Skip to main content
Journal of Patient-Reported Outcomes
Download PDF
Download ePub

Psychometric evaluation of the Adelphi Adherence Questionnaire (ADAQ©) in adults with osteoarthritis

Journal of Patient-Reported Outcomes volume 8, Article number: 118 (2024) Cite this article

Abstract

Background

Medication non-adherence is a common issue in chronic illness. The World Health Organization has recognized a need for a valid and reliable method of measuring adherence to understand and mitigate non-adherence. This study aimed to psychometrically evaluate the English version of the Adelphi Adherence Questionnaire (ADAQ©), a questionnaire designed to assess patient-reported medication adherence across multiple therapy areas, in patients with Osteoarthritis (OA).

Methodology

Data from the Adelphi OA Disease Specific Programme™, a survey of physicians and their consulting adult patients with OA conducted in the United States, November 2020 to March 2021, was used to assess the psychometric properties of the ADAQ. Patients completed the ADAQ, Adherence to Refills and Medication Scale (ARMS), Western Ontario and McMaster Universities Arthritis Index (WOMAC), and EQ-5D-3L. The measurement model of the 13-item ADAQ was assessed and refined using latent variable modelling (Multiple Indicator Multiple Cause, confirmatory and exploratory factor analyses, item response theory, Mokken scaling, and bifactor analyses). Correlational analyses (Spearman’s rank and polyserial as appropriate) with ARMS, WOMAC, and EQ-5D-3L scores assessed construct validity. Anchor- and distribution-based analyses were performed to estimate between-group clinically important differences (CID).

Results

Overall, 723 patients were included in this analysis (54.5% female, 69.0% aged ≥ 60). Latent variable modelling indicated a unidimensional reflective model was appropriate, with a bifactor model confirming an 11-item essentially unidimensional score. Items 12 and 13 were excluded from scoring as they measured a different concept. The ADAQ had high internal reliability with omega hierarchical and Cronbach’s alpha coefficients of 0.89 and 0.97, respectively. Convergent validity was supported by moderate correlations with items of the ARMS, and physician-reported adherence and compliance. Mean differences in ADAQ score between high and low adherence groups yielded CID estimates between 0.49 and 1.05 points, with a correlation-weighted average of 0.81 points.

Conclusion

This scoring model showed strong construct validity and internal consistency reliability when assessing medication adherence in OA. Future work should focus on confirming validity across a range of disease areas.

Introduction

Medication adherence is the degree to which medication is taken as prescribed [1]. Adherence differs from compliance in that it is an active process that presumes patient and physician agreement on recommendations rather than simply passively obeying advice [2]. Medication non-adherence is a common issue across chronic conditions (reported in up to 50% of patients) [3,4,5,6]. A report by the World Health Organization recognizes that a valid and reliable measure of adherence is required to better understand and mitigate medication non-adherence [7].

All methods of assessing medication adherence have their strengths and limitations. ‘Objective’ methods, such as counting remaining treatment doses, risk over-estimating adherence by not accounting for doses taken late or lost, they do not provide insight into reasons for non-adherence, and are costly to implement in practice [8]. Meanwhile, against regulatory standards, existing self-report measures such as the Adherence to Refills and Medications Scale (ARMS [9, 10]), Medication Adherence Rating Scale (MARS [11]) or the Morisky Medication Adherence Scale (MMAS [12]), to name a few, may not be considered fit to assess medication adherence behaviors and drivers across multiple conditions/medication administration methods. This is because they are often developed in a single disease area, a single country, without patient input, or in a single medication type; therefore, evidence for their wider content validity and reliability as well as relevant assessment concepts in certain therapy areas can be lacking [13]. The Adelphi Adherence Questionnaire (ADAQ©) was developed to address some of these gaps in content and evidence for existing measures.

The ADAQ was developed in accordance with professional and regulatory best practice for patient-reported outcome measures to measure adherence across a range of disease areas and diverse populations [14,15,16,17,18]. Measurement concepts were elicited from a review of qualitative literature and interviews with 57 adults with a wide range of health conditions (hypertension, asthma, multiple myeloma, psoriasis, diabetes, depression, multiple sclerosis, and/or schizophrenia) [19, 20]. These disease areas were chosen to ensure concepts accommodated an array of medication administration methods (e.g., topical, injected, inhaled), dosage regimes (e.g., daily, weekly), and patient demographics (e.g., different ages, cognitive abilities). Based on these measurement concepts and guidance from clinical experts, the ADAQ was developed and qualitatively evaluated among target populations to ensure respondent understanding and concept relevance [20,21,22]. While this research provided strong evidence of content validity, a scoring algorithm still needed developing and psychometric properties, such as reliability and construct validity, evaluating across different disease areas. This study focuses on psychometric evaluation of the ADAQ within an osteoarthritis (OA) setting, the first such psychometric evaluation to be performed (though psychometric evaluations in other health conditions are in progress). The psychometric properties of the ADAQ have not previously been evaluated in an OA population and OA patients were not included during qualitative research.

OA is a highly debilitating disease characterized by joint pain and movement difficulties. In addition to the pain and physical limitations experienced by the individual there is a wider burden on health systems and society; OA is estimated to cost the US approximately $136.8 billion annually in direct and indirect costs [23, 24]. To reduce the burden of OA it is essential that treatments effective at controlling pain and managing OA disability are developed and, importantly, that patients are adherent to prescribed long-term medication (such as non-steroidal anti-inflammatory drugs). Medication non-adherence has demonstrable associations with poorer health outcomes in OA patients [25].

The aim of this study was to assess the psychometric properties of the ADAQ, including the assessment of factorial structure (dimensionality), determination of a scoring algorithm, and evaluation of the validity and internal consistency reliability of the ADAQ in patients with OA based on analysis of real-world data.

