Predicting Long-Term Mortality after Acute Coronary Syndrome Using Machine Learning Techniques and Hematological...

Area of Science:

• Cardiology
• Medical Informatics
• Hematology

Background:

• Hematological indices, including red cell distribution width (RDW) and neutrophil-to-lymphocyte ratio (NLR), are linked to outcomes in acute coronary syndrome (ACS).
• The predictive value of machine learning (ML) for mortality in ACS using these hematological features remains underexplored.

Purpose of the Study:

• To develop an alternative risk assessment tool for ACS mortality prediction.
• To utilize easily obtainable features, specifically hematological indices and inflammation markers.

Main Methods:

• A machine learning classifier was trained on data from 5053 patients hospitalized with ACS over 5 years.
• The model predicted in-hospital, 180-day, and 365-day mortality.
• Performance was compared against the Global Registry of Acute Coronary Events (GRACE) Score 2.0 on a test dataset.

Main Results:

• The ML model achieved a c-statistic of 0.89 for in-hospital mortality (GRACE 2.0: 0.90).
• For six-month mortality, the ML model achieved a c-statistic of 0.77 (GRACE 2.0: 0.73).
• RDW and NLR were independently associated with all-cause mortality (P < 0.001).

Conclusions:

• Hematological markers, including neutrophil count and RDW, are strongly associated with all-cause mortality post-ACS.
• A machine-learned model utilizing these parameters offers accurate long-term mortality predictions.
• This ML approach demonstrates comparable or superior accuracy to established risk scores.