The simplest solution can give the best results

Health care is only a small part of a person’s overall well-being and health, but it is difficult to understand and use external factors such as socio-economic status and access to health. food within the health care system. is. Machine learning models can help fill this gap. Richard M. of the Regenstrief Institute and Indiana University. A study from the Fairbanks School of Public Health used the most basic raw data to create the best risk prediction model for primary care physicians to identify patients who might need help. paddy field.

“There is a lot of data available on the social determinants of health. healthBut the challenge is to turn it into something Health provider Project leader Joshua R. Researchers from Vest, Ph.D., MPH, Regenstrief and professors at IU Fairbanks School of Public Health said: to predict model Perhaps the most effective. “

The research team created an algorithm to predict patient referrals by comparing the performance of six approaches to measuring the social determinants of health at the community level. Social worker Hospitalization after a primary care visit. They tested the algorithm with data from encounters with approximately 210,000 patients at Eskenazi Health, a safety net health system in Indianapolis, Indiana.

They found that models using several individual community-level components, such as raw data on housing, income, and education, produced the most accurate predictive models. Other models combined the case with other components to create a single large image. This is more complex and less precise in identifying people who may be at risk.

“This study shows that using raw data is a viable solution for creating benchmark models,” said Dr Best. “In the future, we may need to develop more complex models, which can be a powerful starting point for health systems, especially those with few resources, to start taking advantage of the social determinants of health. health. There is sex. “

The next step in this study is to determine if this approach works in situations other than social work referrals and hospitalization forecasts, and to examine additional data that can be added to this algorithm.

“Choosing a Measurement Approach for Local Social Determinants of Health and Risk Prediction Model Performance” is published online. Informatics for health and social..