Materials and methods

Data collection

This analysis was based on data from the Adelphi OA Disease Specific Programme (DSP™), a large cross-sectional, real-world, survey of patients with OA and their primary care physician, rheumatologist, or orthopedist. The DSP methodology has been previously published [26, 27], validated [28], and shown to be consistent over time [29]. Data were collected from November 2020 to March 2021. Patient and physician data were collected concurrently. This analysis included patients from the United States who were diagnosed with OA by their consulting physicians. Physicians were asked to complete a patient record form detailing patient demographics and clinical characteristics; these same patients were then asked to complete a patient self-complete form containing the ADAQ as well as the 12-item ARMS [9, 10], WOMAC [30], and EQ-5D-3L [31]. All patients who completed at least one item of the ADAQ were included in the analysis.

Assessments

The ADAQ was developed as a 13-item patient-reported questionnaire designed to measure medication adherence [20,21,22]. A conceptual framework is provided in Table 1. Questions consisted of 4 items assessing medication adherence behaviors (e.g. missing a dose, taking less medication than prescribed), 8 items assessing drivers of medication adherence (e.g. forgetting, perceptions of treatment necessity), and one item assessing overall adherence to medication. Possible ordinal item responses for each question are shown in Supplementary Table 1. Item 9 asked about skipping taking medication or taking less due to cost and is designed for countries where patients pay out-of-pocket or have insurance (such as the United States) as opposed to more socialized healthcare systems (such as England); this item can be omitted if irrelevant in certain countries. An overall adherence summary score can be calculated based on the mean of the first 11 items and can range from 0 to 4, with lower scores indicating greater medication adherence. Overall scores are calculated if at least 8 items have been completed. This study developed and validated this scoring methodology in an OA population.

Table 1 Adelphi Adherence Questionnaire conceptual framework
Full size table
Fig. 1
figure 1

Illustrative causal/reflective indicator model (1) and fully reflective model for the Adelphi Adherence Questionnaire (2). Notes Not all questions are listed for concision, in the effect indicator model all questions would represent effects of non-adherence. Abbreviations ADAQ Adelphi Adherence Questionnaire

Full size image

The ARMS is a 12-item patient-reported questionnaire assessing adherence. An overall sum score is calculated from all items and may range from 12 to 48, with lower scores indicating better adherence. Two subscales can also be calculated: ‘Taking medication as prescribed’, and ‘refilling medications on schedule’ [10].

The WOMAC is a 24-item patient-reported questionnaire developed for hip and knee OA [30]. It is divided into 3 subscales: Pain (5 items, range 0–20), stiffness (2 items, range 0–8), and physical functioning (17 items, range 0–68); higher scores indicate worse pain, stiffness, and physical functioning.

The EQ-5D-3L is a generic instrument for describing and valuing health in 5 dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression [31]. Each dimension has 3 levels of response: no problems, moderate problems, and extreme problems.

A patient-reported single item measuring ‘satisfaction’ (‘Overall satisfaction with your medicine’) with 5 response categories (‘Extremely dissatisfied’ to ‘Extremely satisfied’), was administered, in addition to physician-completed single items measuring: ‘adherence’, ‘compliance’ and ‘satisfaction’. Details of responses can be found in Supplementary Table 1.

Analyses

Two different measurement models underlying the ADAQ were evaluated. The first assumed a series of causal indicators (‘reasons for non-adherence’) gave rise to the latent construct of ‘adherence’, combined with several reflective indicators of adherence (‘extent of non-adherence’) [32,33,34]. The second potential model assumed all items were reflective indicators of adherence (illustrated in Fig. 1). The adequacy of these proposed measurement models was initially compared by fitting a multiple-indicator multiple-cause (MIMIC) model, and a unidimensional confirmatory factor analysis (CFA) model. The following global fit statistics for CFA models were calculated: standardized root mean square residual (SRMR; <0.10 acceptable), root mean square error of approximation (RMSEA; <0.1 acceptable), and comparative fit index (CFI; >0.95 acceptable) [35,36,37].

Assuming a fully reflective measurement model, various dimensionality analyses were also employed as recommended in the literature [38]. These analyses included evaluation of monotone homogeneity using a Mokken model—a non-parametric item response theory (IRT) model—and assessment of Loevinger’s scalability coefficient (where item clusters with coefficients > 0.5 reflect strong unidimensionality [39]), as well as exploratory graph analysis (EGA) using a glasso model with walktrap community detection algorithm [40,41,42].

Multidimensional structures suggested by any of these methods were subsequently explored through bifactor CFA modelling using a weighted least squares mean- and variance-adjusted (WLSMV) estimator. For the bifactor model, the proportions of explained common variance (ECV) across items by general and specific factors were then summarized (ECV > 0.80 is considered essential unidimensionality) [43]. Based on evaluation of measurement models and dimensionality, ADAQ scoring was then proposed.

Internal consistency of the ADAQ was assessed using McDonald’s omega coefficients and Cronbach’s alpha [44]. Construct validity of the ADAQ was assessed by comparing correlations to pre-defined hypotheses (which were determined based on consideration of the content of the questions for the different scores and whether measures were self- or physician-reported): ARMS total (r ≥ 0.50), ARMS taking medication as prescribed (r ≥ 0.30), ARMS refilling medications on schedule (r ≥ 0.30), physician-reported adherence (r ≥ 0.30), physician-reported compliance (r ≥ 0.30), physician-reported satisfaction (r ≥ 0.30), patient-reported satisfaction (r ≥ 0.30). Divergent hypotheses were also assessed: WOMAC scores (r ≤ 0.30) and EQ-5D-3L dimensions (r ≤ 0.30). To compare correlations between the ADAQ, patient-reported ARMS, and physician-reported adherence Steiger’s modification z using average correlations was utilized [45]. Response distributions were examined for each item to identify any unexpected or inappropriately skewed distributions.

Preliminary thresholds for between-group clinically important differences (CIDs) were estimated using anchor- and distribution-based methods [46, 47]. The anchor-based analysis results were prioritized for interpretation, with the distribution-based analyses considered to provide context for interpreting the anchor-based analysis. Anchors were formed using the ARMS total score, patient-reported satisfaction, and physician reported adherence, satisfaction, and compliance (details in Supplementary Table 2), where those with polyserial correlations with ADAQ score > 0.3 were retained. The difference in mean ADAQ score between the high and low adherence anchor groups were estimated for the calculation of CID, which were triangulated using a correlation-weighted average based on Fisher’s z transformation [48]. Supplementary distribution-based statistics comprised half a standard deviation (SD) and the standard error of measurement (SEM).

