Caveats on Using Firth's Penalization in the Model‐Based Regression Standardization for Rare Diseases
Statistics in Medicine · 2026-07 · 45 (15-17)
DOI: 10.1002/sim.70644 ↗Kurz gefasst
Fachartikel in der Zeitschrift Statistics in Medicine, online erschienen am 23.06.2026. Verfasst von Sotaro Hashibe und 2 weiteren Autor:innen. Beteiligte Einrichtung: Statistics and Decision Sciences Japan, R&D, Janssen Pharmaceutical K.K., Tokyo, Japan. Hauptthemen laut MEDLINE: Rare Diseases, Models, Statistical. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 38 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Japan Society for the Promotion of Science, Japan Agency for Medical Research and Development.
Abstract (Originalsprache)
Model‐based regression standardization, also known as the parametric g‐formula, is widely used to estimate marginal effect measures. However, in rare disease settings, the small number of observed events relative to the number of covariates can lead to (quasi‐)complete separation, resulting in non‐convergent estimates in the regression models. Firth's penalized likelihood estimates are a common solution to this issue, ensuring finite parameter estimates even for separated data. While effective for estimating regression coefficients, Firth's method introduces bias into model‐based regression standardization because of its tendency to shrink the predicted probabilities to 0.5, leading to discrepancies between the predicted and observed event rates. We examined the implications of applying Firth's method to model‐based regression standardization, illustrating its potential bias through an empirical study on surgical site infections (SSI) in orthopedic surgeries. We also proposed two ad hoc corrections (i.e., Firth's logistic regression with intercept correction and added covariate) to mitigate this bias and evaluated these methods via simulation studies, comparing them with propensity score‐based approaches. Finally, we applied the proposed method to assess the association between SSI, a rare disease, and smoking status in a clinical database of orthopedic surgeries.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Statistics and Decision Sciences Japan, R&D, Janssen Pharmaceutical K.K., Tokyo, Japan.
- Department of Information and Computer Technology, Graduate School of Engineering, Tokyo University of Science, Tokyo, Japan.
- Advanced Quantitative Sciences Japan, Development Japan, Novartis Pharma K.K., Tokyo, Japan.
- Interfaculty Initiative in Information Studies, the University of Tokyo, Tokyo, Japan.
- Department of Biostatistics, School of Public Health, Graduate School of Medicine, the University of Tokyo, Tokyo, Japan.
ORCID-Profile: Sotaro Hashibe, Wataru Hongo, Tomohiro Shinozaki
Förderung
- Japan Society for the Promotion of Science
- Japan Agency for Medical Research and Development
Zitieren
APA
Hashibe, S., Hongo, W., & Shinozaki, T. (2026). Caveats on Using Firth's Penalization in the Model‐Based Regression Standardization for Rare Diseases. Statistics in Medicine, 45(15-17). https://doi.org/10.1002/sim.70644
Vancouver
Hashibe S, Hongo W, Shinozaki T. Caveats on Using Firth's Penalization in the Model‐Based Regression Standardization for Rare Diseases. Statistics in Medicine. 2026;45(15-17). doi:10.1002/sim.70644
BibTeX
@article{Hashibe2026,
author = {Hashibe, Sotaro and Hongo, Wataru and Shinozaki, Tomohiro},
title = {Caveats on Using Firth's Penalization in the Model‐Based Regression Standardization for Rare Diseases},
journal = {Statistics in Medicine},
year = {2026},
volume = {45},
number = {15-17},
doi = {10.1002/sim.70644}
}
EUVIMED