Risk markers by sex and age group for in-hospital mortality in patients with STEMI or NSTEMI: an approach based on machine learning

by   Blanca Vazquez, et al.

Machine learning (ML) has demonstrated promising results in the identification of clinical markers for Acute Coronary Syndrome (ACS) from electronic health records (EHR). In the past, the ACS was perceived as a health problem mainly for men and women were under-represented in clinical trials, which led to both sexes receiving the same clinical attention. Although some approaches have emphasized the importance of distinguishing markers, these distinctions remain unclear. This study aims at exploiting ML methods for identifying in-hospital mortality markers by sex and age-group for patients with ST-elevation myocardial infarction (STEMI) and the Non-ST-elevation myocardial infarction (NSTEMI) from EHR. From the MIMIC-III database, we extracted 1,299 patients with STEMI and 2,820 patients with NSTEMI. We trained and validated mortality prediction models with different hyperparameters, clinical sets, and ML methods. Using the best performing model and a game-theoretic approach to interpret predictions, we identified risk markers for patients with STEMI and NSTEMI separately. The models based on Extreme Gradient Boosting achieved the highest performance: AUC=0.92 (95% CI:0.87-0.98) for STEMI and AUC=0.87 (95% CI:0.80-0.93) for NSTEMI. For STEMI, the top markers for both sexes are the presence of hyponatremia, and metabolic acidosis. More specific markers for women are acute kidney failure, and age>75 years, while for men are chronic kidney failure, and age>70 years. In contrast, for NSTEMI, the top markers for both sexes are advanced age, and intubation procedures. The specific markers for women are low creatinine levels and age>60 years, whilst, for men are damage to the left atrium and age>70 years. We consider that distinguishing markers for sexes could lead to more appropriate treatment strategies, thus improving clinical outcomes.


Identification of Pediatric Sepsis Subphenotypes for Enhanced Machine Learning Predictive Performance: A Latent Profile Analysis

Background: While machine learning (ML) models are rapidly emerging as p...

FineEHR: Refine Clinical Note Representations to Improve Mortality Prediction

Monitoring the health status of patients in the ICU is crucial for provi...

Machine learning techniques to identify antibiotic resistance in patients diagnosed with various skin and soft tissue infections

Skin and soft tissue infections (SSTIs) are among the most frequently ob...

Diabetes Mellitus Forecasting Using Population Health Data in Ontario, Canada

Leveraging health administrative data (HAD) datasets for predicting the ...

Please sign up or login with your details

Forgot password? Click here to reset