Measurement models, dimensionality, and correlation comparison analyses were performed in R (version 4.1.1) with the psych [49], lavaan [50], mirt [51], MplusAutomation [52], BifactorIndicesCalculator [53], EGAnet [54] and mokken [55], and concor [56] packages, and Mplus Version 8.1. All other analyses were performed in SAS Version 9.4.

Results

Data from 723 patients formed the analysis population, of which 54.5% (n = 394) were female and 69.0% (n = 499) were aged 60 or over. Most patients were White (76.6%, n = 554), with 11.3% (n = 82) Black, 6.9% (n = 50) Hispanic/Latinx, and 5.2% (n = 37) other/non-white (Table 2). The joints most commonly affected by OA were knees (59.5%, n = 430), upper/middle/lower back (39.0%, n = 282), hips (27.0%, n = 195), and hands/fingers/thumbs (21.9%, n = 158). Clinicians rated 33.5% (n = 242) of patients as having mild OA, 49.4% (n = 357) moderate, and 16.6% (n = 120) severe; severity was not known or reported for 0.5% (n = 4) of patients. Physicians subjectively reported 42.2% (n = 257) of patients as completely adherent to medication (full breakdown in Table 2). Though the majority of treatments prescribed were orally administered (acetaminophen, ibruprofen, celecoxib, meloxicam, naproxen, naproxen/esomeprazole fixed dose combination, tramadol, codeine, morphine, hydrocodone, oxycodone, tapentadol, and the supplement chondroitin), there was some variability in treatment administration methods for diclofenac (n = 69, oral: 43, patch: 4, cream/topical: 22), hyaluronic acid (n = 23, oral: 4, subcutaneous/intramuscular injection: 4, other: 15),the supplement Glucosamine (n = 27, oral: 26, subcutaneous/ intramuscular injection: 1), intra-articular corticosteroid (n = 56, Oral:1, subcutaneous/ intramuscular injection: 25, other: 30), buprenorphine (n = 4, oral: 43, patch: 1, other: 3), and fentanyl (n = 4, patch/plaster: 4).

Table 2 Patient demographics
Full size table
Fig. 2
figure 2

Bifactor model fitted to the 11 items that make up the Adelphi Adherence Questionnaire score. Abbreviations ADAQ Adelphi Adherence Questionnaire, f factor

Full size image
Fig. 3
figure 3

Unidimensional model fitted to the 11 items that make up the Adelphi Adherence Questionnaire score. Abbreviations ADAQ Adelphi Adherence Questionnaire

Full size image

The fit of the MIMIC CFA model was generally poorer (SRMR = 0.1, CFI = 0.65, RMSEA = 0.12) in comparison to the unidimensional CFA model (SRMR = 0.05, CFI = 0.96, RMSEA = 0.16); therefore, a fully reflective measurement model was chosen for exploration of dimensionality. Inter-item correlations ranged from 0.41 to 0.89 (average 0.67), further supporting a fully reflective measurement model.

Mokken scale analysis demonstrated that items 1 to 11 clustered under a Loevinger’s scalability coefficient of 0.5, but items 12 (patient’s confidence in taking medication as intended) and 13 (patient perceived overall adherence) separated to form a separate cluster (See Supplementary Table 3). Bootstrapped EGA (500 iterations) suggested three factors: Items 1, 2, and 4–8 related to factor 1, items 3, and 9–11 to factor 2, and items 12 and 13 to factor 3 (see Supplementary Fig. 1). Unidimensional CFA model showed that factor loadings of items 12 and 13 (\(\:{\lambda\:}_{12}=0.58\) and \(\:{\lambda\:}_{13}=0.61\)) were considerably lower compared to loadings of other items (ranging from 0.80 to 0.91). In summary, Mokken scale analysis, EGA, and CFA all suggested items 12 and 13 reflect a separate construct to the remaining 11 ADAQ items and were thus excluded from further analysis.

A bifactor model was fitted to the first 11 items of the ADAQ, where the use of two specific factors was guided by EGA. Figure 2 shows this model, which demonstrated acceptable fit across SRMR and CFI but not RMSEA (SRMR = 0.02, CFI = 0.99, RMSEA = 0.11) and a high ECV (0.85) for the general factor, supporting essential unidimensionality. This bifactor model and a unidimensional 11-item CFA (items 1 to 11; Fig. 3) were re-estimated using a full information maximum likelihood estimator which allowed direct comparison of both models with statistics penalizing for more complex models such as Akaike Information Criterion (AIC) or Bayesian Information Criterion (BIC). These statistics consistently favored the bifactor model (AIC: bifactor = 12201, unidimensional = 12616; BIC: bifactor = 12503, unidimensional = 12868; sample size adjusted BIC: bifactor = 12214, unidimensional = 12693).

Given the above, the ADAQ score was defined as the unweighted average of items 1 to 11. Items 12 and 13 can be summarized on their own for additional information, but do not contribute to the ADAQ score due to cumulative evidence that they may reflect a different construct to the other items. Item 13 relating to patient perception of adherence is also likely to violate local independence, a key assumption of CFA models. All subsequent analyses are based on the 11-item ADAQ score, which ranges from 0 to 4 where lower scores indicate greater adherence.

Convergent and divergent validity is summarized in Table 3. The majority of correlation size hypotheses were met, with the ADAQ showing strong convergent correlations with all measures of adherence and compliance, particularly the ARMS total score (0.774) and the ARMS taking medication as prescribed sub scale (0.798). The ADAQ showed a greater correlation with the patient-reported ARMS than physician-reported adherence (z = 11.2936, p-value < 0.001). ADAQ scores did not show convergent correlation with either patient- or physician-reported satisfaction.

