Patient non-adherence with medical therapy continues to be a problem in essentially all medical specialties, especially among patients with asymptomatic chronic conditions such as diabetes, hypertension and hypercholesterolemia. Numerous measures of patient medication adherence exist but their relative efficacy is unknown. The Morisky Instrument, a validated four item self reported scale measuring medication-taking behavior and calculation of medication adherence rates using pharmacy records have been incorporated into numerous trials assessing medication non-adherence in a variety of disease states including diabetes, hypertension and hypercholesterolemia. Other studies have shown only modest correlation between self report and pill refill measures of adherence. Additionally, while both measures of adherence are associated with clinical outcomes, no direct comparison between the two exist with regards to strength of the association with these outcomes.
The objective of this study was to directly compare ReComp, a measure of adherence based on pharmacy data, to the Morisky (MSR) , a self reported adherence scale among patients enrolled in a randomized controlled trial of diabetes group medical clinics (GMC) Specific Aims included: 1) to measure the agreement between ReComp and the Morisky self reported measurement of adherence 2) to evaluate the strength of association between ReComp and Morisky with change in A1c, LDL and systolic blood pressure 3) determine if either medication adherence measure is associated with the efficacy of the diabetes GMC intervention.
We performed cross-sectional and longitudinal analyses on data from 239 patients enrolled in a randomized, controlled trial of GMC. Patients had poorly controlled diabetes (DM) (A1c>= 7.5%) and hypertension (SBP) >=140 or diastolic BP >=90). ReComp was calculated from prescription data for oral hypoglycemics (OHA), antihypertensives, and HMG-COA reductase inhibitors (statins) acquired during the year prior to enrollment. There were 212 patients who completed a Morisky survey and had enough data to calculate ReComp. The ReComp score was truncated at 1.0 and then the average was taken at the patient level for each medication type and dichotomized into < 0.80 (not adherent) and >=0.80 (adherent). The MSR was dichotomized into perfect (score=4) and imperfect (score < 4) adherence. McNemar's Test was used to examine differences in the measurement of medication adherence between both measures and an unweighted kappa statistic was calculated to assess agreement. The first modeling step fit separate linear regression models for each outcome (A1c, LDL, SBP) to assess the effect of adherence using either the Recomp or MSR measure on baseline outcomes adjusting for site, age, race, years since DM diagnosis, BMI and if on insulin or not (A1c outcome only). For the longitudinal analyses, we fit linear mixed-effects models (LMM) to account for the correlation in measures over time and to adjust for clustering within each GMC. We examined whether the outcomes changed differently over time between adherent and non-adherent patients for both adherence measures. We then fit LMMs to these outcomes to assess whether they changed differently over time in adherent and non-adherent patients between the intervention and control arms.
Baseline mean LDL was 98 mg/dl, SBP was 153 mmHg, and A1c was 9.2%. Adherence measured by ReComp was greater than MSR (44% vs 35%, p=0.03); Kappa was 0.12 (95% CI -0.01, 0.24). For ReComp and MSR measures respectively, adherent patients had lower baseline LDL (-13 mg/dl, p<0.03 and -12 mg/dl, p<0.04) but no difference in A1c (-0.3 %, p=0.07 and 0.1 %, p=0.7) or SBP (0.1 mmHg, p=0.9 and -0.2 mmHg, p=0.9) compared to non-adherent. Over 12 months, patients who were adherent by ReComp improved LDL by 3 mg/dl, whereas non-adherent patients improved 16 mg/dl (p=0.01). There was no similar difference in either A1c or SBP. At 12 months, patients adherent by MSR improved A1c by 1.1%, whereas non-adherent patients improved 0.5% (p=0.01). There was no difference in either LDL or SBP. Analyses incorporating treatment arm suggest that the differences were largely in the intervention arm but findings were not statistically significant.
There is poor agreement between MSR and ReComp. Patients who are not filling their prescriptions may have more lipid control benefit from a GMC, presumably via improved adherence. In contrast, patients reporting perfect medication adherence may have more glycemic control benefit from a GMC, presumably by improving their medication regimens. Researchers should be cautious when choosing adherence measures in studies. More than one measure may be needed to capture adherence. Additional studies should be undertaken to further delineate the impact GMC have on adherence in relation to chronic disease state management.
None at this time.