Effectively measuring adherence to medications for systemic lupus erythematosus in a clinical setting.код для вставкиСкачать
Arthritis & Rheumatism (Arthritis Care & Research) Vol. 57, No. 6, August 15, 2007, pp. 1000 –1006 DOI 10.1002/art.22898 © 2007, American College of Rheumatology ORIGINAL ARTICLE Effectively Measuring Adherence to Medications for Systemic Lupus Erythematosus in a Clinical Setting SRI KONERU,1 MICHAEL SHISHOV,2 AVIS WARE,1 YOLANDA FARHEY,1 ANNE-BARBARA MONGEY,1 T. BRENT GRAHAM,2 MURRAY H. PASSO,2 J. LAWRENCE HOUK,1 GLORIA C. HIGGINS,3AND HERMINE I. BRUNNER2 Objective. To assess the reliability and concurrent validity of the Medication Adherence Self-report Inventory (MASRI) when used in systemic lupus erythematosus (SLE), to investigate the predictive validity of the MASRI using pharmacy reﬁll information as the criterion standard, and to propose a sensible approach to the screening for nonadherence in a clinical setting. Methods. Adherence to 2 medications (hydroxychloroquine and prednisone) was measured in 55 patients using the MASRI, pill counts, and physician ratings (MD scale). Adherence based on pharmacy reﬁll information served as a criterion standard with nonadherence deﬁned as adherence rates <80%. To determine test–rest reliability of the MASRI, 20 patients completed the measure twice within a 2-week period. Results. Using pharmacy information, 39% of the patients were nonadherent to prednisone and 51% to hydroxychloroquine. The MASRI had acceptable internal consistency (Cronbach’s ␣ 0.7) and good reliability. Irrespective of the drug assessed, MASRI ratings were moderately correlated with patient adherence (pharmacy), supporting the concurrent validity of the MASRI. The combination of adherence estimation by MD scale rating at <85% and by MASRI at <90% was 87% sensitive and 86% speciﬁc for identifying patients who were nonadherent to prednisone. These cutoff values also appeared suitable for identifying nonadherence to hydroxychloroquine. Conclusion. The MASRI is a reliable measure of adherence to medications in SLE. The measure has concurrent and predictive validity. When combined with the MD scale, the MASRI appears to be a useful screening tool for nonadherence in patients with SLE that could be suitable for clinical practice. KEY WORDS. SLE; Adherence; Compliance; Lupus; Adults; Adolescents; MASRI. Adherence to (or compliance with) medication regimens is often deﬁned as the extent to which patients take medications as prescribed by their health care providers (1). Rates of adherence for individual patients are usually reported as the percentage of the prescribed doses of medications that are actually taken by the patients over a speciﬁed period. Lack of adherence is a ubiquitous problem in the management of chronic diseases, including systemic lupus erythematosus (SLE) (1–3). Therefore, given the frequent occurrence of nonadherence and its potential negative effects on patient prognosis (4,5), effective and easy-to-use adherence measures are needed to identify patients at risk. Despite the importance of adherence to the success of medical interventions, there is no universally accepted criterion standard for measuring medication adherence. Various approaches to ascertain medication adherence are Supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases (grants P60-AR-47784, R03AR-049860, and U01-AR-51868) and the Columbus Medical Research Foundation. 1 Sri Koneru, MD, Avis Ware, MD, Yolanda Farhey, MD, Anne-Barbara Mongey, MD, J. Lawrence Houk, MD: University of Cincinnati Medical Center, Cincinnati, Ohio; 2Michael Shishov, MD, T. Brent Graham, MD, Murray H. Passo, MD, Hermine I. Brunner, MD, MSc: Cincinnati Children’s Hospital Medical Center, Cincinnati, Ohio; 3Gloria C. Hig- gins, MD, PhD: Columbus Children’s Hospital, and Ohio State University, Columbus. Address correspondence to Hermine I. Brunner, MD, MSc, Cincinnati Children’s Hospital Medical Center, University of Cincinnati, William Rowe Division of Rheumatology, E 4010, 3333 Burnet Avenue, Cincinnati, OH 452293039. E-mail: email@example.com. Submitted for publication July 13, 2006; accepted in revised form February 12, 2007. INTRODUCTION 1000 Medication Adherence in SLE currently in use but their implementation in