Prediction-based Inference in Electronic Health Record (EHR)-linked Biobanks

Abstract

Electronic health record (EHR)-linked biobank data facilitate large-scale scientific discoveries such as genome-wide association study (GWAS) on a multitude of phenotypic traits and biomarkers routinely captured in EHR. However, heterogeneous missingness in biomarkers may bias analyses when used as phenotypes. Recently proposed prediction-based (PB) inference methods incorporate external machine ML models to impute missing biomarkers and thereby improve the statistical power and estimation precision in association analyses. Yet, it remains unclear that if these methods are still preferable when the outcome $Y$ is generated under a clinically informative observation process. A comprehensive comparative evaluations and theoretical understanding of the existing PB methods under such realistic EHR missingness mechanisms are lacking. In this paper, we conduct a structured review of 8 recently developed PB methods and categorize them based on their frameworks. We systematically evaluate 9 methods, including 4 PB methods and 5 baseline methods from the traditional missing-data approaches, under 10 different outcome observation process models across different missing assumptions for continuous and binary outcomes. Our results show that PB methods can substantially improve statistical power and estimation efficiency when their missingness assumptions hold, but they may require stronger assumptions than CCA to control type I error. We provide theoretical results to characterize the scenarios under which CCA or PB methods may remain valid. Finally, we apply CCA, weighted CCA, and two PB methods to perform GWAS of six laboratory biomarkers in the All of Us data. The results demonstrate that PB methods can replicate known genetic associations while improving efficiency relative to (weighted) CCA. Furthermore, they extend inferential results to a more representative study population.

Figures (13)

• Figure 1: Directed acyclic graphs (DAGs) of the full data-generating process under different observation models. Dashed circles and arrows represent relationships that occur only in the binary-outcome setting. Red arrows indicate the missingness mechanism (i.e., which variables the missingness indicator depends on).
• Figure 2: Decision tree of whether valid type I error control and consistent estimation can be achieved under different scenarios. Abbreviations are as follows: prediction-based (PB) approaches and complete-case (CC) approaches.
• Figure S1: Mean power of the included methods for $\beta_1$ and $\beta_2$ under different quality of imputations in the linear regression setting when the confounders are continuous. Each column in the panel represents a type of observation model. Within each subplot, X-axis represents the power of the selected methods, and y-axis represents the group of each methods under different imputation quality.
• Figure S2: Type I error of the included methods for $\beta_1$ and $\beta_2$ under different quality of imputations in the linear regression setting when the confounders are continuous. Each column in the panel represents a type of observation model. Within each subplot, X-axis represents the power of the selected methods, and y-axis represents the group of each methods under different imputation quality.
• Figure S3: Mean power of the included methods for $\beta_1$ and $\beta_2$ under different quality of imputations in the linear regression setting when the confounders is a categorical variable. Each column in the panel represents a type of observation model. Within each subplot, X-axis represents the power of the selected methods, and y-axis represents the group of each methods under different imputation quality.

Theorems & Definitions (4)

• Theorem 1
• Theorem 2
• Theorem 3
• proof