A phenotypic risk score for predicting mortality in sickle cell disease

Area of Science:

• Cardiology
• Hematology
• Medical Informatics

Background:

• Risk assessment in sickle cell disease (SCD) is complex, relying on physician experience and diverse test results.
• A standardized, data-driven approach is needed to improve prognostic accuracy for SCD patients.

Purpose of the Study:

• To develop and validate a novel risk score for predicting all-cause mortality in adult patients with SCD.
• To integrate clinical, laboratory, and imaging data using machine learning for enhanced risk assessment.

Main Methods:

• Prospective cohort study of 600 adult SCD patients.
• Utilized random survival forest and regularised Cox regression machine learning (ML) to identify mortality predictors from 70 baseline covariates.
• Developed and internally validated multivariable models and a prognostic risk score.

Main Results:

• Identified nine independent predictors of mortality: tricuspid regurgitant velocity, estimated right atrial pressure, mitral E velocity, left ventricular septal thickness, BMI, BUN, alkaline phosphatase, heart rate, and age.
• The developed risk score demonstrated superior performance with a bias-corrected C-statistic of 0.763.
• The model stratified patients into four groups with significantly different 4-year mortality rates (3%, 11%, 35%, 75%).

Conclusions:

• A validated ML-based mortality risk score effectively integrates cardiopulmonary, renal, and liver end-organ damage markers in SCD patients.
• This risk score provides a more objective and reliable method for assessing mortality risk in SCD, aiding clinical decision-making.
• The study highlights the potential of ML in developing personalized risk prediction tools for complex hematological disorders like SCD.