Embed Size (px)
Citation preview
Development of the Short Form Endometriosis Health Profile Questionnaire:
The EHP-5
Georgina Jones1,2,3, Crispin Jenkinson2 & Stephen Kennedy11Nuffield Department of Obstetrics and Gynaecology, University of Oxford; 2Health Services Research Unit,Department of Public Health and Primary Care, University of Oxford, United Kingdom; 3Present address:Institute of General Practice and Primary Care, University of Sheffield (E-mail: [email protected])
Accepted in revised form 17 April 2003
Abstract
Objective: To develop and validate a short form version of the Endometriosis Health Profile-30 ques-tionnaire which consists of a 30-item core questionnaire and a 23 item modular questionnaire. Methods:Three studies were carried out to develop the Endometriosis Health Profile-5. Study 1: a short form versionof the Endometriosis Health Profile-30 core questionnaire was developed. Study 2: the results were verifiedin a new data set. Study 3: a short form version of the 23 item modular questionnaire was produced. Themodular questionnaire contains six dimensions which may not be applicable to every woman with endo-metriosis and is used to supplement the five scales on the core questionnaire when required. Results: Thefinal instrument contained 11 items: five items from the core questionnaire and six items from the modularquestionnaire. The 95% confidence intervals for Study 1 indicated that the scale scores overlapped witheach item score and were confirmed in the second analysis (study 2). Each item was most highly correlatedto its parent scale and less with the remaining four core questionnaire scales which was verified in study 2.In both studies all correlations were significant at the 0.01 level (two tailed test). Conclusion: The Endo-metriosis Health Profile-5 is a reliable and valid short form questionnaire which can also be supplementedwith a short form version of the modular questionnaire when required. It will be especially useful in clinicalsettings where a short and economical endometriosis health status measure is required.
Key words: Endometriosis, Health-related quality of life, Short Form Questionnaire
Introduction
In recent years there has been an increasing de-mand for the development and validation of psy-chometrically based questionnaires that assesspatients’ health-related quality of life. Health-re-lated quality of life is a multi-dimensional conceptencompassing physical, social and psychologicalaspects associated with a particular disease or itstreatment [1]. Recently, we reported on the devel-opment of a new disease specific questionnaire tomeasure the health status of women with endo-metriosis (Endometriosis Health Profile-30) and
reported on the psychometric properties of theinstrument [2]. The instrument consists of a corequestionnaire with 30 items and five scales and sixmodular parts consisting of 23 questions. Al-though the questionnaire demonstrated high reli-ability and validity, it was decided to produce ashort form version of the questionnaire.The rationale for developing a shorter measurewas that it may have advantages over its lengthiercounterpart, especially in certain clinical situationsand also research, i.e. randomised controlled trials[3]. In clinical practice, health status measurementis frequently limited because clinicians often find it
Quality of Life Research 13: 695–704, 2004.� 2004 Kluwer Academic Publishers. Printed in the Netherlands.
695
too time consuming and difficult to interpret theinformation gained by long health status instru-ments and they are too burdensome for the re-spondent to complete [4]. In relation to theapplication of health status measures in researchstudies, for example in large randomised trials, ithas been found that data quality may be affectedwhen instruments which collect large amounts ofinformation are used [5]. Despite this, no shortform questionnaire has been developed to measurethe health status of women with endometriosis.Short form generic questionnaires have beendeveloped, the Short form-12 [6] and the Shortform-6 [7]. However, as they were designed tomeasure health status across a wide variety ofdiseases, they may not be sensitive enough to as-sess changes in health status which are importantto women with endometriosis [8].Consequently, the aim of this research was to de-velop and validate a short form endometriosis-specific questionnaire from the EndometriosisHealth Profile-30 questionnaire using establishedpsychometric methods.
