Mood States Associated With Transitory Changes in Asthma Symptoms and Peak

Expiratory Flow
GLENN AFFLECK, PHD, ANDREA APTER, MD, HOWARD TENNEN, PHD, SUSAN REISINE, PHD, ERIK BARROWS, BA,
ALICE WILLARD, RN, BSN, JENNIFER UNGER, BA, AND RICHARD ZUWALLACK, MD

Objective: This study examined the within-person relations between transitory changes in mood, asthma symptoms,
and peak expiratory flow rate (PEFR). Methods: Thrice-daily for 21 consecutive days, 48 adults with moderate to
severe asthma entered information in palm-top computers about their mood and asthma symptoms. A multidimen-
sional model of mood, ie, the mood circumplex, informed the assessment of mood arousal and mood pleasantness.
At each observation, participants also recorded their PEFR with peak flow meters that stored blinded data. Albuterol
doses were also monitored electronically. Before and after the 21-day study, spirometric measures of airways
obstruction were taken under controlled conditions. Results: Random effects regression models revealed a signif-
icant, but weak, within-person relation between symptoms and PEFR. Changes in mood vectors with an arousal
component were significantly related to PEFR changes, whereas changes in mood vectors with a pleasantness
component tracked changes in asthma symptom reports, even after adjustment for contemporaneous PEFR and after
controlling for time of day and albuterol dosing. Comparison of spirometric assessments with unsupervised PEFR
suggested that part of the relation between mood arousal and PEFR may be attributable to the “effort-dependence”
of peak flow self-monitoring. Conclusions: Different dimensions of mood were associated with transitory changes
in asthma symptoms and PEFR. This may be one reason why individuals with asthma misperceive the severity of
their symptoms in relation to underlying airways obstruction. Key words: asthma, mood, peak flow, symptoms, diaries.

and asthma symptom reports (5– 8). Consequently, ad-

PEFR ⫽ peak expiratory flow rate, ELI ⫽ electronic
interviewer
INTRODUCTION
Despite steady progress in the understanding and
treatment of asthma, the morbidity and mortality of
this disease remain disturbingly high (1). One reason
for the suboptimal control of asthma complications
may lie in patients’ perception of their airways ob-
struction. Those who underestimate its severity may
delay seeking necessary medical care or comply poorly
with therapy (2–3). Conversely, those who overesti-
mate airways obstruction may use too much aerosol
bronchodilator therapy, which in turn could increase
the underlying severity of the disease (4). The likeli-
hood of these perceptual “errors” is suggested by many
studies that have documented only a weak correlation
between objective measures of airways obstruction
From the Departments of Community Medicine (G.A., H.T., J.U.)
and Behavioral Sciences and Community Health (S.R.) and the Gen-
eral Clinical Research Center (E.B., A.W.), University of Connecticut
School of Medicine, Farmington, Connecticut; Section of Allergy
and Immunology (A.A.), University of Pennsylvania School of Med-
icine, Philadelphia, Pennsylvania; and Department of Medicine
(R.Z-W.), St. Francis Hospital and Medical Center, Hartford, Con-
necticut.
Address reprint requests to: Glenn Affleck, PhD, Department of
Community Medicine, University of Connecticut Health Center,
Farmington, CT 06030. Email: affleck@nso1.uchc.edu
Received for publication March 15, 1999; revision received July
19, 1999.
vances in the clinical management of asthma may
come from identifying processes that regulate both
bronchoconstriction and the perception of asthma
symptoms and that could help explain the weak con-
cordance between these phenomena.
The present study highlights the role that transi-
tory changes in mood play in the fluctuation of
airways obstruction and symptom reports in adults
with asthma. Many investigations have documented
significant associations of emotional states such as
panic, fear, anxiety, arousal, and fatigue with asthma
severity, whether measured by symptom reports or
peak flow monitors (9). This literature, however, is
limited to inferences that justifiably can be drawn
from “between-person” research designs, namely,
that individuals who are, for example, more anxious
tend to experience more severe asthma symptoms.
Few have examined whether changes in these vari-
ables covary within individuals over time; in other
words, whether changes in anxious mood actually
track changes in asthma severity in the course of
daily life for any of the individuals under study.
Such time-intensive “daily process” designs can
provide fresh insights through their capacity to min-
imize retrospection errors in symptom and mood
reporting, use subjects as their own controls, and
establish sequential relations from the many com-
parisons available (10). Brown and Moskowitz (11)
anticipate that this approach will advance behav-
ioral and psychosomatic medicine through its ability
to better capture the ebb and flow of the many psy-
chological and somatic processes that exhibit rapid
moments of change.

