Predicting Hypertension by Twenty Years Postpartum Using Perinatal Data
Study Design For a population-based cohort of 90,316 women with deliveries recorded in the Nova Scotia Atlee Perinatal Database between 1988 and 2009, pregnancy-related data were linked with provincial administrative data to identify new diagnoses of hypertension. Cox regression was used to model the risk of hypertension and estimate hazard ratios for the associations between pregnancy factors and hypertension. The model was simplified through backwards elimination, then translated into a risk score. Results Hypertensive disorders of pregnancy were the strongest pregnancy-related predictors of hypertension, particularly if they recurred in more than one pregnancy. Other pregnancy-related predictors included parity, gestational diabetes, and breastfeeding. The ability of this simplified model to discriminate women who developed hypertension from those who did not was moderate (c-statistic = 0.72). Conclusion Individualized predictions of hypertension risk up to 20 years after pregnancy can be generated from clinical information routinely available at the time of pregnancy.