Exploring opportunities to improve health equity with machine learning and artificial intelligence in healthcare epidemiology.

Abstract

Artificial intelligence (AI) and machine learning (ML) have the potential to improve diagnostic accuracy, infection control protocols, and health outcomes at the population level. AI/ML also has the potential to exacerbate existing health disparities, underscoring the necessity of developing, testing, and evaluating models with equity-centered frameworks. In this review, we highlight current work and future directions in implementing AI/ML to improve health equity in healthcare epidemiology. AI/ML models can improve the quality of data collected in electronic health records, especially with regard to race and ethnicity, aid in identifying vulnerable populations via classification modeling, and help guide equitable intervention protocols by identifying disparities through predictive modeling. However, data-, algorithm-, and deployment-centric biases must be considered at every step of model synthesis, utilizing fairness metrics to assess the presence of bias and employing ideal mitigation strategies to create equitable AI/ML models for healthcare epidemiology.