Comparison of statistical and machine learning models for healthcare cost data: a simulation study motivated by Oncology Care Model (OCM) data

Table 1 Model characteristics for Gamma GLM, PLAQR, and Random Forest (RF) models

	Gamma GLM	PLAQR	Random Forest
Distribution assumption	Parametric	Semi-parametric	Nonparametric
Estimate	Mean	Quantile	Mean
Ability to model skewed outcome	Yes	Yes	Yes
Non-linear effect	Needs to be specified through model diagnostics	Needs to be specified (B-spline)	Data-driven detection; pre-specification not needed
Interaction effect	Needs to be specified through model diagnostics	Needs to be specified through model diagnostics	Data-driven detection; pre-specification not needed
Software	R, SAS, STATA	R	R, SAS

ISSN: 1472-6963