Dimensionality Reduction Techniques for Improving Propensity Score Specification: An Application to a Cohort Study Using Claims Data
Dimensionality reduction techniques, particularly autoencoders, outperformed traditional methods in achieving covariate balance when estimating PS in high-dimensional claims data. These methods may offer improved covariate balance in pharmacoepidemiological studies PS-matched designs in large healthcare databases.
Authors
- Stefan Franzén (ORCID: https://orcid.org/0000-0002-9640-0927)
- Lucy Carty (ORCID: https://orcid.org/0000-0003-2849-3145)
- Mario Ouwens (ORCID: https://orcid.org/0000-0002-8357-8080)
- Eva Lesén (ORCID: https://orcid.org/0000-0002-2198-4382)
- Sima Shahsavari (ORCID: https://orcid.org/0000-0002-6377-3284)
- Jonatan Nåtman (ORCID: https://orcid.org/0000-0001-6242-1810)
- Sudhir Venkatesan (ORCID: https://orcid.org/0000-0002-2093-5781)
Institutions
- AstraZeneca (United Kingdom) (GB)
- AstraZeneca (France) (FR)
- AstraZeneca (Brazil) (BR)
- AstraZeneca (Switzerland) (CH)
- AstraZeneca (Italy) (IT)
- AstraZeneca (Sweden) (SE)
Publication Details
- Journal
- Pharmacoepidemiology and Drug Safety
- Published
- 2025-11-24
- DOI
- https://doi.org/10.1002/pds.70258
- Primary Topic
- Statistical Methods and Bayesian Inference
- Type
- article
- Field-Weighted Citation Impact
- 0.00