Sie sind auf Seite 1von 8

Family Practice

Oxford University Press 1996

Vol. 13, No. 1


Printed in Great Britain

A new measure of patient satisfaction with


mammography. Validation by factor
analytic technique
Kirsti Loeken, Siri Steine, Leiv Sandvik*, Even Laerum and
Amstein Finset**

Introduction

By assessing the user's experience, women's perception of quality of this mass examination may be
measured. Such quality measurement is the driving force
behind any improvement effort.4"6
Based on relevant theory and psychometric
criteria7-' a multidimensional questionnaire MGQ for
measuring patient satisfaction with mammography was
developed and described in detail.9 In terms of content, items were selected and categorized in order to
cover women's mammographic experience on the
following four dimensions; the structural settings; the
process (patient-provider interaction); discomfort; and
general satisfaction.

The success of breast screening programmes where


women between the ages of 50 and 70 are asked to
attend every second year1 hinges on women's
adherence. Experiences of pain, discomfort and distress
are negatively associated with patient satisfaction and
subsequent screening attendance.2-3Department of General Practice, Department of Community
Medicine, University of Oslo, PO Box 1130 Blindem, N-0318,
Oslo, Medstat Research, PO Box 210, 2000 Lillestram and
* Department of Behavioral Medicine, University of Oslo, Oslo,
Norway.

67

Downloaded from http://fampra.oxfordjournals.org/ by guest on March 23, 2013

Loeken K, Steine S, Sandvik l_ Laerum E and Finset A. A new measure of patient satisfaction with mammography. Validation by factor analytic technique. Family Practice 1996;
13: 67-74.
Background and objectives. The success of national breast screening programmes hinges
on women's adherence. By monitoring patients' perceptions, potential barriers to attendance may be detected, measured and possibly alleviated. Consequently a new questionnaire MGQ, measuring patients' experience of and satisfaction with mammography, has
been developed. As discomfort is a predictor of non-attendance, a dimension measuring
physical and psychological discomfort was included.
Methods. The internal structure of observed variables was tested using factor analysis as
part of the validation process. The study was conducted in six radiological departments
in Norway including 550 patients presenting for mammography. The analysis suggested
eight factors explaining 56.7% of the variance.
Results. Construct validity was supported since the factor scales covered all hypothesized
dimensions and all but one subdimension. The factors were internally consistent and externally independent, indicating that distinct aspects of patients' experience with mammography may be assessed and thus possibly improved.
Conclusions. A relationship between pain and re-attendance was suggested as pain and
worries about the next mammography belonged to the same factor. This underlines the
importance of including a discomfort dimension when monitoring patient satisfaction with
mammography.
Keywords. Mammography, patient satisfaction, quality control, questionnaires, pain.

68

Family Practicean international journal

The validation process in relation to a theoretical construct requires three steps.10 Two of these, describing
the domain of variables, i.e. the content validity, and
verifying the relationship between external, clinical
criteria and scales scores, were described earlier.9
With factor analysis, the third step of establishing the
internal structure of the observed variables by identifying underlying dimensions is evaluated.11 Factor
analysis may also point to the possibility of reducing
a large set of variables to a more compact set with
minimum loss of information.11
The purpose of the study was to compare the
theoretical construct with the empirical scales emerging from factor analysis. Reliability and convergent and
discriminant validity of the factor analytic scales were
estimated.

Facilities, patients and item construction


Facilities, patients and item construction have been
described in detail elsewhere,9 and what follows is a
brief outline. The study was conducted in six
radiological departments situated in urban and rural
areas of Norway. The subjects included 550 consecutive, eligible patients; 488 (89%) completed the
questionnaire after written informed consent was obtained. Immediately after the examination the new questionnaire MGQ concerning patient satisfaction was
completed by the patients.
Sociodemograpbic and relevant psychological and
medical variables were obtained before the examination.9 Patient age ranged from 23-86 years (mean 59
years); 278 (57%) were referred for screening mammography, while the others presented with breast
symptoms.
All items except three were constructed as statements,
and satisfaction was assessed indirectly by asking about
agreement and disagreement on a 5-point Likert
scale.12 Pain was measured on a 10-point vertically
printed numerical scale.12
Dimensions of the MGQ questionnaire
The dimensionality was based on Donabedian's
model7 and developed as follows:
1) Structure: four items covering convenience/
accessibility and physical surroundings of the
mammographic setting;
2) Process: eleven items covering information
transfer between patient and staff, staff's interpersonal skills and staffs perceived technical competence;
3) Discomfort: six items covering physical and
psychological discomfort;
4) General satisfaction: six items on "satisfaction
now" and "future satisfaction", i.e. items related
to repeat adherence.

