CHSD. Refining Continence Measurement Tools. Centre for Health Service Development

Transcription

1 CHSD Centre for Health Service Development Refining Continence Measurement Tools RefiningContinenceMeasurementToolsRefiningContinenceMeasurementT oolsrefiningcontinencemeasurementtoolsrefiningcontinencemeasurem enttoolsrefiningcontinencemeasurementtoolsrefiningcontinencemeasu rementtoolsrefiningcontinencemeasurementtoolsrefiningcontinenceme asurementtoolsrefiningcontinencemeasurementtoolsrefiningcontinenc emeasurementtoolsrefiningcontinencemeasurementtoolsrefiningcontin encemeasurementtoolsrefiningcontinencemeasurementtoolsrefiningco ntinencemeasurementtoolsrefiningcontinencemeasurementtoolsrefinin gcontinencemeasurementtoolsrefiningcontinencemeasurementtoolsref iningcontinencemeasurementtoolsrefiningcontinencemeasurementtools RefiningContinenceMeasurementToolsRefiningContinenceMeasurementT oolsrefiningcontinencemeasurementtoolsrefiningcontinencemeasurem enttoolsrefiningcontinencemeasurementtoolsrefiningcontinencemeasu rementtoolsrefiningcontinencemeasurementtoolsrefiningcontinenceme asurementtoolsrefiningcontinencemeasurementtoolsrefiningcontinenc emeasurementtoolsrefiningcontinencemeasurementtoolsrefiningcontin encemeasurementtoolsrefiningcontinencemeasurementtoolsrefiningco ntinencemeasurementtoolsrefiningcontinencemeasurementtoolsrefinin UNIVERSITY OF WOLLONGONG Centre for Health Service Development July, 26

2 Jan Sansoni Nick Marosszeky Emily Sansoni and Graeme Hawthorne Department of Psychiatry The University of Melbourne Suggested citation Sansoni J, Marosszeky N, Sansoni E and Hawthorne G (26) Refining Continence Measurement Tools (Final Report). Centre for Health Service Development, University of Wollongong and the Department of Psychiatry, The University of Melbourne.

3 Table of Contents 1 EXECUTIVE SUMMARY Aims Methods Results Urinary Incontinence Scales Faecal Incontinence Scales Conclusions and Recommendations INTRODUCTION Introduction and Background The Continence Outcome Measurement Suite Project The Measuring Incontinence in Australia Project Description of Continence Measures Included in SAHOS Short Form of the UDI The Incontinence Severity Index (ISI) Wexner Faecal Continence Grading Scale (Wexner FCGS) 1 3 INCONTINENCE MEASURES: FINDINGS FROM RECENT REPORTS Measures of Urinary Incontinence Measures of Faecal Incontinence RATIONALE OF STUDY Aims Overview of Methods DATA ANALYSIS: CLASSICAL APPROACH Item Properties Item Endorsement and Discrimination Item-total Correlations and Internal Consistency Reliability Exploratory Factor Analysis Relationships with Other Health Variables Norm Tables Summary Revised Urinary Incontinence Scale (RUIS) Items Revised Faecal Incontinence Scale (RFIS) Items 35 6 DATA ANALYSIS: ITEM RESPONSE THEORY Introduction Methods

4 6.3 Item Response Theory Basics Urinary Continence Measurement Descriptive Properties of Urinary Incontinence Items Analysis of Urinary Continence Measurement Constructing Improved Measurement Some Validation Tests of the UCA Faecal Continence Measurement Descriptive Properties of Faecal Incontinence Items Analysis of Faecal Continence Measurement Constructing Improved Measurement Some Validation Tests of the FCA Norm Tables CONCLUSIONS AND RECOMMENDATIONS Context Conclusions concerning urinary incontinence measures Classical Test Theory Analysis Modern Test Theory Analysis Relationships with Other Health Variables Conclusions concerning faecal incontinence measures Classical Test Theory Analysis Modern Test Theory Analysis Recommendations REFERENCES 58 APPENDIX 1: THE INSTRUMENTS USED AND THE REVISED INSTRUMENTS 63 Instruments included in 24 SAHOS Urinary Distress Inventory Incontinence Severity Index 63 Wexner Faecal Continence Grading Scale 64 Other Faecal Items (included in 24 SAHOS) 64 Revised Measures for Urinary and Faecal Incontinence Revised Urinary Incontinence Scale 66 Urinary Continence Assessment 66 Revised Faecal Incontinence Scale 67 Faecal Continence Assessment 67 APPENDIX 2: SUPPLEMENTARY FIGURES FOR SECTION 5 68 Additional Figures Demonstrating the Level of Non-endorsement for each Incontinence Item Additional Figures Showing the Distribution of Response for Each Incontinence Item APPENDIX 3: NORM TABLES 95

5 List of Tables Table 1 The Wexner Faecal Incontinence Grading System... 1 Table 2 Total number of subjects according to sex and age group in the SAHOS Data set (N = 2924)* Table 3 Labels, questions and abbreviations for the urinary incontinence items in the SAHOS data set Table 4 Labels, questions and abbreviations for the faecal incontinence items in the SAHOS data set Table 5 The item properties of the urinary incontinency items for males Table 6 The item properties of the urinary incontinency items for females Table 7 The item properties of the faecal incontinency items for males... 2 Table 8 The item properties of the faecal incontinency items for females... 2 Table 9 The item properties of the three incontinence measures (ISI, UDI-6 and Wexner FCGS) for males 21 Table 1 The item properties of the three incontinence measures (ISI, UDI-6 and Wexner FCGS) for females Table 11 Corrected item - total correlations and Cronbach s alpha if the item was deleted for each item of the UDI Table 12 Corrected item - total correlations and Cronbach s alpha if the item was deleted for each item of the Wexner Scale (X4, X5, X6, X8, X1) Table 13 The correlation matrix between the ISI, UDI-6 and Wexner FCGS according to the Pearson correlation coefficient (parametric) [n = 2911] Table 14 Correlation matrix for the 8 urinary incontinence items Table 15 Correlation matrix for the 1 faecal incontinence items... 3 Table 16 Rotated Factor Matrix for the urinary incontinence items W1 W Table 17 Rotated Factor Matrix for the faecal incontinence items X1 X Table 18 Pearson correlations between the ISI, UDI-6 and Wexner FCGS with other measures of health. Correlations.2 or greater are shown Table 19 Urinary continence item frequency distribution and PCA factor loading (item pooled) Table 2 IRT analysis of the UDI and ISI Table 21 Initial IRT analysis of all urinary incontinence items probability thresholds (items pooled)... 4 Table 22 IRT analysis of items psychometric properties (items pooled) Table 23 IRT analysis of items uniform DIF, by gender and age group (items pooled) Table 24 IRT analysis of the UCS measure, derived from the UDI and ISI pooled items Table 25 Preliminary comparison of the UCA with the UDI and ISI Table 26 Suggested cutpoints for interpreting the UCA Table 27 Discrimination of the UCA by stress incontinence and urination frequency Table 28 Faecal continence item frequency distribution and PCA factor loading (items pooled) Table 29 IRT analysis of the Wexner and other items Table 3 IRT analysis of the revised Wexner, excluding #3 and after recoding other items Table 31 IRT analysis of the Faecal Continence Assessment (FCA) scale... 5 Table 32 Suggested cutpoints for interpretation of the FCA... 5 Table 33 Preliminary comparison of the FCA with the Wexner Table 34 Discrimination of the FCA Table 35 Norm tables for males according to age group, using the SAHOS weighted sample Table 36 Norm tables for females according to age group, using the SAHOS weighted sample... 97

6 List of Figures Figure 1 The level of non-endorsement for males and females on Item W1 (Frequent Urination)...22 Figure 2 The level of non-endorsement for males and females on Item W5 (Emptying Bladder)...23 Figure 3 The level of non-endorsement for males and females on Item X4 (Leak Solid)...24 Figure 4 The level of non-endorsement for males and females on Item X6 (Leak Gas)...24 Figure 5 The distribution of responses for females on item W4 (Leak small amount)...25 Figure 6 The distribution of responses for males on item W4 (Leak small amount)...25 Figure 7 The distribution of responses for females on item X4 (Leak solid)...26 Figure 8 The distribution of response for females on item X3 (urgency)...26 Figure 9 Item characteristic curve (ICC) for #1, UDI, frequent urination...37 Figure 1 Probability curves for ISI, #1, How often is there urinary leakage...4 Figure 11 DIF for the UDI, #3, Urine leakage related to activity, by gender...42 Figure 12 Faecal continence #7, number of bowel movements, disordered thresholds...46 Figure 13 Percentage of respondents who answered Not at all for item W1 (UDI-6, Item 1)...68 Figure 14 Percentage of respondents who answered Not at all for item W2 (UDI-6, Item 2)...68 Figure 15 Percentage of respondents that answered Not at all for item W3 (UDI-6, Item 3)...69 Figure 16 Percentage of respondents who answered Not at all for item W4 (UDI-6, Item 4)...69 Figure 17 Percentage of respondents who answered Not at all for item W5 (UDI-6, Item 5)...7 Figure 18 Percentage of respondents who answered Not at all for item W6 (UDI-6, Item 6)...7 Figure 19 Percentage of respondents who answered Never for item W7 (ISI, Item 1)...71 Figure 2 Percentage of respondents who answered None for item W8 (ISI, Item 2)...71 Figure 21 Percentage of respondents who answered Normal for item X Figure 22 Percentage of respondents who answered Never for item X Figure 23 Percentage of respondents who answered Never for item X4 (Wexner, Item 1)...73 Figure 24 Percentage of respondents who answered Never for item X5 (Wexner, Item 2)...73 Figure 25 Percentage of respondents who answered Never for item X6 (Wexner, Item 3)...74 Figure 26 Percentage of respondents who answered Never for item X Figure 27 Percentage of respondents who answered Never for item X8 (Wexner, Item 4)...75 Figure 28 Percentage of respondents who answered Never for item X Figure 29 Percentage of respondents who answered Never for item X1 (Wexner, Item 5)...76 Figure 3 Percentage of males that selected each response option in item W1 for each age group...77 Figure 31 Percentage of females that selected each response option in item W1 for each age group...77 Figure 32 Percentage of males that selected each response option in item W2 for each age group...78 Figure 33 Percentage of females that selected each response option in item W2 for each age group...78 Figure 34 Percentage of males that selected each response option in item W3 for each age group...79 Figure 35 Percentage of females that selected each response option in item W3 for each age group...79 Figure 36 Percentage of males that selected each response option in item W4 for each age group...8 Figure 37 Percentage of females that selected each response option in item W4 for each age group...8 Figure 38 Percentage of males that selected each response option in item W5 for each age group...81 Figure 39 Percentage of females that selected each response option in item W5 for each age group...81 Figure 4 Percentage of males that selected each response option in item W6 for each age group...82

7 Figure 41 Percentage of females that selected each response option in item W6 for each age group Figure 42 Percentage of males that selected each response option in item W7 for each age group Figure 43 Percentage of females that selected each response option in item W7 for each age group Figure 44 Percentage of males that selected each response option in item W8 for each age group Figure 45 Percentage of females that selected each response option in item W8 for each age group Figure 46 Percentage of males that selected each response option in item X1 for each age group Figure 47 Percentage of females that selected each response option in item X1 for each age group Figure 48 Percentage of males that selected each response option in item X2 for each age group Figure 49 Percentage of females that selected each response option in item X2 for each age group Figure 5 Percentage of males that selected each response option in item X3 for each age group Figure 51 Percentage of females that selected each response option in item X3 for each age group Figure 52 Percentage of males that selected each response option in item X4 for each age group Figure 53 Percentage of females that selected each response option in item X4 for each age group Figure 54 Percentage of males that selected each response option in item X5 for each age group Figure 55 Percentage of females that selected each response option in item X5 for each age group Figure 56 Percentage of males that selected each response option in item X6 for each age group... 9 Figure 57 Percentage of females that selected each response option in item X6 for each age group... 9 Figure 58 Percentage of males that selected each response option in item X7 for each age group Figure 59 Percentage of females that selected each response option in item X7 for each age group Figure 6 Percentage of males that selected each response option in item X8 for each age group Figure 61 Percentage of females that selected each response option in item X8 for each age group Figure 62 Percentage of males that selected each response option in item X9 for each age group Figure 63 Percentage of females that selected each response option in item X9 for each age group Figure 64 Percentage of males that selected each response option in item X1 for each age group Figure 65 Percentage of females that selected each response option in item X1 for each age group... 94

8 List of Abbreviations 15D AIHW AQoL COMS CTT DIF EQ5D FCA ICC IIQ IRT ISI MCS MTT PCA PCS PSI RFIS RUIS S.E. mean SAHOS SD SF-36 UCA UDI UDI-6 Wexner FCGS 15D Measure of Health Related Quality of Life Australian Institute of Health and Welfare Assessment of Quality of Life Continence Outcome Measurement Suite Project classical test theory differential item functioning EuroQol, European Quality of Life Measure Faecal Continence Assessment item characteristic curve Incontinence Impact Questionnaire Item Response Theory Incontinence Severity Index Mental Component Summary score of the SF-36 modern test theory Principal Component Analysis Physical Component Summary score of the SF-36 Person Separation Index Revised Faecal Incontinence Scale Revised Urinary Incontinence Scale standard error of the mean South Australian Health Omnibus Survey standard deviation Short Form 36 Item Health Survey (Medical Outcomes Study) Urinary Continence Assessment Urogenital Distress Inventory Urogenital Distress Inventory 6 item, Wexner Faecal Continence Grading Scale

9 1 Executive summary 1.1 Aims A Continence Outcome Measurement Suite Project (COMS) was commissioned by the Australian Government Department of Health and Ageing, National Continence Management Strategy Research Program with the goal of recommending a suite of continence outcome measures to be used by clinicians and researchers in Australia. This project was finalised in early 26 (Thomas et al., 26). Recommendations from this report led to a related project Measuring Incontinence in Australia (Hawthorne, 26). Measuring Incontinence in Australia (26) assessed a number of the recommended measures (Urogenital Distress Inventory 6, Incontinence Severity Index, and the Wexner Faecal Continence Grading Scale) by including them in the autumn 24 South Australian Health Omnibus Survey (Harrison Health Research, 24), which was a community population survey. The Hawthorne (26) study provided Australian prevalence estimates for both faecal and urinary incontinence based on this community survey data. For urinary incontinence, the results suggested that the preferred urinary incontinence measure was the Incontinence Severity Index (ISI). It was found to possess superior measurement properties in comparison with the Urogenital Distress Inventory (UDI-6). Because the UDI-6 measures the impact of urinary incontinence on peoples lives rather than incontinence per se, and may contain items that may be endorsed by those without urinary incontinence, the UDI-6 may overstate incontinence prevalence and the impact of this on peoples lives (defined as their health status and their quality of life). Given its poor psychometric properties, there was a prima facie case for major revision of the UDI-6. Although the ISI was the preferred measure, because it violated the assumptions of classic psychometric theory relating to scale stability as it contains only two items, further research into its properties was also recommended. A main purpose of the current study was to undertake further analysis of the SAHOS dataset to refine these urinary incontinence measures to provide a better instrument for the assessment of urinary incontinence in Australia. Regarding the measurement of faecal incontinence, as Hawthorne (26) indicated the current definition of faecal incontinence by the International Continence Society excludes flatus, yet this is included in the Wexner Faecal Continence Grading System (Wexner FCGS). In addition to this definitional inconsistency, the evidence from Hawthorne (26) and AIHW (26) suggested that the inclusion of the flatus item led to overestimates of faecal incontinence prevalence. It was recommended that further work on the Wexner FCGS be undertaken to remove the flatus item and to improve the measurement properties of this scale. This study examines all three measures for the purpose of refining the measurement of both faecal and urinary incontinence. 1.2 Methods The first approach taken was to examine the items and scales using classical test theory (CTT) (Streiner and Norman, 23). The following steps were undertaken and for these analyses unweighted survey data from 24 SAHOS (Harrison Health Research, 24) were utilized: 1. Explore basic item properties such as means, medians and distributions 2. Item analysis examine item endorsement and discrimination 3. Item analysis examine item total correlations and internal consistency reliability Refining Continence Measurement Tools Page 1

10 4. Content Validity relate the items back to their construct(s) using exploratory factor analysis 5. Select items for refined measures and compare with Item Response Theory (IRT) approaches 6. Examine the relationship of the items and the refined scales with other constructs, namely health utilization, disability and self-reported health status The report on these analyses is to be found in Section 5. The second approach to refining incontinence measures was to use modern test theory (MTT), specifically IRT. In this phase the steps undertaken were: 1. Examine instruments from a logical perspective (e.g. differentiate between items measuring incontinence vs. impact of incontinence) 2. Use IRT analyses to provide more in depth analysis of item data and response options to construct more refined incontinence measures 3. Undertake structural equation modelling to examine and test models that best fit the data 4. Compare findings with the CTT taken above to finalise the refined measures The report on these analyses is to be found in Section Results Urinary Incontinence Scales Sections 5 and 6 provide a discussion of the steps undertaken to derive a refined measure for urinary incontinence. As Hawthorne (26) had earlier indicated, a number of items within the UDI- 6 gained endorsement from conditions that may not be directly associated with urinary leakage or incontinence. An analysis of the factor structure of all urinary incontinence items that were included in the survey indicated that a 2 factor structure resulted (explaining 67 of the variance). This would appear to support this contention. The first factor may best be described as a general urinary incontinence factor whereas the second factor contained higher loadings from UDI-6 items such as frequency of urination (UDI- Item 1), pain or discomfort in the lower abdominal or genital region (UDI-Item 6) and difficulty with emptying your bladder (UDI-Item 5). This factor could be considered to be other urological symptoms. The loadings of these items were lower than for all other items on the general urinary incontinence factor (refer to Table 16). Item total correlations for these items, within the UDI-6 scale were also lower than for the other items (urinary leakage due to urge (UDI-Item 2), urinary leakage due to stress (UDI-Item 3), and small amounts of urine leakage (UDI-Item 4)). For these reasons it is thought a better instrument may result from combining the ISI items (frequency and amount of urinary leakage) with the three items from the UDI-6 that load highly on the general urinary incontinence factor (refer to Table 11 and Table 16). The analyses using CTT methods would suggest a urinary incontinence scale (labelled the Revised Urinary Incontinence Scale (RUIS)) comprised of the following items: 1. Urine leakage related to the feeling of urgency? (UDI-6) 2. Urine leakage related to physical activity, coughing or sneezing? (UDI-6) 3. Small amounts of urine leakage (drops)? (UDI-6) Page 2 Refining Continence Measurement Tools

11 4. How often do you experience urine leakage? (ISI) 5. How much urine do you lose each time? (ISI) Modern Test Theory offers three methods of assessing the psychometric properties of items and scales: (a) the ability to examine non-parametric items response scale performance, thus making sure that the response levels within an item are discriminating as expected (item response theory (IRT), threshold analysis), (b) the opportunity to observe the relationship between item response and respondent characteristics to assess whether known groups differ in their interpretation of an item (IRT, differential item functioning (DIF)), and (c) the capacity to assess the impact on a scale of non-normally distributed items through adding or removing an item. The analysis of the urinary incontinence items using modern test theory (MTT) approaches (Item Response Theory) derived a shorter 3 item scale for urinary incontinence (items 1, 3, and 5 above) than the CTT approach. Item 2 from the UDI (concerning stress incontinence) did not fit the best model due to differential item functioning as it was found that males and females interpret this item quite differently. The response categories for Item 1 from the ISI were also found to be problematic as the response thresholds were disordered, meaning that there was not a graded relationship between incontinence status and endorsed response levels. The resulting 3-item measure, labelled the Urinary Continence Assessment (UCA) scale, is included below: 1. Urine leakage related to the feeling of urgency? 2. Small amounts of urine leakage (drops)? 3. How much urine do you lose each time? These suggested scales, with the proposed response categories, can be found in Appendix 1. It is important to note that these revised scales, in conjunction with the original scales, need to be assessed in clinical field trials as part of their validation. As was the case with the Hawthorne (26) study it was found that there were significant correlations between the urinary and faecal incontinence scales (UDI-6; ISI, Wexner) and relevant domains within the measures of health status / health related quality of life that were included in the survey (SF-36, EQ-15D, AQOL; refer Table 18). This indicates that urinary incontinence is associated with a lessening of one s self-rated health status and Hawthorne (26) has also indicated that health status declines with increasing severity of urinary incontinence. The association of the revised scales (RUIS and UCA, RFIS and FCA) with these variables was further explored and they produced very similar correlations in magnitude and direction to those presented in Table Faecal Incontinence Scales Sections 5 and 6 provide a discussion of the steps undertaken to derive a refined measure for faecal incontinence. The Wexner FCGS and a number of additional items concerning soiling, bowel pattern, frequency and urge were also included in the 24 SAHOS (Harrison Health Research, 24) given the concerns with the Wexner FCGS raised by Thomas et al. (26). The factor analysis of the faecal incontinence items indicated that a 3 factor structure resulted explaining 61 of the variance (Table 17). The items that load highly on the first factor are mainly items concerning soiling / wearing a pad, leakage and the effect of leakage on lifestyle. This factor may be considered to be a general faecal incontinence factor, as all items are concerned with leakage and soiling. The items that loaded highly on the second factor were the flatus leakage item from the Wexner FCGS, a question about type of bowel pattern (e.g. Item X1:normal, constipation, diarrhoea, alternating) and a question concerning faecal urge (Item X3: Do you experience an urgent need to have a bowel movement that makes you rush to the toilet?). These Refining Continence Measurement Tools Page 3

