very
old – Results from the Berlin Aging Study (BASE). Comprehensive Psychiatry,
41, 48-54. DOI:10.1016/S0010-440X(00)80008-5

Segal, D. L., June, A., Payne, M., Coolidge, F. L., & Yochim, B. (2010). Development
and initial validation of a self-report assessment tool for anxiety among older
adults: The Geriatric Anxiety Scale. Journal of Anxiety Disorders, 24, 709-714.
DOI: 10.1016/j.janxdis.2010.05.002

Segal, D. L., Qualls, S. & Smyer, M. A. (2011). Aging and mental health (2nd edition).
Malden, MA: John Wiley and Sons.

Serrano, J. P., Latorre, J. M., Gatz, M., & Montanes, J. (2004). Life review therapy using
autobiographical retrieval practice for older adults with depressive
symptomatology. Psychology and Aging, 19, 272-277. DOI: 10.1037/0882-
7974.19.2.272

Stordal, E., Bjelland, I., Dahl, A. A., & Mykletun, A. (2003). Anxiety and depression in
individuals with somatic health problems: The Nord-Trondelag Health Study
(HUNT). Scandinavian Journal of Primary Health Care, 21, 136-141. DOI:
10.1080/02813430310002030

Strine, T. W., Chapman, D. P., Kobau, R., & Balluz, L. (2005). Associations of self-
reported anxiety symptoms with health-related quality of life and health
behaviors. Social Psychiatry and Psychiatric Epidemiology, 40, 432-438.
DOI:10.1007/s00127-005-0914-1

Suls, J., & Bunde, J. (2005). Anger, anxiety and depression as risk factors for
cardiovascular disease: The problems and implications of overlapping affective
dispositions. Psychological Bulletin, 131, 260-300. DOI:10.1037/0033-
2909.131.2.260

Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th ed.). Boston:
Pearson Edition.

Therrien, Z., & Hunsley, J. (2011). Assessment of anxiety in older adults: A systematic
review of commonly used measures. Aging & Mental Health, 15, 1-16.
DOI:10.1080/13067863.2011.602960

Van Dam, N. T., Earleywine, M., & Forsyth, J. P. (2009). Gender bias in the sixteen-item
Anxiety Sensitivity Index: An application of polytomous differential item
functioning. Journal of Anxiety Disorders, 23, 256-259.
DOI:10.1016/j.janxdis.2008.07.008
 101

van Hout, H. P. J., Beekman, A. T. F., de Beurs, E. Comijs, H., van Marwijk, H., de
Haan, M., et al. (2004). Anxiety and the risk of death in older men. British
Journal of Psychiatry, 185, 399-404. DOI: 10.1192/bjp.185.5.399

Ware, J. E., & Sherbourne, C. D. (1992). The MOS 36-item short-form health survey
(SF-36): Conceptual framework and item selection. Medical Care, 30, 473-483.

Wetherell, J. L., & Areán, P. A. (1997). Psychometric evaluation of the Beck Anxiety
Inventory with older medical patients. Psychological Assessment, 9, 136-144.
DOI:10.1037/1040-3590.9.2.136

Wetherell, J. L., Petkus, A. J., McChesney, K., Stein, M. B., Judd, P. H., Rockwell, E., …
Patterson, T. L. (2009). Older adults are less accurate than younger adults at
identifying symptoms of anxiety and depression. Journal or Nervous and Mental
Disease, 197, 623-626. DOI: 10.1097/NMD.0b013e3181b0c081

Wetherell, J. L., Ayers, C. R., Nuevo, R., Stein, M. B., Ramsdell, J., & Patterson, T. L.
(2010). Medical conditions and depressive, anxiety, and somatic symptoms in
older adults with and without generalized anxiety disorder. Aging & Mental
Health, 14, 764-768. DOI: 10.1080/13607861003713240.

Williams, M. T., Chapman, L. K., Wong, J., & Turkheimer, E. (2012). The role of ethnic
identity in symptoms of anxiety and depression in African Americans. Psychiatry
Research. http://dx.doi.org/10.1016/j.psychres.2012.03.049.

Wolitzky-Taylor, K. B., Castriotta, N., Lenze, E. J., Stanley, M. A., & Craske, M. G.
(2010). Anxiety disorders in older adults: A comprehensive review. Depression
and Anxiety, 27, 190-211. DOI: 10.1002/da.20653

Yesavage, J. A., Brink, T. L., Rose, T. L., Lum, O., Huang, V., Adey, M., & Leirer, V. O.
(1983). Development and validation of a geriatric depression screening scale: A
preliminary report. Journal of Psychiatric Research, 17, 37-49.
DOI:10.1016/0022-3956(82)90033-4

Yerkes, R. M., & Dodson, J. D. (1908). The relation of strength of stimulus to rapidity of
habit formation. Journal of Comparative Neurology and Psychology, 18, 459-
482.

