Leveraging Hematologic Single-Cell Measurements for Patient Triage and Outcome Prediction

Area of Science:

• Hematology
• Clinical Pathology
• Medical Informatics

Background:

• The complete blood count (CBC) is a cornerstone diagnostic tool in medicine.
• Modern hematology analyzers provide extensive cell population data (CPD) beyond standard CBC markers.
• The prognostic utility of CPD in general patient populations remains underexplored.

Purpose of the Study:

• To determine if cell population data (CPD) markers offer prognostic value beyond traditional complete blood count (CBC) markers.
• To assess the association of CPD markers with general patient outcomes.
• To evaluate the impact of incorporating CPD markers into predictive models.

Main Methods:

• Retrospective analysis of CBC and CPD markers from over 10,000 patients.
• Univariate and multivariate analyses of marker associations with inpatient admission, mortality, and length-of-stay.
• Machine learning models were developed using CBC and CPD markers for outcome prediction.

Main Results:

• Numerous CPD markers demonstrated significant associations with mortality, length-of-stay, and emergency department admissions.
• CPD markers maintained prognostic significance across diverse patient demographics and medical specialties, even after adjusting for CBC markers.
• Machine learning models incorporating CPD markers showed improved predictive performance (AUC 0.79 for mortality, 0.81 for admission).
• Two distinct CPD phenotypes were identified: inflammatory (linked to admission) and dysregulatory (linked to mortality).

Conclusions:

• Routinely available cell population data (CPD) can significantly enhance the prognostic capabilities of the complete blood count (CBC).
• CPD markers provide valuable information for evaluating general patient cohorts, complementing standard hematological assessments.
• The findings support the integration of CPD analysis into routine clinical practice for improved patient risk stratification.