Psychosomatic Medicine 62:61– 68 (2000) 61

0033-3174/00/6201-0061
Copyright © 2000 by the American Psychosomatic Society

WITHIN-PERSON STUDIES OF MOOD AND
ASTHMA SEVERITY
Few published studies have inspected relations be-
tween changes in mood and asthma severity at the
within-person level from day to day. Steptoe and
Holmes (12) measured mood and peak expiratory flow
rate (PEFR) four times daily over 20 days in 7 men with
asthma, and 7 nonasthmatic men. Mood ratings in-
cluded “angry-calm,” “relaxed-tense,” and “elated-de-
pressed.” None of the men without asthma, but three
of the asthmatic men, exhibited a significant relation
between one or more mood states and PEFR. Hyland
(13) examined mood-PEFR relations in 10 adults with
asthma by twice-daily observations over 15 days. Par-
ticipants rated an array of 14 negative moods and 10
positive moods. Combining these in a single bipolar
mood scale, Hyland identified three subjects whose
increasing positive mood (and decreasing negative
mood) tracked their changes in PEFR.
Neither of these small sample studies explored the
possibility that there may be different mood correlates
of PEFR as opposed to symptom perceptions. There is
now substantial literature linking affective states with
both minor and chronic illness symptoms (14 –16).
Symptom exacerbations may have emotional conse-
quences, but certain emotions may intensify symp-
toms by increasing attention toward the self. In partic-
ular, variations in sad (or happy) mood may trigger (or
attenuate) both the process of “turning inward”
(17, 18) as well as the amplification of physical symp-
toms (19). For these reasons, Pennebaker (20) con-
cluded that symptom perceptions can have stronger
relations with emotional states than with the underly-
ing physiological processes that create symptoms.
We hypothesize that variations in happy and sad
mood would be most likely to track changes in symp-
tom reports and would do so even after adjusting this
relation for concurrent PEFR. We conducted a prelim-
inary study of this possibility with 21 asthmatic adults
who rated their mood, provided PEFR readings, and
chronicled their symptoms three times a day for 21
days (7). The data were pooled across persons and
observations and analyzed at the within-person level
by fixing the differences between participants in their
mean levels of mood, PEFR, and symptoms. Rising
PEFR scores were most highly related to declining
fatigue and increasing liveliness. Improvements in
symptoms, in contrast, were most likely to correspond
with increasing happiness and decreasing sadness.
These findings stand as preliminary evidence that
short-term fluctuations in symptom perceptions and
bronchoconstriction may correlate with different transi-
tory moods. However, we did not conduct the multivar-
62
G. AFFLECK et al.
iate analyses required to test this hypothesis adequately.
Nor did our “fixed effects” statistical analysis allow us to
generalize our findings to the population from whom the
sample was drawn (21). Time-varying covariates, such as
time of day and medication use, were also not considered
as possible confounds of within-person relations be-
tween mood, symptoms, and peak flow.
The present study extends and refines previous re-
search on this topic in several ways. First, we studied
a larger sample and conducted multilevel random ef-
fects analyses that do permit generalization to the sam-
pled population. Second, we determined whether the
relations between mood and asthma severity could be
specious because they either share the same, but inde-
pendent, circadian pattern or they are both affected by
the use of as-needed (PRN) asthma medications.
Changes in mood (22, 23) and in airway obstruction
(13, 24) across the day have been documented. Albu-
terol dosing is associated with symptom aggravation
and PEFR (25) and may precipitate mood changes.
A third key feature of the study is its reliance on a
theory-derived multidimensional model of mood as-
sessment. Two general models have guided the self-
report measurement of mood: the specific affects ap-
proach and the dimensional approach (26). The first
reflects “the belief that there are many different types
of mood, each with different. . . characteristics and
response patterns.” The second model takes the posi-
tion that “there are a few, usually two, ”core“ dimen-
sions of mood; specific moods are thought to be com-
binations of [these] dimensions” (26, p. 150). For the
present study, we adopted the theory-driven dimen-
sional approach expressed in the circumplex model of
emotion first proposed by Russell (27) and elaborated
by Larsen and Diener (28). This approach places self-
ratings of mood in a two-dimensional circular space,
which distinguishes states according to their degree of
pleasantness (or hedonic tone) and degree of activation
(or arousal tone). Thus, for example, anxious mood
combines high unpleasantness and high arousal,
whereas calm mood combines high pleasantness with
low arousal. We adopted this model principally be-
cause it distinguishes between arousal states, which
may be most likely to correlate with peak flow ratings,
and hedonic states, which may figure most strongly in
symptom reports (7).
Finally, a novel feature of this investigation is its
use of electronic methods for field measurements of all
study variables: mood, symptoms, peak expiratory
flow, and beta-agonist use. These methods assured
participants’ compliance with demanding schedules
for self-reports of mood and symptoms; blinded par-
ticipants to their peak flow records when describing
symptoms; and measured medication use without re-
Psychosomatic Medicine 62:61– 68 (2000)
MOOD, ASTHMA SYMPTOMS, PEFR