Yochim, B. P., Mueller, A. E., June, A., & Segal, D. L. (2011). Psychometric properties
of the Geriatric Anxiety Scale: Comparison to the Beck Anxiety Inventory and
Geriatric Anxiety Inventory. Clinical Gerontologist, 34, 21-33.
DOI: 10.1080/07317115.2011.524600

Yochim, B. P., Mueller, A. E., & Segal, D. L. (2012). Late life anxiety is associated with
decreased memory and executive functioning in community dwelling older adults.
Journal of Anxiety Disorders.
 102

Zumbo, B.D. (1999). A Handbook on the Theory and Methods of Differential Item
Functioning (DIF): Logistic Regression Modeling as a Unitary Framework for
Binary and Likert-Type (Ordinal) Item Scores. Directorate of Human Resources
Research and Evaluation, Department of National Defense; Ottawa, ON.
 103

APPENDIX A

GERIATRIC ANXIETY SCALE – VERSION 1.0

Not at all Sometimes Most of All of the

the time time

1. My heart raced or beat strongly.

2. My breath was short.
3. I had an upset stomach.
4. I felt like things were not real or like I was
outside of myself.
5. I felt like I was losing control.
6. I was afraid of being judged by others.
7. I was afraid of being humiliated or
embarrassed.
8. I had difficulty falling asleep.
9. I had difficulty staying asleep.
10. I was irritable.
11. I had outbursts of anger.
12. I had difficulty concentrating.
13. I was easily startled or upset.
14. I was less interested in doing something I
typically enjoy.
15. I felt detached or isolated from others.
16. I felt like I was in a daze.
17. I had a hard time sitting still.
18. I worried too much.
19. I could not control my worry.
20. I felt restless, keyed up, or on edge.
21. I felt tired.
22. My muscles were tense.
23. I had back pain, neck pain, or muscle
cramps.
24. I felt like I had no control over my life.
25. I felt like something terrible was going to
happen to me.
26. I was concerned about my finances.
27. I was concerned about my health.
28. I was concerned about my children.
29. I was afraid of dying.
30. I was afraid of becoming a burden to my
family or children.
 104

APPENDIX B

SCORING INSTRUCTIONS

Items 1 through 25 are scorable items. Each item ranges from 0 to 3. Each item loads on
only one scale. Items 26 through 30 are used to help clinicians identify areas of concern for the
respondent. They are not used to calculate the total score of the GAS or any subscale.

Total Score = sum of items 1 through 25.

Somatic subscale (9 items) = sum of items 1, 2, 3, 8, 9, 17, 21, 22, 23
Cognitive subscale (8 items) = sum of items 4, 5, 12, 16, 18, 19, 24, 25
Affective subscale (8 items) = sum of items 6, 7, 10, 11, 13, 14, 15, 20

Subscale Item # Item

------------------------------------------------------------------------------------------------------------

Somatic 1 My heart raced or beat strongly.

Somatic 2 My breath was short.
Somatic 3 I had an upset stomach.
Somatic 8 I had difficulty falling asleep.
Somatic 9 I had difficulty staying asleep.
Somatic 17 I had a hard time sitting still.
Somatic 21 I felt tired.
Somatic 22 My muscles were tense.
Somatic 23 I had back pain, neck pain, or muscle cramps.

Cognitive 4 I felt like things were not real or like I was outside of myself.
Cognitive 5 I felt like I was losing control.
Cognitive 12 I had difficulty concentrating.
Cognitive 16 I felt like I was in a daze.
Cognitive 18 I worried too much.
Cognitive 19 I could not control my worry.
Cognitive 24 I felt like I had no control over my life.
Cognitive 25 I felt like something terrible was going to happen to me.

Affective 6 I was afraid of being judged by others.

Affective 7 I was afraid of being humiliated or embarrassed.
Affective 10 I was irritable.
Affective 11 I had outbursts of anger.
Affective 13 I was easily startled or upset.
Affective 14 I was less interested in doing something I typically enjoy.
Affective 15 I felt detached or isolated from others.
Affective 20 I felt restless, keyed up, or on edge.
 105

APPENDIX C

GERIATRIC ANXIETY SCALE-8

Item 1 (My heart raced or beat strongly.)

Item 2 (My breath was short.)
Item 3 (I had an upset stomach.)
Item 6 (I was afraid of being judged by others.)
Item 7 (I was afraid of being humiliated or embarrassed.)
Item 10 (I was irritable.)
Item 11 (I had outbursts of anger.)
Item 17 (I had a hard time sitting still.)
 106

APPENDIX D

GERIATRIC ANXIETY SCALE-10 ITEM VERSON (GAS-10)

Item 10 (I was irritable).

Item 15 (I felt detached or isolated from others).
Item 16 (I felt like I was in a daze).
Item 17 (I had a hard time sitting still).
Item 19 (I could not control my worry).
Item 20 (I felt restless, keyed up, or on edge).
Item 21 (I felt tired).
Item 22 (My muscles were tense).
Item 24 (I felt like I had no control over my life).
Item 25 (I felt like something terrible was going to happen to me).