Risks of mortality associated with common laboratory tests: a novel, simple and meaningful way to set decision limits from data available in the Electronic Medical Record

Clin Chem Lab Med. 2013 Sep;51(9):1803-13. doi: 10.1515/cclm-2013-0167.


Background: Laboratory tests provide objective measurements of physiologic functions, but are usually evaluated by demographic reference-intervals (RI), instead of risk-based decision-limits (DL). We show that hospital electronic medical record (EMR) data can be utilized to associate all-cause mortality risks with analyte test values, thereby providing more information than RIs and defining new DLs.

Methods: Our cohort was 39,964 patients admitted for any reason and discharged alive, during two 1-year periods, at Sarasota Memorial Hospital, Florida, USA. We studied five routinely-performed in-hospital laboratory tests: serum creatinine, blood urea nitrogen, serum sodium, serum potassium, and serum chloride. By associating a mortality odds ratio with small intervals of values for each analyte, we calculated relative risk of all-cause mortality as a function of test values.

Results: We found mortality risks below the population average within these proposed DLs: potassium 3.4-4.3 mmol/L; sodium 136-142 mmol/L; chloride 100-108 mmol/L; creatinine 0.6-1.1 mg/dL; blood urea nitrogen (BUN) 5-20 mg/dL. The DLs correspond roughly to the usually-quoted RIs, with a notable narrowing for electrolytes. Potassium and sodium have reduced upper limits, avoiding a "high-normal" area where the odds ratio rises 2 to 3 times the population average.

Conclusions: Any clinical laboratory test can be transformed into a mortality odds ratio function, associating mortality risk with each value of the analyte. This provides a DL determined by mortality risk, instead of RI assumptions about distribution in a "healthy" population. The odds ratio function also provides important risk information for analyte values outside the interval.

MeSH terms

  • Adult
  • Aged
  • Blood Chemical Analysis / methods*
  • Clinical Laboratory Techniques / methods*
  • Cohort Studies
  • Electronic Health Records
  • Hospital Mortality*
  • Humans
  • Middle Aged
  • Predictive Value of Tests
  • Prognosis
  • Retrospective Studies
  • Risk Factors