Table 3 Correlational analysis of the Adelphi Adherence Questionnaire score with other patient- and physician-reported measures
Full size table

The internal consistency reliability coefficients omega hierarchical and Cronbach’s alpha were 0.89 and 0.97, respectively. Cronbach’s alpha did not increase or notably decrease after deletion of any item from the final 11-item version of the ADAQ, showing that all contributed towards reliability and supporting retention of all items. All anchor measures for score interpretation analyses were sufficiently correlated except for physician-reported satisfaction (r = 0.263) and compliance (r = 0.247). Mean differences in ADAQ score between high and low adherence groups yielded CID estimates between 0.49 and 1.05 points, with a correlation-weighted average of 0.81 points (Table 4). Distribution-based estimates were 0.40 (half SD) and 0.20 (SEM).

Table 4 Preliminary clinically important difference estimates based on anchor measures
Full size table

Discussion

Latent variable modelling was used to evaluate the ADAQ measurement model. This included factor analysis (featuring unidimensional CFA, MIMIC, and bifactor models), EGA, and Mokken scaling. MIMIC, bifactor, and CFA models are subtypes of structural equation measurement models. MIMIC models were used to test the initially proposed conceptual structure of the ADAQ (i.e., that it contains both drivers, and reflections of adherence). EGA is similar to exploratory factor analysis and was used to evaluate the number of factors underlying the ADAQ following unacceptable fit of initially proposed MIMIC models. Mokken scaling (related to item response theory models, but with non-parametric assumptions) was used to help identify potential scales for a unidimensional ADAQ score. Both EGA and Mokken scaling supported the removal of items 12 and 13 from an ADAQ total score, with the latter suggesting the appropriateness of an 11-item unidimensional “adherence” score. CFA and bifactor models were used to evaluate the fit of an 11-item ADAQ score and further assess whether it was appropriate to consider an “essentially unidimensional” structure for the ADAQ (i.e. do all items reflect the overarching concept “adherence”). An essentially unidimensional structure was supported, with a bifactor model (featuring a strong general factor and 2 specific factors providing the best fit), particularly with regard to SRMR, which has been demonstrated to be more appropriate for use in categorical/ordinal factor analyses than RMSEA [37].

The results from the Mokken scale analysis and bifactor modelling supported an essentially unidimensional measurement model for the ADAQ, where the unweighted average of items 1 to 11 of the ADAQ represents the overall ADAQ score. We note that the utility of item weighting of total scores remains a topic of ongoing debate in the methodological literature but was also chosen based on practical considerations (i.e., to facilitate its application in broad research settings, and to multiple disease areas) [57,58,59,60,61,62]. Items 12 and 13 provided additional information but did not contribute to ADAQ scoring because latent variable modelling suggested they represent a different construct. Item 12 assessed the patient’s confidence that they are taking their medication correctly and item 13 provided a patient’s perception of their own adherence. As such, while these items provide useful context to the more in-depth assessment of adherence behaviors provided by items 1–11, these items can still be administered but should be scored separately. Possible scores for the summary score range from 0 to 4 where 0 is full adherence and 4 is complete non-adherence. Preliminary CID estimates indicated that a change in ADAQ score of 0.81 represents a clinical important difference in adherence in the context of OA.

ADAQ scoring showed strong internal consistency reliability with omega hierarchical and Cronbach’s alpha coefficients of 0.89 and 0.97, respectively.

ADAQ scores showed strong convergent correlation with ARMS scores and physician-reported compliance and adherence, and divergent correlation with the WOMAC and EQ-5D-3L scores. Perhaps unsurprisingly, the only areas in which ADAQ scoring did not achieve hypothesized correlation thresholds were with physician- and patient-reported satisfaction. In CFA analysis we found the model that best fit the ADAQ was a unidimensional reflective model meaning ADAQ scores likely represent the effects of non-adherence rather than the causes of non-adherence. It is likely that patient dissatisfaction is a cause of non-adherence rather than an effect of non-adherence [63]. In addition, while one may expect satisfaction to correlate with adherence, there are likely many other factors involved in patient satisfaction [64]. Similarly, there are likely factors other than treatment satisfaction impacting adherence such as the cost of treatment and medication side-effects. The physician-reported satisfaction question particularly focused on level of pain control, which is just one aspect of non-adherence captured by the ADAQ. The ADAQ provides a tool with which physicians can better understand the myriad factors associated with medication adherence.

The fact that the ADAQ showed higher convergent correlation with the patient-reported ARMS than physician-reported medication adherence (tested using Steiger’s modification z using average correlations [45]), highlights one of the key strengths of the ADAQ as a patient-reported measure of adherence. It has been shown in other disease areas that there is often discordance between physician- and patient-reported adherence in chronic conditions [65]. Physicians can only generally report adherence based on clinical test results, symptom assessment, information collected from patients and how frequently patients are collecting their prescriptions, and therefore may not be aware of patients missing doses, losing doses, or not taking medication on time. Clinical measures of adherence also lack the context surrounding non-adherence. The fact that the ADAQ correlates better with patient-reported measures indicates that it is reliably and accurately able to capture the nuance of patient-reported adherence that may be lacking from assessments based on physician judgment.

Response distribution issues could potentially limit ability to distinguish between some patients [66, 67]. Ceiling effects are particularly common among self-report measures of adherence due to social desirability and memory biases [66]. The scoring algorithm here led to nearly 20% of patients scoring 0 on medication adherence potentially indicating a ceiling effect. It has been shown, however, that patients with more severe OA are more likely to be adherent to their medication than those with mild OA [68]. Over half of our cohort were considered moderate to severe by their physician which could explain the high percentage of patients scoring 0 for adherence.