daily clinical practice appears difﬁcult. For example, pill counts are cumbersome, while electronic medication event-monitoring systems are costly; calculating adherence rates based on pharmacy reﬁll data is time intensive, and pharmacologic testing of the exposure to rheumatic medications is often not available, making all of these adherence measures less suitable for use in routine clinical care. Physicians’ estimates of adherence, although readily available in the clinic, are thought to be prone to inaccuracy (6,7). We hypothesized that patient self-report may be an alternative feasible and useful approach to measuring adherence to medications in SLE. Self-report of medication adherence has been explored in the past. For example, Morisky et al (8) published a generic Medication-Taking Scale to predict blood pressure control, and more recently, the disease-speciﬁc Compliance-Questionnaire-Rheumatology (CQR) was developed and validated (9). However, neither the Medication-Taking Scale nor the CQR allow for the determination of speciﬁc medication adherence rates. Conversely, previous research suggests that the Medication Adherence Self-report Inventory (MASRI) is useful for accurately measuring adherence to human immunodeﬁciency virus (HIV) medications in both clinic settings and research (10). The MASRI is appealing for use in other patient populations because it is concise and readily provides a numeric estimate of patients’ adherence rates (0 – 100%). As an initial step toward the validation of the MASRI in rheumatology at large, we assessed its usefulness in patients with SLE (3). The goals of our study were 1) to assess the test–retest reliability and concurrent validity of the MASRI when used in patients diagnosed with SLE, 2) to assess the predictive validity of the MASRI using pharmacy reﬁll information as the criterion standard, and 3) to assess the usefulness of the MASRI alone or in combination with another measure of adherence to effectively screen for drug nonadherence in a clinical setting. PATIENTS AND METHODS Patients. A convenience sample of patients with SLE (3) was recruited from 4 rheumatology clinics afﬁliated with the University of Cincinnati in Cincinnati, Ohio, and Ohio State University in Columbus, Ohio. To be included in the study, patients had to be age ⱖ16 years, have a disease duration of at least 6 months, and be receiving medications for SLE at the time of the study. Patients with a diagnosis of neuropsychiatric SLE syndromes were excluded, as were those with other chronic diseases that might impact adherence as judged by the treating rheumatologist. The protocol was approved by the respective institutional review board of each participating clinic, and all patients signed informed consent documents. Survey procedures and study visit. Dedicated research personnel, rather than professionals participating in the clinical care delivery, supervised the participants in the completion of the study forms. The participants provided information on the doses and types of medications recom- 1001 mended by their rheumatologists. This information was compared with and supplemented with information available in the medical record. The study participants completed the MASRI independently and were assured that their responses would not be shared with the medical personnel involved in their SLE care. To assess the test– retest reliability of the MASRI, the questionnaire was also sent to the ﬁrst 20 study participants 1 week after the initial study interview via US mail. A prestamped envelope was enclosed to facilitate the return of the questionnaire. Reminder telephone calls were placed for questionnaires that were not returned within a 1-week period after the expected receipt of the questionnaire. Medication Adherence Self-report Inventory. The MASRI is a 12-item questionnaire of self-reported medication adherence addressing the frequency (6 items, part A) and the correct timing of medication intake (6 items, part B) (10). Given that, unlike HIV therapy, the intake of medications at certain exact time intervals is an uncommon demand for patients with SLE or patients with other rheumatic diseases, the MASRI part B was not completed by the study participants. For analytical purposes, responses of never