Therapeutics

Association Between the 8-Item Morisky Medication Adherence Scale (MMAS-8) Score and Glycaemic Control Among Chinese Diabetes Patients

The Journal of Clinical Pharmacology XX(XX) 1–9 © 2014, The American College of Clinical Pharmacology DOI: 10.1002/jcph.408

Martin C.S. Wong, MD1, Carmen H.M. Wu, MPH2, Harry H.X. Wang, PhD1,3, Heung Wing Li, MBChB2, Eric M.T. Hui, MBBS2, Augustine T. Lam, MBBS2, Roger Y.N. Chung, PhD1, Benjamin H.K. Yip, PhD1, and Donald E. Morisky, ScD4

Abstract Adherence with oral hypoglycaemic agent is crucial to achieve optimal glycaemic control. The 8-item Morisky Medication Adherence Scale (MMAS-8) has been frequently used, yet the association between MMAS-8 score and glycaemic control among Chinese diabetes patients is largely unknown. Two general out-patient clinics were randomly selected in a district with socio-demographic characteristics representative of the entire Hong Kong population. A consecutive sample of adult type-2 diabetes patients currently taking oral hypoglycaemic agents was included. The glycaemic control was reflected by the level of hemoglobin A1c (HbA1c) taken within the previous 6 months. Factors associated with poor glycaemic control (HbA1c  7.0%) were evaluated by linear regression analysis. From 565 eligible Chinese patients with an average age of 63.2 years (SD 9.7) and male proportion of 46.5%, the average HbA1c was 7.1% (SD 1.1%), and 52.0% had poor glycaemic control. The proportion of poor medication adherence (MMAS-8  6) was 32.2%. After controlling for socio-demographics, lifestyle, medication use, and health characteristics, the MMAS-8 score was correlated with better glycaemic control (beta 0.095; 95%CI 0.164 to 0.026, P ¼ .007). The MMAS-8 score had a weak and negative correlation with HbA1c level. The instrument should be applied with caution when predicting glycaemic control in clinical practice.

Keywords Morisky Medication Adherence Scale, MMAS-8, glycaemic control, oral hypoglycaemic agents, associated factors, primary care

Diabetes mellitus is a global health issue with a worldwide prevalence of 9.8% and 9.2% in adult men and women, respectively.1 Its prevalence and incidence are growing in both Western countries and the Asia Pacific region.2–4 Poor glycaemic control was associated with increased incidence of stroke, all-cause mortality, and cardiovascular deaths.5–7 The World Health Organization (WHO) estimated a rise in the number of diabetic population from 170 million in 2000 to 366 million in 2030,8 when diabetes is predicted to become the seventh leading cause of death.8 Regions with rapidly developed economies such as Hong Kong are encountering such a challenge, where the prevalence of diabetes has been increasing dramatically in young working populations.9 The recent decade has witnessed an emergence of novel oral hypoglycaemic agents,10 yet their full therapeutic benefits could only be realized with optimal adherence to their use.11 A few studies evaluated the levels of medication adherence among diabetes patients,11–14 and the adherence levels were found to be low in general. This highlighted the importance of further research on this clinical issue. Current strategies to assess medication adherence included both direct and indirect methods,11 and self-administered surveys are considered

as one of the valid methodologies in clinical practices.15,16 It is also well recognized that well-designed questionnaire studies could provide adequate sensitivity and specificity in drug adherence assessment.17,18 Previous studies were either small-scale, or provided limited information on the relationship between medication adherence and outcome

1

JC School of Public Health and Primary Care, Faculty of Medicine, The Chinese University of Hong Kong, Hong Kong 2 Department of Family Medicine, New Territories East Cluster, Hospital Authority, Hong Kong 3 General Practice and Primary Care, Institute of Health and Wellbeing, University of Glasgow, Glasgow G12 9LX, UK 4 Department of Community Health Sciences, UCLA Fielding School of Public Health, Los Angeles, CA, USA Submitted for publication 4 June 2014; accepted 3 October 2014. *Corresponding Author: Harry H.X. Wang, BS, MS, PhD, JC School of Public Health and Primary Care, Faculty of Medicine, The Chinese University of Hong Kong, 30-32 Ngan Shing Street, Shatin, N.T., Hong Kong Email: [email protected] Martin C.S. Wong and Carmen H.M. Wu contributed equally to this manuscript.

