Increasingly restrictive definitions of hyperkalemia outcomes in a database study: effect on incidence estimates

PURPOSE: To determine the incidence of hyperkalemia-associated adverse outcomes among ambulatory patients with diabetes newly initiating renin-angiotensin-aldosterone system (RAAS) inhibitor therapy and to examine to what extent increasingly restrictive definitions of hyperkalemia-associated outcomes influenced incidence estimates. METHODS: Retrospective cohort study of 27 355 individuals with diabetes who were new users of RAAS inhibitors at three integrated healthcare systems. RESULTS: Using the least restrictive definition of hyperkalemia-associated outcome, an ambulatory visit (AV), inpatient hospitalization (IP), or emergency department (ED) visit co-occurring within 7 days of coded hyperkalemia diagnosis and/or potassium concentration of > or =5.5 mmol/L, the incidence rate of hyperkalemia was 30.1 per 1000 person-years (p-y). Restricting to only IP or ED visits co-occurring within 24 hours of coded diagnosis and/or potassium > or =6 mmol/L, the incidence rate was 10.2 per 1000 p-y. When this latter definition was further restricted by removing available potassium laboratory values, the incidence rate declined to 9.5 per 1000 p-y. CONCLUSIONS: Modifying the definition of hyperkalemia-associated outcome resulted in up to threefold differences in incidence rate estimates. Our results demonstrate the value of including potassium laboratory results and AVs in identifying hyperkalemia from electronic data. Comparing incidence estimates across published studies requires consideration of differences in outcome definitions.