Abstract
Objective Medication nonadherence is common in patients with systemic lupus erythematosus (SLE) and negatively affects outcomes. To better recognize and address nonadherence in this population, there is a need for an easily implementable tool with interpretable scores. Domains of Subjective Extent of Nonadherence (DOSE-Nonadherence) is a measure that captures both extent of and reasons for nonadherence. We refined and evaluated DOSE-Nonadherence for patients with SLE.
Methods We refined the reasons for the nonadherence domain of DOSE-Nonadherence through rheumatologist feedback and patient cognitive interviewing. We then administered the refined instrument to patients prescribed oral SLE medications and compared the results to the Beliefs About Medicines Questionnaire (BMQ), the Medication Adherence Self-Report Inventory (MASRI), medication possession ratios (MPRs), and hydroxychloroquine (HCQ) blood levels using Pearson correlations.
Results Five rheumatologists provided feedback; 16 patients (median age 43 yrs, 100% female, 50% Black) participated in cognitive interviews and 128 (median age 49 yrs, 95% female, 49% Black, 88% on antimalarials, and 59% on immunosuppressants) completed the refined instrument. Items assessing extent of nonadherence produced reliable scores (α 0.89) and identified 47% as nonadherent. They showed convergent validity with MASRI (r = −0.57), HCQ blood levels (r = −0.55), to a lesser extent MPRs (r = −0.34 to −0.40), and discriminant validity with BMQ domains (r = −0.27 to 0.32). Nonadherent patients reported on average 3.5 adherence barriers, the most common being busyness/forgetting (62%), physical fatigue (38%), and pill fatigue (33%).
Conclusion Our results support the reliability and validity of DOSE-Nonadherence for SLE medications. This refined instrument, DOSE-Nonadherence-SLE, can be used to identify, rigorously study, and guide adherence intervention development in SLE.
Systemic lupus erythematosus (SLE) management requires long-term use of immune-altering medications to prevent disease progression and damage. However, nonadherence is as high as 75% and is associated with increased morbidity and mortality.1-4 Few studies have investigated causes of medication nonadherence in SLE. A 2017 systematic review of SLE adherence literature identified 8 studies conducted in the United States, among which only one examined determinants of nonadherence, finding that depression and polypharmacy were associated with lower adherence.1 Measuring the extent of and identifying modifiable reasons for nonadherence are crucial steps in developing interventions that optimize adherence to lifesaving SLE medications.
Although various methods are used to assess the extent of nonadherence, reasons for nonadherence can be only self-reported. A self-reported measure has the added benefit of providing the necessary information to the clinician at the point of care. To our knowledge, there are no easily implemented self-reported tools to date that assess both the extent of and reasons for nonadherence among patients with SLE.
Previously, a 2-domain, patient-reported instrument had been developed to measure both the extent of and reasons for nonadherence in hypertension, hyperlipidemia, and hepatitis C.5-7 We used this existing measure, Domains of Subjective Extent of Nonadherence (DOSE-Nonadherence), as the starting point to build a SLE-appropriate adherence assessment tool (DOSE-Nonadherence-SLE), incorporating concepts of adherence barriers we discovered from literature review and our prior qualitative results.8-14 We conducted a 2-phase study, first refining the measure specifically for adults diagnosed with SLE, and then quantitatively evaluating its reliability and validity. The purpose of DOSE-Nonadherence-SLE is to determine both the extent of and reasons for nonadherence in patients with SLE.
METHODS
Study setting and population. The Duke Lupus Registry (DLR) is a prospective cohort of patients with SLE followed at the Duke Lupus Clinic. Inclusion criteria for the DLR are age ≥ 18 years, English fluency, absence of cognitive or physical barriers to provide informed consent, and meeting the American College of Rheumatology 1997 or SLE International Collaborating Clinics 2012 SLE classification criteria.15,16 All enrolled subjects provided signed informed consent to participate in research and are followed regularly as clinically indicated. The DLR has been approved by the Duke University institutional review board (study #Pro00008875). The current study was approved as a substudy of the DLR (study #Pro00094645).