This study has several limitations. In terms of methodology, the cross-sectional nature of DSPs means we were unable to study the test-retest reliability of the ADAQ and ability to detect change in adherence, these properties should therefore be addressed in future research. The very high internal consistency coefficients observed could be argued to suggest a degree of redundancy. However, given that all ADAQ items are focused on different adherence behaviors and drivers, it was judged that all items were of value to retain from a content validity perspective. Nevertheless, further item reduction or creation of a short-form could be considered in the future. Given the ADAQ asks patients to report adherence retrospectively, there is the potential for recall bias particularly in patients on medications taken less frequently. Similarly, no recall period was set for the ADAQ to ensure its utility across a wide range of disease and treatment areas with differing dosing schedules (for example, if a 7-day recall period was included, this would not be appropriate for a patient whose treatment is a monthly injection). This does however mean patients could be reporting adherence over differing time periods. Further, as noted, the aim is that the ADAQ should be used across disease areas and treatment modalities including a range of types of treatments and supplements. As such, further validation in other disease populations with a broader range of treatments/supplements is also recommended, to confirm if the measurement properties are consistent across populations. It should also be acknowledged that it could be debated whether considering ‘adherence’ is relevant/appropriate for treatments such as physician-administered intra-articular injections; however, even those require the patient to attend appointments, and so it could be considered a relevant consideration. Moreover, in this study, most patients receiving such treatment were also taking treatments that were self-administered. Despite these limitations our analysis allowed us to develop a robust scoring system in a real-world setting to effectively measure adherence in consulting patients with OA. Data analyzed in this study were collected between November 2020 and March 2021; thus during the COVID-19 pandemic. While these authors see no reason to expect the instrument measurement properties would be different for this reason, it is possible that patients may have been more (perhaps due to simpler routines, less distractions) or less (perhaps due to less contact with healthcare professionals) adherent at that time. Confirmation of the findings presented here in data collected following the pandemic would provide insight/reassurance around this point.

Conclusions

This paper outlines results of a psychometric evaluation of the ADAQ in a large and diverse OA patient population. The analysis demonstrated that the 13-item ADAQ and 11-item scoring algorithm developed herein have strong construct validity and internal consistency reliability when measuring medication adherence in the adult OA population. This version of the ADAQ provides a valid measure of patient-reported adherence to medication and addresses an unmet need for a generic medication adherence measure for use across a variety of conditions in real-world studies and routine clinical practice. To support the use of the ADAQ in the assessment of patient-reported adherence, future work should focus on establishing interpretation guidelines for the ADAQ score and test-retest reliability as well as its validity in other disease areas and confirmation of cross-cultural validity.

Data availability

All data that support the findings of this study are the intellectual property of Adelphi Real World. All requests for access should be addressed directly to Nathan Clarke at nathan.clarke@adelphivalues.com.

Abbreviations

ADAQ©:

Adelphi Adherence Questionnaire

AIC:

Akaike Information Criterion

ARMS:

Adherence to Refills and Medications Scale

BIC:

Bayesian Information Criterion

CFA:

Confirmatory Factor Analysis

CFI:

Comparative Fit Index

CIDs:

Clinically Important Differences

DSP:

Disease Specific Programme

ECV:

Explained Common Variance

EGA:

Exploratory Graph Analysis

IRT:

Item Response Theory

MARS:

Medication Adherence Rating Scale

MIMIC:

Multiple-indicator Multiple-cause

MMAS:

Morisky Medication Adherence Scale

OA:

Osteoarthritis

PRO:

Patient Reported Outcome

RMSEA:

Root Mean Square Error of Approximation

SD:

Standard Deviation

SEM:

Standard Error of Mean

SRMR:

Standardized Root Mean Square Residual

WLSMV:

Weighted Least Squares Mean- and Variance-adjusted Estimator

WOMAC:

Western Ontario and McMaster Universities Arthritis Index

References

  1. National Instititute for Health and Care Excellence (2009) Medicines adherence: involving patients in decisions about prescribed medicines and supporting adherence. Available via https://www.nice.org.uk/guidance/cg76/chapter/introduction. Accessed: 16 Aug 2023

  2. Mir TH (2023) Adherence Versus Compliance. HCA Healthc J Med 4(2):219–220. https://doi.org/10.36518/2689-0216.1513

    Article  PubMed  PubMed Central  Google Scholar 

  3. DiMatteo MR (2004) Variations in patients’ adherence to medical recommendations: a quantitative review of 50 years of research. Med Care Mar 42(3):200–209. https://doi.org/10.1097/01.mlr.0000114908.90348.f9

    Article  Google Scholar 

  4. Jüngst C, Gräber S, Simons S, Wedemeyer H, Lammert F (2019) Medication adherence among patients with chronic diseases: a survey-based study in pharmacies. QJM: Int J Med 112(7):505–512. https://doi.org/10.1093/qjmed/hcz058

    Article  Google Scholar 

  5. Brown MT, Bussell JK (2011) Medication Adherence: WHO Cares? Mayo Clin Proc;86(4):304–314. https://doi.org/10.4065/mcp.2010.0575

  6. Cea-Calvo L, Marín-Jiménez I, de Toro J et al (2020) Different associations of intentional and non-intentional non-adherence behaviors with patient experience with Healthcare and Patient beliefs in medications: a survey of patients with chronic conditions. Patient Prefer Adherence 14:2439–2450. https://doi.org/10.2147/ppa.S281985

    Article  PubMed  PubMed Central  Google Scholar 

  7. Brown MT, Bussell JK (2011) Apr. Medication adherence: WHO cares? Mayo Clin Proc 86(4):304– 14. https://doi.org/10.4065/mcp.2010.0575

  8. Lam WY, Fresco P (2015) Medication adherence measures: an overview. Biomed Res Int 2015:217047. https://doi.org/10.1155/2015/217047

    Article  PubMed  PubMed Central  Google Scholar 

  9. Lomper K, Chabowski M, Chudiak A, Białoszewski A, Dudek K, Jankowska-Polańska B (2018) Psychometric evaluation of the Polish version of the adherence to refills and medications Scale (ARMS) in adults with hypertension. Patient Prefer Adherence 12:2661–2670. https://doi.org/10.2147/ppa.S185305

    Article  PubMed  PubMed Central  Google Scholar 

  10. Kripalani S, Risser J, Gatti ME, Jacobson TA (2009) Development and evaluation of the adherence to refills and medications Scale (ARMS) among low-literacy patients with chronic disease. Value Health Jan-Feb 12(1):118–123. https://doi.org/10.1111/j.1524-4733.2008.00400.x