Methods
The items selected for the Endometriosis HealthProfile-5 (see Appendix 1) were derived from thelong form core questionnaire. The original Na-tional Endometriosis Survey data used to developthe long form core questionnaire (reported else-where) [2] were used to construct the Endometri-osis Health Profile-5 and identify which of the 30items would be included in the shortened versionof the questionnaire (Study 1). To construct theshort form instrument the question which had thehighest item to total correlation in each scale wasselected. Item to total correlation refers to theextent to which there is a linear relationship be-tween the item and its scale total (corrected foroverlap) [9]. To correct for overlap the item whichis to be correlated with the scale is omitted fromthe scale total. The five items selected would thenmake the EHP-5 short form endometriosis healthprofile questionnaire.Next, each scale on the Endometriosis HealthProfile-30 core questionnaire was transformed ona scale of 0 (indicating best possible health statusas measured on the questionnaire) to 100 (indi-
cating worst possible health status). This sameprocedure was then carried out for all 30 items sothat comparisons could be made between eachitem and its parent scale. This was done to assesshow closely the mean score of the scale item on theEHP-5 reflected its scale total. Next 95% confi-dence intervals and ceiling and floor effects werecalculated for each item on the EHP-5 and its scaletotal. It was hypothesised that the 95% confidenceintervals for each item and scale total wouldoverlap indicating that the short form question-naire was producing the same picture of healthstatus as its longer form counterpart. Each item onthe short form questionnaire was then correlatedwith the five core scales of the EndometriosisHealth Profile-30 questionnaire (Spearman rhononparametric coefficient). It was hypothesisedthat the highest correlations should be found be-tween the five items of the Endometriosis HealthProfile-5 and their parent scale.The results of the analyses carried out in Study 1were then verified in the data obtained from thetest–retest survey (reported elsewhere) [2] to see ifthe same pattern of results could be replicated in asmaller data set (Study 2). The same methodologyused for the statistical analysis as described abovefor Study 1 was followed in Study 2.A short form version of the 23 item modularquestionnaire was produced to supplement the useof the Endometriosis Health Profile-5 (Study 3).The original National Endometriosis Survey data,used to develop the Endometriosis Health Profile-5(described above) were used to produce the shortform version of the modular questionnaire. Theprocess followed in the analysis of the modularquestionnaire was the same as in study 1 and 2.The most highly correlated item in each of the sixmodular dimensions was selected. All statisticalanalyses were performed using the SPSS version9.0 for Windows, Oxford University, England.
Results
Of the 1000 questionnaires sent out in the secondpostal survey over a 6 month time period, 371(37.1%) were returned. Of these, 88% (n ¼ 325)were suitable for inclusion in the study; 9%(n ¼ 35) were returned by women who had not yethad their endometriosis surgically confirmed and
696
11 women (3%) no longer had any symptoms andso were unable to complete the questionnaire. Themean age of the sample was 32.5 years (standarddeviation [SD] 7.2, range ¼ 17–58). The meanperiod since diagnosis was 32 months (SD: 44;range: 1–360), [n ¼ 323]. Of the 83 questionnairesadministered to assess test–retest reliability (Study2), 54 (65%) were returned. The mean age of thesample was 33.1 years (SD: 7.2: range: 20–52). Themean period since diagnosis was 65 months (SD:64: range: 1–281).The internal reliability consistency (the extent towhich a scale taps an underlying dimension)
(Cronbach a statistic) [10] and item to total corre-lations of the EHP-30 core questionnaire areshown in Table 1. Items with the highest correla-tion were selected for the new short form. Both ‘feltdepressed’ and ‘had mood swings’ in the emotionalwell-being scale had the same item to total corre-lation. To establish which question would be se-lected, the results of another item reliabilityanalysis were assessed in the data obtained from apilot study involving 141 women recruited from agenetic linkage (OXEGENE) study [11]. In thisanalysis ‘had mood swings’ had the highest item tototal correlation in the scale (0.83) and was there-
Table 1. Internal reliability (Cronbach a) and corrected item to total correlations on the Endometriosis Health Profile-30 Core
Questionnaire
Scale items Corrected item to
total correlation
Cronbach a
Pain 0.93
Felt unable to cope with the pain 0.70
Been unable to do jobs around the home because of the pain 0.78
Found it difficult to stand because of the pain 0.79
Found it difficult to sit because of the pain 0.75
Found it difficult to walk because of the pain* 0.82
Been unable to sleep properly because of the pain 0.61
Found it difficult to exercise or do the leisure activities you would like to do 0.75
Had to go to bed/lie down because of the pain 0.76
Lost appetite/unable to eat because of the pain 0.64
Unable to go to social events because of the pain 0.69
Unable to do the things you want to because of the pain 0.70
Control and powerlessness 0.89
Generally felt unwell 0.60
Felt frustrated as symptoms not getting better 0.64
Felt frustrated as not able to control symptoms 0.74
Felt unable to forget symptoms 0.67
Felt symptoms ruling your life* 0.86
Felt symptoms taking away life 0.74
Emotional well being 0.89
Felt depressed* 0.76
Felt weepy/tearful 0.65
Felt miserable 0.75
Had mood swings* 0.76
Felt bad or short tempered 0.72
Felt violent or aggressive 0.70
Social support 0.86
Felt unable to tell people how you feel 0.63
Felt alone 0.74
Felt others do not understand what you are going through* 0.81
Felt as though others think you are moaning 0.66
Self image 0.83
Felt frustrated as cannot always wear the clothes you would choose to 0.70
Felt your appearance has been affected* 0.75
Lacked confidence 0.58
*Highest item to total correlations in each scale.