Qualitative evaluation and refinement. DOSE-Nonadherence is based on a 7-day recall period. The extent of nonadherence domain contains 3 items (I missed my medicine, I skipped a dose of my medicine, and I did not take a dose of my medicine) on a 5-point Likert scale that are scored by computing their mean, with higher scores indicating greater nonadherence.5 Patients who endorse nonadherence on any of the extent of nonadherence items (score > 1) then go on to complete the reasons for nonadherence domain. The extent of nonadherence items were not meant to be modified as they were designed to be disease-agnostic and have been extensively evaluated through cognitive interviews.5-7 In contrast, the reasons for nonadherence domain was designed to be tailored to specific diseases and medication regimens, and each reason stands alone as a descriptor. Therefore, we focused our efforts on refining the items in this domain.
The first author (KS) formulated a preliminary version of the reasons for nonadherence by augmenting the existing DOSE-Nonadherence items with concepts discovered from literature review and our prior qualitative study, and removing concepts that were not relevant to patients with SLE.8-14 These concepts and preliminary items were reviewed with a standing SLE advisory group generally attended by 2 to 8 patients, 2 to 3 clinicians, and 2 clinic staff, who selected the most salient items (Supplementary Table S1, available with the online version of this article).
The resulting draft questionnaire was reviewed by Duke Lupus Clinic providers (JLR, LGCS, RES, JD, MEBC) who commented on whether the reasons for nonadherence items were comprehensive in addressing nonadherence issues encountered in clinical practice.
Next, we conducted 2 rounds of semistructured individual cognitive interviews with 16 patients in the DLR to evaluate and refine the reasons for nonadherence items, revising them between rounds.17 We purposefully sampled patients prescribed SLE medications for at least 6 months and included 50% Black patients. Common practice and the literature indicated that relatively little new information is learned with > 12 interviews within a relatively homogenous sample of interview participants.18-20 To account for diversity of experiences, we added 4 interviews, yielding a total of 16.
Interviews were conducted via Zoom by 2 interviewers (KS, DRA) using a semistructured interview guide. During the interviews, participants first completed the self-administered DOSE-Nonadherence-SLE online using Research Electronic Data Capture (REDCap).21 The interviewers then shared their screen and reviewed the participant’s response processes and answers together, retrospectively.17 Patients reporting perfect adherence on the extent of nonadherence items were asked to consider hypothetical reasons for missing doses. We used think-aloud and verbal probing techniques to ascertain how an item was interpreted and a response formed. Patients were also encouraged to comment on questionnaire structure, item stem, and missing or redundant concepts. This was to ensure that the instrument comprehensively and parsimoniously identified the most important barriers patients face in taking SLE medications, and that respondents’ understanding of items was consistent with their intended meanings. Patients were not asked to comment on the instructions, 7-day recall period, response scale, or formatting, all of which had been extensively evaluated and were not meant to be modified.5 Interviews were recorded and took approximately 30 minutes each. The interviewers took detailed structured notes to systematically track issues that arose for each item and added, removed, or edited items in response to feedback between rounds. The rationales for revising items were also documented. Matrix analysis was performed to summarize information across interviews.
Quantitative evaluation. After qualitative refinement, we assessed DOSE-Nonadherence-SLE by administering it to patients in the DLR who were prescribed oral medications for SLE. This phase of the study focused on evaluating the reliability and validity of the extent of nonadherence items in patients with SLE.
· Data collection. Cross-sectional data were obtained through questionnaire and electronic medical record (EMR) review. As a result of coronavirus disease 2019 (COVID-19) restrictions on research, questionnaires were initially sent as an email link to all 362 patients in the DLR, and later administered in person during clinic visits to patients who did not respond to the initial email, until the sample size goal was reached. Clinical information was extracted from the most recent visit for patients who completed questionnaires via email, and from the same day encounter for patients who completed questionnaires in person. Self-reported sociodemographic information including race and ethnicity was collected. SLE medications prescribed were obtained through chart review. In addition to the refined DOSE-Nonadherence-SLE, the following instruments and information were obtained.