2 measures such as diabetes control.11–14 Other studies were either conducted among hypertensive patients or based on electronic databases in which the computerized medication prescription records might not accurately reflect patients’ drug-taking behaviors in real life.19–28 It remains largely unknown whether the use of adherence scales from patient interviews (self-administered surveys) could effectively predict one’s level of glycaemic control. The primary objective of this study was to evaluate the association between glycaemic control and medication adherence among Chinese diabetes patients in the clinical setting. We also evaluated factors associated with poor medication adherence and poor glycaemic control.

Methods Ethics Statement Ethics clearance of this study was sought from the Cluster Research Ethics Committee of the Hospital Authority (CRE-2013.590), and the Survey and Behavioural Research Ethics Committee of The Chinese University of Hong Kong. The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki. Study Setting This study adopted a cross-sectional design. Patient interviews were conducted at 2 randomly selected general out-patient clinics, located in a region which has similar age and gender characteristics with that of the entire Hong Kong population (according to the 2011 Hong Kong Population Census29). All patients were recruited by a consecutive sampling design. Inclusion and Exclusion Criteria The study population was recruited from a comprehensive diabetic complication screening programme in which patients were referred by family physicians in the clinics. We included Chinese patients who were (1) aged 18 years or older; (2) diagnosed with type-2 diabetes by physicians; (3) taking at least one oral hypoglycaemic agent on a regular basis in the past 3 months; (4) not taking any insulin injection therapy combined with oral agents; and (5) able to comprehend and communicate in Cantonese. Patients, with mental illnesses, who were unable to independently complete the surveys were excluded. Eligible patients were invited by study investigators to participate in the study after reassurance of the voluntary and confidential nature of the survey. In addition, their participation or nonparticipation would not affect the healthcare services received. Assessment of Medication Adherence The medication adherence was assessed using a validated, self-administered 8-item Morisky medication adherence

The Journal of Clinical Pharmacology / Vol XX No XX (2014)

scale (MMAS-8) survey.30,31 The MMAS-8 has been translated into different languages internationally for different study purposes under various settings.30,32–35 The questionnaire items could be easily understood by patients with low educational level. The survey instrument showed clinically reliability with a Cronbach’s a of 0.83,36 and was shown to be significantly associated with blood pressure control in previous studies.37,38 The MMAS-8 questionnaire adopted a simple and quick scoring algorithm, in which negative response for each question was coded as 1, except for the question asking if the patient took their medications yesterday (where a positive response was coded as 1). The total MMAS-8 score was calculated by summing the values from all the 8 question items. Following the methodology used in the previous literature, optimal adherence was defined as having a MMAS-8 score over 6 out of a total of 8 scores.18,30,31,37 Demographic and Clinical Characteristics of Study Subjects Upon subject recruitment, the socio-demographic information was collected in terms of age, sex, household income, educational level, smoking habits, alcohol drinking, body mass index (BMI), lifestyle measures on physical activity and dietary control, as well as the current number of antihypertensive drugs, lipid-lowering agents, and antiplatelet medications taken. For the measurement of body weight, participants were in light clothing without wearing shoes. The measurement was performed using reliable weight scales assessed by a wall-mounted stadiometer which was regularly calibrated over time by the clinical staff. BMI was further classified into 3 groups: (1) underweight ( 6) BMI (kg/m2) Mean (SD) Underweight ( 6). The average BMI was 25.9 kg/m2 (SD 3.9) with the majority being overweight or obese (79.5% [449/565]). The prevalence of concomitant use of antihypertensive agents (78.2%), lipid-lowering agents (62.1%), and antiplatelets (9.0%) was high. Most were non-smokers (72.7%) and non-drinkers (82.1%). Most of the patients had fair dietary compliance (86.4%) and their adherence with exercise habits was rated as fair or poor (both 37.9%) (Table 1). The Correlation Between the MMAS-8 Score and HbA1c Level The Spearman correlation test showed a negative correlation between the level of HbA1c and MMAS-8 scores (Spearman rho ¼ 0.087, P ¼ .038). Point-biserial correlation was calculated as sensitivity analysis, performed by treating MMAS-8 score as dichotomous variables using different cut-off points. Similar trend of associations (coefficient ¼ 0.127, P ¼ .003 for MMAS-8 score of 7 as the cut-off point; coefficient ¼ 0.092, P ¼ .028 for MMAS-8 score of 6 as the cut-off point) was observed between HbA1c and MMAS-8 scores with different cut-off points (Supplementary Table S1). The Profile of Medication Non-Adherence A total of 383 patients (67.8%) were adherent with their oral hypoglycaemic agents. The proportion of medication adherence was higher among those with lower income (monthly household income HK$10,000; 73.0% vs. 63.7% in other groups); patients with optimal glycaemic control (HbA1c < 7%; 71.4% vs. 63.8%); subjects who were concomitantly using antihypertensive agents (69.5% vs. 61.8%); lipid-lowering agents (68.7% vs. 66.4%); antiplatelet drugs (72.5% vs. 67.3%); as well as those with good dietary compliance (78.4% vs. 57.5%–67.8%) and regular exercise habits (76.6% vs. 58.9%–71.0%). Current smokers (60.9% vs. 67.8%–68.9%) and current drinkers (60.7% vs. 68.8%–83.3%) were less likely to adhere to their medication regimens (Table 2). Factors Associated With Poor Glycaemic Control and Medication Adherence From multiple linear regression analysis with HbA1c as a continuous outcome, it was found that younger age (B ¼ 0.011, 95%CI 0.023–0.00, P ¼ .049), use of