lying on self-report. Taken together, these procedures sequence of three requests for data produced a missing entry for that
mitigate a formidable threat to the validity of experi- time period.
Interview questions were presented one at a time in a fixed order
ence sampling studies.
on a liquid crystal display (2 lines by 20 characters per line). Par-
ticipants replied to each question by scrolling across fixed-response
METHODS AND PROCEDURES options with backward and forward arrows and then pressing an
“enter” button to save the response and its time stamp on a EEPROM
Characteristics of the Sample data pak (which could not be erased without exposing it deliberately
to 30 minutes of ultraviolet light). The response option appearing
Sixty-three adults with moderate to severe asthma being treated
first on the screen with each new question was randomized to
at the allergy and pulmonary clinics of a university and an affiliated
minimize response set. The project’s research associate trained par-
community hospital were asked to participate in the study, and 50
ticipants on a demonstration version of ELI and gave them a manual
accepted the invitation. These participants were required to have a
which reviewed its functions and features.
prebronchodilator forced expiratory flow rate in 1 second (FEV1) of
The mood interview. The 16 items for the mood interview were
less than 80% of predicted for their gender, weight, and height. In
drawn from the mood circumplex items supplied by Larsen and
addition, all demonstrated a 15% or greater increase in FEV1 after
Diener (28). This model classifies affect as pleasant or unpleasant, as
bronchodilator administration within the past year. All participants
aroused or unaroused, or as a mixture of these dimensions. Figure 1
had at least four of the following pretreatment characteristics for
diagrams the resulting mood octants and the specific mood adjec-
National Heart, Lung, and Blood Institute criteria for moderate asth-
tives used to measure them.
ma: exacerbations of cough/wheeze more than once or twice per
To reduce the number of analyses required, four bipolar mood
week; cough and wheeze between acute exacerbations often present;
vector scores comprising four adjectives apiece were derived from
reduced exercise tolerance; PEFR ⬍80% predicted with a reversibil-
the circumplex. The first was the degree of mood pleasantness/
ity of 20% to 30%; nocturnal symptoms at least two to three times
unpleasantness with no arousal distinction (ie, a happy-sad vector)
per week; school or work attendance compromised by asthma; sys-
and was calculated by subtracting the two unpleasant mood adjec-
temic steroids usually necessary to treat exacerbations; and regular
tives in octant 7 from the two pleasant mood adjectives in octant 3.
use of oral or inhaled corticosteroids or cromolyn required daily for
The second was the degree of mood arousal/unarousal with no
more than 2 months of the year before entry into this study. All were
pleasantness distinction (ie, an active-passive vector) and was cal-
currently nonsmokers with a less than a 10 pack-year history of
culated by subtracting the unaroused mood adjectives in octant 5
tobacco use. None had significant pulmonary disease other than
from the aroused mood adjectives in octant 1. The third comprised
asthma. During the initial orientation, each participant was required
the opposite poles of unaroused pleasant mood and aroused un-
to demonstrate reliable use of all electronic devices. Two partici-
pleasant mood (ie, a calm-anxious vector) and was calculated by
pants were excluded from subsequent analyses because of technical
subtracting the adjectives in octant 2 from those in octant 6. The
malfunctions with one of the electronic devices.
fourth combined the opposite poles of aroused unpleasant mood and
The remaining 48 patients (64.6% women; 81.3% white; 54.2%
unaroused pleasant mood (ie, a peppy-drowsy vector) and was
married) had a mean age of 42.1 years (SD ⫽ 14.8) and a mean of 14.4
scored by subtracting the adjectives in octant 4 from those in octant
years of formal education (SD ⫽ 2.1). The average participant had been
8.
diagnosed with asthma 22.8 years earlier (SD ⫽ 15.9). Since then, they
We confirmed the appropriateness of this scoring procedure by
reported having had a mean of 4.6 asthma-related hospitalizations
conducting principal components analyses of the disaggregated data
(SD ⫽ 14.9) and 4.4 asthma-related emergency room visits (SD ⫽ 9.33).
set. The mood vectors in Figure 1, which are orthogonal to each
They recalled having missed a mean of 4.3 days of work during the past
other in theory, should have empirical separation as well. In other
year (SD ⫽ 17.6) because of their asthma symptoms.
words, the four adjectives composing the pleasant/unpleasant vector
should form a component separate from that for the four adjectives
Electronic Interviews of Mood and Asthma
Symptoms
For 21 consecutive days, participants carried a palm-top com-
puter programmed as an electronic interviewer (ELI), which asked
them about their current asthma symptoms and mood three times a
day at randomly selected times— once each during the morning
(between 9:45 AM and 11:15 AM), afternoon (between 2:45 PM and
4:15 PM), and evening (6:45 PM and 9:15 PM). The computer was a
programmable battery-powered Psion Organizer II (Psion, Concord,
MA) with dimensions of 1.4 cm ⫻ 7.8 cm ⫻ 2.9 cm and weight of
250 g. This device has amply demonstrated its feasibility and reli-
ability in prospective daily studies of pain, mood, and fatigue
(29, 30); drinking and smoking (31, 32); asthma symptoms (33); and
sleep quality (30, 34).
Some procedures for the ELI protocol parallel those designed by
Shiffman and colleagues (31) for their electronic diary studies of
cigarette smokers. The data entry procedure for each ELI request
proceeded from the user’s termination with a keystroke of an audible
beep to her choice to answer the interview then 5 minutes later, or
15 minutes later. The auditory signal lasted 60 seconds; if not
answered within this time, it was repeated 5 minutes later, and if not Fig. 1. Mood circumplex octants and adjectives used for measure-
answered again, another 5 minutes later. Failure to answer this ment.

