Research Article | Open Access
Volume 2025 - 4 | Article ID 299 | https://dx.doi.org/10.51521/IJMCCR.2024.e4-2-114
Academic Editor: John Bose
Habeeb Abolaji Bashir *1, George Paul
Komolafe2, Deborah Idowu Akinwolemiwa3
1Department of
Statistics and Data Science, University of Kentucky, Kentucky, USA, ORCID:
0009-0008-2881-2154
2Department of
Computer Science, Boston University. Massachusetts, USA, ORCID:
0009-0001-0413-241X
3Department of Economics, Wayne State University, Detroit, USA, ORCID: 0009-0000-1045-4506
Corresponding Author: Habeeb Abolaji Bashir, Department of Statistics and Data Science,
University of Kentucky, Kentucky, USA, ORCID: 0009-0008-2881-2154.
Citation: Habeeb
Abolaji Bashir, George Paul Komolafe, Deborah Idowu Akinwolemiwa, (2024)
Identifying Patient Subgroups for Personalized Treatments with Model Based
Recursive Partitioning. Int J Med Clin Case Rep, 4(2), 1-6.
Copyright: ©
2024 Habeeb Abolaji Bashir, et al. This is an open-access article
distributed under the terms of the Creative Commons Attribution 4.0
International License, which permits unrestricted use, distribution and
reproduction in any medium, provided the original author and source are
credited.
Abstract
Advances in
precision medicine have highlighted the inadequacy of traditional “one size
fits all” treatment approaches in the face of patient heterogeneity. Different
patients often respond variably to the same therapy due to genetic, clinical,
or environmental factors. This heterogeneity poses a challenge in identifying
which subgroups of patients benefit from a given treatment. We introduce model
based recursive partitioning (MOB) as a statistical approach to automatically
detect patient subgroups with differential treatment effects, addressing gaps
in current subgroup analysis methods. This study uses model based
recursive partitioning (MOB) to identify patient subgroups with
differential treatment effects in a simulated randomized controlled trial. A
dataset of 600 patients, each with baseline demographics, a biomarker and a
severity score, was generated; patients were randomly assigned in approximately
equal numbers to treatment and control (the final counts were 318 and 282
due to chance). Linear regression models were embedded into a recursive
partitioning algorithm to detect effect modifiers. Model performance was
compared against global linear and interaction models as well as a standard
CART tree. MOB correctly recovered the programmed treatment effect
heterogeneity and produced an easily interpretable decision tree. Cross
validated analyses showed that accounting for heterogeneity improved predictive
performance relative to a simple global model. The resulting subgroup rules
could guide personalized treatment strategies and inform future trial designs.
Keywords: Personalized Medicine, Patient stratification, Model based recursive
partitioning, Treatment heterogeneity, Subgroup analysis, Decision trees, Precision medicine
.png)
.gif)
.jpg)
