Shalmali Joshi

Predictive Models for Behavioral Health Disengagement

Abstract:

We will explore the application of Machine Learning predictive models in forecasting behavioral health disengagement. This remains a persistent problem in 50-70% of patients and is a factor in the deterioration of outcomes and increase in healthcare costs. This study will solve the shortcomings of conventional manual and subjective disengagement identification techniques by evolving evidence-based prediction models that can detect at-risk individuals at an earlier stage. The study analyses multiple machine learning algorithms/methods, such as the use of Random Forests, Support Vector machines, Neural Networks, and Gradient Boosting, using a multi-source dataset of over 10,000 behavioral health patients. Preprocessing, engineering of features, and cross-validation of the model were done in a structured way to guarantee that the model is reliable. All findings indicate that the fully optimized model had a validated accuracy of 92%, which is sustained with high precision, recall, and AUC measures, and it is superior to the rule-based and clinician-directed models of assessment. The keynote will emphasize the practical importance of integrating ML-based early-warning systems into the clinical workflow in order to facilitate proactive outreach, minimize missed appointments, and retain patients. 

Profile:

I have been working as an Senior Advanced Analytics Analyst with Elevance Health specifically as part of the Behavioral Health team. I have completed a Master's in Data Science and Analytics from Georgia State University. I have a sound knowledge of ETL pipeline development with Snowflake, Teradata , Hive. I have good hands on experience handling non relational databases like Mongo and DocumentDB. I have expertise  in handling automation for relational as well as non relational databases using services like AWS Glue and EMR clusters. With total IT experience spanning 15+ years, I have worked on Data engineering teams for Fortune 500 companies. Along with data science,  I have sound experience with agile project implementation and technical leadership.