or more than a month ago were grouped together. Of note, only the visual analog scale (VAS) item of the MASRI is used to determine a numeric estimate of the adherence rate (0 –100%). The other items are added to help patients develop this adherence estimate and are possibly useful for identifying speciﬁc situations that interfere with medication adherence. Prior to the use of the MASRI in this study, the questionnaire was piloted among members of a local lupus support group. Other measures of adherence to medications. Pharmacy reﬁll. Adherence estimates based on pharmacy reﬁll data were considered the criterion standard for measuring adherence, the rationale being that pharmacy reﬁll data provide longer-term information about patient adherence and do not rely on the participants remembering to bring medication supplies to the study visit (less bias) (11), and that physician ratings are considered less accurate than other methods of adherence measurement (6,7). Speciﬁcally, the participants provided contact information for the pharmacies where they ﬁlled their prescriptions. These pharmacies were contacted and information about the 4 most recent medication reﬁlls was obtained. Adherence based on pharmacy reﬁll was deﬁned as the percentage of prescribed doses taken and was calculated as follows: (total number of medication doses dispensed/total number of prescribed doses) ⫻ 100%. For example, a patient obtained 90 doses of medication with the last 3 reﬁlls and 120 doses of medication with the last 4 reﬁlls, but he or she was still taking pills dispensed with the fourth reﬁll supply. The number of days between the ﬁrst reﬁll date and the fourth reﬁll was 110 days. Thus, for the 110 days between reﬁlls the patient had only 90 medication doses and therefore his or her adherence would be 90/110 ⫻ 100% ⫽ 82%. Physician rating of patient adherence. The treating rheumatologist provided estimates of the participant’s ad- 1002 herence to each medication separately on an 11-point ordinal scale anchored with 0% (none of the time) and 100% (all the time) as the lower and upper bounds, respectively, and spaces for check marks in 10% intervals (MD scale). The scale was presented with the sentence stem “How frequently do you think the patient is taking the medication as prescribed during the preceding month?” Pill counts. Participants were instructed to bring their current supply of medications prescribed for SLE to the study visit. Besides the physician’s prescription instructions and the number of remaining medication doses, adherence rates considered the time interval between the study visit and the medication reﬁll dates. Statistical analysis. Calculation of adherence rates. Adherence to medication was expressed as a percentage value and deﬁned as the ratio of the number of medication doses taken (numerator) and the number of doses prescribed for the time between the study visit and the reﬁll date of the medication (denominator). Thus, adherence rates are a continuum where 0% means that the participant never takes the medication and 100% means that the medication is taken as prescribed at all times. Deﬁnition of sufﬁcient adherence and nonadherence. For the purpose of the analysis, we categorized adherence as follows: patients with adherence rates ⱖ80% were considered to be sufﬁciently adherent to the medication, and those with adherence rates ⬍80% were considered to be nonadherent to the medication. These cutoff values have been used in the past but are, in essence, chosen arbitrarily (9). Adherence rates and the presence of sufﬁcient adherence or nonadherence, respectively, were measured for prednisone and hydroxychloroquine separately because previous research suggests that medication adherence may differ between different drugs taken for SLE (3). Reliability and concurrent validity of the MASRI. Internal consistency of the MASRI items was assessed by Cronbach’s alpha, with values ⱖ0.7 representing acceptable internal consistency (12). Kappa statistics and intraclass correlation coefﬁcients (ICCs) for categorical data and interval data, respectively, were calculated to assess the test–retest reliability of the MASRI. ICC and kappa values were interpreted as follows: ⬍0.01 represented poor reliability, 0.01– 0.20 represented