12 items appear to be tapping other bowel symptoms that do not appear to be related to faecal incontinence / leakage per se. The only item that loaded on the third factor is frequency of bowel motions (Item X2) and this item has extremely low loadings on the other two factors. It appears to be unrelated to faecal leakage or soiling. An analysis of the item total correlations for the Wexner FCGS indicates that the item concerning flatus has a low corrected item-total correlation and that Cronbach s alpha would be improved if this item were removed from the scale (refer to Table 12). Both the Hawthorne (26) and AIHW (26) reports recommended this item should be removed from the Wexner FCGS as it confounds prevalence estimates and is outside the current ICS definition of faecal incontinence (Norton et al., 22;25). The factor analysis results discussed above would also suggest a refined faecal incontinence scale should not include this item. Although clinicians (Moore et al., 26) had suggested the inclusion of a faecal urgency item this item (Do you experience an urgent need to have a bowel movement that makes you rush to the toilet?) only had a low loading on the general faecal incontinence factor and loaded highly on the second other bowel symptoms factor. Its endorsement pattern was quite different from the other faecal incontinence items (refer to Figure 8) with far fewer respondents endorsing never in relation to this item. This might suggest that bowel urgency is a relatively more common symptom in the community and thus may be endorsed by those not experiencing faecal incontinence. For these reasons the faecal urgency item is not considered for inclusion in the refined scale. However, the item Do you leak stool if you don t get to the toilet in time? loads highly on the general faecal incontinence factor and does contain an urge component. The item concerning type of bowel pattern loaded most highly on the second factor (other bowel symptoms) and has a low loading on the general faecal incontinence factor. Similarly, the question concerning frequency of bowel movements loaded poorly on the general factor and thus these items were not included in the revised scale. An issue concerning item redundancy might be considered by comparing the following items: Do you need to wear a pad to protect your underwear from stool? (Wexner FCGS) Does stool leak so that you have to change your underwear? The pad question from the Wexner FCGS has previously been criticized by Vaizey (1999) as it may relate more to patient fastidiousness rather than faecal incontinence / soiling per se. Given these considerations and its similarity to the soiling item above, in loading on the general faecal incontinence factor (.78 vs..71), it is suggested this item be excluded from the Revised Faecal Incontinence Scale. The proposed scale that results from CTT analysis is the Revised Faecal Incontinence Scale (RFIS) and it contains the following items: 1. Do you leak, have accidents or lose control with solid stool? (Wexner FCGS) 2. Do you leak, have accidents or lose control with liquid stool? (Wexner FCGS) 3. Do you leak stool if you don t get to the toilet in time? 4. Does stool leak so that you have to change your underwear? 5. Does bowel or stool leakage cause you to alter your lifestyle? (Wexner FCGS) However, it is noted that for Item 5 of the Wexner FCGS (and item 5 above) that although it loads highly on the general faecal incontinence factor there may be logical grounds for considering its Page 4 Refining Continence Measurement Tools

13 exclusion as it is measuring a consequence of incontinence rather than a symptom of incontinence per se. It is unconventional to include both the symptom and its consequence in the one scale as this in effect can represent a form of double counting (Thomas et al., 26). There were similarities between the CTT and MTT analyses concerning a number of items that should be excluded from a faecal incontinence scale. These included the items of flatus leakage (Wexner) and items concerning the number of bowel movements and bowel pattern which loaded poorly of the general incontinence factor and were also found to be problematic in the IRT analyses. Analyses using Modern Test Theory approaches (IRT) came to a 4 item solution for faecal incontinence which contains items 1-4 above. IRT is used to find the model with the best fit to the data within the minimum number of items and it is a process commonly used to shorten scales. Item 5 from the RFIS above is not included in this solution as the other items fitted the model better as can be seen in Table 29. The Faecal Continence Assessment (FCA) scale resulting from Modern Test Theory Analyses is provided below. 1. Do you leak, have accidents or lose control with solid stool? (Wexner) 2. Do you leak, have accidents or lose control with liquid stool? (Wexner) 3. Do you leak stool if you don t get to the toilet in time? 4. Does stool leak so that you have to change your underwear? An IRT analysis of a modified Wexner Scale (where the flatus item is excluded) also indicated this scale performed well. Table 33 shows a head-to-head comparison of the FCA with the formal Wexner and the modified Wexner without the flatus question. There was a modest Spearman correlation between the Wexner and the FCA, and a good correlation with the modified Wexner. With respect to discrimination between known groups, the FCA appears to be less sensitive to gender differences and more sensitive to age differences. Regarding classification of cases, the FCA and the modified Wexner produce remarkably consistent results. It is suggested that the response categories for the Wexner FCGS are retained for the revised scales. The revised faecal incontinence scales can be viewed in Appendix 1. It is important to note that the revised scales, in conjunction with the original scales, need to be assessed in clinical field trials as part of their validation. 1.4 Conclusions and Recommendations From the analysis of the urinary and faecal incontinences items and scales included in the 24 SAHOS this study has developed some revised scales for urinary and faecal incontinence (RUIS, RFIS, UCA and FCA) which appeared to improve the screening assessment of incontinence when compared with the original measures. However, one of the limitations of using community survey data is that as the data is collected in face to face interviews the data are at the level of subjective reports of incontinence symptoms rather than confirmed diagnoses. This implies that a community survey will also exclude those currently placed in institutional settings (e.g. nursing homes). These considerations mean that in a community survey there will be a limited range of responses to incontinence items particularly those pertaining to more severe levels of symptoms. It will thus be necessary to trial the refined measures in a range of clinical settings in follow-up field trials. Refining Continence Measurement Tools Page 5

14 It is suggested that the incontinence items included in the 24 SAHOS are administered in a clinical field trial to assess whether the findings from these analyses are replicated amongst clinical samples. This would also enable the datasets to be merged for these items which would permit a more comprehensive analysis and allow for a more definitive conclusion concerning the revised measures. This will also assist in indicating whether the refined measures have superior psychometric properties as would be anticipated from this study. Given some issues raised by the IRT analysis concerning item 1 from the ISI (the frequency of urinary leakage item) and the UDI item 2 (concerning stress incontinence) it is thought that some additional items concerning stress incontinence and the frequency of urinary leakage could be included in the clinical dataset. It is thought that to gain acceptability by clinicians, items covering these domains may need to be included. The three item UCA, however, may be a better instrument to use in future prevalence studies than the ISI. Additional questions which have been developed recently could also be included in the proposed clinical field trials. These include patient-rated global assessments of treatment benefit, satisfaction and willingness to continue treatment (Pleil et al., 25) and Patient Global Impression of improvement and severity for incontinence (Yalcin and Bump, 23). A current study by Hawthorne, Sansoni, Hayes and Marosszeky on patient satisfaction measures may also include recommendations concerning the inclusion of patient satisfaction items. It would then be desirable if field trials could be conducted using the recommended measures across a broader range of field settings specialist continence clinics, general practice and community care settings and in residential care settings. It is also noted that as the 24 SAHOS was a community survey there is limited Australian prevalence data available for continence conditions for those in aged care residential settings and thus it would be particularly useful to pilot test the revised measures in these settings. Page 6 Refining Continence Measurement Tools

15 2 Introduction 2.1 Introduction and Background The Continence Outcome Measurement Suite Project A Continence Outcome Measurement Suite Project (COMS) was commissioned by the Australian Government Department of Health and Ageing, National Continence Management Strategy Research Program with the goal of recommending a suite of continence outcome measures to be used by clinicians and researchers in Australia. This project was finalised in early 26 although recommendations from the earlier draft of this report were critical in the development of a related project Measuring Incontinence in Australia (26) on which this study is based. Measuring Incontinence in Australia (26) assessed a number of the recommended measures from the COMS report by including them in the autumn 24 South Australian Health Omnibus Survey (Harrison Health Research, 24) - a community population survey. The COMS project (Thomas et al., 26) undertook a review of all commonly used instruments to assess faecal and urinary incontinence, and related instruments that assessed the health status, health related quality of life and functional measures of patients experiencing incontinence. With regard to the Urinary Incontinence Symptom Measures category, the King s Health Questionnaire, the Urogenital Distress Inventory (UDI) Short Form, the Urogenital Distress Inventory (UDI) Long Form, the Incontinence Severity Index, the Bristol Female Lower Urinary Tract Symptom assessment, the American Urological Association Symptom Index, the International Continence Society Male assessment, and the International Continence Society Male short form assessment, were assessed as suitable. Given their high ratings on a range of measurement criteria (reliability, validity, applicability and practicability) the first two measures, the King s Health Questionnaire and the UDI (in either form) were the recommended tools in this category and the ISI (a 2-item incontinence severity index) was recommended for use in primary care and public health settings (Thomas et al., 26). With regard to measures of faecal incontinence Thomas et al. (26) found there was little in psychometric terms to separate the various measures of Pescatori et al., (1992), Wexner Faecal Continence Grading System (Jorge and Wexner, 1993), American Medical Systems and St Marks Faecal Incontinence Measures (Vaizey et al., 1999). However, the Pescatori Index stood out from the others in the Vaizey et al. (1999) study in terms of it having significantly lower test-retest reliability. Because of the widespread use of the Wexner Faecal Continence Grading System (Wexner FCGS) and its acceptability amongst clinicians its interim adoption was recommended pending the broadening of the evidence base as to the psychometric properties concerning these tools (Thomas et al., 26). It was noted, however, that the Wexner FCGS did not address the issue of faecal urgency. It was thought the inclusion of the Wexner FCGS along with some further items addressing faecal urgency, bowel pattern and soiling in the Autumn 24 South Australian Health Omnibus Survey (Harrison Health Research, 24) may enable some refinement or revision of the Wexner FCGS (Thomas et al., 26) The Measuring Incontinence in Australia Project As a follow up to the COMS project and as part of the NCMS the Australian Government Department of Health and Ageing funded a special version of the South Australian Health Omnibus Survey (SAHOS) in 24 which included a number of recommended tools from the COMS report. In this report Measuring Incontinence in Australia Hawthorne (26) indicated this study was designed to report on four important incontinence issues: (a) it provided current prevalence estimates of incontinence in the Australian general community, (b) it provided psychometric insights into those incontinence assessment instruments recommended in the Refining Continence Measurement Tools Page 7

16 Thomas et al. (26) report, (c) it reported Australian population norms for the leading five utility instruments and the impact of incontinence on respondents' lives, and (d) it provided Australian population norms and Australian-derived weights for the SF-36 Version 2. The 24 survey involved interviews with sampled households throughout South Australia. The total number of participants interviewed was 315, giving a within scope response rate of 72. The obtained data were weighted by Australian Bureau of Statistics population estimates to achieve representativeness. Full details concerning sampling procedures and the instruments and items included in the survey can be found in Hawthorne (26) and in Harrison Health Research (24). Urinary incontinence was measured by the Incontinence Severity Index (ISI) (Sandvik et al., 1993; Sandvik et al., 1995, Sandvik et al., 2) and the Urogenital Distress Inventory Short Form (UDI-6) (Shumaker et al., 1994; Uebersax et al., 1995). Faecal incontinence was assessed by the Wexner Faecal Continence Grading Scale (Jorge and Wexner, 1993). Given some concerns about the coverage of the Wexner FCGS (Thomas et al., 26) five additional questions were included measuring bowel pattern, the number of weekly bowel movements, bowel movement urge and soiling. 2.2 Description of Continence Measures Included in SAHOS Short Form of the UDI The Urogenital Distress Inventory (UDI) and the Incontinence Impact Questionnaire (IIQ) were developed by Shumaker, Wyman, Uebersax, McLeish & Fantl (1994) to assess the impact of urinary incontinence symptoms upon quality of life for women. The original forms of the IIQ and the UDI had 3 and 19 items respectively but work by Uebersax, Wyman, Shumaker, McLeish and Fantl (1995) created a 7 item version of the IIQ and a 6 item version of the UDI. The UDI assesses symptoms of incontinence whereas the IIQ focuses on the impact of these symptoms on everyday life. For the latter reason the UDI-6 was included in the community survey. The UDI-6 items ask the respondent Do you experience and, if so, how much are you bothered by the following list of symptoms: Frequent urination Urine leakage related to the feeling of urgency Urine leakage related to physical activity, coughing or sneezing Small amounts of urine leakage (drops) Difficulty emptying your bladder Pain or discomfort in the lower abdominal or genital area The response scale is: Not at all Slightly Moderately Greatly Most of the measurement properties of the UDI-6 have been assessed with women. Test-retest reliability (Long Form) was assessed by (Hagen et al., 22); validity has been demonstrated by items predictive of urodynamic findings in women (Lemack and Zimmern, 1999); correlation with patient self-reported symptom complaints (Harvey et al., 21); and sensitivity to change (Harvey et al., 21). Clinical data is provided by Lukacz et al. (24) and Harvey et al. (21), with normative data for both men and women provided by Svatek et al. (25). Recently, Rodriguez and Raz (23) have added an item on nocturia and an additional urgency item to create the UDI-8. Lubeck et al. (1999) and Brown et al. (1999) created a 9 item Urge-UDI Page 8 Refining Continence Measurement Tools

17 from the Long Form UDI, which includes 3 items from the UDI-6. However, the UDI-6 is more commonly used in the field. Although the short form of the UDI includes items on type of incontinence (urge/stress) the short form only addresses the frequency of a small amount of urine loss. It does not include the item from the UDI (long form) concerning the frequency of a large amount of urine loss. For this reason it may not be as good a measure of severity (frequency x amount) as other instruments. Given this it was thought useful to also include the ISI in the community survey as another measure of symptom severity The Incontinence Severity Index (ISI) This index was developed by Sandvik et al. (1993) and has been used in Norwegian epidemiological surveys of health problems (Sandvik et al., 1993; Hannerstad et al., 2). Testretest reliability (kappa) at 3-days for the Index was.78 and test-retest reliability for each question was.69 and.83 which was also satisfactory (Hanley et al., 21). Although developed for epidemiological surveys slightly modified versions of ISI have been used in the evaluation of treatment programs in general practice (Seim et al., 1996; Holtedahl et al., 1998; Melville et al., 23). It correlates well with physician assessment (Melville et al., 23) and other incontinence instruments (Melville et al., 23; 25b; Klovning et al., 24; Murphy et al., 24). In the original 1993 study, concurrent validity with a 48-hour pad weighing test was found to be r =.59. In a later study, Sandvik, Seim, Vanvik and Hunskaar (2) reported a similar concurrent validity of.54 again with a 48-hour pad weighing test. The moderate correlation of the Index with objective measurement parameters is a strong feature of this tool, as many other instruments show lower validation correlations (Thomas et al., 26). The Incontinence Severity Index was originally intended for use with women, and has been validated with women, but there are no reasons why it cannot be used with men as can be seen below the items do not appear to be gender specific. Its inclusion in the SAHOS 24 survey provided some opportunity to obtain data for males for this Index. It provides a useful short form diagnostic severity measure for urinary incontinence, although clearly the ISI does not measure either urgency or the type of urinary incontinence. As Hawthorne (26) indicated the Incontinence Severity Index (ISI) originally consisted of two items, one with 4 response levels and the other with two response levels (Sandvik et al., 1993). In 2 the instrument developers altered the second item s response scales from 2 to 3 levels, known as the four-level severity index (Sandvik et al., 2). The four-level severity index is reported here whereas the earlier version of this instrument is reviewed by Thomas et al. (26) in the COMS report. The instrument has the following structure: How often do you experience urine leakage? Never = Less than once a month = 1 A few times a month = 2 A few times a week = 3 Every day and/or night = 4 How much urine do you lose each time? None = Drops = 1 Small splashes = 2 More = 3 Severity index = (points for frequency) x (points for amount) Refining Continence Measurement Tools Page 9

18 Scoring is through multiplication of endorsed response levels giving a score range from to 12. Higher scores denote more severe urinary incontinence (Hanley et al., 21). Sandvik et al. (2) recommended that when using the four-level severity index the interpretations were scores 1-2 a slight problem, 3-6 moderate, 7-9 severe and 1-12 very severe. The standard scoring system did not discriminate between those with no incontinence symptoms and those with slight symptoms. The ISI was therefore modified through inclusion of a never category; thus the categories became Never / Less than once a month / 1-several times a month / 1-several times a week / Every day and / night which has added an extra category,, describing those with no symptoms. This procedure is that recommended by Sandvik et al. (2). In the Hawthorne (26) study where classification was needed, ISI scores were recoded into (score range) None () / Slight (1 to 2) / Moderate (3 to 6) / Severe (7 to 9) / Very Severe (1 to 12) levels Wexner Faecal Continence Grading Scale (Wexner FCGS)1 The Jorge and Wexner faecal continence grading scale was developed to provide clinicians with a means of assessing faecal incontinence severity (Jorge and Wexner, 1993). The Wexner scale requires assessment on leakage / accidental faeces for solid, liquid, and gas, the need to wear a pad and alterations to lifestyle. Vaizey et al. (1999) report the test-retest reliability of the Wexner score as r =.75 which falls within an acceptable range (Streiner and Norman, 23). Table 1 The Wexner Faecal Incontinence Grading System Type of incontinence Frequency Never Rarely < 1/month Sometimes < 1/week 1/month Usually < 1/day 1/week Solid Liquid Gas Requires pad Lifestyle Always 1/day Q1. In the past four weeks: do you leak, have accidents or lose control with solid stool? Q2. In the past four weeks: do you leak, have accidents or lose control with liquid stool? Q3. In the past four weeks: do you leak, have accidents or lose control with gas (flatus or wind)? Q4. In the past four weeks: do you need to wear a pad to protect your underwear from stool? Q5. In the past four weeks: does bowel or stool leakage cause you to alter your lifestyle? Each item is assessed on a Guttman scale (Never / Rarely / Sometimes / Usually / Always). Scores are determined by a simple summation of endorsements. The range is from to 2 and the higher the score the worse the faecal incontinence. Scores can also be classified into categorical levels (faecal incontinence Never / Rarely (1 episode in past month) / Sometimes (2-4 episodes) / Weekly (>1 week - <1 day episodes) / Daily (1 or more daily episodes)). The construct validity of the Wexner FCGS is supported by significant correlations with a disease specific measure of psychosocial functioning (Deutekom et al., 25b); and correlation with items from the Euro-Qol 5D - usual activities, anxiety / depression, and mobility (Deutekom et al., 25a); patient s subjective feelings of relief after physiotherapy treatment (Deutekom et al., 25b); and health related quality of life the Gastrointestinal Quality of Life Index (GIQLI) and MOS SF-2 - (Rothbarth et al., 21). Recently, Deutekom et al. (25b) found that change scores on the Wexner correlate.94 with change scores on the Vaizey Scale (Deutekom et al., 25b). The Wexner also correlates well with clinical assessment (.78) and the clinical 1 Wexner has started calling this the Cleveland Clinic Florida Fecal Incontinence Score. Since this measure is not widely known by this name it is not used here. Page 1 Refining Continence Measurement Tools