The postulated theoretical dimensionality of 27 items


was investigated by maximum likelihood factoring and
principal axis methods followed by orthogonal
(varimax) and oblique rotation.
Two extractions and two rotations were used in order
to assess the robustness of the factor structure."-"
Both maximum likelihood factoring and principal axis
methods take into account the presence of common and
unique variance.11 Orthogonal rotation assumes uncorrelated factors, oblique rotation assumes correlated
factors.11
Appropriateness of the factor model was tested with
the Kaiser-Meyer-Olkin measure and Bartlett's test of
sphericity.13 Only factors with an eigenvalue equal to
or greater than one were retained and rotated. A scree
plot was also used as guideline for inclusion of
factors.13
Maximum likelihood factoring is especially useful in
confirmatory factor analysis when a hypothesis about
the number or nature of factors measured is stated, as
in the present study.11
After initial factor analysis only items with loadings
0.40 or greater on factors were retained.13 Regarding
items with a significant loading on more than one factor, only the item's largest loading was used when constructing the factor scales.
Scales were computed by summing items, and
transforming sums on to a scale from 0-100, as described elsewhere.9
The emerging scales were analysed with respect to
convergent and discriminant validity by examining the
item to own scale correlation corrected for overlap, and
comparing this with other scales. 16 Convergent
validity was accepted if an item correlation with own
scale exceeded its correlation with other scales.16
Cronbach's alpha was applied as a measure of internal consistency.12 The degree of intercorrelation between scales was measured by Spearman's correlation
coefficient and compared with their reliability
estimates.16 Maximum allowable interscale correlation
was also assessed in relation to the internal consistency
of the scale (Cronbach's alpha). The correlation coefficient should be less than this measure.16
The item intercorrelations were assessed by Spearman's correlation coefficient. Parametric versions of
these tests, i.e. Pearson's correlation coefficient were
also calculated.
The proportion of missing scores for individual items
ranged from one to five. The handling of missing data
was described in detail elsewhere.9 The study was approved by the regional ethics committee.

Results
With both methods of extraction, before rotation, the
items describing pain, discomfort and awkwardness

Downloaded from http://fampra.oxfordjournals.org/ by guest on March 23, 2013

Methods

Data analysis

MGQ questionnaire factor analysis

showed high loadings on the first two unrotated factors which covered 26% of total measured variance.
Extraction with maximum likelihood factoring and
orthogonal rotation resulted in eight factors, in which
19 of the original items were included. This accounted
for 56.7% of total measured variance.
The factor structure was confirmed, since both
estimates and rotations resulted in a similar structure
with eight factors. The factors are presented in Table
1, including items which did not load clearly on any
factor. All items except item 24 loaded on one factor
only. The eight factors shown in Table 1 corresponded with the postulated dimensions as follows.

Process dimension
Three factors, all made up of items from the proposed
process dimension, emerged and were interpreted as
representing this dimension.
Factor m represented 6.1% of explained variance and
was based on two related, but still different items, one
measuring 'subjective' waiting time, the other measuring time in minutes. Factor three was labelled' 'waiting
time".
Factor IV was composed of three items and conformed to the subdimension "information transfer between staff and patient". This was labelled "information
transfer", and accounted for 3.1% of explained
variance.
Factor VIII included a single item explaining 1.7%
of the variance. The factor corresponded to the
hypothesized subdimension "staffs technical skills"
and was labelled "staffs technical skills".
Structure dimension
Factor V was interpreted as representing the proposed
structure dimension and labelled "physical surroundings". This was made up of two items describing the
waiting and examination rooms, it represented 3.5%
of explained variance and was identical to the hypothesized subdimension "physical surroundings".

