Psychometric characteristics of the health care empowerment questionnaire in a sample of patients with arthritis and rheumatic conditions

Abstract Background Patient empowerment can improve health‐related outcomes and is important in chronic conditions, such as arthritis. This study aimed to validate the Health Care Empowerment Questionnaire (HCEQ), a patient‐reported experience measure of empowerment, for use with patients with arthritis and other rheumatic diseases. Methods The HCEQ measures Patient Information Seeking (or Involvement in Decisions) and Healthcare Interaction Results (or Involvement in Interactions) and asks respondents to answer questions in two ways: whether they feel something happened and its importance to them. Face validity was assessed through qualitative data (n = 8, nominal group technique; n = 55, focus groups). Measure structure was assessed through confirmatory factor analysis (CFA); internal consistency was also assessed (n = 9226). Test‐retest reliability was assessed with sub‐sample of participants (n = 182). Results We found adequate face validity of the HCEQ for patients with arthritis. The CFA indicated good fit to the data for the two‐factor structure of the HCEQ (RMSEA = 0.075; CFI = 0.987; TLI = 0.978; SRMR = 0.026). Internal consistency was strong (α=0.94 for both subscales). Test‐retest reliability was moderate for Patient Information Seeking (ICC=0.67) and good for Healthcare Interaction Results (ICC=0.77). Conclusions The HCEQ, with modifications, demonstrated promising psychometric properties within this sample, laying the foundation for further assessment. This work supports the HCEQ as an appropriate instrument for examining experiences with and perceived importance of empowerment in individuals with arthritis and other rheumatic conditions. Patient Contribution Patients contributed to the assessment of face validity. As a measure of patient empowerment, the HCEQ’s use can enable further participation of patients in health care.


| INTRODUC TI ON
Patient-reported outcome measures (PROMs) are valuable in increasing patient engagement in clinical care by tracking self-reported outcomes and informing treatment decisions. 1 Patient-reported experience measures (PREMs) fulfil a similar purpose to PROMs, but they are different in that they measure what happened while care was provided to a patient, and the patient's perspective of their experience, 2 rather than health status or disease progression.
One aspect of patient experience with health care is patient empowerment, defined by the World Health Organization as 'a process through which people gain greater control over decisions and actions affecting their health'. 3 This concept focuses on an individual's own behaviours and beliefs in interactions with their health-care teams, rather than the influence of the health-care team and provider on a patient's experience. Studies support that patient empowerment, such as increased patient involvement in medical decisions, can lead to improved health outcomes, quality of life and satisfaction with health care. 4-7

| Health care empowerment questionnaire
There are a number of measures that purport to assess the multidimensional concept of patient empowerment. One instrument, the Health Care Empowerment Questionnaire (HCEQ), was developed with a sample of ageing adults in Canada. 8 The HCEQ measures three aspects of patient empowerment that had been proposed in the literature: degree of control, or consideration of who is involved in making decisions 9 ; involvement in interactions, or the ability and opportunity to communicate needs and initiate requests with a health-care provider 10 ; and involvement in decisions, or actively obtaining the information necessary to make rational decisions. 11 A concept analysis conducted more recently defines patient empowerment in similar terms, defining key attributes as the ability to effect change based on personal behaviour and the social environment, self-determination and ability for autonomy, and a process for obtaining self-management tools that enables empowerment. 12 With its original population of adults aged 75 and older in Canada, the HCEQ demonstrated good test-retest and internal consistency reliability on its three scales. 8 Specific health issues in the original sample were not provided, though participants were included if they were expected to experience a functional decline. Its measure structure was tested through an exploratory factor analysis (EFA) and validated with a confirmatory factor analysis (CFA), providing evidence for a three-factor structure and support for construct validity in their sample. A Persian version of the HCEQ has been validated with a sample of reproductive age women in Iran, 13 but validation studies of the English version with specific populations, beyond healthy adults, have not been published.

