Establishing an evidence base for critical laboratory value thresholds.

Abstract

Critical values denote laboratory test results indicating a life-threatening situation. The outcomes of this premise have not been rigorously evaluated. Five years of inpatient admissions were examined for critical or "near-critical" results (total admissions = 165,066; total test results = 872,503). In-hospital mortality was examined as a function of time and degree of test result abnormality. Some critical value thresholds appropriately identified patients at risk for death (eg, elevated potassium). Other thresholds were too conservative (elevated hematocrit, hemoglobin) or not conservative enough (elevated lactate). Mortality risk for most critical values was time dependent, but some critical values showed no temporal effect on mortality (elevated activated partial thromboplastin time [APTT], international normalized ratio [INR], and glucose). Following an initial critical result, further worsening was associated with increased mortality. Prior hospital admission within 30 days was a predictor of lower mortality for some (elevated APTT, INR, potassium, and sodium; low glucose, hematocrit, hemoglobin, and potassium) but not other critical values (elevated lactate, glucose, hematocrit, and hemoglobin; low sodium). Only a subset of laboratory critical value thresholds was optimally chosen for increased risk of in-hospital mortality, with a time urgency for most but not all critical values. For many tests, a prior hospital admission imparted a decreased risk of in-hospital death.