    Article  Google Scholar 

  11. Thompson K, Kulkarni J, Sergejew AA (2000) Reliability and validity of a new Medication Adherence Rating Scale (MARS) for the psychoses. Schizophr Res May 5(3):241–247. https://doi.org/10.1016/s0920-9964(99)00130-9

    Article  Google Scholar 

  12. Morisky DE, Green LW, Levine DM (1986) Concurrent and predictive validity of a self-reported measure of medication adherence. Med Care Jan 24(1):67–74. https://doi.org/10.1097/00005650-198601000-00007

    Article  CAS  Google Scholar 

  13. Culig J, Leppée M (2014) From Morisky to Hill-bone; self-reports scales for measuring adherence to medication. Coll Antropol Mar 38(1):55–62

    Google Scholar 

  14. Patrick DL, Burke LB, Gwaltney CJ et al (2011) Content validity—establishing and reporting the evidence in newly developed patient-reported outcomes (PRO) instruments for medical product evaluation: ISPOR PRO good research practices task force report: part 1—eliciting concepts for a new PRO instrument. Value Health 14(8):967–977

    Article  PubMed  Google Scholar 

  15. Committee for Medicinal Products for Human Use (2005) Reflection paper on the regulatory guidance for the use of health-related quality of life (HRQL) measures in the evaluation of medicinal products. Eur Med Agency, London

  16. FDA (2019) Patient-focused drug development: methods to identify what is important to patients guidance for industry, food and drug administration staff, and other stakeholders [DRAFT]. US Food and Drug Administration

  17. FDA (2020) Patient-focused drug development: collecting comprehensive and representative input; guidance for industry, food and drug administration staff, and other stakeholders. US Food and Drug Administration

  18. FDA (2012) Guidance for Industry. Patient-Reported Outcome Measures: Use in Medical Product Development to Support Labeling Claims

  19. Willis GB (2004) Cognitive interviewing: a tool for improving questionnaire design. sage

  20. Morgan L, Exall E, Bentley S et al (2020) AD2 qualitative interviews to explore drivers and Behaviours Associated with Medication NON-Adherence across a range of diseases, Treatment modalities, and Countries. Value Health 23:S399

    Article  Google Scholar 

  21. Bentley S, Morgan L, Exall E et al (2022) Qualitative interviews to Support Development and Cognitive Debriefing of the Adelphi Adherence Questionnaire (ADAQ©): a patient-reported measure of Medication Adherence developed for Use in a range of diseases, Treatment modalities, and Countries. Patient Prefer Adherence 16:2579–2592. https://doi.org/10.2147/ppa.S358046

    Article  PubMed  PubMed Central  Google Scholar 

  22. Bentley S, Exall E, Morgan L et al Improving Assessment of Adherence behaviors and drivers: targeted Literature Review and Concept Elicitation interviews in multiple countries and Disease populations. Patient Prefer Adherence. 2024/12/31 2024;18(null):1231–1242. https://doi.org/10.2147/PPA.S433662

  23. United States Bone and Joint Initiative. The Burden of Musculoskeletal Diseases in the United States (BMUS). In: Fourth ed. Rosemont, IL (2018): https://www.boneandjointburden.org/fourth-edition. Accessed 16 Mar 2023

  24. Hunter DJ, Schofield D, Callander E (2014) The individual and socioeconomic impact of osteoarthritis. Nat Rev Rheumatol 2014;10(7):437–441. https://doi.org/10.1038/nrrheum.2014.44

  25. Park KK, Choi CH, Ha CW, Lee MC (2016) The effects of Adherence to non-steroidal anti-inflammatory drugs and factors influencing drug adherence in patients with knee osteoarthritis. J Korean Med Sci May 31(5):795–800. https://doi.org/10.3346/jkms.2016.31.5.795

    Article  CAS  Google Scholar 

  26. Anderson P, Benford M, Harris N, Karavali M, Piercy J (2008) Real-world physician and patient behaviour across countries: Disease-Specific programmes - a means to understand. Curr Med Res Opin Nov 24(11):3063–3072. https://doi.org/10.1185/03007990802457040

    Article  CAS  Google Scholar 

  27. Anderson P, Higgins V, Courcy J et al (2023) Real-world evidence generation from patients, their caregivers and physicians supporting clinical, regulatory and guideline decisions: an update on Disease Specific Programmes. Curr Med Res Opin Dec 39(12):1707–1715. https://doi.org/10.1080/03007995.2023.2279679

    Article  Google Scholar 

  28. Babineaux SM, Curtis B, Holbrook T, Milligan G, Piercy J (2016) Evidence for validity of a national physician and patient-reported, cross-sectional survey in China and UK: the Disease Specific Programme. BMJ Open 6(8):e010352. https://doi.org/10.1136/bmjopen-2015-010352

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  29. Higgins V, Piercy J, Roughley A et al (2016) Trends in medication use in patients with type 2 diabetes mellitus: a long-term view of real-world treatment between 2000 and 2015. Diabetes Metab Syndr Obes 9:371–380. https://doi.org/10.2147/dmso.S120101

    Article  PubMed  PubMed Central  Google Scholar 

  30. Bellamy N, Buchanan WW, Goldsmith CH, Campbell J, Stitt LW (1988) Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to antirheumatic drug therapy in patients with osteoarthritis of the hip or knee. J Rheumatol Dec 15(12):1833–1840

    CAS  Google Scholar 

  31. EuroQol Group (1990) EuroQol–a new facility for the measurement of health-related quality of life. Health Policy 16(3):199–208. https://doi.org/10.1016/0168-8510(90)90421-9

    Article  Google Scholar 

  32. Voils CI, Maciejewski ML, Hoyle RH et al (2012) Initial validation of a self-report measure of the extent of and reasons for medication nonadherence. Med Care Dec 50(12):1013–1019. https://doi.org/10.1097/MLR.0b013e318269e121

