This observational study looked at 3,018 adults discharged from a large academic medical center. Researchers tested computer models that use both medical records and social information. They wanted to see if these tools could guess which patients might return to the hospital within 30 days. This approach combines clinical data with factors like where people live.
The computer models worked better than the older math methods used before. The best model correctly identified readmission risks with good accuracy. It found that past hospital visits, health conditions, neighborhood income, and family makeup mattered most for predictions. These social factors helped the computer understand risks better than medical data alone.
This research was based on past records, so it shows links, not causes. The study did not measure if patient health actually improved or if costs went down. It only showed prediction accuracy. Doctors use this to plan care, not to cure illness.
Patients should talk to doctors about their specific risks. These tools might help identify high-risk patients earlier. However, this is not a treatment. It is a way to plan care better.