Comorbidity Exposure-Window Definitions and Multidimensional Disparities in Long COVID Risk: Evidence from a U.S. National Cohort (2020-2024)

Long COVID (LC) affects millions of individuals worldwide, particularly those with preexisting comorbidities. However, when assessing comorbidity-related risk factors for LC, it remains unclear whether preexisting comorbidities should be defined prior to SARS-CoV-2 infection or LC diagnosis. This methodological choice may substantially bias estimates of LC risk. In this study, leveraging Electronic Health Records (EHR) from 6,130,413 adults with documented COVID-19 across 49 U.S. states in the National Clinical Cohort Collaborative (N3C) from 2020 to 2024, we evaluated the impact of different comorbidity exposure-window definitions on LC risk estimates. We developed ensemble cross-fitted double/debiased machine learning to adjust for individual- and county- level confounders. We found that defining comorbidities before infection rather than before LC diagnosis yielded 8%-314% higher adjusted attributable risks and 3%-34% higher adjusted relative risks across 16 major comorbidities. Comorbidity-associated risks generally declined from 2020 to 2024, while substantial demographic, socioeconomic, geographic, and multimorbidity-related disparities persisted. Exposure-window specification can be a major source of bias in estimates of LC risk, particularly for comorbidities prone to delayed clinical recognition or underascertainment. More broadly, our results highlight how temporal misclassification of exposures in longitudinal EHR studies can distort risk attribution and population-level inference for risk stratification for LC and prioritization of prevention strategies in clinical medicine and public health, respectively.