Mhd Wasem Alsabbagh1, Lisa M Lix, Dean Eurich, Thomas W Wilson, David F Blackburn. 1. *School of Pharmacy, Faculty of Science, University of Waterloo, Waterloo, ON †Department of Community Health Sciences, Faculty of Medicine, University of Manitoba, Winnipeg, MB ‡Department of Public Health Sciences, School of Public Health, University of Alberta, Li Ka Shing Center for Health Research Innovation, Edmonton, AB §Department of Medicine, Royal University Hospital, Saskatoon Health Region ∥College of Pharmacy and Nutrition, University of Saskatchewan, Saskatoon, SK, Canada.
Abstract
INTRODUCTION: Low socioeconomic status (SES) should be a robust predictor of medication nonadherence because it shares key features with the theoretical origins of this phenomenon. However, population-based studies have demonstrated weak associations overall, possibly because SES is inadequately represented. We compared the performance of multiple versus single-domain measures of SES as predictors of statin adherence. METHODS: This retrospective cohort study used population-based administrative data mapped to area-level census information of individuals who received a statin medication following a hospitalization for coronary heart disease. One-year adherence was calculated by dividing the sum of all tablets dispensed by the total number of days in the observation period (365 d following the first statin dispensation). Logistic regression models were constructed and the relative impact of each SES measure was assessed by its adjusted odds ratio (OR) and improvement over the predictive accuracy of a reference model that included non-SES factors only. RESULTS: More than two thirds (ie, 68.8%; 6517/9478) of eligible individuals exhibited optimal adherence (ie, ≥80%). The estimated impact of SES on optimal adherence differed depending on the SES measure tested. The highest performing single-domain measure, household income (OR=0.75; 95% confidence interval, 0.63-0.90; model c-statistic improvement 0.5%, P=0.04) generated a similar result to the multiple-domain measure (adjusted OR=0.74; 95% confidence interval, 0.62-0.88; model c-statistic improvement 0.7%, P=0.01). CONCLUSION: Multidomain measurements of SES using administrative databases mapped to census data are not associated with better performance in predicting statin medication adherence compared with single-domain measures such as household income.
INTRODUCTION: Low socioeconomic status (SES) should be a robust predictor of medication nonadherence because it shares key features with the theoretical origins of this phenomenon. However, population-based studies have demonstrated weak associations overall, possibly because SES is inadequately represented. We compared the performance of multiple versus single-domain measures of SES as predictors of statin adherence. METHODS: This retrospective cohort study used population-based administrative data mapped to area-level census information of individuals who received a statin medication following a hospitalization for coronary heart disease. One-year adherence was calculated by dividing the sum of all tablets dispensed by the total number of days in the observation period (365 d following the first statin dispensation). Logistic regression models were constructed and the relative impact of each SES measure was assessed by its adjusted odds ratio (OR) and improvement over the predictive accuracy of a reference model that included non-SES factors only. RESULTS: More than two thirds (ie, 68.8%; 6517/9478) of eligible individuals exhibited optimal adherence (ie, ≥80%). The estimated impact of SES on optimal adherence differed depending on the SES measure tested. The highest performing single-domain measure, household income (OR=0.75; 95% confidence interval, 0.63-0.90; model c-statistic improvement 0.5%, P=0.04) generated a similar result to the multiple-domain measure (adjusted OR=0.74; 95% confidence interval, 0.62-0.88; model c-statistic improvement 0.7%, P=0.01). CONCLUSION: Multidomain measurements of SES using administrative databases mapped to census data are not associated with better performance in predicting statin medication adherence compared with single-domain measures such as household income.