Professional Documents
Culture Documents
Wang 2013 Psychometric Properties and Practic
Wang 2013 Psychometric Properties and Practic
Wang 2013 Psychometric Properties and Practic
S-C. Yen, PT, PhD, Department of Methods. Using a 2-parameter logistic item response theory (IRT) procedure and
Physical Therapy, Northeastern
University, Boston, Massachusetts.
the graded response model, the UIQ was assessed for unidimensionality and local
independence, differential item functioning (DIF), discriminating ability, item hier-
J.E. Mioduski, MS, Focus On Ther- archical structure, and test precision.
apeutic Outcomes, Inc, Knoxville,
Tennessee.
†
Results. Four items were dropped to improve unidimensionality and discriminat-
Dr Hart died April 11, 2012. ing ability. Remaining UIQ items met IRT assumptions of unidimensionality and local
[Wang Y-C, Hart DL, Deutscher D, independence. One item was adjusted for DIF by age group. Item difficulties were
et al. Psychometric properties and suitable for patients with PFD with no ceiling or floor effect. Item difficulty param-
practicability of the self-report Uri- eters ranged from ⫺2.20 to 0.39 logits. Endorsed items representing highest difficulty
nary Incontinence Questionnaire
levels were related to control urine flow, impact of leaking urine on life, and
in patients with pelvic-floor dys-
function seeking outpatient reha- confidence to control the urine leakage problem. Item discrimination parameters
bilitation. Phys Ther. 2013;93: ranged from 0.48 to 1.18. Items with higher discriminating abilities were those
1116 –1129.] related to impact on life of leaking urine, confidence to control the urine leakage
© 2013 American Physical Therapy problem, and the number of protective garments for urine leakage.
Association
P
elvic-floor dysfunction (PFD) tutes of Health,11 Food and Drug and collect pilot data for the initial
affects a substantial proportion Administration,12 and World Health item bank. The additional validated
of individuals, mostly wom- Organization6 are encouraging the surveys included the PFDI, PFIQ, Pel-
en.1–3 It is estimated that up to one medical research community to use vic Floor Prolapse/Urinary Inconti-
third of adults experience one or PROs to support intervention effec- nence Sexual Function Question-
more PFD conditions during their tiveness13–15 and monitor patient naire (PISQ), and Pain Disability
lifetime.2,3 To improve functional management.16 Index (PDI).
outcomes and reduce PFD symp-
toms, many patients seek outpatient In 1998, the first International Con- The psychometric properties of the
pelvic-floor physical therapy.4 In a sultation on Incontinence (ICI) was initial FOTO PFD item bank have not
previous longitudinal cohort of held,6 and the ICI Scientific Commit- been studied. The purpose of the
2,452 patients with PFD receiving tee recognized the need to develop a current study was to evaluate psy-
more surgeries related to the condi- UIQ respectively) for those category
tion being treated. Number of The UIQ was designed to evaluate choices and challenges in analyzing
comorbid conditions was assessed urinary function in patients with PFD responses with 2-digit width.
using a list of 29 conditions common seeking outpatient physical therapy
to patients entering an outpatient services. The UIQ consists of 21 Unidimensionality and local inde-
rehabilitation clinic (eg, arthritis, items: 17 related to urinary leakage pendence. To assess IRT assump-
asthma, diabetes, heart attack, AIDS, problems, 2 related to frequency tions of unidimentionality and local
sleep disturbance, cancer).24,25 Exer- problems, and 2 related to retention independence, we conducted
cise history prior to receiving ther- problems. Each item has its own Lik- exploratory factor analyses (EFAs) of
apy was categorized as exercising 3 ert rating scale structure and opera- latent trait variables, followed by
times a week or more, exercising 1 tional definition (Appendix). confirmatory factor analyses (CFAs)
to 2 times a week, or exercising sel- utilizing Mplus (Muthén & Muthén,
Because the minimum covariance ware (University of Washington, from the item pool because of low
coverage was not fulfilled for all Seattle, Washington).41 The difwith- discriminating ability.
