Episodic Disability Questionnaire (EDQ) measurement properties among adults living with HIV in Canada, Ireland, United Kingdom, and United States

Background The Episodic Disability Questionnaire (EDQ) is a generic 35-item patient-reported outcome measure of presence, severity and episodic nature of disability. We assessed the measurement properties of the Episodic Disability Questionnaire (EDQ) with adults living with HIV. Methods We conducted a measurement study with adults living with HIV in eight clinical settings in Canada, Ireland, United Kingdom, and United States. We electronically administered the EDQ followed by three reference measures (World Health Organization Disability Assessment Schedule; Patient Health Questionnaire; Social Support Scale) and a demographic questionnaire. We administered the EDQ only 1 week later. We assessed the internal consistency reliability (Cronbach’s alpha; >0.7 acceptable), and test-retest reliability (Intra Class Correlation Coefficient; >0.7 acceptable). We estimated required change in EDQ domain scores to be 95% certain that a change was not due to measurement error (Minimum Detectable Change (MDC95%)). We evaluated construct validity by assessing 36 primary hypotheses of relationships between EDQ scores and scores on the reference measures (> 75% hypotheses confirmed indicated validity). Results 359 participants completed the questionnaires at time point 1, of which 321 (89%) completed the EDQ approximately 1 week later. Cronbach’s alpha for internal consistency ranged from 0.84 (social domain) to 0.91 (day domain) for the EDQ severity scale, and 0.72 (uncertainty domain) to 0.88 (day domain) for the EDQ presence scale, and 0.87 (physical, cognitive, mental-emotional domains) to 0.89 (uncertainty domain) for the EDQ episodic scale. ICCs for test-retest reliability ranged from 0.79 (physical domain) to 0.88 (day domain) for the EDQ severity scale and from 0.71 (uncertainty domain) to 0.85 (day domain) for the EDQ presence scale. Highest precision was demonstrated in the severity scale for each domain (MDC95% range: 19–25 out of 100), followed by the presence (MDC95% range: 37–54) and episodic scales (MDC95% range:44–76). Twenty-nine of 36 (81%) construct validity hypotheses were confirmed. Conclusions The EDQ possesses internal consistency reliability, construct validity, and test-retest reliability, with limited precision when administered electronically with adults living with HIV across in clinical settings in four countries. Given the measurement properties, the EDQ can be used for group level comparisons for research and program evaluation in adults living with HIV.


Participants
We recruited adults (18 years of age or older) living with HIV from each site using a recruitment poster asking interested individuals to contact the local study investigator. Participants consented to participate in the study by checking 'yes I consent to participate in the study' at the initial information and consent page of the questionnaire administration.

Data Collection
We electronically administered the EDQ followed by three criterion measures (World Health Organization Disability Assessment Schedule (WHO-DAS 2.0), [24,25] Patient Health Questionnaire (8-item) (PHQ-8) [26] and MOS-Social Support Scale) [27] and a demographic questionnaire using the web-based software Qualtrics [28]. Participants completed the questionnaires in-person via a tablet at the clinical site or remotely via a link in an email or Short Message Service (SMS) text. We administered the EDQ only, again 1 week later. At this time, we asked whether participants had a major change in their health status since their last EDQ completion and if yes, to describe the change in their health.

Questionnaires
Episodic Disability Questionnaire: The EDQ is a newly developed patient-reported outcome measure (PROM) re ned from the SF-HDQ [17], comprised of 35-items spanning the six domains: i) physical (10 items); ii) cognitive (3 items); iii) mental-emotional health challenges (5 items), iv) uncertainty or worry about the future (5 items), v) di culties carrying out day-to-day activities (5 items), and vi) challenges to social inclusion (7 items). For each item, individuals are asked to indicate to what extent they are living with a given health-related challenge on that day (severity scale of 0 to 4), and whether that challenge uctuated in the past week (episodic score yes or no). The presence score is derived by dichotomizing severity as present (severity 1-4) or absent (severity of 0). World Health Organization Disability Assessment Schedule: The WHODAS 2.0 is a 36-item selfadministered generic questionnaire of functioning and disability applicable across cultures in adult populations, and directly linked to the International Classi cation of Functioning, Disability, and Health (ICF) [24,25]. The WHODAS 2.0 assesses di culty in performing speci c functions over the previous 30 days across six disability domains: i) cognition, ii) mobility, iii) self-care, iv) getting along, v) life activities and vi) participation). Individuals provide an answer for each question on a 5-point Likert scale (range 0-4) with higher scores indicating increasing di culty completing the task [29]. The WHODAS possesses internal consistency and test-retest reliability and validity and cross-cultural applicability spanning 19 countries [29,30]. The WHODAS is validated in patients with chronic diseases [31] and people living with HIV [32].

