Clinical Action Curves: Measuring the Magnitude of Physician Response to Abnormal Laboratory Results.

Abstract

While reference limits are foundational to interpreting clinical laboratory tests, they may not correspond to the actual values triggering clinical response. We propose to measure this using clinical action curves, which plot test values against an indicator of clinical action. We selected repeat test ordering as a quantifiable, objective, useful measure that is readily calculable using available laboratory data. Using all results in Calgary in 2010-2011 for eight analytes, clinical action curves for each analyte were plotted as the relationship between index test value and retesting hazard, modeled using Cox proportional hazards with restricted cubic splines. Clinical action limits were defined where retesting hazard rose 38% above baseline (25%-50% considered). In general, clinical action increased before the reference limits, and clinical action limits were narrower than reference limits. However, some reference limits showed no increased clinical action and may thus be ignored in practice. Clinical action curves and limits provide practical, objective tools for describing physician responses to test values. Results suggest that many normal results are treated as abnormal and vice versa; such discrepancies require further scrutiny and ultimately reconciliation via altered reference ranges or altered practice patterns.