Accuracy of aspirin prophylaxis for preeclampsia prevention documentation within a large administrative dataset
Lauren S. Tailor, Renee G. Fajardo, Joel G. Ray, Isabelle Malhamé, Sonia M. Grandi- Pediatrics, Perinatology and Child Health
- Epidemiology
Abstract
Background
Low‐dose aspirin prophylaxis is recommended for women at risk of preeclampsia. Capturing aspirin prophylaxis within administrative databases can be challenging since it is an over‐the‐counter medication. The Better Outcome Registry and Network (BORN) database, a perinatal health registry in Ontario, Canada, includes a formal variable that captures aspirin prophylaxis for preeclampsia. This variable has not been formally validated.
Objectives
To assess the accuracy of the aspirin prophylaxis variable in the BORN database against an electronic medical record (EMR).
Methods
This validation study comprised 200 randomly selected women who had a livebirth at St. Michael's Hospital (SMH) in Toronto, Ontario, from January 2018 to July 2022. Recorded aspirin prophylaxis in pregnancy and maternal sociodemographic characteristics were independently extracted by two abstractors. Accuracy of aspirin prophylaxis use in the BORN database was compared to that in the SMH EMR, expressed as sensitivity, specificity, positive (PPV) and negative predictive values (NPV), Cohen's kappa (κ), and overall percent agreement, with 95% confidence intervals (CI). Sensitivity analyses were performed to account for missing or unclear aspirin prophylaxis use.
Results
Among 200 women, 24 (12.0%) received aspirin prophylaxis – 12.5% within the SMH EMR and 8.0% in the BORN database. Women using aspirin were older (37.0 vs 33.0 years) and had higher median gravidity (3 vs. 2). Sensitivity and specificity of the BORN aspirin prophylaxis variable were 62.5% (95% CI 40.6, 81.2) and 100.0% (95% CI 97.3, 100.0), respectively. The corresponding positive and negative predictive values were 100.0% (95% CI 78.2, 100.0), and 93.8% (95% CI 88.6, 97.1), respectively. Cohen's κ was 0.74 (95% CI 0.58, 0.90), and overall percent agreement was 94.4% (95% CI 87.1, 100.0).
Conclusions
Aspirin use within the BORN database, based on a standard variable field, appears accurate enough for the potential use in epidemiological studies of aspirin prophylaxis for preeclampsia or as a covariate in related studies.