slight reliability, 0.21– 0.40 represented fair reliability, 0.41– 0.60 represented moderate reliability, 0.61– 0.80 represented substantial reliability, and 0.81–1 represented almost perfect reliability (13). Spearman’s correlation coefﬁcients (rsp) and ICCs were used to assess the concurrent validity of the MASRI, comparing MASRI VAS estimates with adherence rates based on pharmacy reﬁll data (criterion standard), pill counts, and physician global assessments (MD scale). Spearman’s correlation coefﬁcients were interpreted as follows: unrelated: rsp ⬍ 0.2; weak: 0.2 ⱕ rsp ⬍ 0.4; moderate: 0.4 ⱕ rsp ⬍ 0.6; strong: rsp ⱖ 0.6. Receiver operating characteristic curve analysis. Receiver operating characteristic (ROC) curve analysis was performed (14), with adherence rates by pharmacy reﬁll data serving as the external standard. SAS proc logistic and contingency table analyses (SAS Institute, Cary, NC) Koneru et al were used to determine the sensitivity and speciﬁcity of the various measures of adherence as well as the most appropriate cutoff values of the MASRI VAS that best discriminated between adherent and nonadherent patients. Sensitivity (to identify nonadherence) of a measure was calculated as follows: number of correctly identiﬁed nonadherent patients divided by number of all nonadherent patients. Speciﬁcity of a measure (to identify nonadherence) was calculated as follows: number of correctly identiﬁed adherent patients divided by number of all adherent patients. ROC curves were constructed, e.g., graphs in which the x-axis depicts the values of (1 ⫺ speciﬁcity) and the y-axis shows the sensitivity of the measure. We repeated the analysis assessing the measurement properties of the MASRI when used in combination with other measures of adherence used in this study (MD scale, pill count). RESULTS Participants. Fifty-ﬁve patients (96% women) with a mean ⫾ SD age of 31.7 ⫾ 14 years and mean ⫾ SD SLE duration of 7.2 ⫾ 5.2 years participated in the study. Most participants were white (n ⫽ 27 [49%]), 26 were African American (47%), and 2 were Asian (4%). Forty-one patients were treated with prednisone and 37 with hydroxychloroquine. Twenty-three of the 55 patients were treated with both hydroxychloroquine and prednisone. Adherence estimates by pharmacy reﬁll data. Nonadherence to prednisone, e.g., adherence of ⬍80% based on pharmacy reﬁll data, was common and occurred in 16 (39%) of 41 patients. Nonadherence to hydroxychloroquine was present in 19 (51%) of 37 patients. All patients receiving prednisone completed MD scales and the MASRI, but MASRI ratings were unavailable for 8 patients treated with hydroxychloroquine. Only 28 patients brought their medication supply to the study visit for the pill count (Figure 1). Irrespective of the medication assessed, adherence rates based on the MD scale or the MASRI had a considerable ceiling effect, and both scales were negatively skewed. Using prednisone as an example, physicians assigned an adherence rate of 100% to 16 (39%) of the 41 patients (skewnessMD scale ⫽ ⫺1.97), whereas self-report (MASRI) suggested that 20 (49%) of the 41 patients were fully adherent to prednisone (skewnessMASRI ⫽ ⫺2.2). Conversely, pharmacy data supported that 100% adherence was only present in 11 (27%) of the 41 patients (Figure 1). Reliability of the MASRI. The internal consistency reliability of the MASRI was acceptable (Cronbach’s ␣ ⫽ 0.7) with a minimum alpha coefﬁcient between variables at 0.4, suggesting that none of the included items could be easily removed to improve the scale. Information from 20 patients was available to assess the intrarater reliability of the MASRI. The reliability of the MASRI was moderately good, with a mean kappa value for the Likert scale items of 0.5 (item 1: ⫽ 0.8, item 2: ⫽ 0.3, items 3 and 5: ⫽ 0.5); the test–retest reliability of the MASRI VAS was excellent (ICC ⫽ 0.93). Medication Adherence in SLE 1003 criterion standard of adherence (rsp ⱖ 0.55). Neither pill counts nor the MD scale was as closely correlated with the criterion standard as the MASRI (Table 1). There was moderate agreement (ICC) between the MASRI and the criterion standard (Table 2). Agreement between the MASRI and the criterion standard was moderate for prednisone (ICC ⫽ 0.48) and somewhat