Prescribing history to identify candidates for chronic condition medication adherence promotion.

Poor adherence to long-term prescription medication is a frequent problem that undermines pharmacological control of important risk factors such as hypertension. A medication possession ratio (MPR) can be calculated from Practice Management System (PMS) data to provide a convenient indicator of adherence. We investigate how well prior MPR predicts later MPR, taking MPR<80% as indicative of 'non-adherence,' to assess the potential value of MPR calculation on PMS data for targeting adherence promotion activities by general practices. We examine PMS data for two New Zealand metropolitan general practices, one with a predominantly Pacific caseload, across 2008 and 2009. We find prevalence of non-adherence in 2009 to be 51.63% (95% confidence interval [CI] 47.9-55.3) for patients at the Pacific practice and 28.09% (95% CI 25.0-31.1) at the other practice for patients who are demonstrably active with the practice in 2009. The positive predictive value (PPV) of 2008 non-adherence for 2009 non-adherence is 71.80% (95% CI, 66.5-77.1) and negative predictive value (NPV) 61.52% (95% CI 56.9-66.1) for the Pacific practice; PPV is 61.38% (95% CI 54.6-68.2) and NPV is 82.19% (95% CI 79.2-85.2) for the other practice. The results indicate good potential for decision support tools to target adherence promotion.