Predicting Intensity of Received Care with Pre-Treatment Patient Characteristics

Hausmann, M.L. (2022) Predicting Intensity of Received Care with Pre-Treatment Patient Characteristics. Bachelor thesis, Psychology.

	Text Thesis ML Hausmann.pdf Restricted to Repository staff only Download (478kB)
	Text Exam Committee Answer.pdf Restricted to Repository staff only Download (91kB)

Abstract

Precision psychiatry is an upcoming field that focusses on the prediction of relevant patient outcomes to enhance clinical decision-making. We applied the framework of precision psychiatry to the relatively unexplored outcome of the intensity of care that patients received. Our setting was a large academic psychiatry centre in which most patients fill out an intake questionnaire, which captures a range of clinically relevant patient characteristics. We used the data from this transdiagnostic sample (n = 2664) to train two statistical models, namely lasso regression and a random forest, to predict the summed-up time of all treatment activities in the documentation system per patient. Cross-validation showed that both models were able to explain 8% and 7% of the variation in the data, respectively. The most relevant predictors from both models were living situation, burden from psychiatric symptoms (especially suicidality), disability from health condition, and medication use. When compared to previous studies, the prediction accuracy of our models was low. Our results also show that more sophisticated statistical models do not necessarily perform better than simpler ones. Using data that was gathered during routine clinical care, our study highlights the importance of selecting relevant variables when building statistical models for precision psychiatry.

Item Type:	Thesis (Bachelor)
Supervisor name:	Booij, S.H. and Bringmann, L.F.
Degree programme:	Psychology
Differentiation route:	None [Bachelor Psychology]
Date Deposited:	14 Jul 2022 07:21
Last Modified:	14 Jul 2022 07:21
URI:	http://gmwpublic.studenttheses.ub.rug.nl/id/eprint/936

Actions (login required)

View Item