Practical laboratory-based clinical decision tools and associations with short-term bleeding and mortality outcomes

Abstract

Background: The red cell distribution width (RDW) predicts mortality in numerous populations. The Intermountain Risk Scores (IMRS) predict patient outcomes using laboratory measurements including RDW. Whether the RDW or IMRS predicts in-hospital outcomes is unknown.

Methods: The predictive abilities of RDW and two IMRS formulations (the complete blood count [CBC] risk score [CBC-RS] or full IMRS using CBC plus the basic metabolic profile) were studied among percutaneous coronary intervention patients at Intermountain (males: N = 6007, females: N = 2165). Primary endpoints were a composite bleeding outcome and in-hospital mortality.

Results: IMRS predicted the composite bleeding endpoint (females: χ² = 47.1, odds ratio [OR] = 1.13 per +1 score, p < 0.001; males: χ² = 108.7, OR = 1.13 per +1 score, p < 0.001) more strongly than RDW (females: χ² = 1.6, OR = 1.04 per +1%, p = 0.20; males: χ² = 11.2, OR = 1.09 per +1%, p < 0.001). For in-hospital mortality, RDW was predictive in females (χ² = 4.3, OR = 1.13 per +1%, p = 0.037) and males (χ² = 4.4, OR = 1.11 per +1%, p = 0.037), but IMRS was profoundly more predictive (females: χ² = 35.5, OR = 1.36 per +1 score, p < 0.001; males: χ² = 72.9, OR = 1.40 per+1 score, p < 0.001). CBC-RS was more predictive than RDW but not as powerful as IMRS.

Conclusions: The IMRS, the CBC-RS, and RDW predict in-hospital outcomes. Risk score-directed personalization of in-hospital clinical care should be studied.

Keywords: Decision tool; Electronic health record; Laboratory-based risk tool; Learning health system; RDW; Risk prediction.