| Patient empowerment in arthritis, a chronic condition
In 2018, the authors (EK, KES and RLB) collaborated with the Arthritis Foundation (including authors EC and GSE) to develop the Live Yes! INSIGHTS survey, which uses PROMs and PREMs to longitudinally track member experiences. 14 The INSIGHTS survey is administered quarterly, and results are used to guide regional and national programming, resources, advocacy and research as part of the Arthritis Foundation's larger Live Yes! Arthritis Network, 15 its patient-facilitated network to support an estimated 20 million individuals with arthritis.
In developing the Live Yes! INSIGHTS programme, the Arthritis Foundation was interested in measuring patient empowerment to understand the experiences of patients with different characteristics (eg types of arthritis, race/ethnicity, socio-economic status) and from different geographic regions. By understanding needs and potential differences related to patient empowerment, the Foundation could then tailor their programming and patient education, which may enhance patients' empowerment in their care and in decision making. 16 Patient empowerment is particularly important in populations of patients with chronic conditions, such as arthritis, as these patients interact frequently with the health-care system and therefore have more opportunity for interactions with their providers. Arthritis and other rheumatic conditions are one of the leading causes of chronic pain in the United States 17 and rank seventh in the 2010 National Hospital Discharge Survey as the first-listed diagnosis on the hospital discharge. 18 In the National Ambulatory Medical Survey, osteoarthritis alone accounted for 11 147 000 primary care visits. 19 Similarly, in a 13-year longitudinal study, patients with rheumatoid arthritis were significantly more likely than controls to utilize services provided by general or specialty care physicians (OR = 1.75), particularly early in the disease. 20 Ostensibly, more frequent interactions with the health-care system indicate that patient empowerment is a critical construct to understand within populations with chronic conditions such as arthritis, particularly when evidence has shown that patient empowerment can lead to better outcomes. 4,7 However, the importance patients place on patient empowerment, including information seeking and involvement in decision making, depends on the patient and the decision being made. 21

| Validation of the HCEQ with a population of patients with arthritis
Given the importance of patient empowerment, patient input was used to develop the Live Yes! INSIGHTS programme, which resulted in the selection of the HCEQ (see Schifferdecker et al for more information). 14 The purpose of this study was to assess the psychometric properties of the HCEQ in a sample of patients with arthritis. This is critical as this is the first study to evaluate the HCEQ with a sample like this in the United States.

| ME THODS
We conducted our psychometric validation study in two phases.
In phase 1, we assessed face validity using qualitative methods. In phase 2, we used data from completed INSIGHTS surveys to assess measure structure, internal consistency and test-retest reliability.
Both phases of the study were deemed not human subjects research by our Institutional Review Board.

| Phase 1: Qualitative data to assess face validity
Details of the qualitative study are described in detail elsewhere. 14 Briefly, we used a modified Delphi and virtual nominal group technique (NGT; n = 8), and then six focus groups (n = 55), to get input on the empowerment measure to use in INSIGHTS. Participants primarily included adults with arthritis, though a health-care provider and measurement expert also participated. Participants in the NGT were asked to rate three patient empowerment or self-advocacy measures, one of which was the HCEQ, selected through a critical literature review emphasizing psychometric quality. 23 Ratings were summarized and used to facilitate discussion of the measures among NGT participants, including whether the measures adequately captured patient empowerment or selfadvocacy. After the HCEQ was selected, information about how likely patients would be to complete the survey at multiple points over time was gathered from the focus groups. We recorded and obtained transcripts for both the NGT and the focus groups and assessed face validity through descriptive ranking data collected during the NGTs, along with a content analysis of the NGT discussion and focus group transcripts.

| Study design and participants
For the quantitative psychometric analyses in the current study, we Surveys were administered through an online platform (Qualtrics™, Provo, UT). There were two ways in which participants could provide data: through an anonymous URL provided on the Arthritis Foundation website or through an email invitation using contact lists from people who had previously participated in Arthritis Foundation programming.
Participants were encouraged to take the survey as often as they saw fit, and some of these administrations occurred within the seven-to 14-day window that was determined to be sufficient for assessing test-retest reliability, based on previous research with other patient experience measures. [24][25][26][27] The first survey completed by each eligible participant was used for this study (or two survey administrations occurring seven to 14 days apart for test-retest reliability).
To confirm measure structure and assess internal consistency, we extracted a sample of 9226 individuals from the INSIGHTS population. Participants' data were included in this study if they were at least 18 years old, English speaking, had completed a survey between Participants answered questions using a Likert scale response from one to four. Respondents were asked to answer each of the seven questions in two ways: with regard to their experience ('did you feel that…') and their perceptions of the importance of the item ('how important is it that…'). As per the original article, we created scale scores by obtaining the cross-product of the Feelings and Importance responses for each of the items and summing these items within the Patient Information Seeking and Healthcare Interaction Results scales. See Table 1 for details about the survey and its items.
In addition to the HCEQ and demographic items, the full survey in-