697
fore selected. The five items which had the highestitem to total correlations, comprising the Endo-metriosis Health Profile-5 short form questionnairewere; ‘found it difficult to walk because of the pain’(0.82), ‘felt symptoms are ruling your life’ (0.86),‘had mood swings’ (0.76), ‘felt others do not un-derstand what you are going through’ (0.81) and‘felt your appearance had been affected’ (0.75).These five items were then standardised on a scaleof 0 (indicating best possible health status asmeasured on the questionnaire) to 100 (indicatingthe worst possible health state). As shown in Ta-ble 2(panel a) (Study 1) the descriptive statisticsfor these five items produced very similar results totheir parent scale totals which had also beenstandardised on a scale from 0 to 100. These re-sults were also verified in Study 2 as shown inTable 2(panel b).With the exception of one item (emotional well-being) the results of the 95% confidence intervalsfor Study 1 indicated that the scale scores over-lapped with each item score as shown in Ta-
ble 3(panel a). These results were confirmed in thesecond analysis carried out in Study 2 as shown inTable 3(panel b). Ceiling and floor effects wereevaluated. Overall these were found to be low.However in both study 1 and study 2, the largestceiling effect was seen for the control and power-lessness item (38.8%) and (25.0%) respectively. InStudy 1, the largest floor effect was seen for thepain item (12.7%) which was also verified in Study2 (20.8%).As hypothesised the highest correlations werefound between the five items of the EndometriosisHealth Profile-5 and their parent scale. The resultsindicated that each item was most highly corre-lated to its parent scale and less with the remainingfour core questionnaire scales which were alsoverified in Study 2 (Table 4). The square of thecorrelation co-efficients as shown in Table 4, alsoshow the estimate of the explained variance of theshorter questionnaire (EHP-5) compared to thelong questionnaire (EHP-30). The results indicatethat most information is retained in going from the
Table 2. (a) Study 1: Descriptive statistics for the five core scales of the Endometriosis Health Profle-30 Questionnaire with the
Endometriosis Health Profile-5 items, (b) Study 2: descriptive statistics for the five scales of the Endometriosis Health Profile-30 core
questionnaire with the EHP-5 item
Mean (SD) 95% CI % min % max
(a) Study 1
Pain scale (n ¼ 311) 53.03 (21.51) 50.63–55.43 1.3 0.6
Pain item (n ¼ 322) 52.10 (28.07) 49.02–55.17 12.7 6.5
Control scalea (n ¼ 322) 72.72 (21.04) 70.42–75.03 0.3 8.1
Control item (n ¼ 325) 72.38 (27.93) 69.34–75.43 2.8 38.8
Emotional scale (n ¼ 318) 55.21 (20.33) 52.97–57.46 0.6 3.1
Emotional item (n ¼ 324) 61.11 (25.12) 58.37–63.86 5.2 13.0
Social support scale (n ¼ 323) 61.42 (24.42) 58.74–64.09 1.9 7.7
Social support item (n ¼ 325) 64.92 (27.44) 61.93–67.92 4.3 24.0
Self-image scale (n ¼ 321) 57.22 (26.74) 54.28–60.15 4.4 9.0
Self-image item (n ¼ 322) 59.63 (30.64) 56.27–62.99 9.9 20.8
(b) Study 2
Pain scaleb (n ¼ 47) 42.31 (24.71) 35.06–49.57 14.9 2.1
Pain item (n ¼ 48) 44.79 (28.69) 36.46–53.12 20.8 6.3
Control scale (n ¼ 48) 61.72 (29.84) 53.05–70.38 10.4 6.3
Control item (n ¼ 48) 60.42 (33.75) 50.62–70.22 14.6 25.0
Emotional scaleb (n ¼ 48) 47.14 (23.07) 40.44–53.83 8.3 2.1
Emotional item (n ¼ 48) 51.56 (25.48) 44.16–58.96 8.3 4.2
Social support scaleb (n ¼ 46) 52.17 (25.73) 44.53–59.81 8.7 6.5
Social support item (n ¼ 46) 58.15 (29.37) 49.43–66.87 10.9 13.0
Self-image scale (n ¼ 48) 47.74 (28.75) 39.39–56.09 12.5 4.2
Self-image item (n ¼ 48) 48.96 (32.60) 39.49–58.42 18.8 12.5
a The range for all scales and items was 0–100 except for the control scale (range ¼ 8–100).b The range for all scales and items was 0–100 except for the pain scale (range ¼ 0–91), emotional scale (range ¼ 0–92), and social
support scale (range ¼ 0–88).