1. Existing measures of adherence. To establish convergent validity of the extent of nonadherence domain, or the degree to which its score correlates with other instruments meant to measure the same construct, we assessed related constructs. The Medication Adherence Self-Report Inventory (MASRI) part A is a 6-item instrument that asks about the amount of medication taken in the past month and provides a numerical estimate of adherence from 0 to 100% on a visual analog scale. In patients with SLE, this measure showed internal consistency, reliability, and concurrent and predictive validity when compared to pharmacy refill data.22 We used a cut-off of 90% indicating adherence, given the known ceiling effect.22 Medication possession ratios (MPRs) for all SLE medications 3 months prior to the date of questionnaire completion were obtained by EMR review and supplemented by phone calls to pharmacies. MPR was calculated as the proportion of days covered by total days’ supply dispensed,23 and we used a cut-off of 80% indicating adherence.3,4,24 Available hydroxychloroquine (HCQ) blood levels measured using liquid chromatography coupled with mass spectrometry within 2 weeks of questionnaire completion were included.25,26 Therapeutic range is 500 ng/mL to 2000 ng/mL. Reduced levels have been associated with worse SLE disease activity, supporting its validity in measuring adherence.25-29
2. Medication beliefs. To establish discriminant validity, or the degree to which the scores diverge from instruments meant to measure different constructs, we administered the 18-item Beliefs About Medicines Questionnaire (BMQ), which measures beliefs about the necessity of prescribed medication (Specific-Necessity); beliefs about the danger of dependence, long-term toxicity, and disruptive effects of medication (Specific-Concerns); beliefs that medicines are harmful and/or addictive and should not be taken continuously (General-Harm); and beliefs that medicines are overused by doctors (General-Overuse).30,31 Scores range from 5 to 25 for Specific-Necessity and Specific-Concerns, and from 4 to 20 for General-Harm and General-Overuse, with higher scores indicating stronger beliefs (Table 1).
Comparison measures collected with DOSE-Nonadherence-SLE and their score interpretations.
Statistics. Descriptive statistics were performed. A score > 1 on any of the 3 extent of nonadherence items was considered to indicate nonadherence, and scores of the 3 items were averaged. We used Cronbach α to assess the reliability of the extent of nonadherence items and examined their relationships using inter-item correlations.32 We computed Pearson correlations between the extent of nonadherence item scores and comparison measures (MASRI, MPR, HCQ blood levels, and BMQ subscale scores) to examine convergent and discriminant construct validity. Correlation > 0.50 was considered evidence of convergent validity,6 which was expected between the extent of nonadherence scores and MASRI and HCQ blood levels. Given that taking and refilling medications are different adherence constructs and given the difference in the time period measured (7 vs 90 days),33 we expected a weaker correlation (0.3-0.5) between the extent of nonadherence scores and MPRs. We expected a weak correlation (< 0.30) between the extent of nonadherence scores and BMQ subscale scores as evidence of discriminant validity because medication beliefs may affect, but are distinct from, medication-taking behavior.34
A target sample size of 123 provides 80% power and 5% α error to detect the significance of a correlation of 0.25, which estimates the correlation strength for discriminant validity. This sample size is consistent with ones previously shown to be sufficient in validating patient-reported measures in SLE.35-38 Statistical analyses were performed using Stata (version 14.2; StataCorp).
RESULTS
Qualitative evaluation and refinement. Reactions to the initial draft of DOSE-Nonadherence-SLE from the 5 Duke Lupus Clinic providers (mean age 43 yrs, 80% female, 80% White, average 7 yrs in practice) were overall positive. Providers queried whether items should be combined to reduce response burden or be asked as separate questions. The items, “I forgot” or “I was busy,” “I was physically too tired” or “I was too stiff,” and “I felt too depressed” or “overwhelmed” were queried in particular. These questions were explored in the cognitive interviews with patients.
