Machine learning and SHAP values explain the association between social determinants of health and post-stroke depression

Comprehensive Summary

This study aims to create a machine learning (ML) model that finds a correlation between social determinants of health (SDoH) with levels of post-stroke depression (PSD). The data for this study was acquired from the National Health and Nutrition Examination Survey. Logistic regression was used to determine the correlation between SDoH and PSD, while Cox Regression was used to assess the correlation between SDoH and all-cause mortality in PSD. Four ML models were used (CatBoost, Logistic, Multilayer Perceptron, and Random Forest) to predict the effectiveness and clinical applicability of these ML models. In the results, the logistic regression revealed a positive correlation between SDoH and PSD. Of the four models, CatBoost had the best predictive performance, achieving an AUC of 0.966 and was also shown to have considerable clinical utility and predictive efficacy. In the discussion, the authors speak about how their findings strongly align with current knowledge on the topic.

Outcomes and Implications

This research is important as it can help clinicians further understand how SDoH effect PSD. By identifying these factors, healthcare professionals can identify signs of PSD earlier which can help with more effective interventions. For future longitudinal studies, the researchers recommend that they focus on whether interventions targeting SDoH can reduce the levels of PSD and explore other ways interventions can help improve the overall health of stroke patients.