698
Table 3. Descriptive statistics for the five scales of the EHP-30 core questionnaire and individual items from (a) study 1, (b) study 2
Scale and item N 25th
percentile
Median 75th
percentile
Mean SD % min % max
Pain 311 39 55 71 53.03 21.51 1.3 0.6
Been unable to go to social events 324 25 50 75 45.68 28.92 18.8 5.2
Been unable to do jobs around the home 321 25 50 75 50.08 26.48 12.8 4.7
Found it difficult to stand 323 50 50 75 55.73 27.12 9.9 8.0
Found it difficult to sit 321 25 50 75 49.07 27.65 13.4 4.7
Found it difficult to walk 322 25 50 75 52.10 28.07 12.7 6.5
Found it difficult to exercise/leisure activities 318 50 75 75 60.22 28.32 7.9 16.7
Lost appetite/unable to eat 320 25 50 69 42.66 28.56 19.7 5.0
Been unable to sleep properly 323 50 75 75 61.15 25.76 5.9 13.6
Had to go to bed/lie down 323 50 50 75 57.28 26.89 9.0 9.9
Been unable to do the things you want to 324 50 50 75 56.56 27.42 9.3 11.7
Felt unable to cope with the pain 325 50 50 75 54.85 28.29 9.5 12.3
Control and powerlessness 322 58 75 92 72.72 21.04 0.3 8.1
Generally felt unwell 324 50 75 100 70.68 24.19 3.1 25.9
Felt frustrated as symptoms not getting better 324 50 75 100 72.53 23.64 2.5 28.4
Felt frustrated as not able to control symptoms 324 75 75 100 76.39 22.81 1.5 36.4
Felt unable to forget symptoms 325 50 75 100 71.38 26.58 3.7 32.9
Felt symptoms ruling your life 325 50 75 100 72.38 27.93 2.8 38.8
Felt symptoms taking away life 325 50 75 100 62.62 31.16 8.6 27.4
Emotional well-being 318 42 54 71 55.21 20.33 0.6 3.1
Felt depressed 322 50 50 75 58.70 24.29 6.2 9.0
Felt weepy/tearful 324 50 75 75 60.11 23.61 4.0 9.3
Felt miserable 324 50 75 75 63.27 22.11 2.5 12.3
Had mood swings 324 50 75 75 61.11 25.12 5.2 13.0
Felt bad or short tempered 324 25 50 75 52.39 25.35 6.8 7.4
Felt violent or aggressive 323 0 25 50 35.99 30.00 29.7 4.6
Social support 323 44 63 81 61.42 24.42 1.9 7.7
Felt unable to tell people how you feel 324 50 75 75 59.72 28.13 6.8 16.4
Felt alone 324 25 50 75 56.79 31.42 10.2 19.4
Felt others do not understand 325 50 75 75 64.92 27.44 4.3 24.0
Felt as though others think you are moaning 325 50 75 75 63.54 29.42 7.1 24.0
Self-image 321 42 58 75 57.22 26.74 4.4 9.0
Felt frustrated as cannot always wear the clothes 322 25 50 75 55.05 32.48 15.2 17.7
Felt your appearance has been affected 322 50 62 75 59.63 30.64 9.9 20.8
Lacked confidence 322 50 50 75 56.99 29.82 9.6 17.1
Pain 47 21 48 59 42.31 24.71 14.9 2.1
Been unable to go to social events 47 0 50 50 35.64 26.45 27.7 2.1
Been unable to do jobs around the home 48 6 50 50 39.58 27.21 25.0 2.1
Found it difficult to stand 48 25 50 50 41.15 28.48 22.9 4.2
Found it difficult to sit 48 6 50 69 40.10 29.96 25.0 4.2
Found it difficult to walk 48 25 50 50 44.79 28.69 20.8 6.3
Found it difficult to exercise/leisure activities 47 25 50 75 50 31.71 21.3 10.6
Lost appetite/unable to eat 47 0 50 50 35.11 28.86 29.8 4.3
Been unable to sleep properly 48 25 50 75 47.92 27.69 14.6 6.3
Had to go to bed/lie down 48 6 50 50 39.58 28.17 25.0 2.1