The Journal of Clinical Pharmacology / Vol XX No XX (2014)

lipid-lowering agents (B ¼ 0.225, 95%CI 0.032–0.419, P ¼ .022), and poorer medication adherence (MMAS-8 score; B ¼ 0.095 to 0.164 to 0.026, P ¼ .007) were associated with higher HbA1c levels (Table 3). From an additional regression model with medication adherence (MMAS-8 scores) as the outcome, patients who had good exercise habits (B ¼ 0.222, 95%CI 0.369 to 0.075, P ¼ .003) and lower HbA1c levels (B ¼ 0.137, 95%CI 0.236 to 0.037, P ¼ .007) were associated with higher MMAS scores (Table 4). There was neither interaction nor multicollinearity among all the covariates in both regression analyses, implying the robustness of the regression models.

Discussion Statement of Principal Findings From a total of 565 diabetes patients, the level of optimal medication adherence and good glycaemic control was 67.8% and 52.0%, respectively. More optimal control of diabetes was associated with better medication adherence in both univariate and multiple regression analyses. A cutoff score at 6 to define poor medication adherence for the MMAS-8 survey among diabetes patients was most predictive of glycaemic control. Additional factors associated with poor glycaemic control included younger age and use of lipid-lowering agents, whereas fair or poor exercise habit was associated with poorer medication adherence. Strengths and Weaknesses of the Study This is the first study of this scale which evaluated the association between HbA1c level and medication adherence using the MMAS-8 instrument among Chinese diabetes patients. The existence of an association between MMAS-8 scores and HbA1c level was largely consistent with previous studies conducted in the western population.43 The other socio-demographic and clinical variables were accurately captured, including standardized assessment of lifestyle habits; measurement of BMI using reliable procedures; and history of medication use which was retrieved from validated computerized systems. Study subjects were also assured that their responses to the MMAS-8 survey would not be disclosed to their healthcare providers, thus minimizing the possibility of social desirability bias. Nevertheless, some limitations should be addressed. Firstly, the questionnaire was collected in 2 general out-patient clinics only, and therefore interpretation of the study findings should take into account the representativeness of the study population. Secondly, other confounders that were not measured in this study might potentially influence the association between medication adherence and glycaemic control, such as patients’ knowledge on diabetes, the severity of type-2 diabetes,

5

Wong et al Table 2. Patient Characteristics According to Adherence With Oral Hypoglycaemic Agents (N ¼ 565) Poor Adherence (MMAS  6)