Items from the proposed "convenience and accessibility" subdimension had loadings less than 0.40.
and accordingly not included in the factor structure.
General satisfaction dimension
Factor VI contained three items and accounted for 2.3%
of explained variance. In one item surprise at having
to undress before mammography was described.
Another item referred to how rough the technician was
perceived as being, while the third item related to
whether the service could be recommended to friends.
Since all items expressed dissatisfaction and one item
indicated future mammographic behaviour, the factor
was labelled "future satisfaction".
Factor VII was based on two items derived from the
hypothesized subdimension named "satisfaction now",
and covered 2.4% of explained variance. This factor
was labelled "satisfaction now".
Discriminant validity was mostly supported by correlation analysis (Table 2). Correlation coefficients corrected for overlap were higher between own than
between other dimensions for all but two items.
The interfactor correlations were less than 0.60 and
acceptable according to accepted guidelines. In fact,
only factor II and factor VII, i.e. psychological discomfort and satisfaction now, had a correlation mat exceeded
0.40 (Table 3).
Thus the eight-factor solution provides evidence for
scoring and interpreting distinct scales that differentiate
various aspects of womens' subjective mammographic
experiences.
The reliability estimates of internal consistency
measured as Cronbach's alpha were all except one above
0.50; they ranged from 0.49-0.75 (Table 3). Reliability
measured as Cronbach's alpha of the 19-item scale was
0.77.
Scree plot analysis indicated the inclusion of eight
factors, and the factor model was appropriate according to the Kaiser-Meyer-Olkin measure (0.78), and
Bartlett's test of sphericity (0.00000). The 27 items were
weakly to moderately intercorrelated; the correlation
coefficients ranged from 0.08-0.55 (complete results
not presented), suggesting that orthogonal rotation was
acceptable.

Discussion
Evidence of construct validity of the new questionnaire
MGQ was supported since the factor analysis resulted
in eight internally consistent factors covering all proposed dimensions. The factor structure was considered
robust because it was highly similar in two different
analyses. The amount of total variance accounted for
was satisfactory compared with other questionnaires.14"17 Analysis of discriminant validity of the factors demonstrated a solid psychometric basis for

Downloaded from http://fampra.oxfordjournals.org/ by guest on March 23, 2013

Discomfort dimension
Two factors corresponded to the proposed discomfort
dimension since each was made up of two items from
this dimension. In addition both factors included one
item from the proposed general satisfaction dimension.
Factor I accounted for 11.4 % of explained variance
and covered physical pain and discomfort together with
worry about the next mammographic examination. This
factor was labelled "discomfort, physical".
Factor II explained 9.9% of the variance. This scale
was made up of two items describing negative
psychological experiences such as feeling awkward and
embarrassed, and a third item worded towards altering the service. Factor II was labelled "discomfort,
psychological".

69

70
TABLE

Family Practicean international journal


1 Factor loadings: results of maximum likelihood factoring factor analysis of mammography questionnaire. MGQ. Varimax rotation

Factors

Item no.

I.

Loading
on factor

Factors and abbreviated Item content


Discomfort, physical

11.4

12.

Did you find the examination painful?

0.86

22.

Did the examination cause discomfort?

0.82

23.

I will not dread another mammographic examination.

0.44

II.

Discomfort, psychological

9.9

8.

The examination situation made me feel awkward.

0.70

4.

The examination ""** me feel embarrassed.

0.62

HI.

Certain things should have been done otherwise.

0.43

Staff's interpersonal skills

6.1

19.

I sat too long before being examined.

0.94

20.

How long did you have to wait?

0.66

IV.

Information transfer

3.1

7.

The staff told me all I wanted to know.

0.75

3.

I felt free to ask about anything.

0.49

The staff did not explain what was to be done with me.

0.46.

11.
V.

Physical surroundings

3.5

16.

The examination room was unpleasant

0.54

18.

The waiting room was pleasant

0.51

VI.

General satisfaction, future

2.3

1.

I was surprised that I had to undress.

0.54

2.

The examiner was too rough with me.

0.50

I would advise others not to have the examination.

0.48

26.
VTI.

General satisfaction, now

2.4

24.

The staff did anything to ensure my comfort

0.52

25.

I found nothing to complain about.

0.65

VHI.

Staff's technical skills


5.

The examiner seemed to be professionally capable.


Items with primary loadings < 0.40

6.

The examination mxi* me uneasy.

1.7
0.62

Downloaded from http://fampra.oxfordjournals.org/ by guest on March 23, 2013

27

% Variance
explained by factor

MGQ questionnaire factor analysis


TABLE

71

1 (com.)
The staff used words that were easy to understand.

9.
10.

I was worried in case my body could be injured.

13.

I was able to undress undisturbed.

14.

I had to wait too long before getting an appointment.

15.

The staff 'pushed' me quickly through.

17.

I was treated worse than expected.

21.

The examination was too expensive.


TABLE

I
Discomfort
(physical)

IV

Discomfort
(psychol.)

