Using natural language processing to classify social work interventions

Objectives Health care organizations are increasingly employing social workers to address patients' needs. However, work (SW) activities in health settings largely captured as text data within electronic records (EHRs), making measurement and analysis difficult. This study aims extract classify, from EHR notes, interventions intended needs using natural language processing (NLP) machine learning (ML) algorithms. Study design Secondary of a longitudinal cohort. Methods We extracted 815 SW encounter notes the system federally qualified center. reviewed literature derive 10-category classification scheme for interventions. applied NLP ML algorithms categorize documented according scheme. Results Most (n = 598; 73.4%) contained at least 1 intervention. The most frequent offered by included coordination (21.5%), education (21.0%), financial planning (18.5%), referral community services (17.1%), supportive counseling (15.3%). High-performing kernelized support vector (SVM) (accuracy, 0.97), logistic regression 0.96), linear SVM 0.95), multinomial naive Bayes classifier 0.92). Conclusions can be utilized automated identification EHRs. administrators leverage this approach gain better insight into needed patient population served their organizations. Such information managerial decisions related staffing, resource allocation,