Alternatives

Issues with hdPS

Univariate selection of many proxies

• Recurrent covariates selected separately / univariately
• can be correlated (coming from same patient) and cause multicollinearity
• may inflate variance
• General overfitting problem. Too many adjustment variables?

(Franklin et al. 2015; Schuster, Lowe, and Platt 2016; Karim, Pang, and Platt 2018)

Potential ways to improve

• Multiple recurrent covariates could provide same information, may not be useful anymore in the presence of others. Multivariate structure could be good to consider in a single model.
• Machine learning variable selection methods could be useful to combat multicollinearity.
• Sample splitting methods could be useful in combating overfitting in high dimensions.

Franklin, Jessica M, Wesley Eddings, Robert J Glynn, and Sebastian Schneeweiss. 2015. “Regularized Regression Versus the High-Dimensional Propensity Score for Confounding Adjustment in Secondary Database Analyses.” American Journal of Epidemiology 182 (7): 651–59.

Karim, Mohammad Ehsanul, Menglan Pang, and Robert W Platt. 2018. “Can We Train Machine Learning Methods to Outperform the High-Dimensional Propensity Score Algorithm?” Epidemiology 29 (2): 191–98.

Schuster, Tibor, Wilfrid Kouokam Lowe, and Robert W Platt. 2016. “Propensity Score Model Overfitting Led to Inflated Variance of Estimated Odds Ratios.” Journal of Clinical Epidemiology 80: 97–106.