Process
(staffs
interpers.
skills)

Process
(information
transfer)

Structure
Satisfaction
(physical
(now)
(future)
surroundings)

12

ILfifi

0.14

0.02

0.09

0.11

0.28

0.15

0.10

22

070

0.24

0.09

0.08

0.16

0.23

0.19

0.11

23

(LM

0.29

0.03

0.14

0.11

0.28

0.35

0.20

0.24

049

0.09

0.28

0.08

0.23

0.33

0.14

0.23

(LSI

0.00

0.28

0.19

0.14

0.33

0.12

27

0.27

042*

0.12

0.26

0.26

0.15

0.45

0.25

19

0.08

0.09

QJS1

0.09

0.15

0.10

0.23

0.06

20

0.02

0.05

OM

0.09

0.09

0.07

0.12

0.05

0.13

0.30

0.11

<L51

0.20

0.11

0.34

0.30

0.13

0.25

0.01

040

0.17

0.11

0.31

0.26

11

0.02

0.15

0.13

DJU

0.07

0.00

0.26

0.13

18

0.13

0.11

0.12

0.11

L3ft

0.07

0.26

0.04

16

0.13

0.21

0.09

0.21

03Q

0.11

0.29

0.18

0.06

0.05

0.03

0.11

0.07

QJi

0.04

0.00

0.41

0.16

0.09

0.13

0.12

QJ2*

0.18

0.19

26

0.15

0.17

0.06

0.05

0.02

QJ2

0.04

0.07

24

0.22

0.34

0.17

0.38

0.36

0.18

QJ2

0.23

25

0.25

0.41

0.16

0.26

0.17

0.13

0 52

0.27

0.16

0.18

0.06

0.26

0.19

0.07

0.33

1.00

scales

VI

vm

vn
i

Process
(staffs tech.
skills)

Item no.

Probably scaling errors.


Items belonging to same scale are underlined in the table.

Downloaded from http://fampra.oxfordjournals.org/ by guest on March 23, 2013

Factor

2 Item-factor scale correlations (corrected for overlap) for summated rating scales

Scales
number

No. of
items

Discomfort
(physical)

Discomfort, physical (D-PH)

(0.75)

D.

Discomfort, psychological (D-PS)

0.27

(0.66)

m.

Process, staffs interpersonal skills (SIS)

0.06

0.07

(0.76)

IV.

Process, information transfer (IT)

0.13

0.31

0.10

(0.6 0)

Structure, physical surroundings (PS)

0.15

0.20

0.13

0.20

(0.59)

VI.

Satisfaction, future (SF)

0.32

0.17

0.08

0.15

0.05

(0.49)

VII.

Satisfaction, now (SN)

0.27

0.43

0.18

0.37

0.31

0.17

(0.69)

Process, staffs technical skills (STS)

0.16

0.17

0.06

0.26

0.19

0.07

0.29

I.

V.

Factor scales

Discomfort
(psychol.)

Diagonal entries in parentheses are reliability estimates of internal consistency (Cronbach's alpha).

Process
(staffs
interpers.
skills)

Process
(information
transfer)

Structure
(physical
surroundings)

Downloaded from http://fampra.oxfordjournals.org/ by guest on March 23, 2013

TABLE 3 Intercorrelation offactor scales and reliabUity estimates

Satisfaction
(future)

Satisfaction Process
(now)
(staffs tech.
skills)

73

MGQ questionnaire factor analysis

may further both practical improvements and research


by allowing standards to be set for these subjective experiences. This may forward improvements connected
to breast screeening programmes as well as knowledge
about factors predicting women's reattendance rates and
physicians' referral rates.

Acknowledgement
This study was supported by the Norwegian Medical
Association's Quality Assurance Fund.

References
1

Smigel K. NCI's proposed breast cancer screening guidelines.


JNatl Cancer Inst 1993; 85: 1627.
2
Miller AB. Mammography: Reviewing the evidence.
Epidemiology aspect Can Fam Physician 1993; 39: 85.
3
Orton M, Fitzpatrick R, Fuller A, Mant D, Mfynek C,
Tborogood M. Factors affecting women's response to an invitation to attend for a second breast cancer screening examination. BrJ Gen Pract 1991; 41: 320.
4
Baines CJ, To T, Wall C. Women's attitudes to screening after
participation in the national breast screening study. Cancer
1990; 65: 1663.
5
Fitzpatrick R. Surveys of patient satisfaction: IImportant
general considerations. Br MedJ 1991; 302: 887.
6
Fitzpatrick R. Surveys of patient satisfaction: IIDesigning a
questionnaire and conducting a survey. BrMedJ 1991; 302:
1129.
7
Donabedian A. 77K? definitions of quality and approaches to its
assessment. Explorations in quality assessment and monitoring. Vol. 1. Michigan: Health Administration Press, 1980.
* Crocker L, Algina J. Introduction to classical and modem test
theory. Chicago: Holt, Rinehart & Winston, 1986.
9
Loeken K, Steinc S, Laerum E, Sandvik L. A new instrument
to measure patient satisfaction with mammography. Development, reliability and validity. (Submitted for publication).
10
Mainous AG. Factor analysis as a tool in primary care research.
Fam Pract 1993; 10: 330.
11
Rubin HR, Ware JE, Hays RD. The PJHQ questionnaire. Med
Care 1993; 28 Suppl: 22.
n
Streiner DL, Norman GR. Health measurement scales: a practical guide to their development and use. Oxford: Oxford
University Press, 1989.
13
Norusis MJ. SPSS Base system user's guide. Chicago: SPSS
Inc., 1988.
14
Vaeth PA. Women's knowledge about breast cancer. Am J CUn
15