lower for hydroxychloroquine (ICC ⫽ 0.39). Irrespective of the medication considered, none of the other adherence measures (pill count, MD scale) was in higher agreement with the criterion standard than the MASRI. Predictive validity of the MASRI. MASRI adherence rates ⱖ95% were 67% and 61% sensitive for identifying patients who were nonadherent to prednisone and hydroxychloroquine, respectively. The speciﬁcity of the MASRI for rates ⱖ95% was 68% for identifying prednisone nonadherence and 65% for hydroxychloroquine nonadherence (Figure 2). Figure 1. Relationship of adherence measurements for prednisone and hydroxychloroquine. Adherence rates based on all examined measures are depicted in their relationship to the external standard (pharmacy reﬁll data) for A, prednisone and B, hydroxychloroquine. Patient self-report of adherence (Medication Adherence Self-report Inventory [MASRI]) and physician ratings of adherence (MD scale) showed a ceiling effect for both prednisone and hydroxychloroquine. VAS ⫽ visual analog scale. Concurrent validity of the MASRI. Based on the MASRI and irrespective of the type of medication assessed, patient self-report was at least moderately correlated with the Screening for nonadherence in the clinic: MD scale plus MASRI. The combined consideration of adherence rates using the MASRI and the MD scale resulted in a screening tool that was superior to other combinations of adherence measures completed in this study in discriminating adherent patients from nonadherent patients. For prednisone, MD scale adherence rates ⱖ85% and MASRI rates ⱖ90% were 87% sensitive and 86% speciﬁc for distinguishing sufﬁciently adherent patients from those who were nonadherent. The presence of only 1 of these 2 criteria was still highly speciﬁc (87%) but only 60% sensitive for identifying nonadherence. For hydroxychloroquine, these cutoff values were still 63% sensitive but were less speciﬁc (57%) if both criteria were present (1 criterion: sensitivity 35%, speciﬁcity 100%) (Figure 3A). Estimates ⱖ80% for both the MASRI and the MD scale would have been best for identifying patients who were nonadherent to hydroxychloroquine (Figure 3B). DISCUSSION The MASRI is a reliable, valid, and concise self-report measure of adherence to medications in patients with SLE. Table 1. Correlation between adherence measure estimates* MASRI Prednisone Pharmacy MASRI Pill counts Hydroxychloroquine Pharmacy MASRI Pill counts Pill counts Spearman’s correlation P 0.62 0.55 * MASRI ⫽ Medication Adherence Self-report Inventory. † As measured by the 11-point physician rating scale. Physician rating† Spearman’s correlation P Spearman’s correlation P ⬍ 0.001 0.20 0.14 0.38 0.53 0.39 0.26 0.17 0.02 0.11 0.43 0.002 0.35 0.15 0.09 0.58 0.28 0.15 0.32 0.10 0.42 0.13 1004 Koneru et al Table 2. Intraclass correlation coefﬁcients of various adherence measures* Intraclass correlation coefﬁcient† Medication MASRI Pill count Physician rating‡ Prednisone Hydroxychloroquine 0.48 0.39 0.18 0.22 0.30 0.38 * MASRI ⫽ Medication Adherence Self-report Inventory. † Compared with the adherence measured by pharmacy reﬁll data (criterion standard). ‡ As measured by the 11-point physician rating scale. Our results conﬁrm a close relationship between the MASRI and adherence rates estimated by pharmacy reﬁll data. This ﬁnding is in concordance with the ﬁndings of previous studies, where the MASRI strongly correlated with other measures of adherence rates including electronic medication event monitoring systems, pill counts, pharmacy reﬁll information, and biologic tests (15,16). In contrast to unstructured approaches to obtaining patient-reported adherence, e.g., health care providers simply asking a patient “Are you taking your medicine?” (17), the MASRI offers a structured approach to measuring adherence. The Likert scale items of the MASRI help the patient more accurately judge monthly adherence on the MASRI VAS. Besides being an aid to correctly estimating adherence on the MASRI, the Likert scale items may be useful to address speciﬁc problems with the adherence to a particular medication regimen, such as difﬁculties with taking medications more than once a day, or on weekends. Therefore, different from unstructured approaches, structured patient self-report