Psychosomatic Medicine 62:61– 68 (2000) 63


composing the aroused/unaroused vector. Similarly, the four adjec-
tives composing the aroused unpleasant/unaraoused pleasant vector
should form a component separate from that for the adjectives com-
posing the aroused pleasant/unaroused unpleasant vector.
Two principal components analyses with varimax rotation con-
firmed these predictions. Each mood score was first centered around
the person’s own mean to examine the within-person covariance
structure free of between-person differences in mean levels. The
covariance pattern among the eight adjectives included in the pleas-
ant/unpleasant vector and the aroused/unaroused vector yielded
two components with eigenvalues greater than 1.0 and accounting
for 67.2% of the variance. Each adjective loaded highly on its pre-
dicted vector (⬎0.70 or ⬍⫺0.80). The same was found for the eight
adjectives contained in the other two orthogonal vectors. Two com-
ponents with eigenvalues greater than 1.0 accounted for 67.1% of
the variance, and each adjective loaded highly on its predicted
vector (⬎0.80 or ⬍⫺0.80).
The asthma symptom interview. The ELI asked about four current
asthma symptoms: coughing, wheezing, chest tightness, and short-
ness of breath. Each was rated on a 0 – 6 scale, anchored verbally at
0 ⫽ none, 2 ⫽ mild, 4 ⫽ moderate, and 6 ⫽ severe. Asthma symptom
severity for that interview was scored as the sum across these rat-
ings. Using the mean-centering procedure described previously,
Cronbach ␣ internal consistency estimates for the asthma symptom
severity composite were 0.94 for 79 for the morning reports, 0.96 for
the afternoon reports, and 0.93 for the evening reports.
Electronic Assessments of PEFR
Participants used an electronic peak flowmeter (PeakLog, Medt-
rac Technologies, Inc., Lakewood, CO) to measure their airways
obstruction. This instrument, which is 13 cm ⫻ 6 cm x 2.5 cm and
weighs 138 g, measures airflow with a hot-wire anemometer. Accu-
racy was tested by the manufacturer using recommendations by the
National Asthma Education Program (35). This involved injecting
nine standardized wave forms into each of 10 devices five times
using a computer-driven pulmonary wave form generator. All in-
struments were within 10% of the target values and were within 5%
of each other. A built-in barometer allows for self-correction for
fluctuations in barometric pressure. The device received FDA ap-
proval on November 1, 1994.
PEFR was recorded along with the time and date of measurement
by the electronic peak flowmeter after each electronic interview.
Participants were given standardized instructions and observed in
the use of the device during prestudy training. After prompting by
the electronic diary (with the completion of the symptom interview),
patients were instructed to stand, take a deep breath, fill their lungs
completely, place the mouthpiece in the mouth past the teeth with
lips tightly closed around the mouthpiece, and blow as hard and fast
as possible in a single exhalation. Three such expiratory maneuvers
were obtained, and the highest PEFR was reserved for data analysis.
The validity of peak flow monitor actuations in the field can be
strengthened by establishing their correlation with assessments un-
der controlled conditions. At the beginning and end of the self-
monitoring period, spirometry was performed on all participants
according to criteria of the American Thoracic Society (36) and the
scores averaged across time. Mean daily peak flow ratings were
correlated with spirometric measures of PEFR (r ⫽ 0.81, p ⬍ .001)
and FEV1 (r ⫽ 0.69, p ⬍ .001).
Electronic Recording of Beta-Agonist Use
The only short-acting beta-agonist allowed in the study was
albuterol by metered dose inhalation, prescribed on an as-needed
64
G. AFFLECK et al.
basis. The canister was housed in an MDI Chronolog (Medtrac Tech-
nologies, Inc., Lakewood, CO) which recorded time and date of each
actuation. For our data analyses, we calculated the number of doses
during each morning, afternoon, and evening period dictated by the
electronic interviewer protocol.
RESULTS
Descriptive Findings
Of the 3024 symptom/mood interviews requested,
2947 (97.5%) were completed, and of an equal number
of electronic peak flowmeter actuations requested,
2834 (93.7%) were completed. The mean PEFR across
persons and observations was 347.9 (SD ⫽ 95.1). On
scales that could range from ⫺12 to 12, with higher
numbers representing more pleasant or more aroused
mood states, the mean happy-sad score was 5.9 (SD ⫽
2.2); the mean active-passive score was 1.2 (SD ⫽ 1.7);
the mean calm-anxious score was 5.0 (SD ⫽ 2.2); and
the mean peppy-drowsy score was 1.8 (SD ⫽ 2.5). The
mean asthma symptom score, on the 0 – 6 point scale,
was 3.8 (SD ⫽ 2.9). On average, there were 1.92 albu-
terol doses administered for each 24-hour period
(SD ⫽ 1.42).
Between-Person Relations
Correlations between participants’ average values
on these variables were calculated to examine be-
tween-person associations. Individuals with higher ag-
gregate PEFR (transformed as the ratio between mean
raw scores and those predicted by the subject’s weight,
height, and age) reported more severe symptoms (r ⫽
⫺0.30, p ⬍ .05). As Table 1 indicates, mean PEFR was
unrelated to mean mood vector scores. However, those
with more severe asthma symptoms scored lower on
the happy-sad mood vector, on the calm-anxious mood
vector, and on the peppy-drowsy mood vector. After
controlling for PEFR, mean asthma symptoms corre-
lated significantly with all mood vector scores. Albu-
terol use was higher for those reporting more severe
TABLE 1. Between-Person Relations of Mean Mood Vector
Scores with Mean PEFR and Mean Asthma Symptoms
Mean Asthma
Mean Mean Asthma Symptoms
Mean Mood
PEFR Symptoms Controlling for
Mean PEFR
Happy—sad .12 ⫺.30* ⫺.29*
Active—passive ⫺.12 ⫺.26 ⫺.32*
Calm—nervous .21 ⫺.36* ⫺.32*
Peppy—drowsy .05 ⫺.42** ⫺.42**
* p ⬍ .05, 46 df.
** p ⬍ .01, 46 df.
Psychosomatic Medicine 62:61– 68 (2000)
MOOD, ASTHMA SYMPTOMS, PEFR