Among the 16 patients who participated in cognitive interviews, the median age was 43 years, all were female, 50% were Black, and 31% reported perfect adherence. Supplementary Table S2 (available with the online version of this article) shows concepts covered in the original DOSE-Nonadherence questionnaire, new concepts incorporated that are relevant for patients with SLE, and actions taken based on feedback from the advisory group and cognitive interviews.
During cognitive interviews, most patients indicated that being “too stiff” and “too tired” to take medications were different concepts; therefore, these became separate items. One patient indicated that “I still felt so bad I thought the medicine was not working” and “when I skip a dose, I don’t feel any difference” were redundant. Exploring this further, several patients felt that they were different, and therefore these remained separate items.
Most participants indicated that “I forgot” and “I was busy” go hand in hand and required similar types of intervention, and therefore these were combined into 1 descriptor.
Most participants indicated that being “depressed” and “overwhelmed” were distinct ideas, but a sense of being overwhelmed is reflected by being “tired of taking medicines.” Therefore, “overwhelmed” was removed.
In the first round of interviews, 1 participant suggested adding, “it is hard to open pill bottles” and “I had no one to help me.” During the second round of interviews, most participants indicated that “hard to open bottles” is a concept covered by “too stiff” and “I had no one to help me,” so it was not included.
One participant suggested adding an open-ended question to identify any additional barriers, and several agreed. Therefore, we added a final free-response question. Another participant suggested having patients mark their top 3 barriers. Several participants agreed because it would show that the doctor cares and allows patients to prioritize which barriers to address. Therefore, we added the instruction, “Of the reasons listed above, please mark 3 that you feel are the biggest challenges for you.”
All participants understood all items as intended and indicated that all items were relevant to patients with SLE. All except for one indicated that the questionnaire was the appropriate length. Results became redundant by approximately 14 interviews. The resulting SLE-specific measure containing 25 reasons for nonadherence was used in the quantitative evaluation (Supplementary Table S3, available with the online version of this article).
Quantitative evaluation. Among 128 participants enrolled, 85% completed questionnaires by email and 15% completed them in person. Median age was 49 years, 95% were female, 49% were Black, and 40% had Medicaid/Medicare insurance. On average, patients completed the questionnaire 42 days from their most recent clinic visit. Median disease duration was 14 years. Patients were on average prescribed 2 SLE medications: 88% were using antimalarials, 59% were using synthetic disease modifying anti-rheumatic drugs (DMARD; 30% mycophenolate mofetil, 15% azathioprine, 13% methotrexate, and 2% leflunomide), and 7% were using belimumab (Table 2).
Sociodemographic and clinical characteristics of survey participants (n = 128).
· Extent of nonadherence. The 3 items assessing extent of nonadherence produced reliable scores (α 0.89) and had inter-item correlations between 0.83 and 0.96. These items classified 47% of patients as nonadherent. In comparison, 21% of patients were nonadherent based on the MASRI cut-off of 90%, and 43% based on the MPR cut-off of 80% (Table 3).
Rates of nonadherence and scores of comparison measures among survey participants.
· Reasons for nonadherence. Patients defined as nonadherent based on the extent of nonadherence items on average had 3.5 reasons for missing doses (IQR 1-5).
The most common reasons were “I forgot or I was busy” (n = 37, 62%), “I was too tired” (n = 23, 38%), and “I was tired of taking medicines every day” (n = 20, 33%). These 3 reasons were also most frequently chosen as most challenging to overcome, and together were present in 75% of nonadherent patients. Barriers with the highest scores, indicating greatest impact on adherence, were “I forgot or I was busy” (median 4.0), “I was too tired” (median 4.0), and “I do not have a regular schedule” (median 4.0; Table 4).
Reasons for nonadherence reported by 60 patients who were considered nonadherent based on the extent of nonadherence items.
The least common reasons for missing doses were “I heard someone had a bad experience taking them” (n = 1, 2%), “my family or friends suggested I not take them” (n = 2, 3%), and “the medication instructions were hard to follow” (n = 2, 3%).