19 assessment of improvement (.87) (Vaizey et al., 1999); as well as clinical variables using anal manometry (Deutekom et al., 25b; Nazir et al., 22). It is also sensitive to change (Oliveria et al., 1996; Devesa et al., 22; Deutekom et al., 25b; Vaizey et al., 1999). The discriminative validity of the Wexner score is supported by its ability to discriminate between patients who rated their health much better after physiotherapy treatment than those who rated their situation worse or equal or better (Deutekom et al., 25b). Oliveria et al. (1996) found a good correlation between Wexner scores for those patients with successful sphincter repair surgery based on patient, clinical and physiological evaluation after surgery. Significant improvement was found for those patients with a successful repair, while no improvement was found for those patients whose repair had failed. Clinical Data is provided by papers from Oliveria et al. (1996); Rothbarth et al. (21), Kairaluoma et al. (24), Vaizey et al. (1999) and Nazir et al. (22). Normative Data for the Wexner FCGS in women is provided by Melville et al. (25a), with faecal incontinence defined as loss of liquid or solid stool occurring at least monthly. However, Thomas et al. (26) and Hawthorne (26) have both commented that an obvious difficulty with the Wexner FCGS is that it is unconventional to sum symptoms and symptom effects; a procedure that gives rise to double counting. In the Wexner FCGS, stool leakage and its consequence on lifestyle are both counted. Vaizey et al. (1999) has also criticized the Wexner FCGS for the pad wearing question, arguing this was a measure of patient fastidiousness or urinary comorbidity. There are also no items included in the scale that address the issue of urge incontinence (Deutokom et al., 25a; Hawthorne, 26; Thomas et al., 26; Vaizey et al., 1999). Another issue with the Wexner FCGS is the inclusion of an item on flatus which is equally weighted with other items in the scale. Thus a person who endorses daily flatus and no other item will receive the same score as another who endorses only daily incontinence for liquid stool which might be considered a more severe condition. Thus its measurement of severity may be confounded by flatus. This problem associated with the Wexner FCGS, and indeed a number of other faecal incontinence scales, has been commented on by both Hawthorne (26) and AIHW (26). It should be noted that the ICS definition of faecal incontinence specifically excludes flatus (Norton et al., 22; 25). The impact of the inclusion of flatus items in faecal incontinence scales is that estimates of prevalence will be inflated (AIHW, 26). Indeed the AIHW report notes that of the 199 persons in the SAHOS 24 that report problems with any faecal incontinence symptoms that the majority of the 82 / 199 are only reporting flatus. Similar findings for clinical samples were noted by Boreham et al. (25). Following a recommendation from Thomas et al. (26) Hawthorne (26) included both the Wexner FCGS and additional items on bowel patterns, faecal urgency and soiling in the autumn 24 SAHOS as it was thought that following psychometric analysis of these items it may be possible to further refine the Wexner FCGS to address these issues. Refining Continence Measurement Tools Page 11

20 3 Incontinence Measures: Findings from Recent Reports 3.1 Measures of Urinary Incontinence Hawthorne (26) in Measuring Incontinence in Australia reported that in his analysis of the 24 SAHOS (Harrison Health Research, 24) the reliability of the UDI-6 was Cronbach α =.78. Two items, the last two questions How much are you bothered by pain or discomfort in the lower abdominal or genital area? and Do you have difficulty emptying your bladder? did not fit well with the other items (the item-total correlations were r =.31 and r =.37 respectively). Deletion of these items would have improved the Cronbach α to.81 and the explained variance from 49 to 66 (Hawthorne, 26). The reliability of the ISI was Cronbach α =.89 (Hawthorne, 26). The relationship between the two items of the ISI was examined using kappa (κ) and found to be.74 indicating good agreement between the two items (Landis and Koch, 1977). Hawthorne (26) report there was perfect agreement between the two items for 9 of all respondents. Thus it appears the ISI gained its high level reliability through replication; the r s =.96 and the proportion of explained variance was 92 (Hawthorne, 26). Although this analysis suggested the ISI has excellent measurement properties that it consists of just two items suggests that it violates classical test theory which postulates that at least 3 items are needed for stable measurement interpretation (Hawthorne, 26; Tabachnick and Fidell, 21). Regarding the relationship between the two estimates of urinary incontinence (the ISI and UDI-6) this was examined after conversion to McCall s T-scores to compensate for the different scale ranges used by the two measures (Hawthorne, 26). The correlation was r S =.75 (n = 35, p <.1). This suggested that the ISI and UDI-6 are measuring similar and related aspects of incontinence. However, Hawthorne (26) indicated that the UDI-6 was more sensitive for those with minor or moderate symptoms and that the ISI classified more cases at the floor and ceiling of the instrument. The ISI classified 76 of cases as continent compared with 54 of cases for the UDI-6. The prevalence estimate for urinary incontinence derived from the ISI was 24 as contrasted with the estimate from the UDI-6 of 47 which indicated these measures were not fully compatible (Hawthorne, 26). When the cases were dichotomized into no symptoms / any symptoms there was agreement for only 77 of cases. Of the 23 of cases where there was disagreement nearly all of these were where no symptoms were reported on ISI but were classified as symptomatic by the UDI (Hawthorne, 26). The UDI measures a broader range of urological symptoms and their impact / bother whereas ISI measures frequency and volume of urinary leakage. It has earlier been noted that some items in the UDI (frequency of urination and discomfort in lower abdominal or genital area) may be gaining endorsement from those with non-urinary incontinence conditions. When Hawthorne (26) excluded these UDI items from the classification analysis the agreement between the two measures rose to 87. Excluding these items from the UDI also reduced the prevalence estimate for UDI to 36. Given consideration of the above factors the prevalence estimates derived from the ISI are generally preferred (Hawthorne, 26; AIHW, 26). 3.2 Measures of Faecal Incontinence Given concerns that the Wexner FCGS does not include an item on faecal urgency and may require revision, a number of other items were also included in the SAHOS 24 survey. These included items on faecal urgency, bowel patterns and soiling. As indicated above the Wexner FCGS includes an item on flatus that is equally weighted with other faecal incontinence items in this scale. However, the current ICS definition specifically excludes flatus (Norton et al., 22; 25). In the 24 SAHOS it was found that of the 199 Page 12 Refining Continence Measurement Tools

21 persons that endorsed any faecal incontinence symptom 892 were only endorsing the flatus item (AIHW, 26). Hawthorne (26) indicates this can have a profound effect on prevalence estimates for faecal incontinence. When the flatus item is included in the Wexner FCGS the prevalence is estimated at 35. However, if the flatus item is excluded, to be in line with the ICS definition, the prevalence estimate drops to 8 and the latter figure is more consistent with other prevalence estimates (AIHW, 26). Hawthorne (26) notes that the prevalence estimate of 35 based on the standard Wexner FCGS is higher than that for urinary incontinence as measured by the ISI (24) a situation which he notes is inconsistent with the incontinence literature. The AIHW in their 26 report Australian Incontinence: Data Analysis and Development also used the SAHOS 24 data for their estimates of the burden of faecal incontinence. They examined the flatus question from the Wexner FCGS (X6) in relation to AQOL scores for different severities of faecal incontinence and found that very few with flatus incontinence only had a decreased AQOL (AIHW, 26) They also examined the flatus item in relation to a question on elimination (Question Y8: problems with bowel/ bladder) from the 15D health related quality of life measure (Sintonnen and Pekurinen, 1993; Sintonnen, 1994; 1995; 21). Of the 82 people reporting flatus incontinence only, 84 considered their bowel and bladder worked normally (AIHW, 26). For these reasons they considered it unhelpful to analyse the flatus incontinence only group with those experiencing incontinence for solid or liquid stool. Thus they analysed this group as a separate category and applied a zero severity weight so that it made no contribution to the overall burden of incontinence. AIHW also noted there was considerable overlap amongst the remaining items in the Wexner FCGS (AIHW, 26). The people who use pads or have their lifestyle affected by bowel or stool leakage are largely a subset of those who report problems with leaking or losing control of solid and liquid stool (AIHW, 26). Earlier it was noted that in the Wexner FCGS, that stool leakage (the symptom) and its consequence (the symptom effect e.g. wearing a pad or effect on lifestyle) are both included which presents a form of double counting (Thomas et al., 26 and Hawthorne, 26). AIHW (26) formed an index calculated on just using the liquid and solid stool questions and compared this with an index based on four questions (liquid, solid, pads, lifestyle). The correlation between the two indexes was r =.94 and they concluded there was little additional information value in including the pad and lifestyle questions. They based their prevalence estimate (7.1) on just the incontinence for liquid stool and incontinence for solid stool questions from the Wexner FCGS. Hawthorne (26) examined the relationship between soiling and faecal urge incontinence. When compared with those that reported no faecal urge those who reported at least monthly faecal urge were twelve times more likely to report soiling and those who reported urge often / daily were 3 7 times more likely to report soiling. Faecal urge appears to be a good predictor of soiling (its consequence) and it may be useful to include an item on urgency in any modification of the Wexner FCGS. Refining Continence Measurement Tools Page 13

22 4 Rationale of Study 4.1 Aims The discussion in the preceding sections indicates there is a good case to revise the recommended instruments (UDI-6, ISI and Wexner FCGS) to improve their psychometric performance and appropriateness for use in assessing and monitoring incontinence. Given the findings concerning the UDI-6 and the ISI this study is concerned with analysing the items from these measures to see if a better overall measure can be developed for urinary incontinence. The ISI may be too short for stable measurement interpretation and does not include items concerning type of incontinence which would be particularly useful if it is to be used in primary and community care settings. The UDI-6, although including items concerning the type of incontinence, contains a number of problem items which may be gaining endorsement from nonurinary incontinence conditions. The item concerning urinary leakage in the UDI-6 is restricted to a small amount of volume and it may not appear to be as good a measure of severity as the ISI. The goal of this study is to develop a urinary incontinence measure which is based on a blend of the best items from the UDI-6 and the ISI drawn from an examination of SAHOS 24 data. Similarly it has been identified above that there are a number of problems with the Wexner FCGS. The scale appears to be confounded by flatus which is explicitly excluded by the current ICS definition of faecal incontinence (Norton et al., 25). There appears to be a double counting of elements as symptoms and their consequences are both counted in the total score. The Wexner FCGS also contains no item concerning faecal urgency which is considered important by clinicians in the field (Deutekom et al., 25a; Vaizey et al., 1999; Moore et al., 26). The aim of this study is to examine the psychometric performance of items within the Wexner FCGS along with other faecal incontinence items included in the 24 SAHOS (items concerning bowel patterns, urgency and soiling) to develop a refined faecal incontinence measure. 4.2 Overview of Methods The first approach taken will be to examine the items and scales based on classical test theory (Streiner and Norman, 23). The following steps will be undertaken and for these analyses unweighted survey data from 24 SAHOS will be utilized. Proposed steps in examining the urinary and faecal incontinence items: 1. Explore basic item properties such as means, medians and distributions 2. Item analysis examine item endorsement and discrimination 3. Item analysis examine item total correlations and internal consistency reliability 4. Content Validity relate the items back to their construct(s) using exploratory factor analysis 5. Select items for refined measures and compare with IRT approaches 6. Examine the relationship of the items and the refined scales with other constructs, namely health utilization, disability and self-reported health status The report on these analyses is to be found in Section 5. The second approach to refining incontinence measures was to use modern test theory (MTT), specifically IRT. In this phase the study will: Page 14 Refining Continence Measurement Tools

23 7. Examine instruments from a logical perspective (e.g. differentiate between items measuring incontinence vs. impact of incontinence) 8. Use IRT analyses to provide more in depth analysis of item data and response options and undertake exploratory and confirmatory item level factor analyses, to assist with refining of incontinence measures 9. Undertake structural equation modelling to examine and test models that best fit the data 1. Compare findings with the Classical Approach taken above to finalise the refined measures The report on these analyses is to be found in Section 6. Norm tables for the refined measures according to age and sex are presented in Appendix 3. In Section 7 recommendations will be discussed. One of the limitations of community survey data is that as the data is collected in face to face interviews the data are at the level of subjective reports of incontinence symptoms rather than confirmed diagnoses. A community survey will also exclude those currently placed in institutional settings (e.g. nursing homes) and as such the prevalence estimates for the elderly are likely to be underestimated (Hawthorne, 26). These considerations mean that in a community survey there will be a limited range of responses to incontinence items particularly those endorsing more severe levels of symptoms. It will thus be necessary to trial the refined measures in a range of clinical settings in follow up field trials. Refining Continence Measurement Tools Page 15

24 5 Data Analysis: Classical Approach This psychometric analysis of the continence items in the 24 SAHOS dataset is in three parts. It includes a basic work-up (item properties), followed by Item Analysis and Reliability Analysis (item endorsement and discrimination; item-total correlations and internal consistency reliability); and finally an examination of Content Validity (exploratory factor analysis; relationships with other health variables). This analysis is in line with the guidelines set out in the standard text in health outcomes measurement by Streiner and Norman, 23. The focus of this work is to determine whether the items are homogeneous or internally consistent. That is, are the items tapping the same concept (i.e. faecal or urinary incontinence)? 5.1 Item Properties This secondary data analysis was conducted using an unweighted sample of adults, 18 years of age and over (age groupings were based on those used in the ABS National Health Surveys [see ABS, 1997; ABS 26]; namely year olds; year olds; year olds; year olds; year olds; year olds; and 75 years and over). Table 2 shows the total number of males and females within each age-group. In the 24 SAHOS dataset all the urinary incontinence items were contained in section W and all faecal incontinence items were contained in section X. The urinary incontinence items contained the six items from the UDI-6 (items W1 - W6) and the two items from the Incontinence Severity Index (W7 - W8). The faecal incontinence items included the 5 items from the Wexner FCGS (X4, X5, X6, X8, X1) and 5 other items addressing bowel pattern, urge and soiling (X1, X2, X3, X7, X9). NB: Adolescents (i.e. those under the age of the 18 years) were not included in this analysis of adult survey participants. Table 2 Total number of subjects according to sex and age group in the SAHOS Data set (N = 2924)* Age Group Males Females Total * = This table includes SAHOS survey participants who either refused to answer some of the items or had missing data. It should be noted that small changes in sample size across the various figures used in this section are due to the removal of participants who refused to answer certain items or who had missing data. Examination of Table 2 shows that there are at least 1 people in each cell of the analysis. This suggests that there is an adequate number of scores to statistically analyse. Another key assumption of this correlation analysis is that the Pearson correlation coefficient is robust enough to cope with deviations from the normal distribution (Streiner and Norman, 23). Page 16 Refining Continence Measurement Tools

25 Table 3 and Table 4 outline the items and questions used in the SAHOS data set. Table 3 Labels, questions and abbreviations for the urinary incontinence items in the SAHOS data set Item Question Response Abbreviation W1 Do you experience and, if so, how much are you bothered by: frequent urination? 1 = Not at all 2 = Slightly 3 = Moderately 4 = Greatly Frequent Urination W2 Do you experience and, if so, how much are you bothered by: urine leakage related to the feeling of urgency (a sudden desire to urinate)? 1 = Not at all 2 = Slightly 3 = Moderately 4 = Greatly Urgency Leakage W3 Do you experience and, if so, how much are you bothered by: urine leakage related to physical activity, coughing or sneezing? 1 = Not at all 2 = Slightly 3 = Moderately 4 = Greatly Stress Leakage W4 Do you experience and, if so, how much are you bothered by: small amount of urine leakage (drops)? 1 = Not at all 2 = Slightly 3 = Moderately 4 = Greatly Leak Small Amount W5 Do you experience and, if so, how much are you bothered by: difficulty emptying your bladder? 1 = Not at all 2 = Slightly 3 = Moderately 4 = Greatly Emptying Bladder W6 Do you experience and, if so, how much are you bothered by: pain or discomfort in the lower abdominal or genital area? 1 = Not at all 2 = Slightly 3 = Moderately 4 = Greatly Pain Lower Abdominal W7 How often is urine leakage experienced? = Never 1 = Less than once a month 2 = Several times a month 3 = Several times a week 4 = Every day or night Frequency Leakage W8 How much urine is lost each time? = None 1 = Drops 2 = Small splashes 3 = More Amount Leakage NB: Numerical values for the response categories of these items were recoded in order to match the scoring used in published articles. Codes for those participants who refused to answer the item were also removed. Refining Continence Measurement Tools Page 17

26 Table 4 Labels, questions and abbreviations for the faecal incontinence items in the SAHOS data set Item Question Response Abbreviation X1 In the past four weeks: How do you describe your usual bowel pattern? 1 = Normal 2 = Constipated 3 = Diarrhoea 4 = Alternating constipation and diarrhoea Bowel Pattern X2 In the past four weeks: How many bowel movements do you usually have in a week? 1 = 1 or less 2 = 2 3 = = = = = 27 or more Bowel Movements X3 In the past four weeks: Do you experience an urgent need to have a bowel movement that makes you rush to a toilet? = Never 1 = Rarely 2 = Sometimes 3 = Often or usually 4 = Always* Urgency X4 In the past four weeks: Do you leak, have accidents or lose control with a solid stool? = Never 1 = Rarely 2 = Sometimes 3 = Often or usually 4 = Always* Leak Solid X5 In the past four weeks: Do you leak, have accidents or lose control with a liquid stool? = Never 1 = Rarely 2 = Sometimes 3 = Often or usually 4 = Always* Leak Liquid X6 In the past four weeks: Do you leak, have accidents or lose control with gas (flatus or wind)? = Never 1 = Rarely 2 = Sometimes 3 = Often or usually 4 = Always* Leak Gas X7 In the past four weeks: Do you leak stool if you don t get to a toilet in time? = Never 1 = Rarely 2 = Sometimes 3 = Often or usually 4 = Always* Leak Stool / Urgency X8 In the past four week: Do you need to wear a pad to protect your underwear from stool? = Never 1 = Rarely 2 = Sometimes 3 = Often or usually 4 = Always* Wear Pad X9 In the past four weeks: Does stool leak so that you have to change your underwear? = Never 1 = Rarely 2 = Sometimes 3 = Often or usually 4 = Always* Leak / Change Underwear X1 In the past four weeks: Does bowel or stool leakage cause you to alter your lifestyle? = Never 1 = Rarely 2 = Sometimes 3 = Often or usually 4 = Always* Alter Lifestyle *Rarely: less than once in the past four weeks Sometimes: less than once a week, but once or more in the past four weeks Often or usually: less than once a day but once a week or more Always: once or more per day or whenever you have a bowel movement NB: Numerical values for the response categories of these items were recoded in order to match the scoring used in published articles. Codes for those participants who refused to answer the item were also removed. Page 18 Refining Continence Measurement Tools

27 The following tables present the item properties for each item (mean, standard deviation, median, and range) for males and females. Table 5 The item properties of the urinary incontinency items for males Males Abbreviation N Mean SD Median Range W1 Frequent urination W2 Urgency leakage W3 Stress leakage W4 Leak small amount W5 Emptying bladder W6 Pain lower abdominal W7 Frequency leakage W8 Amount leakage NB: The range is the actual range of scores obtained for each item, not the range of the item s response categories. Table 6 The item properties of the urinary incontinency items for females Females Abbreviation N Mean SD Median Range W1 Frequent urination W2 Urgency leakage W3 Stress leakage W4 Leak small amount W5 Emptying bladder W6 Pain lower abdominal W7 Frequency leakage W8 Amount leakage NB: The range is the actual range of scores obtained for each item, not the range of the item s response categories. Refining Continence Measurement Tools Page 19

28 Table 7 The item properties of the faecal incontinency items for males Males Abbreviation N Mean SD Median Range X1 Bowel pattern X2 Bowel movements X3 Urgency X4 Leak solid X5 Leak liquid X6 Leak gas X7 Leak stool / urgency X8 Wear pad X9 Leak / change underwear X1 Alter lifestyle NB: The range is the actual range of scores obtained for each item, not the range of the item s response categories. Table 8 The item properties of the faecal incontinency items for females Females Abbreviation N Mean SD Median Range X1 Bowel pattern X2 Bowel movements X3 Urgency X4 Leak solid X5 Leak liquid X6 Leak gas X7 Leak stool / urgency X8 Wear pad X9 Leak / change underwear X1 Alter lifestyle NB: The range is the actual range of scores obtained for each item, not the range of the item s response categories. Page 2 Refining Continence Measurement Tools