TABLE 2. Within-person Relations of Mood Vector Scores with Peak Expiratory Flow and Asthma Symptoms

Asthma Symptoms
PEF Asthma Symptoms
(Controlling for PEF)
Mood Vector
ba t p b t p b t p

Happy—sad b
.62 1.87 Ns ⫺.08 ⫺3.33 ⬍.01 ⫺.06 ⫺2.75 ⬍.01
Active—passivec 1.38 3.99 ⬍.001 .00 .20 Ns .02 .84 Ns
Calm—nervousd ⫺.34 ⫺1.11 Ns ⫺.07 ⫺3.54 ⬍.001 ⫺.08 3.50 ⬍.001
Peppy—drowsye 1.68 4.39 ⬍.001 ⫺.05 ⫺2.65 ⬍.01 ⫺.03 ⫺1.95 Ns

a
Unstandardized maximum likelihood estimate.
b
Pleasant, unpleasant mood vector, measured by happy, cheerful, sad, and blue.
c
Aroused, unaroused mood vector, measured by active, lively, passive, and quiet.
d
Unaroused/pleasant, aroused/unpleasant mood vector, measured by calm, relaxed, anxious, and nervous.
e
Aroused/pleasant, unaroused/unpleasant mood vector, measured by peppy, stimulated, drowsy, and tired.

symptoms (r ⫽ 0.40, p ⬍ .001) but was unrelated to
average peak flow ratings or to average mood scores.
Approach to Within-Person Data Analysis
1A valid within-person analysis of the relations
among PEFR, asthma symptoms, and mood requires a
multilevel modeling strategy that partitions the two
sources of variation in the person-observation data set:
differences between persons in the mean levels of the
observations and differences within persons in their
own data over time. To generalize our findings to both
the population of individuals from which the sample
was drawn and to the population of days from which
their daily experiences were sampled, we used a ran-
dom effects regression model, which allowed between-
person differences in intercepts (means) and slopes
(within-person relations) to vary randomly when cal-
culating parameters for mood-PEFR-symptom relations
(37). This procedure provides average estimates of with-
in-person relations regardless of differences between per-
sons in levels of the variables under study. The PROC
MIXED procedure in SAS (38) furnished model parame-
ters in the form of maximum likelihood estimates.1
Within-person changes in symptoms significantly
tracked changes in PEFR (b ⫽ ⫺0.005, p ⬍ .001). Yet,
only 3.7% of the within-person variance in symptoms
could be explained by peak flow actuations at the time
symptoms were reported.2 This suggests that much of
1
Because equal interval time-dependent data are likely to be
nonindependent, leading to autocorrelated residuals and the mises-
timation of standard errors, an error structure assuming a higher
correlation between consecutive error terms, ie, an AR (1) structure,
was fit to each of the statistically significant models. None of the
significant findings reported here were altered by this procedure.
2
The HLM estimates of within-person variance explained in these
models should not be confused with those provided in OLS regres-
sion (39). These estimates refer to the ability of explanatory variables to
reduce the random variance components of their respective equations.
the differences in the average person’s symptom re-
ports from morning to afternoon and afternoon to
evening each day are not because of changes in bron-
choconstriction, at least as they are measured by PEFR
in the field. This leaves substantial room for associa-
tions between mood and asthma symptoms when
PEFR is controlled and the opportunity to examine
symptom perceptions independent of an objective
measure of bronchoconstriction.3
Within-Person Relations
Table 2 summarizes the within-person relations be-
tween the four mood vector scores and PEFR. Rising
scores on the active-passive and the peppy-drowsy
mood vectors were associated with increasing PEFR.
When examined jointly as predictors of PEFR, each
was also an independent correlate of PEFR changes
(active-passive, b ⫽ 0.70, p ⬍ .05; peppy-drowsy, b ⫽
1.28, p ⬍ .001). Together, these two mood vectors
accounted for 4% of the within-person variance in
PEFR.
Table 2 also lists the parameter estimates for within-
person relations between mood and symptoms, with
and without controlling for that observation’s PEFR.
Two mood vectors— happy-sad and calm-nervous—
remained related to fluctuating symptoms regardless
of PEFR changes. When examined together as predic-
tors, both vectors remained significantly correlated
with asthma symptoms (happy-sad, b ⫽ ⫺0.07, p ⬍
.05; calm-nervous, b ⫽ ⫺ 0.08, p ⬍ .01, respectively).
Together, these two mood vectors explained 4.6% of
3
A companion set of analyses examined the ability of mood to
predict the next within-day observation for symptoms or peak flow,
as well as the reverse sequence. Controlling for contemporaneous
associations, there were no significant lagged relations for any of
these analyses.