Been unable to do the things you want to 48 25 50 75 47.92 28.17 16.7 4.2
Felt unable to cope with the pain 48 25 50 75 43.75 29.85 20.8 6.3
Control and powerlessness 48 47 69 82 61.72 29.84 10.4 6.3
Generally felt unwell 48 50 50 75 59.38 29.90 10.4 18.8
699
longer to the shorter questionnaire. For example,for the pain item (study 1 and 2) r ¼ 0.84, whichimplies that 70% of the variance in the pain scalecan be explained by the pain item. For both of
these two studies all correlations were significantat the 0.01 level (two tailed test).Similar analyses were then undertaken to producea shorter version of the modular questionnaire
Table 3. (Continued)
Scale and item N 25th
percentile
Median 75th
percentile
Mean SD % min % max
Frustrated as symptoms not getting better 48 50 75 100 64.06 34.56 14.6 29.2
Felt frustrated as not able to control symptoms 48 50 75 100 67.19 33.08 12.5 33.3
Felt unable to forget symptoms 48 50 75 100 65.63 32.46 12.5 29.2
Felt symptoms ruling your life 48 50 75 94 60.42 33.75 14.6 25.0
Felt symptoms taking away life 48 25 50 75 53.65 36.10 22.9 20.8
Emotional well-being 48 33 48 67 47.14 23.07 8.3 2.1
Felt depressed 48 25 50 75 46.88 28.07 14.6 6.3
Felt weepy/tearful 48 25 50 75 50.52 27.04 12.5 4.2
Felt miserable 48 50 50 75 53.65 24.73 8.3 2.1
Had mood swings 48 25 50 75 51.56 25.48 8.3 4.2
Felt bad or short tempered 48 50 50 75 53.13 25.07 8.3 4.2
Felt violent or aggressive 48 0 25 50 27.08 27.69 41.7 14.6
Social support 46 38 59 70 52.17 25.73 8.7 6.5
Felt unable to tell people how you feel 48 25 50 75 50.52 30.29 16.7 6.3
Felt alone 48 25 50 75 48.44 28.90 16.7 4.2
Felt others do not understand 46 50 75 75 58.15 29.37 10.9 13.0
Felt as though others think you are moaning 48 31 63 75 54.69 32.48 18.8 12.5
Self-image 48 25 50 67 47.74 28.75 12.5 4.2
Felt frustrated as cannot always wear the clothes 48 25 50 75 47.92 33.75 22.9 12.5
Felt your appearance has been affected 48 25 50 75 48.96 32.60 18.8 12.5
Lacked confidence 48 25 50 75 46.35 30.07 18.8 6.3
Table 4. Correlation coefficients (Spearman’s rho) of EHP-30 dimensions with items of the EHP-5 for study 1 (I) and study 2 (II).
Pain Item CP item EM item SS item SI item
Pain
I 0.84 (n = 311) 0.60 (n = 311) 0.35 (n = 311) 0.34 (n = 311) 0.36 (n = 311)
II 0.84 (n = 48) 0.61 (n = 47) 0.53 (n = 47) 0.43 (n = 45) 0.50 (n = 47)
Control
I 0.45 (n = 319) 0.92 (n = 322) 0.41 (n = 321) 0.44 (n = 322) 0.41 (n = 319)
II 0.72 (n = 48) 0.87 (n = 48) 0.48 (n = 48) 0.51 (n = 46) 0.46 (n = 48)
Emotional
I 0.33 (n = 315) 0.45 (n = 318) 0.83 (n = 318) 0.55 (n = 318) 0.36 (n = 315)
II 0.50 (n = 48) 0.55 (n = 48) 0.87 (n = 48) 0.43 (n = 46) 0.46 (n = 48)
Social support
I 0.13 (n = 320) 0.46 (n = 323) 0.51 (n = 323) 0.90 (n = 323) 0.33 (n = 320)
II 0.42 (n = 46) 0.49 (n = 46) 0.47 (n = 46) 0.90 (n = 46) 0.44 (n = 46)
Self image
I 0.32 (n = 319) 0.43 (n = 321) 0.35 (n = 321) 0.40 (n = 321) 0.88 (n = 321)
0.36 (n = 48) 0.52 (n = 48) 0.39 (n = 48) 0.39 (n = 46) 0.93 (n = 48)
All the correlations are significant at the 0.01 level (two-tailed).
