Applying machine learning and text analysis to identify factors that may predict hypertensive heart disease patient outcomes in home healthcare

Abstract

This research focuses on predicting the patient discharge disposition with initial patient assessment and therapy data as well as determining which therapy intervention text had positive impacts on hypertension heart disease patients in home healthcare environments. Older adults prefer to stay in their home, which is known as aging in place. Home healthcare is the last line of defense before advancing to other expensive healthcare options. This research used aggregate transactional data from 2,181 home healthcare patients in the United States (U.S.) from 2016-2022. We used the Centers for Disease Control and Prevention (CDC) Patient Driven Groupings Model and focused on the cardiac circulatory patient’s subcategory of hypertensive heart disease. Data was analyzed from Activity of Daily Life (ADL) assessment scores, the number of disease diagnosis codes per patient, the number of additional cardiac comorbidities, gender, age, standardized hospitalization risks, number of medications per patient, number of interventions per patient, and the length of stay in home healthcare. Machine learning and advanced text analysis were applied to determine which factors and therapy intervention text had the biggest impact on hypertensive heart disease patient outcomes. This research also identified those interventions with the best Signal to Noise (SN) ratios that are currently being piloted in home healthcare settings.