Six patients provided an answer to the final free-response question. One respondent reported that they forgot, and one each reported reflux, headache, could not keep anything down, stomach pain, and felt ill due to other reasons.
· Construct validity. As expected, the extent of nonadherence scores was most strongly correlated with MASRI (r = −0.57, n = 118) and HCQ blood levels (r = −0.55, n = 14), moderately correlated with antimalarial MPR (r = −0.39, n = 113) and DMARD MPR (r = −0.40, n = 76), and weakly correlated with BMQ subscales (r = −0.27 to 0.32, n = 128). In comparison, the MASRI was moderately correlated with antimalarial MPR (r = 0.34) and weakly correlated with BMQ Specific-Necessity (r = 0.23) and Specific-Concerns (r = −0.25) but was not statistically significantly correlated with other comparison measures (Table 5).
Evaluating construct validity of extent of nonadherence items by correlating with other adherence measures and comparing with MASRI.
DISCUSSION
We developed, refined, and psychometrically tested DOSE-Nonadherence-SLE, the first self-reported instrument, to our knowledge, that assesses both the extent of and reasons for nonadherence in patients with SLE. By incorporating feedback from SLE clinic stakeholders in the questionnaire refinement process, the SLE-specific reasons for nonadherence items were perceived to cover the concepts they purport to measure (ie, exhibited face validity). Quantitative evaluation of the disease-agnostic extent of nonadherence items showed that the items produced consistent results and measure what they intend to measure in the SLE population, supporting their reliability and construct validity.
Although self-reported measures are known to underestimate nonadherence, our instrument identified more nonadherent patients than the MASRI, suggesting greater sensitivity. It also performed better than the MASRI in psychometric testing when compared to HCQ blood levels and pharmacy refills, supporting its use in patients with SLE. Further, it has an advantage over existing measures used in rheumatic diseases, including the MASRI and MPR, which only assess degree of nonadherence without considering barriers.
Our study addresses a dearth in the literature in quantifying barriers to adherence among patients with SLE. Hardy et al also assessed adherence barriers in patients with SLE, but used a questionnaire developed in cardiovascular disease prevention that was not adapted or validated for SLE.39 Common adherence barriers discovered in the study overlapped with our findings, including busyness or forgetting and concerns about side effects. Additionally, a common barrier found by Hardy et al was “having personal reasons for not taking medications,” which can be interpreted broadly and thus has unclear implications. In contrast, our reasons for nonadherence items, having been refined for SLE, are more concrete and specific, and therefore more actionable.
DOSE-Nonadherence-SLE could be valuable in both clinical and research settings. Clinically, it offers a structured approach to ascertaining the degree of and barriers to adherence at the point of care, which can guide individualized adherence plans and improve patient-centered care. In research, it can be used to study adherence barriers on a population level, such as identifying racial differences in barriers, as patients from racial and ethnic minority backgrounds are more likely to be nonadherent.40 It can also be used to identify the most impactful adherence barriers and serve as an intermediate outcome measure in adherence intervention trials. For example, in our cohort, the 3 most common and most challenging barriers reported by patients with SLE were busyness or forgetting, being physically too tired, and being tired of taking medications. The prevalence and impact of these barriers make them attractive targets on which to focus intervention efforts, which may include use of reminders, social support, and simplifying the medication regimen.
Our results suggest possible changes that can be made to the reasons for nonadherence items as we work to improve this measure. For example, only 1 to 2 nonadherent patients reported “I heard someone had a bad experience taking [SLE medications],” “my family or friends suggested I not take [SLE medications],” and “the medication instructions were hard to follow.” All patients except for one who reported “I do not have a regular schedule” also reported “I forgot or I was busy,” suggesting that the former may be redundant. These items can be deleted to reduce response burden if larger studies confirm our findings. Considering the trade-off between comprehensiveness and response burden, we suggest retaining the free-response question to determine rarer barriers, particularly if those aforementioned items were removed. Additionally, based on the responses to the free-response question, we decided to retain the item “I missed my lupus pills because I was feeling too sick.” We also suggest retaining the instruction for patients to mark the most challenging adherence barriers. The ability to focus on the most impactful barriers would increase the clinical utility of this questionnaire, particularly in resource-limited settings or when patients report similar scores on multiple adherence barriers.