Patient Health Questionnaire
The PHQ-8 is an 8-item measure of depression severity. Items are rated using a Likert-type scale from 0-3, with a total score range of 0-24. A score of 10 or greater is considered major depression, 20 or more is severe major depression [26]. The PHQ-8 is reliable and valid for use with people living with HIV [33].

Social Support Survey Questionnaire
The MOS-SSS is a self-administered 20-item questionnaire designed to measure ve different dimensions of social support among patients with chronic illness (emotional/informational support, tangible support, positive social interaction and affectionate support) using 5 response options ranging "none of the time" to "all of the time." [27] Higher scores indicate higher levels of social support. The MOS-SSS possesses construct validity and reliability with people living with HIV [34].

Demographic Questionnaire
The demographic questionnaire included 26 items comprised of demographic (e.g. age, sex, gender, race), HIV (e.g. date of HIV diagnosis, viral load), and health characteristics (e.g. concurrent health conditions, general health status).

Analysis
We calculated median (interquartile ranges (IQR)) EDQ scores. Severity and presence domain scores were calculated using the algorithm developed through Rasch analysis (score range: 0-100) [17]. Episodic scores included a simple sum transformed on a scale of 0-100. Higher scores indicated greater presence, severity and episodic nature of disability. We calculated median WHODAS 2.0 domain scores, PHQ-8 scores and MOS-SSS domain scores as per guidelines. For the demographic questionnaire we calculated descriptive statistics including frequencies (%) for categorical variables and median and interquartile ranges (IQR) for continuous variables.

Test-Retest Reliability
We calculated Intra Class Correlations (ICCs) with 95% CIs using T1 and time 2 (T2) EDQ scores estimated from Shrout and Fleiss' ICC (2,1) (lower bound CI of > 0.7 acceptable) [35]. We calculated ICCs with the entire sample of participants who indicated that they did not have a change in their health status. We then estimated ICCs based on mode of administration either remote (independently via SMS or email link) or in-person (tablet). Our test-retest assessment focused on presence and severity scales of the EDQ as the episodic scale refers to uctuations in disability in the past week, hence we did not expect consistency in this scale. where: is the test-retest reliability; is the level of con dence; and is the standard deviation of the measure at baseline.

Construct Validity
We examined correlations for 36 primary a priori hypotheses theorizing relationships between EDQ and the WHODAS 2.0 criterion measure subscales, and EDQ scores and known groups of participants completing the EDQ on a good versus bad day. We examined an additional 44 exploratory a priori hypotheses theorizing relationships between EDQ and the PHQ-8 and MOS-SSS criterion measure subscales, self-rated general health status, and known groups of participants living with ≥ 2 versus ≤ 1 concurrent health conditions in addition to HIV (80 hypotheses total). Our construct validity assessment focused on presence and severity scales of the EDQ as the criterion measures do not capture the episodic nature of disability. We derived the a priori hypotheses from earlier construct validity assessments of the HDQ and SF-HDQ [12][13][14]. Spearman correlation coe cients of | ≥ 0.30|, | ≥ 0.50| and | ≥ 0.70|, were de ned as 'weak', 'moderate' and 'strong,' respectively [37]. We interpreted the lower and upper bound of con dence intervals when assessing the hypotheses. Construct validity was de ned as > 75% con rmed hypotheses [35,38].

Sample Size
To detect a weak correlation |r = 0.20|, between EDQ and criterion scores, with a power of 0.90 and alpha of 0.05 required 259 participants [41]. To account for questionnaires with missing responses and loss to follow-up at T2, our targeted sample size was 75 adults living with HIV in each of the ve cities for a total of 375 participants [42]. Results Three hundred fty-nine participants completed the questionnaires at T1, of which 321 (89%) completed the EDQ at T2. Most participants (80%) completed the T2 EDQ within two weeks of T1 completion.
Of the 321 participants who completed T2, 46 (15%) 274 participants (85%) reported no change in health status and were included in test-retest reliability assessment. The characteristics of those who did and did not complete the EDQ at T2 were similar (see Additional le 1).

Characteristics of Participants
See Table 1 of characteristics of participants by site. Most participants identi ed as men (83%), median age 51 years, living with a median of 4 concurrent health conditions in addition to HIV. There were differences in participant characteristics (Table 1) and EDQ scores (see Additional le 2) across cities, re ecting the types of populations and services provided across the sites. Participants at Casey House, which is a day health program in Toronto tended to be living with more concurrent health conditions, were less likely to be employed, and more likely be on income support compared to participants from the other sites (Table 1).

Mode of Administration
The mode of administration at the Ireland, UK, and US sites were primarily remote whereby participants completed the questionnaires independently online by accessing the link to the questionnaire via SMS text or email (76%), whereas at the Canadian site, most participants completed the questionnaires inperson using a tablet (24%). There were differences in characteristics based on mode of administration at T1 across cities, given the mode of administration was dependent to each city (see Additional le 3).