items using the original data set due par software examines 3 ordinal
to missing values, for the purposes of logistic regression (OLR) models Item hierarchical structure. Item
assessing unidimensionality and for each item and each demo- difficulty hierarchical order was
local independence of the 21 UIQ graphic category selected for ana- inspected via estimated item diffi-
items, we generated a set of data lysis: sex (female and male), age culty parameters. Item difficulty
where imputed values supplanted group (18 – 44, 45– 64, and ⱖ65 parameters were expressed in logits
missing responses, as described by years), symptom acuity (acute, sub- with higher positive values indicat-
Hart et al.31 To generate the imputed acute, and chronic), and number ing a more challenging task that usu-
values, the original data set, which of PFD comorbid conditions (1⫽ ally is accomplished or endorsed by
contained actual responses and miss- patient reported only one urinary patients with higher functioning.
Missing 14
The remaining 18-item set was rean-
alyzed. All remaining items met the
Payer source (%)
evaluation criteria. The first 3 eigen-
PPO 44
values were 7.81, 1.20, and 1.02,
Medicare part B 16 with the first 3 factors explaining
HMO 5 43%, 7%, and 6% of data variance. Fit
Medicaid 2 statistics for 1-, 2-, and 3-factor mod-
Indemnity insurance 2
els were CFI values of 0.88, 0.94,
and 0.96, respectively, TLI values of
Medicare part A 2
0.97, 0.98, and 0.99, respectively,
Other 25
and RMSEA values of 0.07, 0.05,
Missing 4 and 0.04, respectively, supporting
a
HMO⫽health maintenance organization, PPO⫽preferred provider organization. unidimensionality.
b
Functional comorbidities are medical conditions shown to affect physical functioning.
DIF
After removing items 2, 11, and 18,
the results of DIF analysis using the
18 UIQ items with real data values Item Hierarchical Structure potential items for single-item
were suggestive of no DIF by sex, Item hierarchical structure of the screening purposes. However, the
age group, acuity, and number of final UIQ items is presented in Table TIF curve shifted slightly toward the
PFD comorbid conditions, except 2. The numbers of patients who left (lower ability measures), which
the presence of nonuniform DIF by responded to specific items are listed implied more difficult items were
sex for item 12 (What type of pro- in the “Frequency Count” column. needed to increase test information
tection do you use for your urine Items are ranked based on the item and thus reduce the measurement
leakage?) (P⬍.0001) and uniform difficulty parameter, with more diffi- error at the high-functioning level.
and nonuniform DIF by age group cult items on the top. Item difficulty
for item 15 (To what extent do you parameters ranged from ⫺2.20 to Discussion
feel your sex life has been affected 0.39 (logits). Items representing The purpose of this study was to
by urine leakage?) (P⬍.0001 and more difficult tasks to be endorsed evaluate psychometric properties
Table 2.
Item Characteristics of the Urinary Incontinence Questionnaire (UIQ) Itemsa
Category Parameter
21. Control urine flow 440 0.39 0.09 0.61 0.04 1.57 ⫺0.38 ⫺1.18
after starting to urinate
17. Control urine leak 996 0.19 0.04 0.92 0.03 1.71 1.17 0.74 0.34 ⫺0.36 ⫺0.69 ⫺1.12 ⫺1.79
(0–10 points)
14. Leaking urine 996 0.15 0.03 1.18 0.04 1.49 0.44 ⫺0.31 ⫺1.61
interferes with your
life
9. Urine leak when you 1,004 0.09 0.07 0.48 0.02 2.22 0.84 0.38 ⫺0.89 ⫺2.55
16. Level of confidence 996 0.07 0.04 1.00 0.03 1.61 0.09 ⫺1.70
(4 levels)
1. Urine leak while 1,326 ⫺0.09 0.03 0.95 0.03 2.53 0.19 ⫺0.24 ⫺0.86 ⫺1.62
awake
20. Delay urination after 439 ⫺0.24 0.08 0.65 0.03 2.05 1.11 ⫺0.08 ⫺0.85 ⫺2.23
feeling the urge
12. Type of protection 993 ⫺0.55 0.05 0.79 0.03 1.18 0.36 ⫺1.54
7. Urine leak before you 1,006 ⫺0.57 0.04 0.96 0.03 2.13 0.44 0.03 ⫺0.79 ⫺1.80
can get to the toilet
