biostudies-literatureKhurshid SAmerican Heart AssociationFondation LeducqAmerican Heart Association-American Stroke AssociationNHLBI NIH HHSU.S. Department of Health & Human Services | NIH | National Heart, Lung, and Blood InstituteNINDS NIH HHSU.S. Department of Health & Human Services | NIH | National Heart, Lung, and Blood Institute (NHLBI)American Heart Association (American Heart Association, Inc.)47https://www.ebi.ac.uk/biostudies/studies/S-EPMC8993873biostudies-literatureUnknown5(1)Electronic health record (EHR) datasets are statistically powerful but are subject to ascertainment bias and missingness. Using the Mass General Brigham multi-institutional EHR, we approximated a community-based cohort by sampling patients receiving longitudinal primary care between 2001-2018 (Community Care Cohort Project [C3PO], n = 520,868). We utilized natural language processing (NLP) to recover vital signs from unstructured notes. We assessed the validity of C3PO by deploying established risk models for myocardial infarction/stroke and atrial fibrillation. We then compared C3PO to Convenience Samples including all individuals from the same EHR with complete data, but without a longitudinal primary care requirement. NLP reduced the missingness of vital signs by 31%. NLP-recovered vital signs were highly correlated with values derived from structured fields (Pearson r range 0.95-0.99). Atrial fibrillation and myocardial infarction/stroke incidence were lower and risk models were better calibrated in C3PO as opposed to the Convenience Samples (calibration error range for myocardial infarction/stroke: 0.012-0.030 in C3PO vs. 0.028-0.046 in Convenience Samples; calibration error for atrial fibrillation 0.028 in C3PO vs. 0.036 in Convenience Samples). Sampling patients receiving regular primary care and using NLP to recover missing data may reduce bias and maximize generalizability of EHR research.NPJ digital medicineCohort design and natural language processing to reduce bias in electronic health records research.PMC8993873K24HL1057801R01HL092577T32HL007208R01HL134893R01 NS103924K24 HL10578014CVD01U01NS06967318SFRN34110082R01 HL139731K24HL153669K01HL148506R01HL128914R01HL140224T32 HL007208K23 HL159243R01NS103924R01 HL134893R38HL150212R01 HL0925771R01HL139731K01 HL14850618SFRN34250007R38 HL150212K24 HL153669R01 HL140224R01 HL12891421SFRN812095Klarqvist MDRMielke JGhadessi MHarrington LXHaimovich JSDi Achille PKhurshid SAnderson CDSingh PAshburner JMPhilippakis AASarma GAl-Alusi MAWang XDiedrich CLubitz SATurner ACEllinor PTFriedman SFCunningham JWBatra PLau ESDiamant NMcElhinney ADerix AAtlas SJReeder CHo JEEilken HMfalseCohort design and natural language processing to reduce bias in electronic health records research.Electronic health record (EHR) datasets are statistically powerful but are subject to ascertainment bias and missingness. Using the Mass General Brigham multi-institutional EHR, we approximated a community-based cohort by sampling patients receiving longitudinal primary care between 2001-2018 (Community Care Cohort Project [C3PO], n = 520,868). We utilized natural language processing (NLP) to recover vital signs from unstructured notes. We assessed the validity of C3PO by deploying established risk models for myocardial infarction/stroke and atrial fibrillation. We then compared C3PO to Convenience Samples including all individuals from the same EHR with complete data, but without a longitudinal primary care requirement. NLP reduced the missingness of vital signs by 31%. NLP-recovered vital signs were highly correlated with values derived from structured fields (Pearson r range 0.95-0.99). Atrial fibrillation and myocardial infarction/stroke incidence were lower and risk models were better calibrated in C3PO as opposed to the Convenience Samples (calibration error range for myocardial infarction/stroke: 0.012-0.030 in C3PO vs. 0.028-0.046 in Convenience Samples; calibration error for atrial fibrillation 0.028 in C3PO vs. 0.036 in Convenience Samples). Sampling patients receiving regular primary care and using NLP to recover missing data may reduce bias and maximize generalizability of EHR research.2022-01-01T00:00:00Z2022 Apr2025-04-21T16:54:35.359Z2025-04-21T16:54:35.359ZS-EPMC89938733539645410.1038/s41746-022-00590-0