| Data analysis
To assess measure structure (and construct validity), we ran a confirmatory factor analysis (CFA), applying the same measurement structure to this sample of adults with arthritis and other rheumatic conditions, as was applied in the original HCEQ psychometric development. 8 A CFA was used since the aim of this analysis was to confirm an existing factor structure presented in previous research 8 with a new population of patients with arthritis and other rheumatic diseases. 28,29 We examined two of three factors presented in the original model: Involvement in Decisions (Patient Information Seeking) and Involvement in Interactions (Healthcare Interaction Results). In addition, we examined an alternative model separating out the Feelings and Importance responses, rather than using the cross-products, for two reasons: 1) to determine whether other models demonstrated better fit to the data and 2) because previous literature using patient empowerment measures included items similar to the HCEQ Feelings responses and examined the impact of importance separately. 21,22 Because patients may feel differently about the importance of patient empowerment, 21 These analyses were also run on the two additional models using the Feelings and Importance responses separately.

| RE SULTS
Demographic information for each sample is presented in Table 2.

| Face validity
After discussion during the virtual NGT in phase 1, participants selected the HCEQ as their preferred measure of patient empowerment or self-advocacy, pointing to its ability to 'provide much more valuable information' as compared to other surveys, lending support to its face validity. Overall, on their initial ratings of the HCEQ and its items, more than half of participants reported that they thought the HCEQ was moderately to extremely useful. The present models all used maximum-likelihood estimation with Satorra-Bentler correction. For verification, this model was also estimated using asymptotic distribution-free (ADF) estimation, which makes no assumption of normality, and the results were comparable.

Feelings responses prompt Importance responses prompt Items
b All of the Ns provided are for the present sample (not for the original validation article). *P < .001 (for present sample).

| Construct validity and measure structure
Due to the results of the NGT, we removed the third factor in the original model (Degree of Control). The items that loaded onto the two remaining scales through an EFA and confirmed with a CFA in the original article were specified in the same way in the current model. Thus, a two-factor model was fit, with three items on Patient Information Seeking and four items on Healthcare Interaction Results.
Missingness was low (n = 508, 5.21%), and thus, cases with missing values were deleted listwise. Since the assumption of multivariate normality was not met for both the Patient Information Seeking and Healthcare Interaction Results scales, model fit was examined using maximum-likelihood (ML) estimation with the Satorra-Bentler adjustment, which adjusts for non-normal data. Model fit indices are provided in Table 3, alongside the results from the original development and validation article. Table 3  Another set of models was run to determine whether separating out the Feelings and Importance responses demonstrated better fit to the data (see Table 3 for more information on items included). Two separate models were specified: one with the Feelings responses and one with the Importance responses, both using the two scales: Patient Information Seeking (three items) and Healthcare Interaction Results (four items). Scale scores were obtained by summing the individual items in each scale, rather than summing the cross-products.
Using ML estimation with the Satorra-Bentler adjustment, the model for Feelings (n = 9226) indicated good overall fit to the data, RMSEA  Figure 2. While these separate models did indicate good fit, they were not a substantive improvement over the original combined model.

| Internal consistency reliability
Cronbach's alphas were run, and missing data were deleted listwise as missingness was low. For the combined model using cross-products, Cronbach's alpha for Patient Information Seeking (α = 0.94) and Healthcare Interaction Results (α = 0.94) were good (Table 3).
When the responses were separated, Cronbach's alphas for the Feelings responses were somewhat lower, but acceptable, for Patient Information Seeking (α = 0.91) and Healthcare Interaction Results (α = 0.93). Cronbach's alphas for the Importance responses were still within the acceptable range for Patient Information Seeking (α = 0.92) but were slightly lower than 0.9 for Healthcare Interaction Results (α = 0.89).

| Test-retest reliability
We used ICCs to assess test-retest reliability using 182 eligible cases ( Test-retest reliability for the two groups of Importance responses was low for both Patient Information Seeking (ICC =0.57) and Healthcare Interaction Results (ICC =0.59).