Bold indicates highest correlations between the five selected items and five scales.
700
which could be used to supplement the short formversion of the core questionnaire when necessary.Table 5 shows the results of the internal reliabilityconsistency and item to total correlations for themodular questionnaire. Both items in the ‘rela-tionship with children’ dimension achieved thesame item to total correlations (0.95). The sameitem to total correlations occur in every scalewhich is made up of two items [8]. Consequently, itwas decided to select the item which had the leastmissing responses and ‘not relevant’ responses.Both items had the same number of missing re-sponses 0.3% (n ¼ 1). However, the results of the‘not relevant’ responses were slightly higher for theitem ‘been unable to play with children’ 74.2%(n ¼ 241) compared to the item ‘found it difficultto look after children’ 73.2% (n ¼ 238). The sixitems which were therefore selected to make the
short form version of the modular questionnaireare shown in Table 6.
Discussion
Tools assessing effects upon quality of life are beingused infrequently in research and clinical practiceand. For example, in a systematic review on themeasurement of HRQoL in women with commonbenign gynaecologic conditions, it was found thatin only 17% of the studies had a standardised in-strument been used in randomised trials [12]. Inaddition, often the instruments administered inresearch just measure one dimension of illness e.g.psychological health, rather than using broadermeasures of health status which identify other ar-eas of well-being affected by disease [13].
Table 5. Internal reliability (Cronbach a) and corrected item to total correlations on the Endometriosis Health Profile-30 Modular
Questionnaire
Scale items Corrected item
to total correlation
Cronbach
a
Intercourse (n = 259) 0.96
Experienced pain during or after intercourse? 0.82
Felt worried about having intercourse because of the pain?* 0.94
Avoided intercourse because of the pain? 0.93
Felt guilty about not wanting to have intercourse? 0.86
Felt frustrated because you cannot enjoy intercourse? 0.90
Work (n = 264) 0.95
Had to take time offwork because of the pain? 0.88
Been unable to carry out duties at work because of the pain?* 0.90
Felt embarrassment about symptoms at work? 0.82
Felt guilty about taking time off work? 0.85
Felt worried about not being able to do your job? 0.85
Medical Profession (n = 239) 0.95
Felt doctor(s) seen not doing anything for you? 0.89
Felt doctor(s) think it is all in your mind?* 0.92
Felt frustrated at doctor(s) lack of knowledge about endometriosis? 0.87
Felt like you are wasting the doctor(s) time? 0.85
Infertility (n = 150) 0.92
Felt worried about the possibility of not having children/more children? 0.85
Felt inadequate because you may not/have not been able to have children/more children? 0.82
Felt depressed at the possibility of not having children/more children?* 0.87
Felt as though the possibility/inability to conceive has put a strain on personal relationships? 0.85
Relationship with children (n = 86) 0.97
Found it difficult to look after your child/children?* 0.95
Been unable to play with your child/children?* 0.95
Treatment (n = 247) 0.79
Felt frustrated because treatment is not working?* 0.67
Found it difficult coping with side effects of treatment? 0.60
Felt annoyed at the amount of treatment you have had to have? 0.64
* Highest item to total correlations in each scale.