29 The item properties of the three incontinence measures for males and females are outlined below. Table 9 The item properties of the three incontinence measures (ISI, UDI-6 and Wexner FCGS) for males N Mean SD Median Range ISI UDI Wexner FCGS W1 + W2 + W3+ W4+ W5+ W6 = UDI-6; W7 + W8 = ISI X4 + X5 + X6 + X8 + X1 = Wexner FCGS NB: The range is the actual range of scores obtained for each measure, not each measure s score range. (The score range for each measure is ISI: 12; UDI-6: 6 24; Wexner FCGS: 2) Table 1 The item properties of the three incontinence measures (ISI, UDI-6 and Wexner FCGS) for females N Mean SD Median Range ISI UDI Wexner FCGS W1 + W2 + W3+ W4+ W5+ W6 = UDI-6; W7 + W8 = ISI X4 + X5 + X6 + X8 + X1 = Wexner FCGS NB: The range is the actual range of scores obtained for each measure, not each measure s score range. (The score range for each measure is ISI: 12; UDI-6: 6 24; Wexner FCGS: 2) These tables show the item properties and the low level of endorsement for these items. This is what you would expect in a community population survey. However, low levels of endorsement do restrict the range of the scores that can be analysed. Further examination of the endorsement issue follows. 5.2 Item Endorsement and Discrimination The figures in Appendix 2 demonstrate the level of non-endorsement (i.e. the inverse of endorsement) for each of the urinary and faecal incontinence items, according to age group and Refining Continence Measurement Tools Page 21

30 sex. This allows for the analysis of the each item s endorsement in the community (this can be viewed as each item s ability to discriminate between groups in the community). Examining the urinary incontinence items (see Appendix 2) one finds: Different patterns of endorsement for males and females; with females endorsing the Not at all / Never / None category less than males. This indicates the presence of urinary incontinence is more common in females Declining levels of endorsement of the Normal or Not at all / Never / None category across age groups; with younger age groups endorsing the Normal or Not at all / Never / None category more. This indicates the presence of urinary incontinence is more common in older adults There is a large (approximately 2 4 ) difference between high and low endorsing age groups for males and females for most urinary incontinence items (see items W1, W2, W3, W4, W7, W8); while items W5 (emptying bladder) and W6 (pain lower abdominal) show a smaller difference across age groups To illustrate these points two figures are presented as examples. Figure 1 The level of non-endorsement for males and females on Item W1 (Frequent Urination) W.1 Do you experience and, if so, how much are you bothered by: frequent urination? (n = 2923) Male Female Age Group Page 22 Refining Continence Measurement Tools

31 Figure 2 The level of non-endorsement for males and females on Item W5 (Emptying Bladder) W.5 Do you experience and, if so, how much are you bothered by: difficulty emptying your bladder? (n = 2922) Male Female Age Group Examining the faecal incontinence items (see Appendix 2) figures one finds: Similar patterns of endorsement of the Normal or Never category for males and females (within 1 percent). This indicates that the presence of faecal incontinence is not common amongst males or females Across all age groups a high proportion of survey participants (greater than 8 or 9) endorse the Normal or Never category for most faecal incontinence items (see items X1, X4, X5, X7, X8, X9, X1). This indicates the presence of faecal incontinence is not common across age groups There were lower levels of responding with Normal or Never (around 5 or 6) for items X3 (urgency) and X6 (leak gas). This would indicate that these symptoms are more common in the community than the other faecal incontinence leakage symptoms (The endorsement of Never for Item X6 also appears to decline with age). To illustrate these points two figures are presented as examples. Refining Continence Measurement Tools Page 23

32 Figure 3 The level of non-endorsement for males and females on Item X4 (Leak Solid) X.4 In the past four weeks: Do you leak, have accidents or lose control with solid stool? (n = 2921) Age Group Male Female Figure 4 The level of non-endorsement for males and females on Item X6 (Leak Gas) X.6 In the past four weeks: Do you leak, have accidents or lose control with gas (flatus or wind)? (n = 2921) Age Group Male Female From this analysis, it appears that most urinary incontinence items appear to exhibit different levels of endorsement according to age group and sex. While the levels of endorsement for the faecal incontinence items do not appear to exhibit these differences, with uniform high levels of non-endorsement across most faecal incontinence items. Items X3 (urgency) and X6 (leak gas) are the only exceptions here. The additional figures in Appendix 2 show the proportion of males and females endorsing each response category for each item and each age group. These figures visually display the distribution of responses to each question. Page 24 Refining Continence Measurement Tools

33 An examination of these figures demonstrates very similar patterns of response across the urinary incontinence items, particularly when you examine the Slightly / Less than once a month / A few drops response category. Females endorse these categories more often than males. There is also a greater endorsement of this category in the older age groups when combined. This can be seen in the following figures: Figure 5 The distribution of responses for females on item W4 (Leak small amount) W.4 Do you experience and, if so, how much are you bothered by: small amount of urine leakage (drops)? (n = 1716) Not at all Slightly Moderately Greatly Refused Age Group FEMALES Figure 6 The distribution of responses for males on item W4 (Leak small amount) W.4 Do you experience and, if so, how much are you bothered by: small amount of urine leakage (drops)? (n = 126) Not at all Slightly Moderately Greatly Refused Age Group MALES Refining Continence Measurement Tools Page 25

34 While most of the faecal incontinence items show very low levels of response for this second category for both sexes (i.e. around 1). Items X3 (urgency) and X6 (gas leak) have high rates of responding for the Rarely i.e. less than once in the past four weeks option indicating these symptoms are more common in the community. As an illustration of the previous point, the following figure provides a typical example of the distribution of responses for most faecal incontinence items. Figure 7 The distribution of responses for females on item X4 (Leak solid) X.4 In the past four weeks: Do you leak, have accidents or lose control with solid stool? (n = 1716) Never Rarely Sometimes Often or usually Always Refused Age Group FEMALES Compare this figure to the distribution of response for item X3 (urgency) below: Figure 8 The distribution of response for females on item X3 (urgency) 1 X.3 In the past four weeks: Do you experience an urgent need to have a bowel movement that makes you rush to a toilet? (n = 1716) Never Rarely Sometimes Often or usually Always Refused Age Group FEMALES Page 26 Refining Continence Measurement Tools

35 5.3 Item-total Correlations and Internal Consistency Reliability The overall internal consistency reliability data for the three incontinence measures are as follows: ISI =.83 (Pearson s correlation was used because there were only two items); UDI-6 =.78; and.57 for the Wexner FCGS. The results for the ISI and the UDI-6 are in the acceptable range of.7.9 while Wexner FCGS is in the unacceptable range (Streiner and Norman, 23). The internal consistency results for the UDI-6 compare to the published figures by Harvey et al., 21 and Lukacz et al., 24 who found alpha values of.52 and.73 respectively. Further analysis of the homogeneity of the items was undertaken by examining the corrected item - total correlations and the effect on Cronbach s alpha (the internal consistency of the scale) if each item were deleted. For both these coefficients, higher values closer to 1., are required if one wishes to produce scales with items that are internally consistent or homogenous. The item - total correlation is corrected in the sense that the effect of the original item is removed from the total score. Table 11 and Table 12 outline the corrected item - total correlations and the Cronbach s alpha coefficients if the item was deleted for the UDI-6 and Wexner FCGS. Table 11 Corrected item - total correlations and Cronbach s alpha if the item was deleted for each item of the UDI-6 Item Corrected Item Total Correlation Cronbach s Alpha if Item Deleted W1 (Frequent Urination) W2 (Urgency Leakage).7.7 W3 (Stress Leakage) W4 (Leak Small Amount) W5 (Emptying Bladder) W6 (Pain Lower Abdominal) This table shows that for the UDI-6 that the corrected item - total correlations for items W5 (emptying bladder) and W6 (pain lower abdominal) are low, just above.2 which is at the lower end of the acceptable range (Streiner and Norman, 23). The Cronbach s alpha data shows that if items W5 (emptying bladder) and W6 (pain lower abdominal) were deleted from the UDI it would not affect the overall internal consistency of the scale. Indeed if both these items are removed Cronbach s alpha increases slightly to.81. Refining Continence Measurement Tools Page 27

36 Table 12 Corrected item - total correlations and Cronbach s alpha if the item was deleted for each item of the Wexner Scale (X4, X 5, X6, X8, X1) Item Corrected Item Total Correlation Cronbach s Alpha if Item Deleted X4 (Leak Solid) X5 (Leak Liquid) X6 (Leak Gas) X8 (Wear Pad).39.5 X1 (Alter Lifestyle).42.5 This table shows that item X6 (leak gas) has a low corrected item - total correlation, just above the acceptable range of.2 (Streiner and Norman, 23). The Cronbach s alpha data also suggests that if item X6 (leak gas) is deleted then Cronbach s alpha moves to an acceptable level of.77. This analysis shows that the UDI-6 can be shortened without affecting its internal consistency by removing items W5 (emptying bladder) and W6 (pain lower abdominal). The internal consistency of the Wexner FCGS can be raised to acceptable levels by the removal of item X6 (leak gas). 5.4 Exploratory Factor Analysis Prior to exploratory factor analysis an examination of the relationships or correlations between the three incontinence measures was undertaken. These are presented below using the Pearson s correlation coefficient: Page 28 Refining Continence Measurement Tools

37 Table 13 The correlation matrix between the ISI, UDI-6 and Wexner FCGS according to the Pearson correlation coefficient (parametric) [n = 2911] Correlation Type = Pearson ISI UDI-6 Wexner ISI 1 UDI Wexner X4 + X5 + X6 + X8 + X1 = Wexner FCGS W1 + W2 + W3+ W4+ W5+ W6 = UDI-6; W7 + W8 = ISI This table shows high correlations between the ISI and UDI-6, with a reduced correlation between the faecal and urinary incontinence measures (The magnitude or strength of these relationships remains even when non parametric alternative correlations like Spearman s rho and Kendall s tau_b are used). The proportion of the variance (r 2 ) explained for each of these measures, using the Pearson s correlation coefficient was as follows: Wexner and UDI =.17; Wexner and ISI =.12; and UDI and ISI =.52. Exploratory factor analysis attempts to reduce a correlation matrix into its underlying constructs. It thereby simplifies description and helps examine the relationship between individual items and the underlying common elements. Table 14 and Table 15 are correlation matrices for the urinary and faecal incontinence items. Table 14 Correlation matrix for the 8 urinary incontinence items W1 W2 W3 W4 W5 W6 W7 W8 W1 Frequent Urination 1 W2 Urgency Leakage.56 1 W3 Stress Leakage W4 Leak Small Amount W5 Emptying Bladder W6 Pain Lower Abdominal W7 Leakage Frequency W8 Leakage Amount W1 + W2 + W3+ W4+ W5+ W6 = UDI-6; W7 + W8 = ISI Refining Continence Measurement Tools Page 29

38 Table 15 Correlation matrix for the 1 faecal incontinence items X1 X2 X3 X4 X5 X6 X7 X8 X9 X1 X1 Bowel Pattern 1 X2 Bowel Movements X3 Urgency X4 Leak Solid X5 Leak Liquid X6 Leak Gas X7 Leak Stool / Urgency X8 Wear Pad X9 Leak / Change Underwear X1 Alter Lifestyle X4 + X5 + X6 + X8 + X 1 = Wexner FCGS These matrices show high correlations between certain items, for example W7 (leakage frequency) and W8 (leakage amount) with W 4 (leak small amounts); and X9 (leak / change u nderwear) with X7 (leak stool / urgency). It also demonstrates where there are low correlations between items, for example between W6 (pain lower abdominal) and W8 (leakage amount); as well as X2 (bowel movements) and X8 (wear pad). These patterns of high and low correlations can be further simplified by using the psychometric technique of factor analysis. The method of exploratory factor analysis used was principal components analysis for extraction (eigenvalues > 1.) with varimax rotation. Table 16 and Table 17 provide the rotated factor variables for the urinary incontinence and faecal incontinence items. The urinary incontinence items had a two factor solution, accounting for 67 of the variance, while the factor analysis for the faecal incontinence items created a three factor solution accounting for 61 of the variance. According to Norman and Streiner (23) these percentages represent acceptable solutions. (Further analysis of this data, which included adding the group of year olds to the sample and / or using the sample weights derived from the 24 SAHOS produced very similar factor structures.) Table 16 Rotated Factor Matrix for the urinary incontinence items W1 W8 High loadings >.5 are in bold Factor 1 2 W1 Freq Urination W2 Urgency Leakage W3 Stress Leakage.82.9 W4 Leak Small Amount W5 Emptying Bladder W6 Pain Lower Abdominal.9.75 W7 Leakage Frequency W8 Leakage Amount W1 + W2 + W3+ W4+ W5+ W6 = UDI-6; W7 + W8 = ISI Page 3 Refining Continence Measurement Tools

39 Table 17 Rotated Factor Matrix for the faecal incontinence items X1 X1 High loadings >.5 are in bold Factor X1 Bowel Pattern X2 Bowel Movements X3 Urgency X4 Leak Solid X5 Leak Liquid X6 Leak Gas X7 Leak Stool / Urgency X8 Wear Pad X9 Leak / Change Underwear X1 Alter Lifestyle X4 + X5 + X6 + X8 + X1 = Wexner FCGS For the urinary incontinence items, Rotated Factor 1 accounted for a large proportion of the variance While Rotated Factor 2 accounted for Describing the item loadings on each factor the following points emerge: Items W2 (urgency leakage), W3 (stress leakage), W4 (leak small amount) from the UDI-6 as well as items W7 (leakage frequency) and W8 (leakage amount) from the ISI, load highly on Rotated Factor 1 (weights above.5) Item W1 (frequent urination) from the UFI-6 appears to be complex as it loads on both Rotated Factors 1 and 2 Items W5 (emptying bladder) and W6 (pain lower abdominal) from the UDI-6 load highly on Rotated Factor 2 (weights above.5) In light of this, Rotated Factor 1 appears to represent the common factor of urinary leakage / incontinence, while Rotated Factor 2 seems to reflect specific urological symptom issues like lower abdominal pain and bladder emptying. In terms of urinary incontinence, this analysis suggests items W2, W3, W4, W7, and W8 load on a common central factor. This seems to be defined by the issue of leakage. This factor combines items from the ISI and UDI-6. For the faecal incontinence items, Rotated Factor 1 accounted for a large proportion of the variance 4.6, while Rotated Factor 2 and 3 accounted for 1.7 and 1.24 respectively. Examining the item loadings on each factor the following points emerge: Items X4 (leak solid), X5 (leak liquid), X8 (wear pad), and X1 (alter lifestyle) from the Wexner FCGS, plus X7 (leak stool / urgency) and X9 (leak / change underwear), all load heavily on Rotated Factor 1 (weights above.5) Refining Continence Measurement Tools Page 31

40 Items X6 (gas), X3 (urgency) and X1 (bowel pattern) load highly on Rotated Factor 2 (weights above.5). This seems to reflect a collection of other symptoms like gas, urgency and erratic bowl patterns. This factor is hard to interpret, as these items do not seem to neatly fit together Item X2 (bowel movements) loads only on Rotated Factor 3. This item appears to define this factor almost completely In light of this, Rotated Factor 1 appears to represent the common factor of soiling. While Rotated Factor 2 could represents other bowel / stomach symptoms and Rotated Factor 3 represents the number of bowel movements. In terms of faecal incontinence, this analysis suggests items X4, X5, X7, X8, X9 and X1 load on a common factor. This seems to be defined by leakage and soiling. This factor is made up of most of the items from the Wexner FCGS plus two extra items. An issue concerning item redundancy might be considered by comparing the following items: Do you need to wear a pad to protect your underwear from stool? (Item X8, Wexner FCGS) Does stool leak so that you have to change your underwear? (Item X9) The pad question from the Wexner FCGS has previously been criticized by Vaizey (1999) as it may relate more to patient fastidiousness rather than faecal incontinence / soiling per se. Given these considerations and its similarity to the soiling item above, in loading on the general faecal incontinence factor (.78 vs..71), it is suggested this item be excluded from the Revised Faecal Incontinence Scale. 5.5 Relationships with Other Health Variables This stage of the analysis examines the relationship between the three incontinence measures, the ISI, UDI-6 and Wexner FCGS with other health variables in the SAHOS Data set. The other self-reported health measures include scales like the SF-36 (Ware and Sherbourne, 1992), EQ5D (EuroQoL Group, 199), 15D (Sintonen and Pekurinen, 1993) and AQoL (Hawthorne, Richardson an d Osborne, 1999); as well as the Friendship Scale (Hawthorne and Griffith, 2), a measure of social isolation. Subscales and relevant individual items were also analysed and questions about health utilization (the number and type of services used in the past month) and self-reported disability days (full days and partial days over the past month). Table 18 outlines the Pearson correlation coefficients between the ISI, UDI-6 and Wexner FCGS with other measures of health. For ease of interpretation significant correlations which are of a magnitude of.2 or greater are shown. Page 32 Refining Continence Measurement Tools

41 Table 18 Pearson correlations between the ISI, UDI-6 and Wexner FCGS with other measures of health. Correlations.2 or greater are shown ISI UDI-6 Wexner FCGS ISI 1 UDI Wexner CGS Self-Rated Health (SF-36 Q1) SF-36 PCS SF-36 MCS -.21 SF-36 PF SF-36 RP SF-36 BP SF-36 GH SF-36 VI SF-36 SF -.25 SF-36 RE SF-36 MH -.24 EQ5D EQ5D Mobility Qu EQ5D Look after Qu EQ5D Usual activities Qu EQ5D Pain Qu EQ5D Anxious / Depressed Qu.25 Friendship Scale Score D Q1 Mobility D Q2 Vision 15D Q3 Hearing 15D Q4 Breathing D Q5 Sleeping D Q6 Eating 15D Q7 Speech 15D Q8 Elimination D Q9 Usual activities D Q1 Mental function D Q11 Discomfort and symptoms D Q12 Depression.23 15D Q13 Distress D Q14 Vitality D Q15 Sexual activity D Total AQoL Total AQoL Illness dimension AQoL Independent living dimension AQoL Social relationships dimension AQoL Physical senses dimension AQoL Psychological w ellbeing dimension AQoL Q1 medicines.24.3 Number of types of services used Total Number of Health Utilization Visits ABS Disability Days - Full ABS Disability Days - Partial.2 X4 + X5 + X6 + X8 + X1 = Wexner FCGS W1 + W2 + W3+ W4+ W5+ W6 = UDI-6; W7 + W8 = ISI Refining Continence Measurement Tools Page 33

42 An examination of the direction of the correlations was conducted and they seem to be in the correct order. For those sub-scales of the self-report health status measures where a higher score indicated more problems (e.g. 15D and EQ-5D subscales) there was a positive correlation with incontinence scales where again a high score is indicative of more problems. However, for those self-reported health status measure subscales where a higher score is indicative of fewer problems (e.g. SF-36 subscales), then the correlation is negative. Where the item / sub-scale and total scores produce different signs, for example with the EQ5D, this is because the total score has been inverted. As expected, moderate to low correlations (.2.3) were found on the total scores of the selfreport health status measures SF-36 PCS, EQ5D, 15D and AQoL. This suggests a relationship between continence conditions and health status. A high correlation was also found with 15D question regarding elimination (.44.53). In terms of the subscales and relevant items, the following patterns emerged when correlations greater than.2 were found for all three of the incontinence measures: Correlations with relevant items and subscales concerning physical functioning, role functioning, mobility, usual activities, independent living (SF-36 PF, SF-36 RP, EQ5D Mobility, EQ5D Usual Activities, 15D Mobility, 15D Usual Activities, AQoL Independent living dimension) Correlations with relevant items and subscales concerning general health and vitality (selfrated health SF-36 Q1, SF-36 GH, SF-36 VI, 15D Vitality) Correlations with relevant items and subscales concerning mental functioning and social relationships (15D Mental function; AQoL Psychological well-being dimension, AQoL Social relationships dimension). However, the SF-36 Mental Health, EQ5D Anxious / Depressed, 15D Depression did not correlate with all three incontinence measures. This is also true for the SF- 36 MCS score. Correlations with discomfort and symptoms (15D: Discomfort and Symptoms) and breathing (15D: Breathing). No relationship was found between items regarding vision, hearing, eating and speech (see the items from the 15D and the AQoL Physical Senses dimension) Also no relationship was found between the number and type of services used in the past month; as well as self-reported disability days either full or partial. These patterns supported the correlations that were found for the total scores of the various health status measures. Overall, this correlation analysis suggests that the scores on the continence measures are related to the health status of survey responders. N B: Tables in sections 5.3, 5.4, 5.5 were re-analysed for males and females to examine any possible sex differences. No major differences to the current interpretation emerged from this reanalysis. Page 34 Refining Continence Measurement Tools