F I G U R E 1
Final two-factor model and factor loadings for the Health Care Empowerment Questionnaire (HCEQ), with standardized coefficients

| D ISCUSS I ON
While multiple patient-reported experience measures have been developed and validated with different populations, this is the first study to validate the HCEQ, a measure specifically capturing patient empowerment, with a sample of adults with arthritis and other rheumatic conditions in the United States. This is valuable as patients with arthritis frequently interact with the health-care system, [18][19][20] and the ability to assess patient empowerment using a reliable and valid tool could identify opportunities for increasing patient empowerment, and thus potentially related health outcomes. 4,7 Our study provides preliminary support for the use of the Considering the two answer types (Feelings and Importance) separately may also be a beneficial use of the HCEQ. Though the CFA assessing the original model (using the cross-products of Feelings and Importance responses) produced good fit to the data, so did separate models for the Feelings and Importance responses.
If used separately, relationships, or the lack thereof, could be explored between a patient's experiences (Feelings) and the importance they attribute to having that experience (Importance).

| Limitations and strengths
Though efforts were made to address issues with the data, some limitations are worth noting. First, we used a sample of convenience that was gathered through the efforts of a national foundation. The demographics of this sample indicate that it is not entirely representative of patients with arthritis and rheumatic conditions in the United States; thus, generalizability should be considered. However, these methods allowed us to recruit over 9,000 participants, which is a strength of the study. Additionally, we relied upon self-report of arthritis diagnosis, which we were not able to verify. While confirmation of diagnoses is preferable, a recent meta-analysis suggested that self-reported arthritis type (specifically osteoarthritis and rheumatoid arthritis) was sufficiently accurate for large-scale studies when diagnosis cannot be confirmed. 32 Also, this was a non-incentivized study, so there is no perceived incentive to lie. Further, though we were not able to verify diagnoses, the fact that we include participants with different types of arthritis and rheumatic conditions allows us to speak to the use of this measure across these diagnoses. We were also not able to speak to whether these findings relate to a specific type of provider, as we did not ask participants to report the type of provider.
While we did have an adequate sample of patients who completed the HCEQ twice within seven to 14 days, these data were not collected intentionally for the purpose of assessing test-retest reliability. We aimed to address this as much as possible by ensuring that patients had not changed medications or visited a hospital in between administrations of the HCEQ.

| Future directions and conclusions
This study provides initial support for the use of the HCEQ with populations with arthritis and other rheumatic conditions. These findings support the use of the HCEQ in the Arthritis Foundation's Live Yes! Network as an assessment for the construct of patient empowerment with this population. Based on these findings, future directions and next steps may include further validation of the HCEQ with this population. For instance, it may be beneficial to assess additional psychometric qualities, such as concurrent and predictive validity. In doing so, it may also be beneficial to assess whether suggested changes improve the test-retest reliability of the HCEQ with this population, particularly for the separated Feelings and Importance responses. If so, this would allow for an examination of the relationship between the Feelings and Importance responses.

ACK N OWLED G EM ENTS
We would like to thank Lynn Foster-Johnson, PhD, for her consultation on components of our statistical analysis.

CO N FLI C T O F I NTE R E S T
There are no conflicts of interest to report for any authors.

AUTH O R CO NTR I B UTI O N S
Knight participated in the conception and design process, prepared for data acquisition, developed the analysis plan, led the analysis and interpretation of the data, drafted the manuscript and led the revision process, and provided final approval of the version to be submitted. Carluzzo participated in the conception and design process,

DATA AVA I L A B I L I T Y S TAT E M E N T
Research data are not shared.