Retrospective application of algorithms to improve identification of pregnancy outcomes from the electronic health record

Objective To improve upon the accuracy of ICD codes for identifying maternal and neonatal outcomes by developing algorithms that incorporate readily available EHR data. Study design Algorithms were developed for gestational hypertension (GHTN), pre-eclampsia (PreE), gestational diabetes mellitus (GDM) and were compared to ICD codes and chart review. Accuracy and sensitivity analyses were calculated with their respective 95% confidence limits for each of the comparisons between algorithms, ICD codes alone, and chart review. Results Sensitivity of GHTN ICD codes was 8.1% vs. 83.8% for the algorithm when compared to chart review. In comparison to chart review, sensitivity of ICD codes for PreE was 7.5% vs. 71.4% for the algorithm. GDM had similar sensitivity rates for both ICD codes and the algorithm. Conclusion Application of algorithms, validated by chart review, enhanced capture of several outcomes. Algorithms should be obligatory adjunct tools to the ICD codes for identification of outcomes of interest.