43 5.6 Norm Tables Norm Tables for the revised incontinence scales are presented in Appendix C. 5.7 Summary Based on these analyses examining the correlations between the items, it is suggested that the revised incontinence scales would contain the following items: Revised Urinary Incontinence Scale (RUIS) Items The five preferred urinary incontinence items are: W2 W3 W4 W 7 Do you experience and, if so, how much are you bothered by: urine leakage related to the feeling of urgency (a sudden desire to urinate)? (UDI-6) Do you experience and, if so, how much are you bothered by: urine leakage related to physical activity, coughing or sneezing? (UDI-6) Do you experience and, if so, how much are you bothered by: small amount of urine leakage (drops)? (UDI-6) How often is urine leakage experienced? (ISI) W8 How much urine is lost each time? (ISI) Revised Faecal Incontinence Scale (RFIS) Items The five preferred faecal incontinence items are: X4 Do you leak, have accidents or lose control with a solid stool? (Wexner FCGS) X5 Do you leak, have accidents or lose control with a liquid stool? (Wexner FCGS) X7 Do you leak stool if you don t get to the toilet in time? X9 Does stool leak so that you have to change your underwear? X1 Does bowel or stool leakage cause you to alter your lifestyle? (Wexner FCGS) The performance of these items should now be further examined on actual clinical samples. This would allow for a wider range of responses to the items, as there would be a greater number of people with moderate to severe incontinence. Field testing would also allow validity data to be obtained, as the revised items could be compared to clinical variables (like 24 hour pad tests for urinary incontinence). Refining Continence Measurement Tools Page 35

44 6 Data Analysis: Item Response Theory 6.1 Introduction The report (Hawthorne, In Press) on Measuring Incontinence in Australia, based on the 24 South Australian Health Omnibus Survey (SAHOS), noted some limitations in three incontinence measures, the Incontinence Severity Index (ISI), the Urogenital Distress Inventory (UDI) and the Wexner Continence Grading Scale (Wexner FCGS) and recommended further examination of their psychometric properties. Two approaches toward psychometric examination are possible: classic test theory (CTT) and modern test theory (MTT). CTT is based on examining correlations among items and the use of statistical tests based on normal distributions, although many parametric-based tests (such as analysis of variance, regression and factor analysis) are robust where violations in data distribution occur (Rummel, 197; Tabachnick, 21). MTT does not make these assumptions and instead examines items and scales based on probabalistic models of Guttman scaling. These methods, then, overcome the violation of assumptions behind parametric tests where there are non-normal data distributions (Andrich, 1978; Masters, 1982; Mokken, 1982). Importantly MTT offers three methods of assessing the psychometric properties of items and scales: (a) the ability to examine non-parametric items response scale performance, thus making sure that the response levels within an item are discriminating as expected (item response theory (IRT), threshold analysis), (b) the opportunity to observe the relationship between item response and respondent characteristics to assess whether known groups differ in their interpretation of an item (IRT, differential item functioning (DIF)), and (c) the capacity to assess the impact on a scale of non-normally distributed items through adding or removing an item. In the SAHOS dataset, it was hypothesized that these issues may have affected the measurement of incontinence. The purpose of this study was to investigate these issues in the incontinence items included in the SAHOS for the purpose of reporting whether it was possible to derive improved measures for both urinary and faecal incontinence assessment. 6.2 Methods This analysis draws cases from the 24 SAHOS database. The details are given in Hawthorne (Hawthorne, In Press). Of the 315 cases, 4 refused to complete the urinary incontinence questions, and there were a further 6 cases with some missing data. These cases were all excluded from the analysis leaving 35 cases in the database for the analysis of urinary incontinence. For the faecal incontinence questions, there were 6 cases that had refused to answer any of the questions, and a further 49 with missing data. These 65 cases were excluded from the faecal incontinence data analysis, leaving 2961 available cases. The presence of outliers was ascertained through calculation of the Mahanalobis distance, however none of the identified cases were exerting undue leverage and so all were retained in the database. Item response distributions were analysed by simple frequency counts. Initial item bank unidimensionality was examined through principal component analysis. Partial credit IRT (Masters, 1982) was used to examine the properties of items and to build the measurement scales. During IRT analysis, extreme cases (those reporting no incontinence symptoms at all) were discarded (N=1,471). The appropriateness of sample size across item response distributions was checked against Linacre s criteria (Linacre, 1999) and random samples drawn for the analyses. Page 36 Refining Continence Measurement Tools

45 Data analyses were carried out in SPSS (SPSS, 24) and RUMM22 (Andrich et al., 24). 6.3 Item Response Theory Basics IRT models two different parameters: (a) the probability of endorsing an item response level and (b) the underlying ability of the respondent to endorse an item level. The relationship between these two parameters enables the researcher to judge how well an item is performing. Figure 9 illustrates this relationship. Along the bottom (x-axis) is a person s ability or attribute to respond to the item, such as his / her incontinence severity. This is expressed in logits, which is an approximate z-score o f θ = P where P is the probability of endorsing a response that 1 P accurately reflects the underlying situation. Up the side (y-axis) is the probability of selecting a particular response, based on the available probability thresholds (if there are 4 response levels, then there are 3 available thresholds between levels 1/2 and 2/3 and 3/4). The resulting curve showing the relationship between the two parameters is the item characteristic curve (ICC) which plots the probability of a person selecting a response level. The ICC for the UDI #1 assessing frequent urination is shown as an example (see Figure 9). Figure 9 Item characteristic curve (ICC) for #1, UDI, frequent urination Item characteristic curve (ICC) for #1, UDI, frequent urination When a person shifts from endorsing one response to endorsing another, the probability of the threshold between the two different levels can be calculated. Good fitting IRT models are where there is a graded monotonic relationship between the person ability such that persons with high abilities (e.g. with severe incontinence) endorse high response categories (e.g. frequent urinary urge or leakage). As Figure 9 shows, the UDI #1 measuring frequency of urination has most responses in the lower left-hand quadrant, suggesting it under-discriminates by incontinence severity (the dots are mainly clustered around the item probability threshold of 1., indicating respondents, regardless of incontinence status, tended to select either the first or second response level). Refining Continence Measurement Tools Page 37

46 6.4 Urinary Continence Measurement Descriptive Properties of Urinary Incontinence Items The properties of the items measuring urinary incontinence are given in Table 19. The chief finding across items is the extreme skew: between 72 to 93 of respondents reported no symptoms of incontinence. The Principal Component Analysis (PCA) loadings given in the table suggests that the pooled items formed an initial uni-dimensional scale, thus suggesting the items were amenable to IRT analysis. Table 19 Urinary continence item frequency distribution and PCA factor loading (item pooled) Urinary continence item frequency distribution and PCA factor loading (items pooled) Valid percentage of cases PCA loading UDI Not at Slight Mode-rate Greatly all 1. Frequent urination? Urine leakage related to the feeling of urgency (sudden desire to urinate)? 3. Urine leakage related to physical activity, coughing, or sneezing? 4. Small amounts of urine leakage (drops)? Difficulty emptying your bladder? Pain or discomfort in the lower abdominal or genital area? ISI 7. How often is urine leakage experienced? Never <1 a month Few times a month Few times a week Daily/ nightly How much urine is lost each time? None Drops Small More.87 splashes Statistics: PCA analysis: Eigenvalue: 4.3, variance explained: Analysis of Urinary Continence Measurement Table 2 shows a partial credit IRT analysis of the UDI and ISI. For the six items comprising the UDI, three were misfitting items (#2, 4 and 6), suggesting that these were not actively contributing to the scale (this suggests that the UDI measurement model is primarily characterised by items #1, 3, and 5 which implies that the UDI is mainly concerned with urological symptoms rather than incontinence per se). The Person Separation Index (PSI) suggested the UDI was capable of discriminating between about 2-3 groups. The overall fit of the data to the model was unsatisfactory as shown by the probability of the χ 2 failing to reach >.5. Page 38 Refining Continence Measurement Tools

47 The ISI was also an unsatisfactory model for similar reasons. Item 7 was a misfit, and although the PSI was excellent (suggesting the ISI could discriminate between several groups of respondents), 2 again the probability of the χ failed to reach >.5. Table 2 IRT analysis of the UDI and ISI IRT analysis of the UDI and ISI UDI Location Fit χ 2 DF Prob (a) SE residual DF 1 Frequent urination? <.1 Urine leakage related to the feeling of 2* urgency (sudden desire to urinate)? Urine leakage related to physical 3 activity, coughing, or sneezing? <.1 <. 1 Small amounts of urine leakage 4* (drops)? <.1 5 Difficulty emptying your bladder? <.1 Pain or discomfort in the lower 6* abdominal or genital area? <.1 Summary statistics for the UDI: Mean item fit residual: (SD = 3.58); Mean person fit residual: -.3 (SD =.75), Person Separation Index:.76, χ 2 = , df = 3, p <.1. ISI How often is urine leakage 7.* experienced? <.1 8. How much urine is lost each time? Summary statistics for the ISI: Mean item fit residual: (SD = 2.7); Mean person fit residual: -.39 (SD =.58), Person Separation Index:.94, χ 2 = 74.74, df = 8, p <.1. a = In logits * = Misfitting items, fit residuals > Constructing Improved Measurement To build a potentially improved measurement model, all items from the UDI and ISI were pooled. Table 21 shows the initial IRT probability threshold analysis of the pooled urinary items. The key feature of the table is that the first of the two ISI items, (How often is urinary leakage experienced?), had disordered thresholds, suggesting that response level 3 was not discriminating between cases but was dominated by the responses < once per month and A few times a week. A graphical plot of this item is given in Figure 1. Collapsing Levels 2 and 3 into a single level solved this problem, as shown in Table 21 (see #7A). All further analyses used this recoding. Refining Continence Measurement Tools Page 39

48 Table 21 Initial IRT analysis of all urinary incontinence items probability thresholds (items pooled) Initial IRT analysis of all urinary incontinence items probability thresholds (items pooled) UDI Logits Probability thresholds Location 1/2 2/3 3/4 4/5 1 Frequent urination? Urine leakage related to the feeling of urgency (sudden desire to urinate)? Urine leakage related to physical activity, coughing, or sneezing? Small amounts of urine leakage (drops)? Difficulty emptying your bladder? Pain or discomfort in the lower abdominal or genital 6 area? ISI 7 How often is urine leakage experienced? A How often is urine leakage experienced? How much urine is lost each time? Note: 7A shows the details after recoding #7 by collapsing levels 2 & 3 into a single response level. Figure 1 Probability curves for ISI, #1, How often is there urinary leakage Probablility curves for ISI, #1, How often is there urinary leakage Key: = Never, 1 = < once per month 2 = A few times a month 3 = A few times a week 4 = Every day/night Page 4 Refining Continence Measurement Tools

49 After the recoding of the ISI #1 item, the fit of items to the partial credit model was assessed. The data are given in Table 22 based on all 8 urinary incontinence questions pooled. The mean item and person fit residuals given in the notes to the table suggest there was poor fit between the model and the data. All 8 items were classified as misfitting, suggesting the items did not form a coherent scale. The ISI #1 item, (How often is urinary leakage experienced?), was the item with the largest misfit. Table 22 IRT analysis of items psychometric properties (items pooled) IRT analysis of items psychometric properties (items pooled) UDI Location (a) SE Fit residual DF χ 2 DF Prob 1* Frequent urination? <.1 Urine leakage related to the feeling of 2* urgency (sudden desire to urinate)? <.1 Urine leakage related to physical 3* activity, coughing, or sneezing? <.1 Small amounts of urine leakage 4* (drops)? <.1 5* Difficulty emptying your bladder? <.1 Pain or discomfort in the lower 6* abdominal or genital area? <.1 Summary statistics for the UDI: Mean item fit residual: (SD = 3.58); Mean person fit residual: -.3 (SD =.75), Person Separation Index:.76, χ 2 = , df = 3, p <.1. ISI How often is urine leakage 7A* experienced? <.1 8* How m uch urine is lost each time? <.1 Summary statistics for the ISI: Mean item fit residual: (SD = 2.7); Mean person fit residual: -.39 (SD =.58), Person Separation Index:.94, χ 2 = 74.74, df = 8, p <.1. Summary statistics for all 8 items pooled: Mean item fit residual: -2.5 (SD = 7.21); Mean person fit residual: -.29 (SD =.95), Person Separation Index:.87, χ 2 = 1,261.44, df = 64, p <.1. a = In logits * = Misfitting items, fit residuals >2.5. Table 23 shows uniform DIF analysis for the pooled 8 items. For items 1, 3 and 5 from the UDI there was DIF across all person logit categories of incontinence. This was also the case for both the ISI items. The UDI Item 3 is given in Figure 11 as an example of a uniform DIF item. It shows that females across all levels of incontinence were significantly more likely than males to endorse higher levels of urinary leakage consequent upon physical activity, coughing or sneezing. Table 23 also shows uniform DIF by age group, revealing that it was present in 6 of the 8 items. Refining Continence Measurement Tools Page 41

50 Table 23 IRT analysis of items uniform DIF, by gender and age group (items pooled) IRT analysis of items uniform DIF, by gender and age group (items pooled) Gender (Male, Female) Age group (15-29, 3-59, 6+) MS F DF Prob MS F DF Prob UDI 1 Frequent urination? < <.1 2 Urine leakage related to the feeling of urgency (sudden desire to urinate)? 3 Urine leakage related to physical activity, coughing, or sneezing? 4 Small amounts of urine leakage (drops)? 5 Difficulty emptying your bladder? 6 Pain or discomfort in the lower abdominal or genital area? ISI 7A How often is urine leakage experienced? 8 How much urine is lost each time? Notes: Because of the number of tests, for significance p < < < < < < < < < <.1 Figure 11 DIF for the UDI, #3, Urine leakage related to activity, by gender DIF for the UDI, #3, Urine leakage related to activity, by gender Given the slight superiority of the ISI over the UDI, the two items from the ISI were pooled. To this pool each of the UDI items was added iteratively; the best model with 3 items in it (2 from the ISI and 1 from the UDI) was accepted - where the criteria were the best fit statistics (i.e. those given in the Tables 21, 22 and 23). To this 3-item model, the remaining UDI items were iteratively added Page 42 Refining Continence Measurement Tools

51 using the same procedures and criteria. Again, the best fitting model was accepted, with 4 items. This model was then assessed by iteratively removing from the model each item in turn and re- the model properties against the criteria. This resulted in the removal of the ISI #1 item assessing leaving a model with three items. The properties of this final model are given in Table 24. For convenience the model was labelled the Urinary Continence Assessment (UCA) measure. None of the final 3 items were misfitting, the item locations covered a broad range of logits suggesting that the measure would have applicability to a wide audience, the mean fit residuals were reasonable and the χ 2 value indicated that the data fitted the model well. Table 24 IRT analysis of the UCS measure, derived from the UDI and ISI pooled items IRT analysis of the UCA measure, derived from the UDI and ISI pooled items Location (a) SE Fit residual DF χ 2 DF Prob 1 Urine leakage related to the feeling of urgency (sudden desire to urinate)? <.1 2 Small amounts of urine leakage (drops)? <.1 3 How much urine is lost each time? Summary statistics: Mean item fit residual:.6 (SD = 1.63); Mean person fit residual: -.41 (SD =.97), Person Separation Index:.87, χ 2 = 21.46, df = 64, p =.12. a = In logits * = Misfitting items, fit residuals > Some Validation Tests of the UCA Table 25 shows a head-to-head comparison of the UCA with the UDI and the ISI. This table suggests that the UCA has superior measurement properties than either the UDI or ISI. It also suggests, based on very preliminary tests of discrimination, that it is more sensitive than the ISI but less sensitive than the UDI. Given that the report (Hawthorne, In Press) concluded that the UDI classified too many cases with incontinence, these findings are probably appropriate. It is emphasized that these results are preliminary and that further validation work on the UCA is needed. Refining Continence Measurement Tools Page 43

52 Table 25 Preliminary comparison of the UCA with the UDI and ISI Preliminary comparison of the UCA with the UDI and ISI UCA UDI ISI N. items N. misfitting items 3 1 Adequacy of model fit (P(χ 2 ).5).12 <.1 <.1 Score range score range used Spearman correlation UDI.79 ISI Gender (Male/Female) Kruskall-Wallis χ Age (15-29/3-59/6+) Kruskall-Wallis χ Urinary classification None Mild Moderate Severe/Very severe Preliminary cut-points for the UCA were determined through logical criteria based on the wording of the response levels to the three items. Suggested cut-points are given in Table 26. Table 26 Suggested cutpoints for interpreting the UCA Value Label UCA raw score Possible item response patterns Notes None (,,) All 3 items are endorsed at the top level (level ) 1 Slight/Mild 1 (,,1) Any item is endorsed at level 1 2 (,,2)(,1,1) Any item is endorsed at level 2, or any two items are endorsed at level 1 3 (1,1,1)(,1,2)(,,3) All three items are endorsed at level 1, two items are endorsed at level 1 and 2, or any item is endorsed at level 3 2 Moderate 4 (,1,3)(1,1,2)(,2,2) Any item is endorsed at level 3 and another item at level 1, any item is endorsed at level 2 and the other two items at level 1, or any two items are endorsed at level 2 5 (,2,3)(1,1,3)(1,2,2) Any item is endorsed at level 3 and another item at level 2, any item is endorsed at level 3 and both other items at level 1, or any two items are endorsed at level 2 and the third item at level 1 3 Severe 6 (,3,3)(1,2,3)(2,2,2) Any two items endorsed at level 3, or any item endorsed at level 3, an item at level 2 and an item at level 1, or all three items endorsed at level 2 7 (2,2,3)(1,3,3) An item endorsed at level 3 and two other items endorsed at level 2, or two items items endorsed at level 3 and the other item at level 1 8 (2,3,3) Two items endorsed at level 3 and the third at level 2 9 (3,3,3) All items endorsed at level 3 Page 44 Refining Continence Measurement Tools

53 For example, a case who reported he/she had moderate urine leakage related to urge (Item 1 Level 2), who was slightly bothered by small amounts of urine leakage (Item 2, Level 1) and who reported that when they lost urine he/she lost a few drops (Item 3, Level 1) would have a urinary incontinence state of (2, 1, 2) which would be classified as Moderate. Based on these cut-points, the data distribution in Table 25 was obtained. This suggests that the proportions classified by the UCA were similar to those of the UDI and ISI standard scoring algorithms. Finally, the three items identified in this study do not include direct measures of either stress incontinence or frequency of urination. The UDI item measuring stress incontinence was found to be psychometrically difficult, in that it was the item with the large DIF (Table 23, also see Figure 11) suggesting that males and females reacted to the item quite differently. Similarly, the UDI item measuring frequency of urination also suffered from DIF by both gender and age group (Table 23). The difficulty with the ISI item measuring frequency was that the probability thresholds were disordered (see Figure 1). Even after recoding, models incorporating these three items resulted in item misfit (not shown) suggesting they were potentially redundant items. These items do, however, raise the question of whether the three items in the UCA measure frequency and stress incontinence. To investigate this, the UCA was assessed against each level of each of these three items. The results are given in Table 27. This shows that the UCA highly discriminated by these forms of incontinence suggesting that it assesses these incontinence problems extremely well. Table 27 Discrimination of the UCA by stress incontinence and urination frequency Discrimination of the UCA by stress incontinence and urination frequency UCA scores N Mean sd Stress incontinence UDI, item 3: Urine leakage related to physical Not at all activity, coughing or sneezing Slight Moderate Greatly Statistics: Kruskall-Wallis χ 2: 1,532.38, df = 3, p <.1 Frequency of urination UDI, item 1: How much are you bothered by Not at all frequent urination? Slight Moderate Greatly Statistics: Kruskall-Wallis χ 2: , df = 3, p <.1 Frequency of urine leakage ISI, item 1: How often is urine leakage Never experienced? <1 a month Few times a month Few times a week Daily/Nightly Statistics: Kruskall-Wallis χ 2: 2,418.59, df = 4, p <.1 Refining Continence Measurement Tools Page 45

