(163) Methodological framework to conduct outcome validation in administrative or insurance claims and/or electronic healthcare record databases: a practical guide for pharmacoepidemiologists

Background: Misclassification of outcome leads to bias when assessing the effects of medicinal products in structured coded healthcare databases such as administrative or insurance claims and electronic healthcare record databases. Medical chart extraction and review is considered the gold standard outcome validation method, but usually represents a lengthy and resource intensive process. It may be infeasible, especially for rare outcomes, or when none or few of the relevant medical charts are accessible.

Objectives: Considering these limitations, and especially when rapid safety assessment of medicinal products is needed, alternative approaches may be used to validate outcomes. This work provides a framework to guide pharmacoepidemiologists through a landscape of outcome validation methodologies.

Methods: A targeted review of the literature has been conducted in 2023 to identify methods that have been used so far to validate outcome code algorithms/phenotypes in health care database research. Both peer-reviewed scientific publications and publicly available methodological guidelines have been screened and critically assessed for inclusion.

Results: In addition to the “gold standard” outcome validation method based on medical chart extraction and review, 5 alternative approaches have been identified, namely (1) code algorithm/phenotype validation through electronic health record or administrative/insurance claims longitudinal patient profile review, (2) code algorithm/phenotype validation through generation of a probabilistic reference standard, (3) validation through database linkage, (4) extrapolation of a previous validation study, and (5) sensitivity analysis using broad and narrow code algorithms/phenotypes. All methods are integrated in a practical decision tree, which includes key feasibility assessment criteria. Strengths and limitations of each method are discussed.

Conclusions: In recent research, several alternative approaches to validate outcomes identified in healthcare databases were demonstrated to be feasible and recommended. Upon feasibility assessment, these methods may enable future validation efforts when medical chart review is infeasible or unnecessary, and/or under time and resource constraints.