Psychosomatic Medicine 62:61– 68 (2000) 65


the within-person variance in symptoms not already
accounted for by variance in PEFR.
These analyses take account of all observations in
the data set. But in so doing, they may miss effects that
are revealed only when asthma symptoms or airways
obstruction are most severe. To examine this possibil-
ity, we recomputed the analyses of relations of mood
with asthma symptoms and PEFR by comparing mood
reports at extreme observations. The first compared
mood scores at moments when PEFR was relatively
high for that individual (⬎1 SD above that person’s
mean) with those when PEFR was relatively low (⬎1
SD below the mean). Only the peppy-tired mood vec-
tor was significantly different for these two sets of
observations (b ⫽ 0.29, p ⬍ .05). The second analysis
compared mood scores at moments when asthma
symptoms were relatively severe and relatively mild.
Only the happy-sad mood vector differed significantly
between these two sets of observations (b ⫽ ⫺0.30, p ⬍
.04). These two significant findings echo significant
effects reported in Table 1 for the full spectrum of
observations. However, two other significant effects
appearing in Table 1 were missed by this analysis,
suggesting that mood may play a more prominent role
in explaining the total variation of PEFR and asthma
symptoms across the day than it does in the ability to
differentiate between daily episodes of comparatively
severe vs mild symptomatology.
Next, we examine three possible sources of con-
founding that might spuriously inflate the associations
of mood with PEFR and symptoms: a shared time
course across reporting occasions within the day; as-
needed administration of albuterol; and the effort-de-
pendence of unsupervised peak flow monitoring.
Time of Day Effects?
The relations summarized to this point could sim-
ply be due to sharing the same time course across daily
observations. To address this possibility, each variable
was modeled for within-person changes across the
day. Asthma symptoms did not differ by time of day,
but PEFR did. As the day progressed, PEFR declined
(b ⫽ ⫺5.94, p ⬍ .001).4 So, too, did the mood states
that were significantly associated with PEFR (active-
4
This finding may at odds with the common clinical observation
that asthmatics’ bronchoconstriction is worse after awakening and
better in the evening. The first PEFR in the present analysis was
taken in the late morning each day. We had also measured PEFR
one-half hour after awakening, but omitted these data from this
study because mood was not measured on this occasion. These
earlier morning PEF ratings were indeed lower than those in the
evening.
66
G. AFFLECK et al.
passive, b ⫽ ⫺0.24, p ⬍ .001; peppy-drowsy, b ⫽
⫺0.47, p ⬍ .001). However, both mood dimensions
remained significantly associated with PEFR when
time of day was entered along with them in the model
predicting PEFR, ruling out confounding by time of day.
Albuterol Use Effects?
Participants’ self-administration of albuterol doses
during each observation period was unrelated to that
observation’s PEFR, but more doses were administered
during times when they reported more severe symp-
toms (b ⫽ 0.79, p ⬍ .001). Albuterol dosing explained
3.7% of the within-person variance in symptoms.
Higher use of albuterol was also associated with lower
scores on the active-passive mood dimension (b ⫽
⫺0.28, p ⬍ .05) and on the peppy-drowsy dimension
(b ⫽ ⫺0.43, p ⬍ .001). Inspection of the relations
presented in Table 1 disclose only one relation that
could be confounded by albuterol use, namely the
association between symptoms and peppy-drowsy
mood. However, even controlling for albuterol dosing,
this relation remained statistically significant.
Differential Effort in Peak Flow Monitoring?
The specific moods associated with PEFR signal
differences in arousal states. Thus, it could be argued
that these relations are less a function of intrinsic
relations between mood and PEFR than they are of the
relations between mood and the effort required to use
the PEFR monitor to measure bronchoconstriction ac-
curately. For example, might the relation between
drowsiness and peak flow be due to the effect of
drowsiness on suboptimal effort with the peak flow-
meter?
This is a challenging problem for field studies. To
understand this possibility, we computed a crude in-
dex of each participant’s overall effort as the ratio of
his or her mean PEFRs across the 21 days with the gold
standard of average PEFRs measured before day 1 and
after day 21 with supervised spirometric assessments.
Higher ratios could be construed as consistent with a
more accurate or effortful use of PEFR monitors at
home.
This ratio did not correlate with the mean mood
scores for either the active-passive vector (r ⫽ 0.05) or
the peppy-drowsy vector (r ⫽ 0.14). However, it was
significantly associated with the magnitude of the
within-person relations between these moods and
PEFR. This was revealed by a random effects regres-
sion analysis that examined the ratio’s ability to pre-
dict variation in the mood-PEFR slopes. Participants
with higher ratios, who arguably were more effortful in
Psychosomatic Medicine 62:61– 68 (2000)