54 6.5 Faecal Continence Measurement Descriptive Properties of Faecal Incontinence Items The properties of the items measuring urinary incontinence are given in Table 28. The chief finding across items is the extreme skew: between 54 to 98 of respondents reported no symptoms of incontinence. Of the 1 items available for analysis, in 6 the skew was so extreme that the presence of sparse data violated the sample size axioms for IRT (Linacre, 1999). Regarding items from the Wexner, of the 5 items four violated these axioms for the categories Often and Always. These categories were therefore collapsed such that these Wexner items response scales became Never/Rarely/Other (Sometimes, Often and Always). Items 9 and 1 in Table 28 had the extreme categories collapsed for similar reasons. The sole exception to this pattern of extremely skewed sparse data was the faecal incontinence item #7 (How many bowel movements do you usually have?), where 74 of respondents indicated 5-12 movements a week. Understandably, this item loaded poorly on the PCA and was considered for immediate removal. An attempt to deal with the data was made through recoding this item: the first two levels were collapsed, and the last two levels were collapsed. The resulting response set can be seen in Figure 12. Another item considered for immediate removal was the Wexner item 3 (Do you leak, have accidents or lose control with gas (flatus or wind)?). There were two grounds for considering its removal. Flatus is not included in the International Continence Society definition of faecal incontinence which states that faecal incontinence is the involuntary loss of liquid or solid stool that is a social or hygienic problem (Norton et al., 25). This definition is consistent with that of other researchers, such as the Royal College of Physicians (Whitehead et al., 1999; Royal College of Physicians, 1995). This item was also considered for immediate exclusion because of the poor PCA loading reported in Table 28. Figure 12 Faecal continence #7, number of bowel movements, disordered thresholds Faecal continence #7, number of bowel movements, disordered thresholds Key: = <=2 a week 1 = 3-4 a week 2 = 5-12 a week 3 = 13+ a week Page 46 Refining Continence Measurement Tools

55 The other item that was also considered for discarding at this point was the faecal item #6 (How do you describe your usual bowel pattern?). This was on the grounds that the response scale was difficult to interpret. It is possible for a person who experiences, say, chronic constipation to describe his / her bowel movements as being normal for him / her. The PCA loadings given in the table suggests that the pooled items for faecal incontinence may have formed an initial uni-dimensional scale, thus suggesting the items were perhaps amenable to IRT analysis, with the exception of #7. Table 28 Faecal continence item frequency distribution and PCA factor loading (items pooled) Faecal continence item frequency distribution and PCA factor loading (items pooled) Wexner 1. Do you leak, have accidents or lose control with solid stool? Valid percentage of cases PCA loading Never Rarely Someti Often Always mes Do you leak, have accidents or lose control with liquid stool? 3. Do you leak, have accidents or lose control with gas (flatus or wind)? 4 Do you need to wear a pad to protect your underwear from stool? 5. Does bowel or stool leakage cause you to alter your lifestyle? Other faecal incontinence questions 6. How do you describe your usual bowel pattern? Normal Constipated Diarrhoea Both How many bowel movements do you usually have in a week? (a) Never Rarely Sometimes Often Always 8. Do you experience an urgent need to have a bowel movement that makes you rush to a toilet? 9. Do you leak stool if you don t get to a toilet in time? Does stool leak so that you have to change your underwear? Statistics: PCA analysis: Eigenvalue: 3.83 variance explained: Refining Continence Measurement Tools Page 47

56 6.5.2 Analysis of Faecal Continence Measurement Table 29 shows a partial credit IRT analysis of the Wexner. The results suggest that the Wexner, at least in this population sample, does not form a scale in the psychometric sense of the term. All items were misfitting and, clearly, had little in common. The other faecal incontinence items examination is also presented in Table 29. As with the Wexner, these were a collection of items that did not form a scale in any sense of the word. Table 29 IRT analysis of the Wexner and other items IRT analysis of the Wexner and other items Location (a) SE Fit residual DF χ 2 DF Prob Wexner 1* Do you leak, have accidents or lose control with solid stool? <.1 2* Do you leak, have accidents or lose control with liquid stool? <.1 3* Do you leak, have accidents or lose control with gas (flatus or wind)? <.1 4* Do you need to wear a pad to protect your underwear from stool? <.1 5* Does bowel or stool leakage cause you to alter your lifestyle? <.1 Summary statistics: Mean item fit residual: -5. (SD = 1.3); Mean person fit residual: -.56 (SD =.59), Person Separation Index:.67, χ 2 = , df = 2, p <.1. Summary statistics (excluding #3): Mean item fit residual: -.61 (SD = 1.12); Mean person fit residual: -.34 (SD =.92), Person Separation Index:.86, χ 2 = 43.82, df = 16, p =.1. Other faecal items How do you describe your usual bowel 6 pattern? <.1 How many bowel movements do you 7* usually have in a week? (a) <.1 Do you experience an urgent need to have a bowel movement that makes <.1 8* you rush to a toilet? Do you leak stool if you don t get to a 9* toilet in time? <.1 Does stool leak so that you have to 1* change your underwear? <.1 Summary statistics: Mean item fit residual: (SD = 5.2); Mean person fit residual: -.57 (SD =.6), Person Separation Index:.15, χ 2 = , df = 25, p <.1. a = In logits * = Misfitting items, fit residuals > Constructing Improved Measurement The first step in constructing a measurement model was to discard the Wexner flatus item (#3) and re-run the Wexner scale analysis. As shown in Table 3 this immediately transformed the Wexner scale from a collection of items into a psychometric scale with reasonable fitting properties. Two items, however, had disordered thresholds: #4 and #5. These were recoded and ordered thresholds obtained only after dichotomising these items to ensure there were sufficient cases in each category for a stable analysis. The re-analysis of the revised Wexner (excluding #3, flatus, Page 48 Refining Continence Measurement Tools

57 and with #4 and #5 recoded) is shown in Table 3. Although these data suggest an improved measurement model, there were issues relating to the inclusion of item #5, (Does bowel or stool leakage cause you to alter your lifestyle?). This item measures the general consequences of incontinence, not the incontinence itself. As shown, this analysis indicated that in this population the Wexner data did not fit the IRT model. Table 3 IRT analysis of the revised Wexner, excluding #3 and after recoding other items IRT analysis of the revised Wexner, excluding #3 and after recoding other items Response levels (a) Location (b) SE Fit residual DF χ 2 DF Prob 1 Do you leak, have 1. Never accidents or lose 2. Rarely control with solid stool? 3. Sometimes/Often/Always 2 Do you leak, have 1. Never accidents or lose 2. Rarely control with liquid stool? 3. Sometimes/Often/Always 4 Do you need to wear a 1. Never pad to protect your underwear from stool? 2. Rarely/Sometime/Often/Always 5 Does bowel or stool 1. Never leakage cause you to alter your lifestyle? 2. Rarely/Sometime/Often/Always Summary statistics: Mean item fit residual: -.77 (SD = 1.3). Mean person fit residual: -.5 (SD =.84), Person Separation Index:.82, χ 2 = 25.6, df = 12, p =.1 a = after recoding b = In logits At this point all remaining items were pooled and item psychometric properties examined. Item 7 (How many bowel movements do you usually have in a week) was removed because even after recoding as described above, the probability thresholds were completely disordered (see Figure 12). Item 6 (How do you describe your usual bowel pattern?) was dichotomised in order to remove disordered thresholds ( = normal, 1 = constipated, diarrhoea, both). An iterative procedure was followed, whereby each item was removed from the base model until the best model was accepted. The iteration was then repeated backwards until the most parsimonious model was accepted. The results are shown in Table 31. Four items survived the analysis, two were from the Wexner and two were new items. None of the four items had disordered thresholds, none were misfitting, and none had significant DIF when tested by gender and age group (not shown). The PSI indicated the four items were capable of a high level of discrimination. For convenience, the four items were labelled the Faecal Continence Assessment (FCA) scale. Refining Continence Measurement Tools Page 49

58 Table 31 IRT analysis of the Faecal Continence Assessment (FCA) scale IRT analysis of the Faecal Continence Assessment (FCA) 1 Do you leak, have accidents or lose control with solid stool? 2 Do you leak, have accidents or lose control with liquid stool? Do you leak stool if you don t get to a 3 toilet in time? Does stool leak so that you have to 4 change your underwear? Location (a) SE Fit residual DF χ 2 DF Prob Summary statistics: Mean item fit residual:.39 (SD =.45); Mean person fit residual: -33 (SD = 1.25), Person Separation Index:.87, χ 2 = 22.93, df = 2, p =.29. a = In logits * = Misfitting items, fit residuals > Some Validation Tests of the FCA Preliminary cut-points for the FCA were determined through logical criteria based on the wording of the response categories for the four items. Suggested cut-points are shown in Table 32. Table 32 Suggested cutpoints for interpretation of the FCA Value Label FCA raw score Possible item response patterns None (,,,) All items endorsed at level 1 Slight/Mild 1 (,,,1) 2 (,,,2)(,,1,1) 2 Moderate 3 (,1,1,1)(,,,3)(,,1,2) 4 (,,,4)(,,1,3)(,,2,2)(,1,1,2)(1,1,1,1) 5 (,,1,4)(,,2,3)(,1,1,3)(1,1,1,2)(,1,2,2) 3 Severe 6 (,,2,4)(,,3,3)(,1,1,4)(,1,2,3)(1,1,2,2)(1,1,1,3) 7 (,,3,4)(,1,2,4)(1,1,1,4)(,1,3,3)(1,1,2,3)(1,2,2,2)(,2,2,3) 8 (,,4,4)(,1,3,4)(,2,2,4)(1,1,2,4)(1,2,2,3)(1,1,3,3)(2,2,2,2) 9 (,1,4,4)(1,1,3,4)(,2,3,4)(,3,3,3)(1,2,3,3) 1 (,2,4,4)(1,1,4,4)(2,2,3,3)(1,3,3,3) 11 (,3,4,4)(1,2,4,4)(2,3,3,3)(2,2,3,4) 12 (,4,4,4)(1,3,4,4)(2,2,4,4)(3,3,3,3) 13 (1,4,4,4)(2,3,4,4)(3,3,3,4) 14 (2,4,4,4)(3,3,4,4) 15 (3,4,4,4) 16 (4,4,4,4) All items endorsed at level 4 Table 33 shows a head-to-head comparison of the FCA with the formal Wexner and the modified Wexner without the flatus question. There was a modest Spearman correlation between the Wexner and the FCA, and a good correlation with the modified Wexner. With respect to discrimination between known groups, the FCA appears to be less sensitive to gender differences and more sensitive to age differences. Regarding classification of cases, the FCA and the modified Wexner produce remarkably consistent results. Notes Page 5 Refining Continence Measurement Tools

59 Table 33 Preliminary comparison of the FCA with the Wexner Preliminary comparison of the FCA with the Wexner FCA Wexner Wexner-Mod (a) N. items N. misfitting items 5 4 Adequacy of model fit (P(χ 2 ).5).29 <.1 <. 1 Score range score range used Spearman correlation Wexner.4 Wexner-Mod.64.5 Gender (Male/Female) Kruskall-Wallis χ Age (15-29/3-59/6+) Kruskall-Wallis χ Faecal classification None Rarely/Mild Sometimes/Moderate Weekly/Daily/Severe a = Exclu ding the flatus item Finally, Table 34 provides a test of discriminatory power against the only item with known good psychometric properties that was not used in either the Wexner or the FCA. This suggests that the FCA is highly sensitive to faecal urge incontinence. Table 34 Discrimination of the FCA Discrimination of the FCA FCA scores N Mean sd Urge Urgent need to have a bowel movement that Never makes you rush to a toilet Rarely Sometimes Often Always Statistics: Kruska ll-wallis χ 2 : , df = 4, p < Norm Tables Norm Tables for the revised incontinence scales are presented in Appendix C. Refining Continence Measurement Tools Page 51

60 7 Conclusions and Recommendations 7.1 Context This study examined the psychometric performance of a number of measurement scales for urinary and faecal incontinence (the UDI -6, the ISI, the Wexner FCGS) that were included in the 24 SAHOS a community population survey (Harrison Health Research, 24). Earlier studies (Hawthorne, 26; AIHW, 26) had indicated that there was a case for the revision of these scales. The UDI-6 appeared to contain some items that were being endorsed by those without urinary incontinence. The ISI has problems as it is only a two item scale (which raises questions about its measurement stability), and it was noted the Wexner FCGS includes an item on flatus which confounds faecal incontinence prevalence estimates derived from this scale. A more detailed discussion of these issues has been provided in Section 3.2. The data was analysed using both Classical Psychometric Analyses (refer to Section 5) and by Modern Test Theory (MTT, including Item Response Theory, IRT; refer to Section 6). Following classical approaches basic item properties were examined (such as the endorsement / non endorsement rates, the distribution of responses for each item, the ability of each item to differentiate between those with continence and incontinence etc). Such analyses can be helpful in selecting the better performing items for a revised scale. The corrected item-total correlations for each of the scales were examined in turn as this provides information on how much contribution each item makes to the scale total. Items that make little contribution may be candidates for non-inclusion in a revised scale as they may be measuring other dimensions or gaining endorsement from those with non-continence conditions. The internal consistency reliability of the existing measures, and whether it can be improved by the deletion of particular items, was also examined. Factor analyses were undertaken of all urinary incontinence items and all faecal incontinence items respectively. Factor analysis is useful in that it indicates the underlining dimensions that are measured by this pool of items. It identifies the correlation of items with each of these dimensions or factors. For example, if an individual item is loading on an incontinence factor that is different to the other items in the measure, it may be useful to exclude this individual item as it may be measuring a different dimension. Modern Test Theory, using IRT, can perform a more in depth analysis of items in relationship to the underlying condition of respondents in this case incontinence state. It can be used to assess, for example, how well participants select response options and whether items are performing differently with different groups, for example males and females. It can also provide detailed examination of the relationship between items, thus identifying those which should be retained in a scale. Modern Test Theory is theoretically driven, so a model or a range of models are identified (these hypothesize causal relationships between the items) and the models are tested to see whether each model could generate the data that has been found and which model is best to describe the set of data found. These approaches can be used to revise measures and a discussion of these approaches was outlined in Section 6. Page 52 Refining Continence Measurement Tools

61 7.2 Conclusions concerning urinary incontinence measures Classical Test Theory Analysis As would be expected in a community survey, as contrasted with a clinical sample, the endorsement rates indicate a large proportion of the sample endorse responses such as not at all/never/none for most of the continence items. The UDI-6 has a score range of 6-24 and the ISI has a score range between -12.The means and standard deviation rates for the UDI for males and females are presented in Table 9 and Table 1 and the endorsement rates and the distribution of responses for each of these items are found in Figure 1 and Figure 2 and Appendix 2. These indicate that the presence of urinary incontinence is more common in females and in older adults. The internal consistency reliability of the UDI-6 was r =.78 and for the ISI was r =. 83. These are considered of an appropriate magnitude (Streiner and Norman, 23). The corrected item-total correlations for the UDI-6 are reported in Table 11. It can be seen that the item - total correlations for UDI Items 5 and 6 (difficulty emptying bladder and pain in lower abdominal or genital region) were lower than for other items (urinary leakage due to urge (UDI-Item 2), urinary leakage due to stress (UDI-Item 3), and small amounts of urine leakage (UDI-Item 4)). Given these findings for UDI items 5 and 6 it is suggested they are not included in the revised scale. A factor analysis of the urinary incontinence items indicated there was a 2 factor solution which explained 67 variance (an acceptable result according to Norman and Streiner, 23). The first factor may best be described as a general urinary incontinence factor whereas the second factor contained higher loadings from UDI-6 items such as frequency of urination (UDI-Item 1), pain or discomfort in the lower abdominal or genital region (UDI-Item 6) and difficulty with emptying your bladder (UDI-Item 5). This factor could be considered to be other urological symptoms. The loadings of these items were lower than for all other items on the general urinary incontinence factor (refer to Table 17). For these reasons it is thought a better instrument may result from combining the ISI items (frequency and amount of urinary leakage) with the three items from the UDI-6 that have higher item-total correlations and load more highly on the general urinary incontinence factor (refer below and to Table 11 and Table 16). The proposed refined measures derived from Classical Test Theory Analyses will be referred to as the Revised Urinary Incontinence Scale (RUIS) and will contain the following items: 1. Urine leakage related to the feeling of urgency 2. Urine leakage related to physical activity coughing or sneezing 3. Small amounts of urine leakage (drops) 4. How often do you experience urine leakage? 5. How much urine do you lose each time? The Revised Urinary Incontinence Scale (RUIS) can be viewed in Appendix 1. It is important to note that the refined scale, in conjunction with the original scales, needs to be assessed in clinical field trials as part of its validation. Refining Continence Measurement Tools Page 53

62 7.2.2 Modern Test Theory Analysis Modern Test Theory offers three methods of assessing the psychometric properties of items and scales: (a) ability to examine the items response scale performance, thus making sure that the response levels within an item are discriminating as expected (item response theory (IRT), threshold analysis), (b) the opportunity to observe the relationship between item response and respondent characteristics to assess whether known groups differ in their interpretation of an item (IRT, differential item functioning (DIF)), and (c) the capacity to assess the impact on a scale of non-normally distributed items through adding or removing an item. The analysis of the urinary incontinence items using modern test theory (MTT) approaches derived a shorter 3 item scale for urinary incontinence (items 1, 3, and 5 above) than the Classical Test Theory analyses. Item 2 from the UDI (concerning stress incontinence) did not fit the best model due to differential item functioning as it was found that males and females interpret this item quite differently. The response categories for Item 1 from the ISI were also found to be problematic as a couple of response levels dominated and one response option was rarely chosen. The Urinary Continence Assessment (UCA) scale is included below: 1. Urine leakage related to the feeling of urgency? 2. Small amounts of urine leakage (drops)? 3. How much urine do you lose each time? The Urinary Continence Assessment Scale (UCA) can be viewed in Appendix 1. It is important to note that the refined scale, in conjunction with the original scales, needs to be assessed in clinical field trials as part of its validation Relationships with Other Health Variables As was the case with the Hawthorne (26) study it was found that there were significant correlations between the urinary and faecal incontinence scales (UDI-6, ISI, Wexner) and relevant domains within the measures of health status / health related quality of life that were included in the survey (SF-36, EQ-15D, AQOL; refer Table 18). This indicates that urinary incontinence does lessen one s self-rated health status and Hawthorne (26) has also indicated that health status declines with increasing severity of urinary incontinence. The association of the revised scales (RUIS and UCA, RFIS and FCA) with these variables was further explored and they produced very similar correlations in magnitude and direction to those presented in Table Conclusions concerning faecal incontinence measures Sections 5 and 6 provide a discussion of the steps undertaken to derive a refined measure for faecal incontinence. The Wexner FCGS and a number of additional items concerning soiling, bowel pattern, frequency and urge were also included in the 24 SAHOS (Harrison Health Research, 24) given earlier concerns with the Wexner FCGS raised by Thomas et al. (26) Classical Test Theory Analysis As would be expected in a community survey, as contrasted with a clinical sample, the endorsement rates indicate a large proportion of the sample endorse responses such as not at all/never/none for most of the faecal incontinence items. This is even more the case than for the urinary incontinence items. The Wexner FCGS has a score range of -2 and the means and standard deviation rates for the Wexner FCGS for males and females are presented in Table 9 and Table 1 and the endorsement rates and the distribution of responses for each of the faecal incontinence items are found in Figure 3, Figure 4 and Appendix 2. These indicate that the Page 54 Refining Continence Measurement Tools