5. Leak urine after 1,028 ⫺0.78 0.06 0.62 0.02 2.07 0.29 ⫺0.12 ⫺0.72 ⫺1.52
finished urinating
10. Level of activity that 865 ⫺0.86 0.06 0.76 0.04 0.81 ⫺0.09 ⫺0.72
causes urine leakage
15. Sex life affected (age 294 ⫺0.87 0.11 0.67 0.06 0.83 0.02 ⫺0.85
group 1)
13. No. of protective 724 ⫺1.38 0.05 0.97 0.04 1.19 0.51 ⫺0.32 ⫺1.37
garments
15. Sex life affected (age 451 ⫺1.40 0.11 0.60 0.05 0.93 ⫺0.09 ⫺0.84
group 2)
3. Urine leak when 1,024 ⫺1.71 0.06 0.82 0.04 1.39 0.38 ⫺0.12 ⫺0.51 ⫺1.14
asleep
8. How much urine 791 ⫺1.93 0.07 0.58 0.03 1.65 0.61 ⫺2.26
leaks before getting
to the toilet
4. How much urine 349 ⫺1.99 0.12 0.53 0.04 1.51 ⫺1.51
leaks while sleeping
6. How much urine 651 ⫺2.20 0.10 0.52 0.03 1.52 ⫺1.52
leaks after urinating
a
Items were ranked based on the item difficulty parameter, with more difficult items on the top. The first column lists the item number as listed in the
Appendix. Freq Count⫽number of patients who have responded to a specific item, Diff⫽item difficulty parameter, SE⫽standard error, slope⫽item
discrimination parameter. Items 2, 11, 18, and 19 were removed from the analysis, and item 15 was split into 3 items by age group: age group 1 (age 18 –
44 years), age group 2 (age 45– 64 years), and age group 3 (age ⱖ65 years). For item 15, age group 3 was dropped due to low frequency count and
unstable parameter estimations.
erties of urinary incontinence ques- stable and precise estimates of item impact of urinary incontinence on
tionnaires: Handa and Massof18 parameters for patients with PFD in social life (eg, hobbies, ability to do
(N⫽27 women with stress urinary general. Comparing our results with household chores, going on vaca-
incontinence) and Bower et al43 the findings of these 2 studies was tion),18 whereas the UIQ emphasizes
(N⫽156 children with bladder dys- difficult because the questionnaires urinary urgency and frequency, as
function). Compared with these 2 used were related to the quality of well as severity of the urinary
studies,18,43 our larger and more life in children (eg, body image, fam- symptoms.
diverse sample should produce more ily and home, self-esteem)43 or the
sets (with 25%, 50%, and 100% In the process of developing the unadjusted and fully adjusted ability
records supplemented with imputed questionnaire, we administered the estimates was 0.999, similar to the
values), all 3 analyses demonstrated same questionnaire to both male finding by Crane et al,45 suggesting
that one factor was sufficient for ade- and female participants. In the no practical DIF.
quate model fit. For 25%, 50%, and future, as we continue to collect
100% imputed data sets, respec- more data, we intend to develop sex- We used the GRM measurement
tively, there were local dependence specific surveys because urinary and model to perform the initial exami-
relationships among 35 (15%), 21 bowel structures and sex functions nation of the psychometric proper-
(10%), and 3 (1%) item pairs (out of are very different between sexes. To ties of the UIQ items because it is a
210 item pairs), with absolute corre- examine the sex factor, we used a model for polytomous ordinal data39
lation residuals higher than desired method developed by Crane et al36 and it is a 2-parameter model con-
(⬎0.10). Because the data set with for DIF detection by sex. Results sup- taining both item difficulty and dis-
we were not in control of the data in analyzing responses with 2-digit Yen provided consultation (including review
collection procedure, and there was width (ie, item 17 with response of manuscript before submission).
no specific timetable for patients to categories of 1–11), we collapsed The institutional review boards of Focus On
be assessed, as no training was given 2 of the lowest and highest Therapeutic Outcomes, Inc and the Univer-
to therapists prior to the data collec- responses. Although the real impact sity of Wisconsin–Milwaukee approved the
study procedures.