    Article  Google Scholar 

  33. Voils CI, Hoyle RH, Thorpe CT, Maciejewski ML, Yancy WS Jr (2011) Improving the measurement of self-reported medication nonadherence. J Clin Epidemiol Mar 64(3):250–254. https://doi.org/10.1016/j.jclinepi.2010.07.014

    Article  Google Scholar 

  34. Bollen K, Lennox R (1991) Conventional wisdom on measurement: a structural equation perspective. Psychol Bull 110:305–314

  35. Kline RB (2005) Principles and practice of structural equation modeling, 2nd ed. Principles and practice of structural equation modeling, 2nd ed. New York, NY, US: Guilford Press; p. xviii, 366-xviii, 366

  36. Hu L-t, Bentler PM (1999) Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct Equ Model 6:1–55. https://doi.org/10.1080/10705519909540118

    Article  Google Scholar 

  37. Shi D, Maydeu-Olivares A, Rosseel Y (2020) Assessing Fit in Ordinal Factor Analysis Models: SRMR vs. RMSEA. Struct Equ Modeling 27(1):1–15. https://doi.org/10.1080/10705511.2019.1611434

  38. Stochl J, Fried EI, Fritz J et al (2022) On dimensionality, Measurement Invariance, and suitability of Sum scores for the PHQ-9 and the GAD-7. Assess Apr 29(3):355–366. https://doi.org/10.1177/1073191120976863

    Article  Google Scholar 

  39. Stochl J, Jones PB, Croudace TJ (2012) Mokken scale analysis of mental health and well-being questionnaire item responses: a non-parametric IRT method in empirical research for applied health researchers. BMC Med Res Methodol Jun 11:12:74. https://doi.org/10.1186/1471-2288-12-74

    Article  Google Scholar 

  40. Christensen AP, Garrido LE, Guerra-Peña K, Golino H (2020) Comparing community detection algorithms in psychological data: a Monte Carlo simulation. https://doi.org/10.31234/osf.io/hz89e

  41. Friedman J, Hastie T, Tibshirani R (2008) Sparse inverse covariance estimation with the graphical lasso. Biostatistics 9(3):432–441. https://doi.org/10.1093/biostatistics/kxm045

    Article  PubMed  Google Scholar 

  42. Golino HF, Epskamp S (2017) Exploratory graph analysis: a new approach for estimating the number of dimensions in psychological research. PLoS ONE 12(6):e0174035. https://doi.org/10.1371/journal.pone.0174035

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  43. Reise SP, Moore TM, Haviland MG (2010) Bifactor models and rotations: exploring the extent to which multidimensional data yield univocal scale scores. J Pers Assess Nov 92(6):544–559. https://doi.org/10.1080/00223891.2010.496477

    Article  Google Scholar 

  44. McDonald RP (1999) Test theory: a unified treatment. Lawrence Erlbaum Associates Publishers, vol xi. p 485

  45. Steiger JH (1980) Tests for comparing elements of a correlation matrix. Psychol Bull 87(2):245–251

  46. Revicki D, Hays RD, Cella D, Sloan J (2008) Recommended methods for determining responsiveness and minimally important differences for patient-reported outcomes. J Clin Epidemiol Feb 61(2):102–109. https://doi.org/10.1016/j.jclinepi.2007.03.012

    Article  Google Scholar 

  47. Mamolo C, Bushmakin A, Cappelleri J (2014) Application of the itch severity score in patients with moderate-to-severe plaque psoriasis: clinically important difference and responder analyses*. J Dermatolog Treat 04/10(26). https://doi.org/10.3109/09546634.2014.906033

  48. Trigg A, Griffiths P (2021) Triangulation of multiple meaningful change thresholds for patient-reported outcome scores. Qual Life Res Oct 30(10):2755–2764. https://doi.org/10.1007/s11136-021-02957-4

    Article  Google Scholar 

  49. Revelle W (2021) psych: Procedures for Personality and Psychological Research. Northwestern University, Evanston, Illinois, USA. https://CRAN.R-project.org/package=psych Version = 2.1.9

  50. Rosseel Y (2012) Lavaan: an R Package for Structural equation modeling. J Stat Softw 48(2):1–36. https://www.jstatsoft.org/v48/i02/

    Article  Google Scholar 

  51. Philip Chalmers R (2012) Mirt: a Multidimensional Item Response Theory Package for the R environment. J Stat Softw 1–29. https://doi.org/10.18637/jss.v048.i06

  52. Hallquist MN, Wiley JF (2018) MplusAutomation: an R Package for Facilitating large-scale latent variable analyses in Mplus. Struct Equ Model 25(4):621–638. https://doi.org/10.1080/10705511.2017.1402334

    Article  Google Scholar 

  53. Dueber D (2021) BifactorIndicesCalculator Bifactor Indices Calculator. R package version 0.2.2. https://CRAN.R-project.org/package=BifactorIndicesCalculator

  54. Golino H, Christensen AP (2021) EGAnet: exploratory graph analysis– a framework for estimating the number of dimensions in multivariate data using network psychometrics. R package version 0.9.9

  55. van der Ark LA (2012) New Developments in Mokken Scale Analysis in R. J Stat Softw 05/24 48(5):1–27. https://doi.org/10.18637/jss.v048.i05

    Article  Google Scholar 

  56. Suda T, Traxler A (2020) concorR: CONCOR and Supplemental Functions. R package version 0.2.1, https://github.com/ATraxLab/concorR

  57. Streiner DL, Norman GR, Cairney J (2015) Health measurement scales: a practical guide to their development and use, 5th ed, vol xiii. Oxford University Press, p 399

  58. Widaman KF, Revelle W (2024/08/01 2023) Thinking About Sum Scores Yet Again, Maybe the Last Time, We Don’t Know, Oh No.: A Comment on. Educ Psychol Meas 84(4):637–659. https://doi.org/10.1177/00131644231205310

  59. McNeish D, Wolf MG (2020) Thinking twice about sum scores. Behav Res Methods Dec 52(6):2287–2305. https://doi.org/10.3758/s13428-020-01398-0