63 presence of faecal incontinence is not common amongst males or females and that the presence of faecal incontinence is not common across age groups. The internal consistency reliability of the Wexner FCGS was r =.57 which is considered to be in the unacceptable range (Streiner and Norman, 23) and is less than reported elsewhere (Vaizey, 1999). The item total correlations for the Wexner FCGS are reported in Table 12. It can be seen that the item total correlation for the Wexner FCGS flatus item (Do you leak, have accidents or lose control with gas (flatus/wind)?) was lower than for the other Wexner FCGS items (Table 12). This table also indicates that Cronbach s alpha would be improved to a satisfactory level if this item was deleted from the Wexner FCGS. This finding is consistent with earlier concerns raised concerning the inclusion of the flatus item in this scale. Hawthorne (26) and AIHW (26) have suggested this item should be removed from the Wexner FCGS as it confounds prevalence estimates and is outside the current ICS definition of anal incontinence (Norton et al., 22; 25). The factor analysis results discussed above would also suggest a refined faecal incontinence scale should not include this item. The factor analysis of the faecal incontinence items indicated that a 3 factor solution, explained 61 of the variance (an acceptable result according to Norman and Streiner, 23). The items that load highly on the first factor are mainly items concerning soiling / wearing a pad, leakage and the effect of leakage on lifestyle. This factor may be considered to be a general faecal incontinence as all items are concerned with leakage and soiling (refer to Table 17). The items that load highly on the second factor are the flatus leakage item from Wexner FCGS, a question about type of bowel pattern (e.g. Item X1: normal, constipation, diarrhoea, alternating) and a question concerning faecal urge (Item X3: Do you experience an urgent need to have a bowel movement that makes you rush to the toilet?). These items appear to be tapping other bowel symptoms that do not appear to be related to faecal incontinence / leakage per se. The only item that loads on the third factor is frequency of bowel motions (Item X2) and this item has extremely low loadings on the other two factors. It appears to be unrelated to faecal leakage or soiling. Although clinicians (Moore et al., 26) had suggested the inclusion of a faecal urgency item this item only had a low loading on the general faecal incontinence factor and loaded highly on the second other bowel symptoms factor. Its endorsement pattern is quite different from the other faecal incontinence items (refer to Figure 8 and see Appendix 2) with far fewer in the survey endorsing never in relation to this item. This might suggest that bowel urgency is a relatively more common symptom in the community and thus may also be endorsed by those not experiencing faecal incontinence. For these reasons this faecal urgency item is not considered for inclusion in the refined scale. However, the item Do you leak stool if you don t get to the toilet in time? loads highly on the general faecal incontinence factor and does contain an urge component. The item concerning type of bowel pattern loads most highly on the second factor (other bowel symptoms) and has a low loading on the general faecal incontinence factor. Similarly, the question concerning frequency of bowel movements loads poorly on the general factor and thus these items were not included in the revised scale. An issue concerning item redundancy might be considered by comparing the following items: Do you need to wear a pad to protect your underwear from stool? (Wexner FCGS) Does stool leak so that you have to change your underwear? The pad question from the Wexner FCGS has previously been criticized by Vaizey (1999) as it may relate more to patient fastidiousness rather than faecal incontinence / soiling per se. Given these considerations and its similarity to the soiling item above, in loading on the general faecal Refining Continence Measurement Tools Page 55

64 incontinence factor (.78 vs..71), it is suggested this item be excluded from the Revised Faecal Incontinence Scale. The proposed scale that results from Classical Test Theory Analyses is the Revised Faecal Incontinence Scale (RFIS) and it contains the following items: 1. Do you leak, have accidents or lose control with solid stool? (Wexner FCGS) 2. Do you leak, have accidents or lose control with liquid stool? (Wexner FCGS) 3. Do you leak stool if you don t get to the toilet in time? 4. Does stool leak so that you have to change your underwear? 5. Does bowel or stool leakage cause you to alter your lifestyle? (Wexner FCGS) However, it is noted that for Item 5 of the Wexner FCGS (and item 5 above) that although it loads highly on the general faecal incontinence factor there may be logical grounds for considering its exclusion as it is measuring a consequence of incontinence rather than a symptom of incontinence per se. It is unconventional to include both the symptom and its consequence in the one scale as this in effect can represent a form of double counting (Thomas et al., 26). However, it is thought that clinicians may prefer the retention of the lifestyle question and this issue can be further addressed in clinical trials. There were similarities between the CTT and MTT analyses (refer to the section below) concerning a number of items that should be excluded from a faecal incontinence scale. These included the items of flatus leakage (Wexner) and items concerning the number of bowel movements and bowel pattern which loaded poorly of the general incontinence factor reported here and were also found to be problematic in the IRT analyses. It is suggested that the response categories for the Wexner FCGS are retained. The Revised Faecal Incontinence Scale (RFIS) can be viewed in Appendix 1. It is important to note that the refined scale, in conjunction with the original scales, needs to be assessed in clinical field trials as part of its validation Modern Test Theory Analysis Analyses using Modern Test Theory approaches (IRT) came to a 4 item solution for faecal incontinence which contains items 1-4 above. IRT is used to find the model with the best fit to the data within the minimum number of items and it is a process commonly used to shorten scales. Item 5 from the RFIS above is not included in this solution as the other two items concerning soiling fit the model better as can be seen in Table 29. The Faecal Continence Assessment (FCA) scale resulting from this analysis is provided below. 1. Do you leak, have accidents or lose control with solid stool? (Wexner) 2. Do you leak, have accidents or lose control with liquid stool? (Wexner) 3. Do you leak stool if you don t get to the toilet in time? 4. Does stool leak so that you have to change your underwear? An IRT analysis of a modified Wexner Scale (where the flatus item is excluded) also indicated this scale performed well. Table 33 shows a head-to-head comparison of the FCA with the formal Wexner and the modified Wexner without the flatus question. There was a modest Spearman Page 56 Refining Continence Measurement Tools

65 correlation between the Wexner and the FCA, and a good correlation with the modified Wexner. With respect to discrimination between known groups, the FCA appears to be less sensitive to gender differences and more sensitive to age differences. Regarding classification of cases, the FCA and the modified Wexner produce remarkably consistent results. It is suggested that the response categories for the Wexner FCGS are retained. The Faecal Co ntinence Assessment scale (FCA) can be viewed in Appendix 1. It is important to note that the revised scale, in conjunction with the original scales, needs to be assessed in clinical field trials as part of its validation. 7.4 Recommendations From the analysis of the urinary and faecal incontinences items and scales included in the 24 SAHOS this study has developed two revised scales for urinary and faecal incontinence respectively (RUIS and UCA, RFIS and FCA). However, one of the limitations of using community survey data is that as the data is collected in face to face interviews the data are at the level of subjective reports of incontinence symptoms rather than confirmed diagnoses. A community survey will also exclude those currently placed in institutional settings (e.g. nursing homes) and as such the prevalence estimates for the elderly are likely to be underestimated (Hawthorne, 26). These considerations mean that in a community survey there will be a limited range of responses to incontinence items particularly those pertaining to more severe levels of symptoms. It will thus be necessary to trial the refined measures in a range of clinical settings in follow up field trials. The following recommendations are made: It is suggested that the incontinence items included in the 24 SAHOS are administered in a clinical field trial to assess whether the findings from these analyses are replicated amongst clinical samples. This would also enable the datasets to be merged for these items which would permit a more comprehensive analysis and allow for a more definitive conclusion concerning the revised measures. This will also assist in indicating whether the refined measures have superior psychometric properties as would be anticipated from this study. Given some issues raised by the IRT analysis concerning item 1 from the ISI (the frequency of urinary leakage item) and the UDI item 2 (concerning stress incontinence) it is thought that some additional items concerning stress incontinence and the frequency of urinary leakage could be included in the clinical dataset. It is thought that to gain acceptability by clinicians, items covering these domains may need to be included. The three item UCA, however, may be a better instrument to use in future prevalence studies than the ISI. Additional questions which have been developed recently could also be included in the proposed clinical field trials. These include patient-rated global assessments of treatment benefit, satisfaction and willingness to continue treatment (Pleil et al., 25) and Patient Global Impression of improvement and severity for incontinence (Yalcin and Bump, 23). A current study by Hawthorne, Sansoni, Hayes and Marosszeky on patient satisfaction measures and incontinence may also include recommendations concerning the inclusion of patient satisfaction items for these trials. It would then be desirable if field trials could be conducted using the recommended measures across a broad range of field settings specialist continence clinics, general practice and community care settings and in residential care settings. It is also noted that as the 24 SAHOS was a community survey there is limited Australian prevalence data available for continence conditions for those in aged care residential settings and thus it would be particularly useful to pilot test the revised measures in these settings. Refining Continence Measurement Tools Page 57

66 8 References AIHW (26) Australian incontinence data analysis and development. AIHW, Canberra. Andrich D (1978) A rating formulation for ordered response categories. Psychometrika. Vol. 43, No. 1, pp.561. Andrich D, Sheridan B, et al. (24) RUMM22. RUMM Laboratory Pty Ltd, Perth. Australian Bureau of Statistics (1997) 1995 National Health Survey: SF-36 Population Norms Australia (4399.). Australian Bureau of Statistics, Canberra. Australian Bureau of Statistics (26) National Health Survey: Summary of Results Australia (4364.). Australian Bureau of Statistics, Canberra. Boreham MK, Richter HE, Kenton KS, Nager CW, Gregory WT, Aronson MP, et al. (25) Anal incontinence in women presenting for gynecologic care: Prevalence, risk factors, and impact upon quality of life. American Journal of Obstetrics and Gynecology, Vol. 192, pp Brown JS, Posner SF and Stewart AL (1999) Urge incontinence: New health-related quality of life measures. JAGS. Vol. 47, No.8, pp.98. Deutekom M, Terra MP, Dobben AC, et al. (25a) Impact of faecal incontinence severity on health domains. Colorectal Disease. Vol. 7, pp.263. Deutekom M, Terra MP, Dobben AC, et al. (25b) Selecting an outcome measure for evaluating treatment in fecal incontinence. Dis Colon Rectum. Vol. 48, pp Devesa JM, Rey A, Hervas PL, et al. (22) Artificial anal sphincter: Complications and functional re sults of a large personal series. Dis Colon Rectum. Vol. 45, pp EuroQol Group (199) EuroQol: a new facility for measurement of health-related quality of life. Health Policy. Vol. 16, pp.199. Hagen S, Hanley J and Capewell A (22) Test-retest reliability, validity and sensitivity to change of the urogenital distress inventory and the incontinence impact questionnaire. Neurourology and Urodynamics. Vol. 21, pp.534. Hanley J, Capewell A and Hagen S (21) Validity study of the severity index, a simple measure of urinary incontinence in women. British Medical Journal. Vol. 322(7294), pp.196. Hannestad YS, Rortveit G, Sandvik H, Hunskaar S (2) A community-based epidemiological survey of female urinary incontinence: the Norwegian EPINCONT study. Epidemiology of Incontinence in the County of Nord-Trondelag. Journal of Clinical Epidemiology. Vol. 53, No. 11, pp.115. Harrison Health Research (24) South Australian Health Omnibus Survey (SAHOS) (Report). South Australian Department of Health, Adelaide. Har vey MA, Kristjansson B, Griffith D, et al. (21) The Incontinence Impact Questionnaire and the Urogenital Distress Inventory: A revisit of their validity in women without a urodynamic diagnosis. Am J Obstet Gynecol. Vol. 185, pp.25. Hawthorne G (26) Measuring incontinence in Australia. Department of Health and Ageing, Canberra. Page 58 Refining Continence Measurement Tools

67 Hawthorne G and Griffith P (2) The Friendship Scale: Development and properties. (Working Paper 114) National Centre for Health Program Evaluation, Melbourne. Hawthorne G, Richardson J, Osborne R (1999) The Assessment of Quality of Life (AQoL) instrument: a psychometric measure of health related quality of life. Quality of Life Research. Vol. 8, pp.29. Hawthorne G, Sansoni J, Hayes L and Marosszeky N (26) Patient Satisfaction and Incontinence (Draft Report) (in press) Holtedahl K, Verelst M, and Schiefloe A (1998) A population based, randomized control trial of conservative treatment for urinary incontinence in women. Act a Obstet Gynecol Scand. Vol. 77, pp.671. Jorge J and Wexner S (1993) Etiology and management of faecal incontinence. Dis Colon Rectum. Vol. 36, No. 1, pp.77. Kairaluoma MV, Raivio P, Aarnio MT, et al. (24) Immediate repair of obstetric anal sphincter rupture: Medium-term outcome of the overlap technique. Dis Colon Rectum. Vol. 47, pp Klovning A, Avery K, Sandvik H, et al. (24) A web-based comparison of two questionnaires for assessing the severity of urinary incontinence: The ICIQ-II SF versus the Incontinence Severity Index. (ICS Conference Paper) Lam TCF, Kennedy ML, Chen FC, et al. (1999) Prevalence of faecal incontinence: Obstetric and constipation-related risk factors: a population-based study. Colorectal Disease. Vol. 1, pp.197. L andis JR, Koch GG (1977) The measurement of observer agreement for categorical data. Biometrics. Vol. 33, pp.159. Lemack GE and Zimmern PE (1999) Predictability of urodynamic findings based on the Urogenital Distress Inventory - 6 Questionnaire. Urology. Vol. 54, pp.461. L inacre JM (1999) Investigating rating scale category utility. Journal of Outcome Measurement. Vol. 3, No. 2, pp13. Lukacz ES, Lawrence JM, Burchette RJ, et al. (24) The use of Visual Analog Scale in urogynecologic research: A psychometric evaluation. Am J Obstet Gynecol. Vol. 191, pp.165. Macmillian AK, Merrie AE, Marshall RJ, et al. (24) The prevalence of fecal incontinence in community-dwelling adults: A systematic review of the literature. Dis Colon Rectum. Vol. 47, pp Masters G (1982) A Rasch model for partial credit scoring. Psychometrika. Vol. 47, pp.149. M elville JL, Fan MY, Newton K, et al. (25a) Fecal incontinence in US women: A populationbased study. Am J Obstet Gynecol. Vol. 193, pp.271. Melville JL, Katon W, Delaney K, et al. (25b) Urinary incontinence in US women. Arch Intern Med. Vol. 165, pp.537. Melville JL, Miller EA, Fialkow MF, et al. (23) Relationship between patient report and physician assessment of urinary incontinence severity. Am J Obstet Gynecol. Vol. 189, pp.76. Refining Continence Measurement Tools Page 59

68 Mokken RJ (1982) A non-parametric approach to the analysis of dichotomous item responses. Applied Psychological Measurement. Vol. 6, No. 4, pp.417. Moore K, Ho MT, Lapsley H, Brown I, et al. (26) Development of a framework for economic and cost evaluation of continence conditions (Final Report). Department of Health and Ageing, Canberra. Murphy M, Culligan P, Arce C, et al. (24) Construct validity of the incontinence severity index. 24. (ICS Conference Paper) Nazir M, Carlsen E, Jacobsen AF, et al. (22) Is there any correlation between objective anal testing, rupture grade, and bowel symptoms after primary repair of obstetric anal sphincter rupture? An observational cohort study. Dis Colon Rectum. Vol. 45, pp Norman GR and Streiner DL (23) PDQ Statistics. B.C. Decker, Toronto. Norton C, Christiansen J, Butler U, Harari D, Nelson RL, Pemberton J, et al. (22) Anal Incontinence: Report of Committee 15. In Abrams P, Cardozo L, Khoury S, Wein A (eds) (22) Incontinence. 2nd International Consultation on Incontinence. (2nd Edition) Health Publications Limited, United Kingdom. Norton C, Whitehead WE, Bliss DZ, Metsola P, Tries J (25) Conservative treatment and pharmacological management of faecal incontinence in adults. (pp. 1521) In Abrams A, Cardozo L, Khoury S, Wein A (eds) (25) Incontinence: Volume 2 - Management. International Continence Society, Paris. O'Keefe EA, Talley NJ, Tangalos EG, et al. (1992) A Bowel Symptom Questionnaire for the E lderly. Journal of Gerontology. Vol. 47, No.4, pp.m116. Oliveira L, Pfeifer J and Wexner SD (1996) Physiological and clinical outcome of anterior s phincteroplasty. British Journal of Surgery. Vol. 83, pp.52. Perry S, Shaw C, McGrother C, et al. (22) Prevalence of faecal incontinence in adults aged 4 y ears or more living in the community. Gut. Vol. 5, pp.48. Pescatori M, Anastasio G, Bottini C and Mentasti A (1992) New grading and scoring for anal in continence. Evaluation of 335 patients. Dis Colon Rectum. Vol. 35, No. 5, pp.482. Pleil AM, Coyne KS, Reese PR, et al. (25) The validation of patient-rated global assessments of treatment benefit, satisfaction, and willingness to continue - The BSW. Value in Health. Vol. 8, Suppl 1, pp.s25. Rodriguez LV and Raz S (23) Prospective analysis of patients treated with a distal urethral polypropylene sling for symptoms of stress urinary incontinence: Surgical outcome and satisfaction determined by patient driven questionnaires. Journal of Urology. Vol. 17, pp.857. Rothbarth J, Bemelman WA, Meijerink WJHJ, et al. (21) What is the impact of fecal incontinence on quality of life? Dis Colon Rectum. Vol. 44, pp.67. Royal College of Physicians (1995) Incontinence: Causes, management and provision of services. A Working Party of the Royal College of Physicians. Journal of the Royal College of Physicians of London Vol. 29, No. 4, pp.272. Rummel R (197) Applied factor analysis. Northwestern University Press, Evanston. Page 6 Refining Continence Measurement Tools

69 Sandvik H, Hunskaar S, Seim A, Hermstad R, Vanvik A and Bratt H (1993) Validation of a severity index in female urinary incontinence and its implementation in an epidemiological survey. Journal of Epidemiology and Community Health. Vol. 47, pp.497. Sandvik H, Hunskaar S, Vanvik A, Bratt H, Seim A and Hermstad R (1995) Diagnostic classification of female urinary incontinence: an epidemiological survey corrected for validity. J ournal of Clinical Epidemiology. Vol. 48, pp.339. Sandvik H, Seim A, Vanvik A and Hunskaar S (2) A severity index for epidemiological surveys of female urninary incontinence: Comparison with 48-hour Pad-Weighing Tests. Neurourology and Urodynamics. Vol. 19, pp.137. Seim A, Siversten B, Eriksen BC and Hunkskar S (1996) Treatment of urinary incontinence in women in general practice: observational study. British Medical Journal. Vol. 312, pp Shumaker S, Wyman J, Uebersax J, McClish D and Fantl J (1994) Health-related quality of life measures for women with urinary incontinence: the Incontinence Impact Questionnaire and the Urogenital Distress Inventory. Continence Program in Women (CPW) Research Group. Quality of Life Research. Vol. 3, No. 5, pp.291. Sintonen H (1994) The 15D measure of health-related quality of life: Reliability, validity and sensitivity of its health state descriptive system. (Working Paper 41) National Centre for Health Program Evaluation, Melbourne. Sintonen H (1995) The 15D measure of health-related quality of life: Feasibility, reliability and validity of its valuation system. (Working Paper 42) National Centre for Health Program Evaluation, Melbourne. Sintonen H (21) The 15D instrument of health-related quality of life: Properties and applications. Annals of Medicine. Vol. 33, pp.328. Sintonen H and Pekurinen M (1993) A fifteen-dimensional measure of health-related quality of life (15D) and its applications. In Walker S, Rosser R (eds) (1993) Quality of Life Assessment. Kluwer Academic Publishers, Dordrecht. SPSS (24) SPSS for Windows, Version 13.. SPSS Inc, Chicago. Streiner DL and Norman GR (23) Health Measurement Scales: A practical guide to their development and use (3rd Ed.) Oxford Medical Publications, Oxford. Svatek R, Roche V, Thornberg J, et al. (25) Normative values for the American Urological Association Symptom Index (AUA-7) and Short Form Urogenital Distress Inventory (UDI-6) in patients 65 and older presenting for non-urological care. Neurourology and Urodynamics. Vol. 24, pp.66. Tabachnick BG and Fidell LS (21) Using Multivariate Statistics. (4th Ed) Harper Collins, New York. Thomas S, Nay R, Moore K, Fonda D, Hawthorne G, Marosszeky N and Sansoni J (26) Continence Outcomes Measurement Suite Project (Final Report). Department of Health and Ageing, Canberra. Uebersax J, Wyman J, Shumaker S, McClish D and Fantl J (1995) Short forms to assess life quality and symptom distress for urinary incontinence in women: the Incontinence Impact Questionnaire and the Urogenital Distress Inventory. Continence Program for Women Research Group. Neurourology and Urodynamics.Vol.14, No. 2, pp Refining Continence Measurement Tools Page 61

70 Vaizey C, Carapeti E, Cahill J, and Kamm M (1999) Prospective comparison of faecal incontinence grading systems. Gut. Vol. 44, No. 1, pp.77. Ware J and Sherbourne C (1992) The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection. Medical Care. Vol. 3, No. 6, pp.473. Whitehead WE, Wald A, et al. (1999) Functional disorders of the anus and rectum. Gut 45, Suppl 2, pp.ii55. Yalcin I and Bump RC (23) Validation of two global impression questionnaires for incontinence. A m J Obstet Gynecol. Vol. 189, pp.98. Page 62 Refining Continence Measurement Tools

