Comparison of Lactation Information from Electronic Health Records with Survey Data Across Five US Health Systems
Recommended Citation
Jansen G, Seburg E, Vazquez Benitez G, Ehresmann K, Mohamed H, Avalos L, Negriff S, Loree A, Howick C, Daida Y, Palmsten K. Comparison of Lactation Information from Electronic Health Records with Survey Data Across Five US Health Systems. Pharmacoepidemiol Drug Saf 2025; 34.
Document Type
Conference Proceeding
Publication Date
8-19-2025
Publication Title
Pharmacoepidemiol Drug Saf
Keywords
adult, breast milk, child, conference abstract, depression, diagnosis, electronic health record, female, health care system, Hispanic, human, infant, lactation, live birth, major clinical study, male, predictive value, pregnancy, prevalence
Abstract
Background: Data on lactation status for research are often collected through surveys. Information on milk feeding collected at routine healthcare visits and stored in electronic health records (EHR) is an emerging source of data for pharmacoepidemiologic studies of lactation-related outcomes. However, validation is needed to ensure the data can be used for research. Objectives: To compare information on lactation status obtained from structured EHR fields with survey data. Methods: We included individuals from five US healthcare systems who participated in the Managing Our Mood survey. Information related to lactation is captured during routine healthcare visits within structured EHR fields at each site. Individuals had a live birth (March 2022-October 2023), depression diagnosis during pregnancy, and = 1 record of milk feeding information in their or their infant's EHR. We compared lactation status from EHR data up to 9 months after delivery with survey data collected 3-4 months after delivery as the reference. We assessed agreement on lactation status (human milk feeding ever and at survey completion) between data sources using percent agreement, Cohen's kappa, sensitivity, specificity, positive predictive value (PPV), and negative predictive value with 95% confidence intervals (CIs) overall and by number of records available with milk feeding information. Results: Among 281 eligible individuals, most were non-Hispanic White (43.3%) or Hispanic (24.7%), and 19.2% had public insurance. The prevalence of human milk feeding ever was 93.2% according to survey data and 91.5% according to EHR data. The median number of infant or birthing parent records with lactation information from EHR was 6. Prevalence of human milk feeding at survey completion was 73.0% and 77.6% according to EHR. Agreement between data sources for human milk feeding ever and at survey completion was = 92% with kappas = 0.77. For human milk feeding ever, sensitivity was 97.3% (95% CI: 94.6%, 98.7%) and PPV was 99.2% (95% CI: 97.2%, 99.8%). For human milk feeding at survey, sensitivity was 98.0% (95% CI: 95.1%, 99.2%) and PPV was 92.2% (95% CI: 87.9%, 95.1%). Among those with = 6 records, sensitivity of human milk feeding at survey was 99.2% (95% CI: 95.4%, 99.9%); PPV was 96.7% (95% CI: 91.8%, 98.7%). Among those with < 6 records sensitivity was 96.6% (95% CI: 90.3%, 98.8%); PPV was 86.6% (95% CI: 78.4%, 92.0%). Conclusions: There was substantial agreement on lactation status between EHR and survey data. Sensitivity and PPV were lower among those with fewer lactation records, suggesting the need for bias analysis according to number of lactation records available to mitigate misclassification. These findings suggest that lactation information from structured EHR may be used for pharmacoepidemiologic research.
Volume
34
