Table 1 A summary of the challenges faced by epidemiologists when conducting electronic health record-based research, their manifestations in terms of threats to validity, and potential solutions
From: A narrative review on the validity of electronic health record-based research in epidemiology
Challenge | Sub-challenge | Example | Threat(s) to validity | Potential solution(s) |
|---|---|---|---|---|
#1 Representativeness | -- | Catchment of a federally qualified health center versus academic medical center | Selection bias and generalizability | Comparison to external data; Inverse probability weighting for selection bias |
#2 Data availability and interpretation | 2.1 Billing versus Clinical versus Epidemiological Needs | Presence or absence of diagnostic codes | Information bias and confounding | Validation study; quantitative bias analysis |
2.2 Consistency in Data and Interpretation | Variations in reported laboratory results | Information bias and confounding | Validation study; quantitative bias analysis | |
2.3 Unstructured Data: Clinical Notes and Reports | Operationalizing phenotypes from the encounter note | Information Bias and confounding | Natural language processing | |
#3 Missing measurements | -- | Socioeconomic status not captured | Information or Selection Bias, Confounding | Imputation, Surrogate Measures, Validation Study |
#4 Missing visits | -- | Lack of longitudinal view of patient | Information or Selection Bias | Imputation, Surrogate Measures, Validation Study |