    Article  Google Scholar 

  60. McNeish D, Wolf MG (2020) Corrigendum to: Thinking twice about sum scores. Behav Res Methods 52(6):2674–2674. https://doi.org/10.3758/s13428-020-01468-3

  61. McNeish D (2023) Psychometric properties of sum scores and factor scores differ even when their correlation is 0.98: a response to Widaman and Revelle. Behav Res Methods Dec 55(8):4269–4290. https://doi.org/10.3758/s13428-022-02016-x

    Article  Google Scholar 

  62. Widaman KF, Revelle W (2023) Thinking thrice about sum scores, and then some more about measurement and analysis. Behav Res Methods 2023;55(2):788–806. https://doi.org/10.3758/s13428-022-01849-w

  63. Barbosa CD, Balp MM, Kulich K, Germain N, Rofail D (2012) A literature review to explore the link between treatment satisfaction and adherence, compliance, and persistence. Patient Prefer Adherence 6:39–48. https://doi.org/10.2147/ppa.S24752

    Article  PubMed  PubMed Central  Google Scholar 

  64. Robinson RL, Schnitzer TJ, Barlow S et al (2022) Satisfaction with medications prescribed for Osteoarthritis: a cross-sectional survey of patients and their Physicians in the United States. Pain Ther Mar 11(1):191–208. https://doi.org/10.1007/s40122-021-00350-0

    Article  Google Scholar 

  65. Sidorkiewicz S, Tran VT, Cousyn C, Perrodeau E, Ravaud P (Sep 2016) Discordance between drug adherence as reported by patients and drug importance as assessed by Physicians. Ann Fam Med 14(5):415–421. https://doi.org/10.1370/afm.1965

  66. Stirratt MJ, Dunbar-Jacob J, Crane HM et al (2015) Self-report measures of medication adherence behavior: recommendations on optimal use. Transl Behav Med Dec 5(4):470–482. https://doi.org/10.1007/s13142-015-0315-2

    Article  Google Scholar 

  67. Lee W-Y, Ahn J, Kim J-H et al (2013) Reliability and validity of a self-reported measure of medication adherence in patients with type 2 diabetes mellitus in Korea. J Int Med Res 41(4):1098–1110

    Article  PubMed  Google Scholar 

  68. Schepman P, Thakkar S, Robinson R, Malhotra D, Emir B, Beck C (2021) Moderate to severe osteoarthritis pain and its impact on patients in the United States: a National Survey. J Pain Res 14:2313–2326. https://doi.org/10.2147/JPR.S310368

  69. European Pharmaceutical Market Research Association (EphMRA). Code of Conduct. Accessed 21 Mar 2022. https://www.ephmra.org/media/4857/ephmra-2020-code-of-conduct-final.pdf

  70. US Department of Health and Human Services. Summary of the HIPAA Privacy Rule. Accessed 21 Mar 2023. http://www.hhs.gov/sites/default/files/privacysummary.pdf

  71. Health Information Technology (HITECH). Health Information Technology Act. Accessed 21 Mar 2023. https://www.healthit.gov/sites/default/files/hitech_act_excerpt_from_arra_with_index.pdf

  72. World Health Association (2013) World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects. JAMA 310(20):2191–2194

Download references

Acknowledgements

Medical writing and editorial assistance were provided by Daniel Green, PhD, of Adelphi Real World (Bollington, United Kingdom).

Funding

Data collection was undertaken by Adelphi Real World as part of an independent survey, entitled the Adelphi Osteoarthritis Disease Specific Programme™. The Disease Specific Programme is a wholly owned Adelphi product. This analysis was funded by Adelphi Real World. ADAQ development was funded by Adelphi Real World.

Author information

Authors and Affiliations

  1. Adelphi Values, Adelphi Mill, Grimshaw Lane, Bollington, Macclesfield, SK10 5JB, UK

    Nathan Clarke, Andrew Trigg, Rob Arbuckle, Jan Stochl & Sarah Bentley

  2. Department of Kinanthropology, Charles University, Prague, Czechia

    Jan Stochl

  3. Adelphi Real World, Bollington, UK

    Victoria Higgins & James Piercy

Authors
  1. Nathan Clarke
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Andrew Trigg
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Rob Arbuckle
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Jan Stochl
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Victoria Higgins
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Sarah Bentley
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. James Piercy
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

NC, VH, SB, AT, and RA were involved in the conceptualization and design of this analysis, analysis and interpretation of data, and critical revision of the paper. AT was also involved in the initial drafting of the manuscript. JP was involved in data acquisition and critical revision of the manuscript. JS was involved in analysis and interpretation of data and critical revision of the manuscript.

Corresponding author

Correspondence to Nathan Clarke.

Ethics declarations

Ethics approval and consent to participate

All patients provided informed consent prior to data collection. Ethical approval was provided by the Western Institutional Review Board (protocol number: 20203287). Data collection was undertaken in line with European Pharmaceutical Marketing Research Association guidelines [69]. Each survey was performed in full accordance with relevant legislation at the time of data collection, including the US Health Insurance Portability and Accountability Act 1996 [70], Health Information Technology for Economic and Clinical Health Act legislation [71], and the Declaration of Helsinki [72]. No personally identifiable data were collected. Physician and patient data were pseudonymized by assigning a code to each response, which allowed for matching of patient self-complete and patient record forms.

Consent to publish

Not applicable as we are not publishing any personally identifiable data.

Competing interests

NC, RA and SB are employees of Adelphi Values. Adelphi Values conducted the study analyses. Adelphi Values were paid consultants to Adelphi Real World in connection with the development of this manuscript. AT was an employee of Adelphi Values at the time of study. JS is paid as a consultant to Adelphi Values. VH, JP are employees of Adelphi Real World who funded the development and validation of the Adelphi Adherence Questionnaire.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary Material 1

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Clarke, N., Trigg, A., Arbuckle, R. et al. Psychometric evaluation of the Adelphi Adherence Questionnaire (ADAQ©) in adults with osteoarthritis. J Patient Rep Outcomes 8, 118 (2024). https://doi.org/10.1186/s41687-024-00789-7

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s41687-024-00789-7

Keywords