Total population (N) Sex Male Female Household income (HK$) 10,000 10,001–20,000 20,001 Educational level Primary or low Secondary Tertiary HbA1c HbA1c < 7% HbA1c  7% BMI (kg/m2) Normal or underweight ( 6)

N

%

N

%

182

32.2

383

67.8

83 99

31.6 32.8

180 203

68.4 67.2

67 70 45

27.0 36.3 36.3

181 123 79

73.0 63.7 63.7

97 76 9

32.8 31.5 32.1

199 165 19

67.2 68.5 67.9

84 98

28.6 36.2

210 173

71.4 63.8

38 144

32.8 32.1

78 305

67.2 67.9

47 135

38.2 30.5

76 307

61.8 69.5

72 110

33.6 31.3

142 241

66.4 68.7

168 14

32.7 27.5

346 37

67.3 72.5

128 25 29

31.1 39.1 32.2

283 39 61

68.9 60.9 67.8

145 35 2

31.3 39.3 16.7

319 54 10

68.8 60.7 83.3

8 157 17

21.6 32.2 42.5

29 331 23

78.4 67.8 57.5

32 62 88

23.4 29.0 41.1

105 152 126

76.6 71.0 58.9

HbA1c, glycated hemoglobin taken 6 months within the survey; MMAS-8, Morisky medication adherence scale, 8 Items; BMI, body mass index.

and physicians’ prescribing practices, which might need further in-depth study to investigate. Thirdly, a crosssectional study design could not establish cause-andeffect relationship due to the possibility of reverse causality. Last but not least, we used HbA1c readings captured within the previous 6 months prior to the date of the survey, and critics might argue that this might bias toward the null hypothesis (ie, HbA1c was not associated with MMAS).

Relationship With Literature The level of optimal medication adherence rate varied among previous studies conducted in different settings. The adherence rate tended to be higher in studies based on clinical database. In 2 retrospective cohort studies previously conducted by our research team among diabetes patients in Hong Kong, we found an adherence rate of 89.6% by using medication possession ratio (MPR) as a proxy measure based on a territory-wide computerized

6

The Journal of Clinical Pharmacology / Vol XX No XX (2014)

Table 3. Factors Associated With Glycaemic Control (HbA1c as a Continuous Variable) Unadjusted Model B (95%CI) Age Sex Male Female Educational level Primary or below Secondary or above BMI (kg/m2) Household income (HK$) 10,000 10,001 or above Use of antihypertensive agents Use of lipid-lowering agents Use of antiplatelets Smoking Non-smoker Ever smoked Drinking Non-drinker Ever drank Dietary compliance Good Fair or poor Exercise Good Fair or poor MMAS-8 score

0.014 ( 0.023,

Adjusted Model P Value

0.004)

.005

B (95%CI) 0.011 ( 0.023, 0.000)

P Value .049

Reference 0.104 ( 0.083, 0.291)

.276

Reference 0.146 ( 0.081, 0.373)

.208

Reference 0.030 ( 0.189, 0.128) 0.020 ( 0.004, 0.044)

.707 .098

Reference 0.071 ( 0.244, 0.103) 0.017 ( 0.008, 0.043)

.425 .181

Reference 0.008 ( 0.092, 0.108) 0.168 ( 0.395, 0.058) 0.223 (0.032, 0.415) 0.006 ( 0.333, 0.320)

.874 .144 .023 .970

Reference 0.027 ( 0.133, 0.078) 0.144 ( 0.385, 0.098) 0.225 (0.032, 0.419) 0.046 ( 0.288, 0.38)

.609 .242 .022 .787

Reference 0.013 ( 0.138, 0.111)

.836

Reference 0.013 ( 0.133, 0.159)

.860

Reference 0.059 ( 0.062, 0.180)

.341

Reference 0.071 ( 0.059, 0.201)

.283

Reference 0.183 ( 0.070, 0.436)

.156

Reference 0.092 ( 0.169, 0.353)

.488

Reference 0.102 ( 0.018, 0.222) 0.109 ( 0.177, 0.041)