appears to provide adherence estimates that moderately to strongly correlate with pharmacy reﬁll information, and the MASRI appears to be a good predictor of the criterion standard in linear models (see below). The MASRI is not the only patient-completed measure of adherence. One measure is the Medication-Taking Scale (8), the usefulness of which has not been validated for antirheumatic therapies, and exact adherence rates cannot be calculated from its raw scores. Another recently validated self-report measure for patients with rheumatic diseases is the CQR. The CQR consists of 19 items, rated on 4-point Likert scales, where reversely scored CQR items can be used to derive a CQR raw score that requires mathematical transformation into a discrete adherence rate between 0% and 100% (9). Although CQR cutoff values can be deﬁned that have moderate sensitivity (62%) and high speciﬁcity (95%) for detecting nonadherence, an advantage of the MASRI may be that its adherence rates are good statistically signiﬁcant predictors of medication adherence (prednisone: R2 ⫽ 46%, df ⫽ 38, F ⫽ 29.5, P ⬍ 0.0001; hydroxychloroquine: R2 ⫽ 22%, df ⫽ 27, F ⫽ 7.24, P ⫽ 0.01), whereas CQR summary scores are unrelated to pharmacy reﬁll-based adherence rates. An added advantage of the MASRI over the CQR may be that the former provides adherence rates that can be instantly read from the MASRI VAS, and no calculations or transformations of scores are necessary, likely adding to the suitability of the MASRI for clinic practice. Nonetheless, a more comprehensive validation of adherence self-report measures for rheumatology is warranted to study beneﬁts and shortfalls of the various self-report tools that have been developed. Therefore, the Figure 2. Receiver operating characteristic (ROC) curve for detecting nonadherence to A, prednisone and B, hydroxychloroquine. ROC curves show the relationship of the sensitivity and (1 ⫺ speciﬁcity) to identify nonadherent patients. Pharmacy reﬁll information was used to deﬁne nonadherence to medications (criterion standard), with sufﬁcient adherence being deﬁned as an adherence rate ⱖ80%, while nonadherence was deﬁned as an adherence rate ⬍80%. Adherence measures whose curves approach the upper left corner of the graph are best for discriminating adherent from nonadherent patients. Medication Adherence in SLE 1005 Figure 3. Receiver operating characteristic (ROC) curve of the Medication Adherence Self-report Inventory (MASRI) and physician (MD) estimates of nonadherence to A, prednisone and B, hydroxychloroquine. ROC curves show the relationship of the sensitivity and (1 ⫺ speciﬁcity) to discriminate between adherent and nonadherent patients. Pharmacy reﬁll information was used to deﬁne adherence and nonadherence to medications (criterion standard), with sufﬁcient adherence being deﬁned as adherence rates ⱖ80% and nonadherence as rates ⬍80%. Deﬁnitions whose curves approach the upper left corner of the graph are most suitable for discriminating adherent from nonadherent patients. VAS ⫽ visual analog scale. MASRI should be compared directly with the MedicationTaking Scale (8) and the CQR (9). At present, adherence measurement in clinical practice is likely sporadic at best. Occasionally used approaches for detecting nonadherence are physician proxy ratings of patient adherence, pill counts, and adherence estimates based on pharmacy reﬁll information; the latter 2 methods are relatively time intensive, with pill counts relying on the patients to bring the medications to the clinic visit. Pill counts in such selected subgroups of patients may confer more optimistic estimates of adherence. There is only limited information about the accuracy of adherence estimates performed by rheumatologists (18). However, when used as the sole measure of adherence, ratings by other groups of physicians are insensitive and nonspeciﬁc for estimating patient adherence (6,7,19). Based on correlation analysis in this study, physician estimates were not as closely correlated with the criterion standard as the MASRI, suggesting that patients’ self-report may provide somewhat more valuable information, and based on ROC analysis rheumatologists’ adherence ratings by themselves cannot be considered proﬁcient measures of patient