Blood pressure (BP) lowering medications have impressive efficacy in reducing cardiovascular and renal events; but low adherence threatens their effectiveness. Analysis of patterns in electronic prescribing from electronic medical records (EMRs) may have the potential to reveal cohorts of patients with significant adherence problems.
We developed a computational framework to identify patient cohorts with poor adherence to long-term medication through analysis of electronic prescribing patterns. A range of quality reporting criteria can be specified (as an XML document). We illustrate the framework by application to the EMRs of a New Zealand general practice with a focus on adherence to angiotensin-converting enzyme inhibitors (ACE-inhibitors) and/or angiotensin II receptor blockers (ARBs) in patients classified with hypertension and diabetes. We analyse medication supply based on Medication Possession Ratio (MPR) and duration of lapse in ACE-inhibitors/ARBs over a 12-month evaluation period. We describe graphical tools to assist visualisation of prescribing patterns and relationship of the analysis outputs to controlled blood pressure.
Out of a cohort of 16,504 patient EMRs, 192 patients were found classified with both hypertension and diabetes and under active ACE-inhibitor and/or ARB management. Of these, 107 (56%) patients had an ACE-inhibitor/ARB MPR less than 80% together with a lapse in ACE-inhibitors/ARBs for greater than 30 days. We find non-adherent patients (i.e. MPR <80% or lapse >30 days) are three times more likely to have poor BP than adherent patients (odds ratio=3.055; p=0.012).
We have developed a generic computational framework that can be used to formulate and query criteria around issues of adherence to long-term medication based on practice EMRs. Within the context of the example we have used, the observed adherence levels indicate that a substantial proportion of patients classified with hypertension and diabetes have poor adherence, associated with poorer rates of blood pressure control, that can be detected through analysis of electronic prescribing. Further work is required to identify effective interventions using the reporting information to reduce non-adherence and improve patient outcomes.
Mabotuwana, Thusitha, Jim Warren, and John Kennelly. "A computational framework to identify patients with poor adherence to blood pressure lowering medication." International Journal of Medical Informatics 78, no. 11 (November 2009): 745-756.