Medication adherence, recall periods and factors affecting it: A community-based assessment on patients with chronic diseases in urban slums.

OBJECTIVE To evaluate medication adherence, the effect of recall periods on self-reported adherence and factors influencing medication adherence among patients of chronic diseases, such as hypertension and diabetes, particularly in the community. METHODS A cross-sectional cohort study was conducted among individuals with hypertension and/or diabetes coming as outpatients in community camps organised in a cluster of urban slums. Responses towards questions regarding self-reported quantitative and qualitative adherence for one week and one month along with information on pill burden, socio-demographic and other factors were recorded using a mobile application. RESULTS Among 379 participants living in urban slum communities, who were prescribed anti-hypertensive or oral anti-diabetic medications previously, mean medication adherence over previous one week was 67.99% (standard deviation (SD) ± 38.32) and 6.87 (SD ± 3.62) on a ten-point numeric scale. The medication adherence for one month showed a strong significantly positive correlation with that of 1 week for both percentage-based (r = +0.910, 95% CI = 0.864 to 0.950, P < .0001) and Likert (ρ = +0.836, 95% CI = 0.803 to 0.863, P < .0001) scales. Age (r = 0.219, 95% CI = 0.120 to 0.313, P = .043) and pill burden (r = -0.231, 95% CI = -0.145 to -0.322, P < .0001) were found to significantly affect medication adherence. The odds of random blood sugar reduction were found to be significant (OR 1.98, 95% CI = 1.30 to 3.00, P = .001) with adequate adherence. A linear regression equation was developed to predict medication adherence percentage for a patient which was found to have 61.8% predictive power using multilayer perceptron modelling. CONCLUSION Overall, medication adherence was sub-optimal. Adherence assessments can be reliably performed using either one week or one month recall periods. With further refinement and validation, the regression equation could prove to be a useful tool for physicians.