adherence. Nonetheless, when combined with the MASRI, physician ratings appear to facilitate the identiﬁcation of nonadherence and, in contrast to performing pill counts and obtaining pharmacy reﬁll information, the MASRI and the physician rating of adherence can be readily performed in clinics. When combined with the MD scale, the MASRI was highly sensitive and speciﬁc for identifying nonadherence among patients, with the prediction of nonadherence to prednisone being somewhat better than that of hydroxychloroquine. For reasons that were not entirely clear, both the physicians and the patients themselves had more difﬁculty estimating adherence rates for hydroxychloroquine than for prednisone. A possible explanation could be that, compared with prednisone, hydroxychloroquine has less immediate effects and side effects, decreasing the clues for the physicians to detect nonadherence and making it more difﬁcult for the patients to perform adherence self-reports. One limitation of the MASRI may be that, in its current form, it does not allow the assessment of medication overuse. Two patients receiving prednisone in our study reported taking 50 –100% more prednisone than was prescribed, and based on our qualitative data 5 of the 41 patients treated with prednisone reported taking more prednisone than prescribed on some days. Importantly, overuse was not reported for any other SLE medication. Therefore an adaptation of the MASRI VAS to allow reporting of medication overuse may even improve the usefulness of the MASRI, especially with respect to prednisone. However, even in its current version the MASRI correlated well with pharmacy reﬁll information for both prednisone and hydroxychloroquine. Additionally, estimates of a measure’s sensitivity and speciﬁcity depend on the prevalence of the condition considered in the population, and are therefore prone to change when the tool is used in other SLE cohorts with more or less frequent adherence problems. However, the prevalence of prednisone and hydroxychloroquine in the 1006 Koneru et al study participants is similar to what has been reported in some other rheumatology cohorts, adding to the validity of our ﬁndings and likely supporting the usefulness of the MASRI (20). A limitation of our study may be that patients completed the MASRI with the assurance that the results would not be shared with the treating physician. This might have led to more accurate self-reported adherence than if patients expected that their rheumatologist would review the MASRI responses. However, this notion is not supported by previous studies using the MASRI in other patient populations (21,22). Based on this pilot study, patient self-report of medication adherence using the MASRI offers a reliable and valid approach to measuring the extent to which patients follow prescribed medication regimens. In combination with physician-estimated adherence, the MASRI has the potential to be a useful measure of adherence suited for daily clinical practice. The MASRI appears to be useful for estimating adherence to different types of medications, and research in HIV suggests that the same holds true for combinations of drugs. Additional research is required to conﬁrm the usefulness of the MASRI in larger populations and its responsiveness to change over time. As is recommended by the American College of Rheumatology, multiple levels and ongoing validation are necessary prior to the routine clinical use of the MASRI for measuring adherence to medications in patients with SLE (23). 7. 8. 9. 10. 11. 12. 13. 14. 15. AUTHOR CONTRIBUTIONS Dr. Brunner had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study design. Koneru, Ware, Farhey, Brunner. Acquisition of data. Koneru, Shishov, Ware, Farhey, Mongey, Graham, Passo, Houk, Higgins. Analysis and interpretation of data. Koneru, Shishov, Passo, Houk, Brunner. Manuscript preparation. Koneru, Graham, Passo, Houk, Higgins, Brunner. Statistical analysis. Shishov, Brunner. 16. 17. 18. 19. REFERENCES 1. Haynes RB, McDonald HP, Garg AX. Helping patients follow prescribed treatment: clinical applications. JAMA 2002;288: 2880 –3. 2. Hochberg MC. Updating the American College of Rheumatology revised criteria for the classiﬁcation of systemic lupus erythematosus [letter]. Arthritis Rheum 1997;40:1725. 