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

Table 3 Relative Contribution^a of Candidate Predictor Groups to Regression Model Fit for resource intensive healthcare outcomes

Candidate Predictor Domain	Emergency Room Visits^b (%) ^c	Inpatient Days^b (%) ^c	Hospital Expenditures^b (%) ^c
Demographics	22.70	24.40	18.38
Adult & Child Health Characteristics	23.99	32.13	43.23
Community	19.71	16.46	15.02
Consumer Expenditures	33.60	27.01	23.37

^aThe relative contributions of variables from each candidate predictor group are assessed by measuring the difference in R² from the full model minus the R² from the reduced model containing variables from 1 of the 4 candidate predictor groups
^bAll outcomes expressed as annual log per capita values from 2017
^cThe percentage from each group represents the percent contribution to the full model, for each outcome

ISSN: 1472-6963