Can diverse population characteristics be leveraged in a machine learning pipeline to predict resource intensive healthcare utilization among hospital service areas?

Table 1 Best Performing Models for resource intensive healthcare outcomes in 2017 among Hospital Service Areas

Outcome (log per capita)	ER Visits (N = 3,153)	Inpatient Days (N = 3,174)	Hospital Expenditures (N = 3,174)
Candidate predictor groups included ^a	4	4	4
Feature Selection	LASSO^d	LASSO^d	LASSO^d
Model Type	Random Forest	LASSO^d	Gradient Boosting Machines
MSE^b	0.003	0.011	0.004
R^{2 c}	0.247	0.184	0.782

^aCandidate predictor groups: 1. Demographics, 2. Adult & Child Health Characteristics, 3. Community Characteristics, and 4. Consumer Expenditure Variables
^bMSE = mean squared error, calculated on test-set
^cCoefficient of determination, calculated on test-set
^dLeast Absolute Shrinkage and Selection Operator

ISSN: 1472-6963