71 Appendix 1: The Instruments Used and the Revised Instruments Instruments included in 24 SAHOS Urogenital Distress Inventory - 6 The UDI-6 items ask the respondent Do you experience and if so how much are you bothered by the following list of symptoms: Frequent urination Urine leakage related to the feeling of urgency Urine leakage related to physical activity, coughing or sneezing Small amounts of urine leakage (drops) Difficulty emptying your bladder Pain or discomfort in the lower abdominal or genital area The response scale is Not at all Slightly Moderately Greatly Incontinence Severity Index How often do you experience urine leakage? Never = Less than once a month = 1 A few times a month = 2 A few times a week = 3 Every day and/or night = 4 How much urine do you lose each time? None = Drops = 1 Small splashes = 2 More = 3 Severity index = (points for frequency) x (points for amount) Refining Continence Measurement Tools Page 63

72 Wexner Faecal Continence Grading Scale Frequency Type of Never Rarely Sometimes incontinence < 1/month < 1/week 1/month Usually < 1/day 1/week Solid Liquid Gas Requires pad Lifestyle Always 1/day Q1. In the past four weeks: do you leak, have accidents or lose control with solid stool? Q2. In the past four weeks: do you leak, have accidents or lose control with liquid stool? Q3. In the past four weeks: do you leak, have accidents or lose control with gas (flatus or wind)? Q4. In the past four weeks: do you need to wear a pad to protect your underwear from stool? Q5. In the past four weeks: does bowel or stool leakage cause you to alter your lifestyle? Other Faecal Items (included in 24 SAHOS) X1 In the past four weeks: how do you describe your usual bowel pattern? Normal Constipated Diarrhoea Alternating constipation and diarrhoea X2 In the past four weeks: how many bowel movements do you usually have in a week? 1 or less or more X3 In the past four weeks: do you experience an urgent need to have a bowel movement that makes you rush to a toilet? Never Rarely, i.e. less than once in the past four weeks Sometimes, i.e. less than once a week, but once or more in the past four weeks Often or usually, i.e. less than once a day but once a week or more Always, i.e. once or more per day or whenever you have a bowel movement X7 In the past four weeks: do you leak stool if you don t get to a toilet in time? Never Rarely, i.e. less than once in the past four weeks Sometimes, i.e. less than once a week, but once or more in the past four weeks Often or usually, i.e. less than once a day but once a week or more Always, i.e. once or more per day or whenever you have a bowel movement Page 64 Refining Continence Measurement Tools

73 X9 In the past four weeks: does stool leak so that you have to change your underwear? Never Rarely, i.e. le ss than once in the past four weeks Sometimes, i.e. less than once a week, but more than once in the pa st four weeks Often or usua lly, i.e. less than once a day but more than once a week Always, i.e. more than once a day or whenever you ha ve a bowel movement Refining Continence Measurement Tools Page 65

74 Revised Measures for Urinary and Faecal Incontinence Revised Urinary Incontinence Scale Do you experience and if so how much are you bothered by: Urine leakage related to the feeling of urgency Urine leakage related to physical activity, coughing or sneezing Small amounts of urine leakage (drops) The response scale for each item is Not at all Slightly Moderately Greatly How often do you experience urine leakage? Never = Less than once a month = 1 A few times a month = 2 A few times a week = 3 Every day and/or night = 4 How much urine do you lose each time? None = Drops = 1 Small splashes = 2 More = 3 Urinary Continence Assessment Urine leakage related to the feeling of urgency Small amounts of urine leakage (drops) The response scale for each item is Not at all Slightly Moderately Greatly How much urine do you lose each time? None = Drops = 1 Small splashes = 2 More = 3 Page 66 Refining Continence Measurement Tools

75 Revised Faecal Incontinence Scale Do you leak, have accidents or lose control with solid stool? Do you leak, have accidents or lose control with liquid stool? Do you leak stool if you don t get to the toilet in time? Does stool leak so that you have to change your underwear? Does bowel or stool leakage cause you to alter your lifestyle? The response scale for each item is Never Rarely, i.e. less than once in the past four weeks Sometimes, i. e. less than once a week, but once or more in the past four weeks Often or usually, i.e. less than once a day but once a week or more Always, i.e. once or more per day or whenever you have a bowel movement Faecal Continence Assessment Do you leak, have accidents or lose control with solid stool? Do you leak, have accidents or lose control with liquid stool? Do you leak stool if you don t get to the toilet in time? Does stool leak so that you have to change your underwear? The response scale for each item is Never Rarely, i.e. less than once in the past four weeks Sometimes, i.e. less than once a week, but once or more in the past four weeks Often or usually, i.e. less than once a day but once a week or more Always, i.e. once or more per day or whenever you have a bowel movement Refining Continence Measurement Tools Page 67

76 Appendix 2: Supplementary Figures for Section 5 Additional Figures Demonstrating the Level of Non-endorsement for each Incontinence Item The following figures demonstrate the level of non-endorsement (i.e. the inverse of endorsement) for each of the urinary and faecal incontinence items, according to age group and sex. Figure 13 Percentage of respondents who answered Not at all for item W1 (UDI-6, Item 1) W.1 Do you experience and, if so, how much are you bothered by: frequent urination? (n = 2923) Male Female Age Group Figure 14 Percentage of respondents who answered Not at all for item W2 (UDI-6, Item 2) W.2 Do you experience and, if so, how much are you bothered by: urine leakage related to the feeling of urgency (a sudden desire to urinate)? (n = 2923) Male Female Age Group Page 68 Refining Continence Measurement Tools

77 Figure 15 Percentage of respondents that answered Not at all for item W3 (UDI-6, Item 3) W.3 Do you experience and, if so, how much are you bothered by: urine leakage related to physical activity, coughing or sneezing? (n = 2922) Age Group Male Female Figure 16 Percentage of respondents who answered Not at all for item W4 (UDI-6, Item 4) W.4 Do you experience and, if so, how much are you bothered by: small amount of urine leakage (drops)? (n = 2922) Age Group Male Female Refining Continence Measurement Tools Page 69

78 Figure 17 Percentage of respondents who answered Not at all for item W5 (UDI-6, Item 5) W.5 Do you experience and, if so, how much are you bothered by: difficulty emptying your bladder? (n = 2922) Male Female Age Group Figure 18 Percentage of respondents who answered Not at all for item W6 (UDI-6, Item 6) W.6 Do you experience and, if so, how much are you bothered by: pain or discomfort in the lower abdominal or genital area? (n = 2922) Age Group Male Female Page 7 Refining Continence Measurement Tools

79 Figure 19 Percentage of respondents who answered Never for item W7 (ISI, Item 1) W.7 How often is urine leakage experienced? (n = 2921) Male Female Age Group Figure 2 Percentage of respondents who answered None for item W8 (ISI, Item 2) W.8 How much urine is lost each time? (n = 2922) Male Age Group Female Refining Continence Measurement Tools Page 71

80 Figure 21 Percentage of respondents who answered Normal for item X1 X.1 In the past four weeks: How do you describe your usual bowel pattern? (n = 2921) Male Female Age Group Due to the nature of item X.2 In the past four weeks: How many bowel movements do you have in a week? an analysis of non-endorsement was not appropriate. Figure 22 Percentage of respondents who answered Never for item X X.3 In the past four weeks: Do you experience an urgent need to have a bowel movement that makes you rush to the toilet? (n = 2921) Male Female Age Group Page 72 Refining Continence Measurement Tools

81 Figure 23 Percentage of respondents who answered Never for item X4 (Wexner, Item 1) X.4 In the past four weeks: Do you leak, have accidents or lose control with solid stool? (n = 2921) Age Group Male Female Figure 24 Percentage of respondents who answered Never for item X5 (Wexner, Item 2) X.5 In the past four weeks: Do you leak, have accidents or lose control with liquid stool? (n = 2921) Age Group Male Female Refining Continence Measurement Tools Page 73

82 Figure 25 Percentage of respondents who answered Never for item X6 (Wexner, Item 3) X.6 In the past four weeks: Do you leak, have accidents or lose control with gas (flatus or wind)? (n = 2921) Age Group Male Female Figure 26 Percentage of respondents who answered Never for item X7 X.7 In the past four weeks: Do you leak stool if you don't get to a toilet in time? (n = 2921) Age Group Male Female Page 74 Refining Continence Measurement Tools

83 Figure 27 Percentage of respondents who answered Never for item X8 (Wexner, Item 4) X.8 In the past four weeks: Do you need to wear a pad to protect your underwear from stool? (n = 2921) Age Group Male Female Figure 28 Percentage of respondents who answered Never for item X9 X.9 In the past four weeks: Does stool leak so that you have to change your underwear? (n = 2921) Age Group Male Female Refining Continence Measurement Tools Page 75

84 Figure 29 Percentage of respondents who answered Never for item X1 (Wexner, Item 5) X.1 In the past four weeks: Does bowel or stool leakage cause you to alter your lifestyle? (n = 2921) Male Female Age Group Page 76 Refining Continence Measurement Tools

85 Additional Figures Showing the Distribution of Response for Each Incontinence Item The following additional figures show the proportion of males and females endorsing each response category for each item and each age group. These figures visually display the distribution of responses to each question. Figure 3 Percentage of males that selected each response option in item W1 for each age group W.1 Do you experience and, if so, how much are you bothered by: frequent urination? (n = 126) Not at all Slightly Moderately Greatly Refused Age Group MALES Figure 31 Percentage of females that selected each response option in item W1 for each age group W.1 Do you experience and, if so, how much are you bothered by: frequent urination? (n = 1717) Not at all Slightly Moderately Greatly Refused 2 1 Age Group FEMALES Refining Continence Measurement Tools Page 77

86 Figure 32 Percentage of males that selected each response option in item W2 for each age group W.2 Do you experience and, if so, how much are you bothered by: urine leakage related to the feeling of urgency (a sudden desire to urinate)? (n = 126) Not at all Slightly Moderately Greatly Refused Age Group MALES Figure 33 Percentage of females that selected each response option in item W2 for each age group W.2 Do you experience and, if so, how much are you bothered by: urine leakage related to the feeling of urgency (a sudden desire to urinate)? (n = 1717) Not at all Slightly Moderately Greatly Refused Age Group FEMALES Page 78 Refining Continence Measurement Tools

87 Figure 34 Percentage of males that selected each response option in item W3 for each age group W.3 Do you experience and, if so, how much are you bothered by: urine leakage related to physical activity, coughing or sneezing? (n = 126) Not at all Slightly Moderately Greatly Refused Age Group MALES Figure 35 Percentage of females that selected each response option in item W3 for each age group W.3 Do you experience and, if so, how much are you bothered by: urine leakage related to physical activity, coughing or sneezing? (n = 1716) Not at all Slightly Moderately Greatly Refused Age Group FEMALES Refining Continence Measurement Tools Page 79

88 Figure 36 Percentage of males that selected each response option in item W4 for each age group W.4 Do you experience and, if so, how much are you bothered by: small amount of urine leakage (drops)? (n = 126) Not at all Slightly Moderately Greatly Refused Age Group MALES Figure 37 Percentage of females that selected each response option in item W4 for each age group W.4 Do you experience and, if so, how much are you bothered by: small amount of urine leakage (drops)? (n = 1716) Not at all Slightly Moderately Greatly Refused Age Group FEMALES Page 8 Refining Continence Measurement Tools

89 Figure 38 Percentage of males that selected each response option in item W5 for each age group W.5 Do you experience and, if so, how much are you bothered by: difficulty emptying your bladder? (n = 126) Not at all Slightly Moderately Greatly Refused Age Group MALES Figure 39 Percentage of females that selected each response option in item W5 for each age group W.5 Do you experience and, if so, how much are you bothered by: difficulty emptying your bladder? (n = 1716) Not at all Slightly Moderately Greatly Refused Age Group FEMALES Refining Continence Measurement Tools Page 81

90 Figure 4 Percentage of males that selected each response option in item W6 for each age group W.6 Do you experience and, if so, how much are you bothered by: pain or discomfort in the lower abdominal or genital area? (n = 126) Not at all Slightly Moderately Greatly Refused Age Group MALES Figure 41 Percentage of females that selected each response option in item W6 for each age group W.6 Do you experience and, if so, how much are you bothered by: pain or discomfort in the lower abdominal or genital area? (n = 1716) Not at all Slightly Moderately Greatly Refused Age Group FEMALES Page 82 Refining Continence Measurement Tools

91 Figure 42 Percentage of males that selected each response option in item W7 for each age group W.7 How often is urine leakage experienced? (n = 126) Never Less than once a month Several times a month Several times a week Every day or night Refused Age Group MALES Figure 43 Percentage of females that selected each response option in item W7 for each age group W.7 How often is urine leakage experienced? (n = 1715) Never Less than once a month Several times a month Several times a week Every day or night Refused Age Group FEMALES Refining Continence Measurement Tools Page 83

92 Figure 44 Percentage of males that selected each response option in item W8 for each age group W.8 How much urine is lost each time? (n = 126) None Drops Small splashes More Refused Age Group MALES Figure 45 Percentage of females that selected each response option in item W8 for each age group W.8 How much urine is lost each time? (n = 1716) None Drops Small splashes More Refused Age Group FEMALES Page 84 Refining Continence Measurement Tools

93 Figure 46 Percentage of males that selected each response option in item X1 for each age group X.1 In the past four weeks: How do you describe your usual bowel pattern? (n = 125) Normal Constipated Diarrhoea Alternating constipation and diarrhoea Refused Age Group MALES Figure 47 Percentage of females that selected each response option in item X1 for each age group X.1 In the past four weeks: How do you describe your usual bowel pattern? (n = 1716) Normal Constipated Diarrhoea Alternating constipation and diarrhoea Refused Age Group FEMALES Refining Continence Measurement Tools Page 85

94 Figure 48 Percentage of males that selected each response option in item X2 for each age group X.2 In the past four weeks: How many bowel movements do you usually have in a week? (n = 125) or less or more Refused Age Group MALES Figure 49 Percentage of females that selected each response option in item X2 for each age group X.2 In the past four weeks: How many bowel movements do you usually have in a week? (n = 1716) or less or more Refused Age Group FEMALES Page 86 Refining Continence Measurement Tools

95 Figure 5 Percentage of males that selected each response option in item X3 for each age group X.3 In the past four weeks: Do you experience an urgent need to have a bowel movement that makes you rush to a toilet? (n = 125) Never Rarely Sometimes Often or usually Always Refused Age Group MALES Figure 51 Percentage of females that selected each response option in item X3 for each age group 1 X.3 In the past four weeks: Do you experience an urgent need to have a bowel movement that makes you rush to a toilet? (n = 1716) Never Rarely Sometimes Often or usually Always Refused Age Group FEMALES Refining Continence Measurement Tools Page 87

96 Figure 52 Percentage of males that selected each response option in item X4 for each age group X.4 In the past four weeks: Do you leak, have accidents or lose control with solid stool? (n = 125) Never 7 Rarely 6 Sometimes 5 Often or usually Always Refused Age Group MALES Figure 53 Percentage of females that selected each response option in item X4 for each age group X.4 In the past four weeks: Do you leak, have accidents or lose control with solid stool? (n = 1716) Never Rarely Sometimes Often or usually Always Refused Age Group FEMALES Page 88 Refining Continence Measurement Tools

97 Figure 54 Percentage of males that selected each response option in item X5 for each age group X.5 In the past four weeks: Do you leak, have accidents or lose control with liquid stool? (n = 125) Never Rarely Sometimes Often or usually Always Refused Age Group MALES Figure 55 Percentage of females that selected each response option in item X5 for each age group X.5 In the past four weeks: Do you leak, have accidents or lost control with liquid stool? (n = 1716) Never Rarely, ie less than once in the past four weeks Sometimes, ie less than once a week, but more than once in the past four weeks Often or usually, ie less than once a day but more than once a week Always, ie more than once a day or whenever you have a bowel movement Refused Age Group FEMALES Refining Continence Measurement Tools Page 89

98 Figure 56 Percentage of males that selected each response option in item X6 for each age group 1 X.6 In the past four weeks: Do you leak, have accidents or lose control with gas (flatus or wind)? (n = 125) Never Rarely Sometimes Often or usually Always Refused Age Group MALES Figure 57 Percentage of females that selected each response option in item X6 for each age group 1 X.6 In the past four weeks: Do you leak, have accidents or lose control with gas (flatus or wind)? (n = 1716) Never Rarely, Sometimes, Often or usually, Always Refused Age Group FEMALES Page 9 Refining Continence Measurement Tools

99 Figure 58 Percentage of males that selected each response option in item X7 for each age group X.7 In the past four weeks: Do you leak stool if you don't get to a toilet in time? (n = 125) Never Rarely Sometimes Often or usually Always Refused Age Group MALES Figure 59 Percentage of females that selected each response option in item X7 for each age group X.7 In the past four weeks: Do you leak stool if you don t get to a toilet in time? (n = 1716) Never Rarely Sometimes Of ten or usually Always Refused Age Group FEMALES Refining Continence Measurement Tools Page 91

100 Figure 6 Percentage of males that selected each response option in item X8 for each age group X.8 In the past four weeks: Do you need to wear a pad to protect yourself from stool? (n = 125) Never Rarely Sometimes Often or usually Always Refused Age Group MALES Figure 61 Percentage of females that selected each response option in item X8 for each age group X.8 In the past four weeks: Do you need to wear a pad to protect yourself from stool? (n = 1716) Never Rarely Sometimes Often or usually, Always Refused 1 Age Group FEMALES Page 92 Refining Continence Measurement Tools

101 Figure 62 Percentage of males that selected each response option in item X9 for each age group X. 9 In the past four weeks: Does stool leak so that you have to change your underwear? (n = 125) Never Rarely Sometimes Often or usually Always Refused Age Group MALES Figure 63 Percentage of females that selected each response option in item X9 for each age group X. 9 In the past four weeks: Does stool leak so that you have to change your underwear? (n = 1716) Never Rarely Sometimes Often or usually Alw ays Refused Age Group FEMALES Refining Continence Measurement Tools Page 93

102 Figure 64 Percentage of males that selected each response option in item X1 for each age group X.1 In the past four weeks: Does bowel or stool leakage cause you to alter your lifestyle? (n = 125) Never Rarely Sometimes Often or usually Always Refused 1 Age Group MALES Figure 65 Percentage of females that selected each response option in item X1 for each age group X.1 In the past four weeks: Does bowel or stool leakage cause you to alter your lifestyle? (n = 1716) Never Rarely, ie less than once in the past four weeks Sometimes, ie less than once a week, but more than once in the past four weeks Often or usually, ie less than once a day but more than once a week Always, ie more than once a day or whenever you have a bowel movement Refused Age Group FEMALES Page 94 Refining Continence Measurement Tools

103 Appendix 3: Norm Tables The following two norm tables are for the revised incontinence scales - Revised Urinary Incontinence Scale, Urinary Continence Assessment, Revised Faecal Incontinence Scale, Faecal Incontinence Assessment (RUIS, UCA, RFIS, FCA) - according to sex and age group, based on weighted sample data used in the SAHOS Health Omnibus survey. (NB: The response categories for items from the UDI-6 have been realigned to a 3 scoring in order to match scoring for the ISI and Wexner FCGS items). Information is provided on the mean score, standard deviation (SD), standard error of the mean (S.E. mean), minimum and maximum score, median score, and scores at the 25th Percentile, 5th Percentile, 75th Percentile, 9th Percentile and 95th Percentile. The n value for each row is based on the number of subjects obtained from the sample weighting. (The score range for each measure is RUIS: 17; UCA: 1; RFIS: 2; FCA: 16) Refining Continence Measurement Tools Page 95

104 Table 35 Norm tables for males according to age group, using the SAHOS weighted sample Males Age Group Scale Mean SD S.E. Mean Min Max Median 25th 5th 75th 9th 95th Percentile Percentile Percentile Percentile Percentile N RUIS UCA RFIS FCA RUIS UCA RFIS FCA RUIS UCA RFIS FCA RUIS UCA RFIS FCA RUIS UCA RFIS FCA RUIS UCA RFIS FCA RUIS UCA RFIS FCA Page 96 Refining Continence Measurement Tools

105 Table 36 Norm tables for females according to age group, using the SAHOS weighted sample Females Age Group Scale Mean SD S.E. Mean Min Max Median 25th 5th 75th 9th 95th Percentile Percentile Percentile Percentile Percentile N RUIS UCA RFIS FCA RUIS UCA RFIS FCA RUIS UCA RFIS FCA RUIS UCA RFIS FCA RUIS UCA RFIS FCA RUIS UCA RFIS FCA RUIS UCA RFIS FCA Refining Continence Measurement Tools Page 97