Predicting Service-Use in Mental Healthcare: A Machine Learning Approach
Keywords
Loading...
Authors
Issue Date
2021-07-01
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
We predicted service-use, particularly High Care Use (HCU), for insurance agreement
purposes, at Pro Persona mental healthcare using a) models by van Mens and colleagues (in
press), b) extensions of those models, and c) adjusted models to predict HCU status of
patients. We trained random forest models on patient data starting care in 2018 (N = 9,816)
and validated it on data from patients starting in 2019 (N = 10,095). The predictive accuracy
matched the accuracy reported in the original study (van Mens et al., in press) although our
extended models were not more accurate. Models outperformed predictions based on the
totals of last year. Our HCU classifier performed best in classifying the top 20% of patients in
terms of service-use though the clinical utility was not sufficient to warrant application. We
conclude the models by van Mens and colleagues can be used at the insurance level within the
Pro Persona context.
Keywords: machine learning, service-use, high care use, mental healthcare, cross validation, random forest
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen