A study published in JAMA Cardiology incorporated social determinants of health into a machine learning model to predict the risk of heart failure and mortality after hospitalization.
The prediction of hospital mortality in patients with heart failure regularly uses logistic regression models. To improve these predictions, the authors of the study incorporated the use of Social Determinants of Health in machine learning models, since traditional models do not take into account this type of data.
The objective of the study was to develop and validate new machine learning models that incorporate social determinants of health, to predict mortality from heart failure. The study seeks to respond to the racial disparities that exist in this type of research that incorporates new technologies such as Artificial Intelligence. In this way, through the social determinants of health, the prediction of mortality from heart failure can be improved, especially in black patients.
The data on hospitalizations for heart failure corresponded to information between January 1, 2010 and December 31, 2020 from a clinic specialized in heart failure. The authors analyzed data between January 2021 and April 2022 and performed an external validation with hospitalization data from a study that used a cohort with data between 2005 and 2014.
Machine learning models were trained on this data, and the authors categorized the patients into African Americans and non-African Americans, where 91% of the patients were white.
Machine learning showed superior performances to traditional models in both racial groups. And by adding the social determinants of health, they improved the prognostic utility of the models in black patients, but not in non-black patients, which highlights the importance of this type of variable in medical research.