.096 .002

Reference 0.032 ( 0.092, 0.155) 0.095 ( 0.164, 0.026)

.615 .007

The bold values represent statistically significant differences (P < .05). Model statistics: R2 ¼ 22.7%. CI, confidence interval; HbA1c, glycated hemoglobin taken 6 months within the survey; MMAS-8, Morisky medication adherence scale, 8 Items; BMI, body mass index.

clinical databases.40,44 In another study which used prescription refill data from a university pharmacy center,13 the medication adherence rate among 810 diabetes patients with low income who were recruited from a university-based internal medicine clinic in rural central Virginia was 79.7%, and the most recent HbA1c level was 8.1%. It was also found that for each 10% increment in adherence with oral hypoglycaemic agents, there was a 0.2% decrease in HbA1c level.13 Nevertheless, a study conducted by Ahmad and colleagues using national self-reported data in Malaysia showed that only 47% of diabetes patients in primary care clinics were medication adherent, as assessed by a self-designed medication compliance questionnaire.11 In another local hospitalbased study among diabetes patients recruited in a pharmacist-managed compliance programme, the baseline rate of medication adherence and the average glycaemic control was 41.3% and 7.4%, respectively.12 Bezie and colleagues evaluated the parameters influencing therapeutic compliance in type-2 diabetes patients admitted to a French general hospital due to uncontrolled diabetes or

change in health status. Approximately 64.9% of patients were found to be adherent measured through systematic patient interviews by pharmacist students.14 Compared to the above-mentioned studies, where self-reported adherence measures were used,11,12,14 the adherence rate of 67.8% found in our current study was relatively higher, and the average glycaemic control (7.1% vs. 7.4%) was also better than that reported by Lee and Leung.12 In both studies using routine clinical database40,44 and patient selfreport data,11 it was found that younger age was associated with poor medication adherence. Nevertheless, apart from the study performed by Schectman and colleagues,13 there have been no studies which correlated the relationship between HbA1c and medication adherence—which represents a new finding among Chinese diabetes patients. Regarding independent factors associated with poor glycaemic control, younger age was also found as an associated factor. The underlying reasons remained speculative, yet Lee and Leung12 proposed that older patients were more likely to have disease progression, which might lead to heightened awareness of diseases and

7

Wong et al Table 4. Factors Associated With Medication Adherence (MMAS-8 as a Continuous Variable) Unadjusted Model B (95%CI) Age Sex Male Female Educational level Primary or below Secondary or above BMI (kg/m2) Household income (HK$) 10,000 10,001 or above Use of antihypertensive agents Use of lipid-lowering agents Use of antiplatelets Smoking Non-smoker Ever smoked Drinking Non-drinker Ever drank Dietary compliance Good Fair or poor Exercise Good Fair or poor HbA1c

Adjusted Model P Value

0.016 (0.004, 0.028)

.007

Reference 0.018 ( 0.244, 0.209)

.879

B (95%CI)

P Value

0.008 ( 0.006, 0.021)

.256

Reference 0.082 ( 0.355, 0.191)

.554

Reference 0.003 ( 0.194, 0.189) 0.002 ( 0.031, 0.027)

.979 .898

Reference 0.055 ( 0.154, 0.263) 0.005 ( 0.025, 0.036)

.606 .732

Reference 0.069 ( 0.190, 0.051) 0.193 ( 0.081, 0.466) 0.133 ( 0.099, 0.365) 0.422 (0.029, 0.814)

.259 .167 .262 .035

Reference 0.034 ( 0.160, 0.093) 0.080 ( 0.209, 0.370) 0.196 ( 0.037, 0.428) 0.336 ( 0.065, 0.736)

.604 .586 .099 .100

Reference 0.005 ( 0.146, 0.155)

.950

Reference 0.002 ( 0.177, 0.174)

.984

Reference 0.105 ( 0.251, 0.042)

.161

Reference 0.105 ( 0.261, 0.051)

.186

Reference 0.343 ( 0.647, -0.038)

.027

Reference 0.295 ( 0.607, 0.017)

.064

Reference 0.264 ( 0.408, 0.121) 0.159 ( 0.258, 0.060)