Internal Consistency Reliability
The EDQ met criteria for internal consistency across domains of the presence, severity and episodic scales (ICC > 0.7   (Table 3).

Construct Validity Hypotheses
Of the 36 primary hypotheses, 29 (81%) were con rmed; and of the 44 secondary exploratory hypotheses, 36 (82%) were con rmed, supporting construct validity for use with adults living with HIV (see Additional le 5).

Discussion
The EDQ scales possess internal consistency reliability, and the EDQ severity and presence scales  [12]. We anticipated lower Cronbach's alpha in this study, given the EDQ has fewer number of items in each domain (3-10 items) compared with the original longform HDQ (3-20 items) [12].
Our assessment for test-retest reliability of the EDQ severity and presence scales demonstrated ICCs > 0.70. Lower bound CIs for all ICCs were > 0.70, with the exception of the physical (0.69) and uncertainty (0.64) domains of the EDQ presence scale (Table 3). While community members living with HIV and clinicians highlighted the utility of the precursor SF-HDQ as an individual assessment of disability [18], this requires higher thresholds for reliability (> 0.80) [43], which we did not achieve in this study. Our assessment of ICCs involving only two points in time, may not account for the potential daily uctuations in disability, which would have in uenced the EDQ scores at T2 interpreted as error, hence the ICCs represented may underestimate the test-retest reliability of the EDQ. Future reliability of the EDQ with more repeated measures over time would be bene cial. Results suggest the EDQ may be positioned for group-based or program evaluation purposes rather than assessing disability or its use as an evaluative measure to assess change at an individual level. Furthermore, we acknowledge the di culty to identify source of measurement error when the mode of EDQ administration was dependent on the recruitment site with a diversity in populations. Ultimately, clinicians must determine what level of error they are willing to tolerate given the EDQ's intended use and the types of decisions that might be made based on the scores (e.g. referral to services, eligibility for disability income support, etc.) in clinical practice.
For test-retest reliability, it is important to consider potential sources of error such as mode of administration (electronic or paper-based), type of administration (interview administered versus selfcompletion), or setting [44]. The primary mode of administration in this study was remote whereby participants received a link via SMS or email to independently complete the questionnaire opposed to inperson administration via tablet at the Toronto site. Lower ICCs reported with in-person (tablet) administration may be attributed to the small sample size (n = 24) (see Additional le 4). Nevertheless, our provision of multiple options for mode of administration is a strength of the study and highlights the utility of the EDQ across different clinical settings.
The EDQ demonstrated lack of precision across the severity, presence and episodic scales (Table 4). This highlights the limitation for the EDQ to measure change in disability as distinct from day-to-day variability. While further work is needed to determine responsiveness to change in disability over time, results suggest the EDQ may only be able to detect large changes in disability that surpass day to day variability and measurement error.

Implications for Practice and Research
The EDQ has potential for use in clinical and community-based settings to describe disability, facilitate Our test-retest reliability and construct validity assessment was with the EDQ presence and severity scales only, as we did not expect consistency in a scale that refers to uctuations in the past week, and the criterion measures did not measure the episodic nature of health challenges. While the episodic scale of the EDQ is an important feature to characterize disability experiences, [16,18], and is unique to other disability measures [24,48], the use of the scale should be descriptive in nature to assist in providing a broader picture of the disability experience at a point in time.
This work directly builds on the SF-HDQ and its utility and sensibility with adults living with HIV in community and clinical settings [18]. Clinicians and persons living with HIV may use the EDQ or its precursor SF-HDQ in clinical practice. Both possess the same number of items, structural validity, and scoring algorithm based on the Rasch logit scale [17]. However, we anticipate the EDQ will be of greater use among adults aging with HIV who may experience disability attributed to other concurrent health conditions and not speci cally HIV. Furthermore, using the EDQ will enable cross comparisons with other chronic conditions in the future. Current research is assessing the utility and properties of the EDQ with other chronic and episodic conditions [49].

Limitations
We were unable to determine whether mode of administration or site in uenced the EDQ properties as we were unable to disassociate mode of administration from clinical site, country, and participant characteristics. Nevertheless, differences in participant characteristics and EDQ scores, re ected the types of populations served at each site and we consider the heterogeneity of the sample across the sites a strength of our study. Only 11% of the sample were women, which under-represents the proportion of women living with HIV in these countries [50,51]. Research is currently assessing the measurement properties of the EDQ speci cally with women living with HIV in the UK [52]. Finally, our results may not be transferable to adults living with HIV in low or middle-income countries [53]. Participants consented to participate in the study by checking 'yes I consent to participate in the study' at the initial information and consent page of the questionnaire administration.

Consent for publication
Not applicable.

Availability of data and materials
The data supporting the conclusions of this article are included within the article and its additional les. The data used and/or analyzed during this study may be available from the corresponding author on reasonable request.

Competing interests
None declared.