Blood cell differential count discretization methods to predict survival in acutely ill adults reporting to the emergency room: a retrospective cohort study in 2020

Abstract

Aims: To assess survival predictivity of baseline blood cell differential count (BCDC) discretization methods in acutely ill adults visiting the emergency room over one-year. Methods: Retrospective cohort study on one-year survival of adults reporting to the emergency room of the A. Manzoni Hospital (Italy) during 2020. Automated BCDC analysis performed at baseline, assessed hemoglobin, red cell mean volume and distribution width (RDW), platelet distribution width (PDW), platelet-hematocrit, absolute red blood cells, white blood cells, neutrophils, lymphocytes, monocytes, eosinophils, basophils, and platelets. Discretization cutoffs were defined by: Benchmark laboratory reference values and Tailored (maximally selected rank statistics for linear and sigmoid-shaped distributed variables; optimal-equal hazard ratio (HR) method for U-shaped distributed variables. Explanatory variables (age, gender, inward admission) were analyzed using Cox multivariable regression. Receiver operating characteristic curves used the sum of Cox-significant variables in each method. Results: Of 11052 patients (median age 67 years, interquartile range (IQR) 51-81, 48% female), 59% (n=6489) were discharged and 41% (n=4563) were admitted. After a 306-day median follow up (IQR 208-417 days), 9455 (86%) patients were alive and 1597 (14%) deceased. Increased HRs were associated with age > 73 years (HR=4.29 CI 3.78-4.87), and hospital admission (HR=2.05, CI 1.83-2.29). Age, sex, hemoglobin, mean corpuscular volume, RDW, PDW, neutrophils, lymphocytes and eosinophils were significant in overall. Benchmark included basophils and platelet count (area under the ROC curve (AUROC) 0.78). Tailored included monocyte counts and PCT (AUROC of 0.82). Conclusions: Tailored discretization of BCDC provided meaningful insight regarding acute patient survival.