3. Shishov M, Koneru S, Graham TB, Houk LJ, Mongey AB, Passo M, et al. The Medication Adherence Self-Report Inventory (MASRI) can accurately estimate adherence with medications in systemic lupus erythematosus (SLE) [abstract]. Arthritis Rheum 2005;52 Suppl 9:S188. 4. Uribe AG, Alarcon GS, Sanchez ML, McGwin G Jr, Sandoval R, Fessler BJ. Systemic lupus erythematosus in three ethnic groups. XVIII. Factors predictive of poor compliance with study visits. Arthritis Rheum 2004;51:258 – 63. 5. Rojas-Serrano J, Cardiel MH. Lupus patients in an emergency unit: causes of consultation, hospitalization and outcome: a cohort study. Lupus 2000;9:601– 6. 6. Wagner JH, Justice AC, Chesney M, Sinclair G, Weissman S, 20. 21. 22. 23. Rodriguez-Barradas M, and the VACS 3 Project Team. Patientand provider-reported adherence: toward a clinically useful approach to measuring antiretroviral adherence. J Clin Epidemiol 2001;54 Suppl 1:S91– 8. Finney JW, Hook RJ, Friman PC, Rapoff MA, Christophersen ER. The overestimation of adherence to pediatric medical regimens. Child Health Care 1993;22:297–304. Morisky DE, Green LW, Levine DM. Concurrent and predictive validity of a self-reported measure of medication adherence. Med Care 1986;24:67–74. De Klerk E, van der Heijde D, Landewe R, van der Tempel H, van der Linden S. The compliance-questionnaire-rheumatology compared with electronic medication event monitoring: a validation study. J Rheumatol 2003;30:2469 –75. Walsh JC, Mandalia S, Gazzard BG. Responses to a 1 month self-report on adherence to antiretroviral therapy are consistent with electronic data and virological treatment outcome. AIDS 2002;16:269 –77. Choo PW, Rand CS, Inui TS, Lee ML, Cain E, CordeiroBreault M, et al. Validation of patient reports, automated pharmacy records, and pill counts with electronic monitoring of adherence to antihypertensive therapy. Med Care 1999;37:846 –57. Cronbach LJ. Essentials of psychological testing. 5th ed. New York: HarperCollins; 1990. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159 –74. Hanley JA, Negassa A, Edwardes MD, Forrester JE. Statistical analysis of correlated data using generalized estimating equations: an orientation. Am J Epidemiol 2003;157: 364 –75. Hugen PW, Langebeek N, Burger DM, Zomer B, van Leusen R, Schuurman R, et al. Assessment of adherence to HIV protease inhibitors: comparison and combination of various methods, including MEMS (electronic monitoring), patient and nurse report, and therapeutic drug monitoring. J Acquir Immune Deﬁc Syndr 2002;30:324 –34. Rapoff MA, Belmont JM, Lindsley CB, Olson NY. Electronically monitored adherence to medications by newly diagnosed patients with juvenile rheumatoid arthritis. Arthritis Rheum 2005;53:905–10. Stephenson BJ, Rowe BH, Haynes RB, Macharia WM, Leon G. Is this patient taking the treatment as prescribed? JAMA 1993; 269:2779 – 81. Deyo RA, Inui TS, Sullivan B. Noncompliance with arthritis drugs: magnitude, correlates, and clinical implications. J Rheumatol 1981;8:931– 6. Liu H, Golin CE, Miller LG, Hays RD, Beck CK, Sanandaji S, et al. A comparison study of multiple measures of adherence to HIV protease inhibitors [published erratum appears in Ann Intern Med 2002;136:175]. Ann Intern Med 2001;134:968 –77. De Klerk E, van der Heijde D, Landewe R, van der Tempel H, Urquhart J, van der Linden S. Patient compliance in rheumatoid arthritis, polymyalgia rheumatica, and gout [published erratum appears in J Rheumatol 2003;30:423]. J Rheumatol 2003;30:44 –54. Duong M, Piroth L, Grappin M, Forte F, Peytavin G, Buisson M, et al. Evaluation of the Patient Medication Adherence Questionnaire as a tool for self-reported adherence assessment in HIV-infected patients on antiretroviral regimens. HIV Clin Trials 2001;2:128 –35. Duong M, Piroth L, Peytavin G, Forte F, Kohli E, Grappin M, et al. Value of patient self-report and plasma human immunodeﬁciency virus protease inhibitor level as markers of adherence to antiretroviral therapy: relationship to virologic response. Clin Infect Dis 2001;33:386 –92. Classiﬁcation and Response Criteria Subcommittee of the American College of Rheumatology Committee on Quality Measures. Development of classiﬁcation and response criteria for rheumatic diseases. Arthritis Rheum 2006;55:348 –52.