MOOD, ASTHMA SYMPTOMS, PEFR
their use of peak flow monitors, were those who ex-
hibited weaker relations between PEFR and active-
passive mood (b ⫽ ⫺3.94, p ⬍ .05) and peppy-drowsy
mood (b ⫽ ⫺0.4.63, p ⬍ .05).
DISCUSSION
This study is the first to identify mood states that are
differentially related to asthma symptoms and peak
expiratory flow. To summarize, two mood vectors that
captured the degree of arousal (active-passive and pep-
py-drowsy) were significantly related to PEFR
changes. Two other mood vectors that reflected degree
of pleasantness (happy-sad and calm-nervous) were
significantly related to asthma symptom reports, after
adjusting these relations for contemporaneous PEFR.
These findings remained significant even after control-
ling for time of day and albuterol dosing.
Several features of our research may be responsible
for these novel findings. These include a) a theory-
driven assessment of mood that separates arousal from
hedonic tone; b) a daily process methodology that
minimizes retrospection error in symptom and mood
reporting; c) use of electronic methods that blind sub-
jects to peak flow monitoring results and encourage
and assess compliance with symptom ratings, and d)
emphasis on the within-person patterning of changes
in mood, symptoms, and peak flow within and across
days.
Although we measured mood, asthma symptoms,
and PEFR three times a day, our analyses failed to
establish temporal priority. Thus, we cannot tell
whether these mood states preceded changes in PEFR
or asthma symptoms or were engendered by them. Our
decision to sample these processes three times a day
was balanced against the burden of a 21 day self-
monitoring period. It may be that readings need to be
taken more frequently within the day to capture se-
quential relations. Alternatively, the relations we un-
covered may arise from simultaneous processes or se-
quential processes that are so ephemeral that they would
be virtually impossible to detect in field studies.
Mood and Asthma Symptoms
A substantial quantity of literature has elucidated
connections between mood and the perception of ill-
ness symptoms. Dysphoric mood may be a conse-
quence of symptom exacerbations, but it could also
amplify symptoms through increasing self-focused at-
tention. Whether they are a consequence or antecedent
of symptom changes, fluctuating levels of pleasant/
unpleasant mood should most likely track changes in
subjective asthma symptoms. Our findings support
Psychosomatic Medicine 62:61– 68 (2000)
this hypothesis. It was variations in happy-sad and
calm-anxious moods that were independently related
to the ebb and flow of symptoms. Most important,
these findings persisted even after controlling for con-
temporaneous peak flow ratings. Thus, within the lim-
its of the reliability and validity of unsupervised peak
flow monitoring as a measure of airways obstruction,
there seems to be a unique relation between these
moods and the individual’s subjective experience of
bronchoconstriction.
Mood and PEFR
Interpretation of the relations between mood states
and PEFR is aided in part by having blinded study
participants to their peak flow ratings. Although symp-
tom reports covaried with mood vectors involving
pleasantness, PEFR covaried with mood vectors in-
volving arousal. This included variations in both ac-
tive-passive and peppy-drowsy mood.
Straightforward interpretation of the mood-PEFR re-
lations, however, is complicated by the effort-depen-
dence of peak flow monitoring reliability. If arousal
affects expiration effort, then our findings may reflect
nothing more than the differential reliability of peak
flow monitoring with and without optimal effort.
Without having frequent supervised actuations as pri-
mary data, it is impossible to dismiss this possibility.
We tried to develop an understanding of this problem
by calculating a gold standard for each participant’s
PEFR under maximum effort with controlled spirom-
etry readings before and after the 21– day self-monitor-
ing study. Comparing this against each participant’s
aggregate PEFR across the 21– day period afforded us a
crude indicator of his or her optimal effort with peak
flow monitoring. Interestingly, this indicator did not
correlate with participants’ aggregate scores on the two
mood states that correlated with PEFR changes. How-
ever, it did affect the magnitude of the within-person
relation between these mood states and changes in
PEFR. Those whose peak flow monitoring was judged
by this indicator to be more effortful exhibited weaker
connections between PEFR and these moods. Thus,
our findings concerning the relation between mood
and PEFR may be due in part to differential effort in
peak flow monitoring. To settle this question, future
investigations should plan more frequent in-home su-
pervised monitoring of airways obstruction.
These caveats notwithstanding, the present study’s
findings that different moods figure in symptom per-
ceptions as opposed to changes in airways obstruction
establish emotions as one factor that may account for
the inaccuracy of some asthma patients’ estimation of
the severity of their illness. Between 4% and 5% of the
67

