Prediction of Future Health Care Utilization Through Note-Extracted Psychosocial Factors

David A. Dorr
Ana R. Quiñones
Taylor King
Melissa Y. Wei
Kellee White
Cosmin A. Bejan
Peer-Reviewed Article
June 2022


Natural language processing of patient visit notes can help providers identify social factors that may lead to health care utilization for older adults with multiple chronic conditions.


Many older adults who have multiple chronic conditions also experience psychosocial factors, such as social isolation, chronic stress, financial insecurity, and housing, that worsen poor health outcomes. Structured assessments of social needs are inconsistently applied in health care settings, so using natural language processing artificial intelligence that understands human language is a potential solution. Natural language processing can identify social and behavioral issues from patient visit notes using keywords. This study evaluated approximately 75,000 patient records using natural language processing to identify psychosocial factors from medical charts and explore how these factors can predict health care utilization.


Patients with identified psychosocial factors were more likely to be hospitalized and have emergency department visits, and equally likely to have office visits, indicating that psychosocial factors can be used as predictors of utilization outcomes as well as mortality.


Natural language processing can help health care providers identify social and behavioral factors that will impact chronic conditions and health care utilization.

Posted to The Playbook on