701
Reasons which have been identified to account forthe limited use of health status instruments inclinical practice are that clinicians often find it tootime consuming and difficult to interpret the in-formation gained by long health status instru-ments and they are too burdensome for therespondent to complete. Consequently, producingshort and simple instruments to measure HRQoLwould have advantages over those that are longer,especially in clinical practice.The short form EHP-5 provides the option of us-ing a very brief instrument in the measurement ofhealth outcomes for women with endometriosiswhere the long form version would not be appro-priate. The similar results obtained by the EHP-5from the analyses carried out in both studies sug-gest this questionnaire is providing the same pic-ture of health status as the longer version.The modular questionnaire of the EHP-30 wasdeveloped to contain items that may not be ap-plicable to every woman with endometriosis in-cluding the domains: work, intercourse, infertility,treatment, relationship with children and themedical profession. The selection of the six itemsfrom the six dimensions of the EHP-30 modularquestionnaire means that the EHP-5 can also besupplemented with a short form version of themodular questionnaire when required. In situa-tions where the EHP-30 may prove too time con-suming to complete and analyse, for example whenadministered in general practice, the EHP-5 mayprove especially useful.It has been argued that it is meaningless to shortena questionnaire if the reliability and validity of an
instrument has not been determined [14]. For ex-ample Coste et al. [4], found that many authorswho have shortened questionnaires in the psy-chometric field have neglected important statisticaland methodological factors in the developmentprocess: in particular, whether the measurementproperties of the longer form are satisfactory.However, we feel that the high internal reliabilityand construct validity of the EHP-30 core ques-tionnaire (reported elsewhere) [2] indicate it waspossible to produce a shorter form instrumentfrom this long form version.The results produced by a shortened version of aquestionnaire should also replicate those obtainedusing the long form instrument, i.e. the criterionvalidity of a measure [4]. Criterion validity refersto the extent to which a health status tool corre-lates with an existing instrument – ideally a ‘goldstandard’ [14]. Usually in health status measure-ment there is no gold standard by which to assessthe criterion validity of a health status measure [8].The best that can usually be assessed is the con-struct validity of the instrument, i.e. testing aquestionnaire to determine whether prior hypoth-eses generated are reflected in the results. By pro-ducing a short form version of the questionnairethe original long form becomes the ‘gold standard’by which criterion validity can be assessed [14].In relation to the Endometriosis Health Profile-5,the results produced from this short form measurewere very similar to those obtained from the longform which suggests that the questionnaire isproviding the same picture of health status as itslong form counterpart. Consequently, the Endo-
Table 6. Selected six items and item to total correlations for the Short Form Modular Questionnaire
Scale Item to total
correlation
Work
Been unable to carry out duties at work because of the pain 0.90
Intercourse
Felt worried about having intercourse because of the pain 0.94
Medical profession
Felt doctor(s) think it is all in your mind 0.92
Infertility
Felt depressed at the possibility of not having children/more children 0.87
Relationship with children
Found it difficult to look after children 0.95
Treatment
Felt frustrated because treatment is not working 0.67
702
metriosis Health Profile-30 can be considered asthe ‘gold standard’ upon which to assess the cri-terion validity of the new short instrument.Another argument against shortening the numberof items on a questionnaire is that the compre-hensiveness and precision of the instrument will beforfeited [15]. However, evidence from other re-search has suggested that shorter questionnairesdo not necessarily lose much information obtainedby their longer form versions [16]. The Endo-metriosis Health Profile-5 obtained very similarresults in both studies compared to its longerversion, which suggests that it can be used withoutsubstantial loss of information in the measurementof health status in women with endometriosis.Overall low ceiling and floor effects for the itemsand scales were found. However, floor effects werehigher for the pain item and ceiling effects for thecontrol and powerlessness item on the EHP-5,perhaps indicating that these items may havelimitations for reporting a deterioration or im-provement in health state in future assessments.Another possible limitation of this research wasthat from the postal survey to the National En-dometriosis Society a response rate of only 37.1%was obtained. One of the reasons for the lowresponse rate may be because to respect the con-
fidentiality of members of the National Endo-metriosis Society it was not possible to follow upnon-responders to the first mailing of the ques-tionnaires. The importance of following up non-responders in postal surveys has been well docu-mented [17]. Most postal surveys require follow upmailings as initial response rates using this methodof data collection are too low [8, 18]. However,because similar results were obtained from thedifferent data set in study 2, this would indicatethat the item structure of the EHP-5 is supported,thus verifying the development of the new shorterquestionnaire.In conclusion, it appears that the EndometriosisHealth Profile-5 is a reliable and valid short formquestionnaire which can also be supplementedwith a short form version of the modular ques-tionnaire when required. It will be especially usefulin clinical settings where a short and economicalendometriosis health status measure is required.
Acknowledgements
The authors thank Pharmacia Corporation, USfor funding this research. Source of Funding: Ed-ucational grant, Pharmacia Corporation, US.
Appendix 1
PART 1
DURING THE LAST 4 WEEKS, HOW OFTEN BECAUSE OF YOUR ENDOMETRIOSIS HAVE YOU...
Never Rarely Sometimes Often Always
1. Found it difficult to walk
because of the pain
2. Felt as though symptoms
are ruling your life?
3. Had mood swings?
4. Felt others do not understand
what you are going through?