tion. Additionally, generalizability of is unknown, we did monitor the
results may be limited because differ- potential influence on the item cali- This research, in part, was presented at the
ences between participating clinics bration of item 17 by comparing the Combined Sections Meeting of the Ameri-
can Physical Therapy Association; February
and clinics that do not collect data results derived from the 2-parameter 8 –12, 2012; Chicago, Illinois.
using FOTO may exist. GRM using Parscale and the results
derived from the PCM using WIN- DOI: 10.2522/ptj.20120134
Because data were collected in rou- STEPS. The results were similar, with
13 Hart DL, Connolly JB. Pay-for-Performance 23 Swinkels IC, van den Ende CH, de Bakker 35 McHorney CA. Ten recommendations for
for Physical Therapy and Occupational D, et al. Clinical databases in physical ther- advancing patient-centered outcomes
Therapy: Medicare Part B Services. Final apy. Physiother Theory Pract. 2007;23: measurement for older persons. Ann
Report. Grant #18-P-93066/9-01: Health & 153–167. Intern Med. 2003;139:403– 409.
Human Services/Centers for Medicare & 24 Groll DL, To T, Bombardier C, Wright JG. 36 Crane PK, Gibbons LE, Jolley L, van Belle
Medicaid Services. Available at: http:// The development of a comorbidity index G. Differential item functioning analysis
www.cms.gov/Medicare/Billing/Therapy with physical function as the outcome. with ordinal logistic regression techniques
Services/downloads//P4PFinalReport06- J Clin Epidemiol. 2005;58:595– 602. DIFdetect and difwithpar. Med Care.
01-06.pdf. 2006;44(11 suppl 3):S115–S123.
25 Hart DL, Werneke MW, Deutscher D, et al.
14 Resnik L, Liu D, Hart DL, Mor V. Bench- Effect of fear-avoidance beliefs of physical 37 Hart DL, Deutscher D, Crane PK, Wang
marking physical therapy clinic perfor- activities on a model that predicts risk- YC. Differential item functioning was neg-
mance: statistical methods to enhance adjusted functional status outcomes in ligible in an adaptive test of functional sta-
internal validity when using observational patients treated for a lumbar spine dys- tus for patients with knee impairments
data. Phys Ther. 2008;88:1078 –1087. function. J Orthop Sports Phys Ther. 2011; who spoke English or Hebrew. Qual Life
15 Resnik L, Hart DL. Using clinical outcomes 41:336 –345. Res. 2009;18:1067–1083.
to identify expert physical therapists. Phys 26 Muthén LK, Muthén BO. Mplus User’s 38 Nilsagård YE, Forsberg A. Practicability
Ther. 2003;83:990 –1002.
Appendix.
Urinary Incontinence Questionnaire (UIQ)a
2 How
5 How often do you leak urine after you thought you had finished urinating?
1 Never
2 Once or less per week
3 More than once a week
4 Once a day
5 Several times a day
6 Every time
6 How much urine usually leaks after you thought you had finished urinating?
1 A few drops
2 Enough to make underpants/pads wet
3 Enough to wet outer clothes
4 Urine runs down legs onto floor
7 How often does urine leak before you can get to the toilet?
1 Never
2 Once or less per week
3 More than once a week
4 Once a day
5 Several times a day
6 Every time
(Continued)
Appendix.
Continued
8 How much urine usually leaks before you can get to the toilet?
1 A few drops
2 Enough to make underpants/pads wet
3 Enough to wet outer clothes
4 Urine runs down legs onto floor
9 How often does urine leak when you are physically active, including coughing or sneezing?
1 Never
2 Once or less per week
3 More than once a week
Appendix.
Continued
16 Describe your level of confidence in your ability to control your urine leakage problem.
1 Complete confidence
2 Moderate confidence
3 Little confidence
4 No confidence
17 How well do you control your urine leakage? (0 being “no control” to 10 being “full control”)
1 0 (no control)
2 1
a
The Urinary Incontinence Questionnaire may not be used or reproduced without written permission from the authors.