16
11
18
19

20

Oncol 1993; 16: 446.


Carey RG, Seibert JH. A patient survey system to measure
quality improvement: questionnaire reliability and validity.
Med Care 1993; 3 1 : 834.
Ware JE, Sherbourne C D . The MOS 36-hem short-form health
survey. Med Care 1992; 30: 473.
Comrey A L . Common methodological problems in factor
analytic studies. / Consult CUn Psychol 1978; 46: 648.
Jackson V P , Lex A M , Smith DJ. Patient discomfort during
screen-film mammography. Radiology 1988; 168: 421.
Vaile MSB, Calnan M , Rutter DR, Wall B. Breast cancer
screening services in three areas: uptake and satisfaction. /
Publ Health Medicine 1993; 15: 37.
Lerman C, Rimer B, Trock B, Balshem A, Engstrom P F . Factors associated with repeat adherence to breast cancer screening. Prevent Med 1990; 19: 279.

Downloaded from http://fampra.oxfordjournals.org/ by guest on March 23, 2013

distinguishing between different aspects of women's


mammographic experience.
This study represents a hetergenous sample of
radiological departments and patients, which supports
generalizability.
The factor structure corresponds to the proposed
model, but differs slightly from predictions. However,
construct validity was regarded as supported since the
emerging factors were clinically meaningful and
represented an internal structure very smiliar if not identical to the hypothesized dimensions.
Two factors pertaining to physical and psychological
discomfort and attitudes towards repeat adherence
emerged. The importance of these factors was suggested
statistically by the high loadings of their items on the
first, large unrotated factors, and empirically by the fact
that discomfort predicts women- and physician-related
barriers to mammography.20-23
The mammographic procedure depends on forced
compression of women's breasts, challenging the technicians' ability to interact with people as well as with
technology. Three factors were identified as related to
separate aspects of humaneness. Differentiating various
aspects may allow for specific improvements.
Waiting time was included as one such aspect. This
is consistent with another study where waiting time was
described as an important predictor of patient satisfaction in a radiological department.23
The proposed structure dimension was represented
in the factor structure by one subdimension; i.e. physical
surrounding while items concerning distance to a mammographic centre or cost of the examination were excluded. The importance of these aspects is likely to
differ in different populations.
The "satisfaction future" factor pertaining to
dissatisfaction and reattendance was regarded as highly
relevant, but due to its borderline internal consistency
it needs strengthening.
Some methodological concern was evoked also by
the "satisfaction now" factor based on items framed
in general terms. Such items are liable to attract a higher
acquiescence response set bias than more specific
items.24 These biasing effects may diminish the
discriminatory power of the "satisfaction now"
factor.9
The eight factors included only 19 items. If the
questionnaire were to include only these items, the internal consistency of the sum scale and each factor-based
short scale would be acceptable, and interpreting and
measuring both global and separate aspects of patient
satisfaction with mammography would be possible. This
points to the possibility of shortening the questionnaire.
In conclusion, even if shortening the instrument may
be possible, evidence has been provided that the new
questionnaire in its present form may be used as a valid
and reliable outcome measure of women's perception
of mammography. An outcome measure for clinical use

74
21

22

Family Practicean international journal


Stein JA, Fox SA, Murata JP. The influence of ethnicity,
socioeconomic status and psychological barriers on use of
mammography. J Health Soc Behav 1991; 32: 101.
Smith LW, Altmaicr EM, Ross RR, Johnson B D , Berberoglu
IS. Patients' expectations of radiology in noninteractive en-

counters. Radiology 1989; 172: 275.


Ware JE. Effects of acquiescent response set on patient satisfection
ratings. Med Care 1978; 16: 327.
M
Hays RD, Ware JE. My medical care is better than yours. Med
Care 1986; 2 4 : 519.

Downloaded from http://fampra.oxfordjournals.org/ by guest on March 23, 2013