within-person variation in symptoms and PEFR were
explained by contemporaneous mood. On balance,
this effect size was equivalent to that captured by the
within-person relation between symptoms and PEFR
and between albuterol use and symptoms. How much
of the unexplained fluctuation in symptoms and PEFR
is simply idiosyncratic or random, and how much
might be because of predictable individual differences
between asthma patients needs to be examined in fu-
ture studies using larger samples to permit adequate
statistical power to address this question.
This study was supported by a grant from the Amer-
ican Lung Association of Connecticut and the Univer-
sity of Connecticut General Clinical Research Center
(National Institutes of Health Grant M01-RR06192).
REFERENCES
1. Weiss K, Wagener D. Changing patterns of asthma mortality:
identifying target populations at high risk. JAMA 1990;264:
1683–7.
2. Li J, O’Connell E. Clinical evaluation of asthma. Ann Allergy
1996;76:1– 4.
3. Kikuchi Y, Okabe S, Tamura G, Hida W, Homma W, Shirato R.
Chemosensitivity and perception of dyspnea in patients with a
history of near-fatal asthma. N Engl J Med 1994;330:1329 –34.
4. Sears M, Taylor D, Print C, Lake D, Li Q, Flannery E, Yates D,
Lucas M, Herbison G. Regular inhaled beta-agonist treatment in
bronchial asthma. Lancet 1990;336:1391– 6.
5. McFadden E, Kiser R, DeGroot W. Acute bronchial asthma:
relationships between clinical and physiologic manifestations.
N Engl J Med 1973;288:221–5.
6. Kendrick A, Higgs C, Whitfield M, Lazlo G. Accuracy of percep-
tion of severity of asthma: patients treated in general practice.
BMJ 1993;307:422– 4.
7. Apter, A, Affleck G, Reisine S, Tennen H, Barrows E, Willard A,
ZuWallack R. Perception of airways obstruction in asthma: se-
quential daily analyses of symptoms, PEFR, and mood. J Allergy
Clin Immunol 1997;99:605–12.
8. Apter A, ZuWallack R, Clive, J. Common measures of asthma
severity lack association for describing its clinical course. J
Allergy Clin Immunol 1994;94:732–7.
9. Lehrer P, Isenberg S, Hochron S. Asthma and emotion: a review.
J Asthma 1993;30:5–21.
10. Tennen H, Affleck G. Daily processes in coping with chronic
pain: methods and analytic strategies. In: Zeidner M, Endler N,
editors. Handbook of coping. New York: Wiley; 1996. p.
151–180.
11. Brown K, Moskowitz D. It’s a function of time: a review of the
process approach to behavioral medicine research. Psychosom
Med 1998;20:109 –17.
12. Steptoe A, Holmes R. Mood and pulmonary function in adult
asthmatics: pilot self-monitoring study. Br J Med Psychol 1985;
58:87–94.
13. Hyland M. The mood-peak flow relationship in adult
asthmatics: a pilot study of individual differences and direction
of causality. Br J Med Psychol 1990;63:379 – 84.
14. Affleck G, Tennen H, Urrows S, Higgins P. Neuroticism and the
pain-mood relation in rheumatoid arthritis: insights from a pro-
spective daily study. J Consult Clin Psychol 1992;60:119 –26.
68
G. AFFLECK et al.
15. Larsen R, Kasimatis M. Day-to day physical symptoms: individ-
ual differences in the occurrence, duration, and emotional con-
comitants of minor daily illnesses. J Pers 1991;59:387– 424.
16. Persson L, Sjoberg L. Mood and somatic symptoms. J Psychosom
Res 1987;31:499 –511.
17. Ingram R, Smith T. Depression and internal focus of attention.
Cogn Therapy Res 1984;8:139 –52.
18. Sedikides C. Mood as a determinant of attentional focus. Cogn
Emot 1992;6:129 – 48.
19. Stalling R. Mood and pain: the influence of positive and negative
affect on reported body aches. J Soc Behav Pers 1992;7:323–34.
20. Pennebaker J. The psychology of physical symptoms. New York:
Springer; 1982.
21. Schwartz J, Stone A. Strategies for analyzing ecological momen-
tary assessment data. Health Psychol 1998;17:6 –16.
22. Clark L, Watson D, Leeka J. Diurnal variation in the positive
affects. Motiv Emot 1988;13:205–34.
23. Egloff B, Tausch A, Kohlmann C, Krohne H. Relationships be-
tween time of day, day of the week, and positive mood: explor-
ing the role of the mood measure. Motiv Emot 1995;19:99 –110.
24. Hetzel M, Clark T. Comparison of normal and asthmatic circadian
rhythms in peak expiratory flow rate. Thorax 1980;35:732– 8.
25. Mann M, Eliasson O, Patel K, ZuWallack R. An evaluation of
severity-modulated compliance with q.i.d. dosing of inhaled
beclomethasone. Chest 1992;102:1342– 6.
26. Stone A. Measurement of affective response. In: Cohen S,
Kessler R, Underwood Gordon L. editors. Measuring stress: a
guide for health and social scientists. New York: Oxford Uni-
versity Press; 1995. p. 148 –174.
27. Russell J. A circumplex model of affect. J Pers Soc Psychol
1980;39:1161–78.
28. Larsen R, Diener E. Promises and problems with the circumplex
model of emotion. In: Clarke, MS, editor. Newbury Park, CA:
Sage Publications; 1992. p. 25–59.
29. Stone A, Broderick J, Porter L, Krupp L, Gnys M, Paty J, Shiffman
S. Fatigue and mood in chronic fatigue syndrome patients: results
of a momentary assessment protocol examining fatigue and mood
levels and diurnal patterns. Ann Behav Med 1994;16:228 –34.
30. Affleck G, Urrows S, Tennen H, Higgins P, Abeles M. Sequential
daily relations of sleep, pain intensity, and attention to pain
among women with fibromyalgia. Pain 1996;68:363– 8.
31. Shiffman S, Fischer L, Paty J, Gnys M, Hickcox M, Kassel J.
Drinking and smoking: a field study of their association. Ann
Behav Med 1994;16:203–9.
32. Swendsen J, Tennen H, Carney A, Affleck G, Willard A, Hromi
A. Mood and alcohol consumption: an experience sampling test
of the self-medication hypothesis. J Abnorm Psychol in press.
33. Hyland M, Kenyon C, Howarth P. Diary keeping in asthma: com-
parison of written and electronic methods. BMJ 1993;306:487–9.
34. Totterdell P, Reynolds S, Parkinson B, Briner R. Associations of
sleep with everyday mood, minor symptoms, and social inter-
action experience. Sleep 1993;17:466 –75.
35. National Heart, Lung, and Blood Institute National Asthma Ed-
ucation Program Expert Panel Report. Guidelines for the diag-
nosis and management of asthma. J Allergy Clin Immunol 1991;
88:425–34.
36. American Thoracic Society. Standardization of spirometry:
1987 update. Am Rev Resp Dis 1987;130:1296 –2009.
37. Bryk A, Raudenbush S. Hierarchical linear models. Newbury
Park, CA: Sage; 1992;.
38. SAS Institute. SAS/STAT Software. Changes and enhancements
through release 6.11. Cary, NC: SAS Institute; 1996.
39. Kreft I, de Leeuw J. Introducing multilevel modeling. London:
Sage publications; 1998.
Psychosomatic Medicine 62:61– 68 (2000)