Our study has several limitations. First, our questionnaire assesses adherence in the past 7 days, which may not represent patients’ behavior over longer periods and may be more strongly influenced by the white-coat adherence effect, where adherence improves immediately surrounding an appointment. This recall period was chosen after previous extensive cognitive interviews showing that the extent of and reasons for nonadherence were more easily and accurately recalled in the last 7 days compared to 14 or 30 days.5,6 The measure could be administered repeatedly to obtain information over longer periods. Second, a small number of the patients who underwent cognitive interviews reported perfect adherence and considered hypothetical reasons for missing doses. This could have minor effects on the results of the questionnaire refinement. Third, most patients completed surveys through email, and those patients may be more engaged with medical care and therefore more adherent than the average patient with SLE, possibly influencing the number or selection of reported adherence barriers, affecting the generalizability of our findings. Generalizability may also be affected by the patient population of our academic SLE clinic located in the Southeastern US. Future studies should consider its evaluation in and adaptation to other organizational and cultural contexts. Fourth, the number of HCQ blood levels taken within 2 weeks of the questionnaire administration was small, as blood work was typically performed during a clinic visit. However, we still detected significant correlation with the extent of nonadherence items. Although MPR and HCQ blood levels are objective measures of adherence, their use may be limited by availability and cost. Additionally, reasons for nonadherence covered in this questionnaire apply to oral medications, and barriers to taking self-injections can overlap with but are distinct from oral medications. However, this questionnaire is still relevant, as only a small percentage of patients with SLE take self-injections, and almost all of them also take oral medications. Last, we were not able to meaningfully assess validity correlations among reasons for nonadherence to identify redundant items or to correlate those reasons with other measures, owing to small numbers of patients reporting each adherence barrier.
In conclusion, our results support the reliability and validity of DOSE-Nonadherence-SLE in assessing the extent of and reasons for nonadherence among patients with SLE. In comparison to the MASRI and MPR, DOSE-Nonadherence-SLE offers a structured approach to recognizing adherence barriers. As such, this instrument can be used to identify and rigorously study the most common reasons for nonadherence in SLE and guide intervention development to target them.
ACKNOWLEDGMENT
The DOSE-Nonadherence questionnaire is copyrighted by Duke University.
Footnotes
Research reported in this publication was supported by the Duke Center for Research to Advance Healthcare Equity (REACH Equity), which is supported by the National Institute on Minority Health and Health Disparities (U54MD012530) and National Center for Advancing Translational Sciences (NCATS) of the National Institutes of Health (NIH 1KL2TR002554). TMC receives research funding from the US Food and Drug Administration (75F40119C10080, 75F40120C00069), Pfizer, Merck (Q-11859), NIH (5R01-CA249568-02), PCORI (HL-2019C1-16059), the Duke Endowment, and a private donation, and consults with Regenxbio. CIV is supported by a Research Career Scientist award (RCS 14-443) from the Health Services Research & Development service of the Department of Veterans Affairs (VA). AEE receives research funding from NIH NCATS (award no. 1KL2TR002554), Pfizer, and Exagen. JLR receives research funding from GSK, Pfizer, and Exagen, and consults for Eli Lilly, Immunovant, and Exagen. MEBC receives research funding from GSK, Exagen, and Pfizer, and consults for GSK and UCB. JD receives research funding from Pfizer. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH, the VA, or the US government.
TMC consults with Regenxbio. AME receives research funding from Pfizer and Exagen. JLR receives research funding from GSK, Pfizer, and Exagen; and consults for Eli Lilly, Immunovant, and Exagen. MEBC receives research funding from GSK, Exagen, and Pfizer, and consults for GSK and UCB. JD receives research funding from Pfizer. The remaining authors declare no conflicts of interest relevant to this article.
- Accepted for publication July 25, 2022.
- Copyright © 2022 by the Journal of Rheumatology
