BackgroundElectronic health records (EHR) present a research opportunity to provide “rapid assays” for building prospective (historical cohort) studies. However, there are significant challenges to extracting and harmonizing data from different hospital and ambulatory care systems that need to be overcome before EHRs can be readily used to study nutrition‐related outcomes.Methoda community‐academic partnership with representatives from NYC Community Health Centers (n=4) and Hospitals (n=4), The Rockefeller University, The Sackler Institute for Nutrition Science and Clinical Directors Network (CDN), a primary care practice‐based research network (PBRN), was established to build a multisite de‐identified database extracted from the EHRs of female adolescents aged 12–21 years and their offspring through 24 months of age, receiving primary care between 2011 and 2015.Representatives convened through 2015 and 2016 to provide guidance on hypothesis development, variable selection, data extraction and review analyses, focusing on nutrition outcomes during pre‐conception, prenatal and postnatal periods.ResultsThe cohort included 32,030 unique female adolescents 2,291 (7.2%) of whom were pregnant adolescents. The majority were from the Bronx (60%). Keywords, ICD‐9 codes and laboratory procedure codes were provided to Health IT experts to extract clinical data on diabetes, hypertension, hyperlipidemia, gestational diabetes, and delivery and birth outcomes. Systolic blood pressure (SBP) was retrieved for 95% of the cohort (with data retrieval in each site ranging from 90% to 98%). Laboratory results for Hemoglobin A1C, total cholesterol, triglycerides, and fasting plasma glucose were retrieved for 11%, 16%, 12%, and 17% of patients respectively, with large variability among the sites (from as low as 0.4% to 50%).Data analysisAmong the cohort, 3% were underweight, 55% normal, 21% overweight and 20% obese. Higher mean SBP and a higher blood glucose levels, hemoglobin A1c, total cholesterol, and triglycerides levels were observed among overweight and obese patients compared to normal weight females (p<0.05). Obese adolescents were less likely to have low birth weight babies compared to normal weight females (OR=0.50, 95% CI: 0.29, 0.87). Underweight and normal weight mothers had significantly higher rates of low birth weight babies and overweight/obese mothers had higher rates of large babies (chi‐square p=0.01).Conclusionthe use of EHR is feasible as a rapid assay to describe nutritional outcomes in under‐studied populations. Preliminary results appear to confirm the directions and size of associations noted in large observational studies, and will help generate additional hypotheses. Taking a highly collaborative approach with clinicians, IT experts, basic science and clinical researchers has been critical to accomplish data extraction, harmonization' analysis and quality improvement.
Read full abstract