5. Felt your appearance
has been affected?
Please check that you have ticked one box for each question before moving onto the next page
703
PART 2
DURING THE LAST 4 WEEKS, HOW OFTEN BECAUSE OF YOUR ENDOMETRIOSIS HAVE YOU...
References
1. Colwell H, Mathias SD, Pasta DJ, Henning JM, Steege JF.
A health-related quality of life instrument for symptomatic
patients with endometriosis: A validation study. Am J
Obstet Gynaecol 1998; 179: 47–55.
2. Jones G, Kennedy S, Barnard A, Wong J, Jenkinson C.
Development of an Endometriosis Quality of life Instru-
ment: The Endometriosis Health Profile-30. Obstet Gynecol
2001; 98: 258–264.
3. Wasson J, Keller A, Rubenstein L, Hays R, Nelson E,
Johnson D, and the Dartmouth Primary care COOP Pro-
ject. Benefits and obstacles of health status assessment in
ambulatory settings: The clinicians point of view. Med Care
1992; 30: MS42–MS49.
4. Coste J, Guillemin F, Pouchet J, Fermanian J. Method-
ological approaches to shortening composite measurement
scales. J Clin Epidemiol 1997; 50: 247–252.
5. Pocock SJ. Clinical Trials: A Practical Approach. Chich-
ester: John Wiley and Sons, 1983.
6. Ware J, Kosinski M, Keller SD. A 12 item short-form
health survey. SF-12: Construction of scales and prelimi-
nary tests of reliability and validity. Med Care 1996; 34:
220–233.
7. Ware JE, Nelson EC, Sherbourne CD, Stewart AL. Pre-
liminary tests of a 6-item general health survey: A patient
application. In: Stewart AL, Ware JE (eds), Measuring
Functioning and Wellbeing: The Medical Outcomes Study
Approach, London: Duke University Press, 1992.
8. Streiner DL, Norman G. Health Measurement Scales: A
Practical Guide to Their Development and Use. 2nd ed.
Oxford: Oxford University Press, 1995.
9. Gandek B, Ware JE, Aaronson NK, et al. Tests of data
quality, scaling assumptions, and reliability of the SF-36 in
eleven countries: Results from the IQOLA Project. Inter-
national Quality of Life Assessment. J Clin Epidemiol 1998;
51: 1149–1158.
10. Cronbach L. Coefficent alpha and the internal structure of
tests. Psychometrika 1951; 16: 297–334.
11. Kennedy S. Is there a genetic basis to endometriosis? Semin
Reprod Endocrinol 1997; 15: 309–318.
12. Jones G, Kennedy S, Jenkinson C. Health-related quality of
life measurement in women with common benign gyneco-
logical conditions: A systematic review. Am J Obstet Gy-
necol 2001; 187: 501–511.
13. Fitzpatrick R. Applications of health status measures. In:
Jenkinson C, (ed.), Measuring Health and Medical Out-
comes, London: UCL Press, 1994: 27–41.
14. Jenkinson C, Fitzpatrick R, Jenkinson C. Health Status
Measurement in Neurological Disorders. Oxford: Radcliffe
Medical Press, 2000.
15. Bowling A. Measuring Health: A Review of Quality of Life
Measurement Scales. Buckingham: Open University Press,
1997.
16. Katz JN, Larson MG, Phillips C, Fossel A, Liang MH.
Comparative measurement sensitivity of short and longer
health status instruments. Med Care 1992; 30: 917–925.
17. de Vaus DA. Surveys in Social Research. London: UCL
Press, 1991.
18. Oppenheim AN. Questionnaire Design, Interviewing and
Attitude Measurement. London: Continuum, 1992.
Address for correspondence: Georgina Jones, Institute of Gen-
eral Practice & Primary Care, Community Sciences Centre,
Northern General Hospital, Herries Road, Sheffield, S5 7AU,
UK
Phone: þ44-114-2715929
E-mail: [email protected]
Never Rarely Sometimes Often Always
A. Been unable to carry out duties
at work because of the pain?
If not relevant please tick here
B. Found it difficult to look
after your child/children?
If not relevant please tick here
C. Felt worried about having
intercourse because of the pain?
If not relevant please tick here
D. Felt doctor(s) think it is all in mind?
If not relevant please tick here
E. Felt frustrated because treatment
is not working?
If not relevant please tick here
F. Felt depressed at the possibility
of not having children/more children?
If not relevant please tick here
704
