Risk Score Development to Predict Postpartum Cardiovascular Disease Incorporating Pregnancy and Nonpregnancy Factors

BACKGROUND: Pregnancy events including severe maternal morbidity are associated with future cardiovascular disease (CVD), yet no models exist to identify at-risk individuals postpartum. OBJECTIVES: The purpose of this study was to develop a risk score with likelihood ratios (LRs) predicting postpartum CVD based on pregnancy and nonpregnancy CVD risk factors. METHODS: Using national administrative hospitalization data from Canada 2008-2021, the authors randomly selected 1 delivery per individual aged 18-55 years without pre-existing CVD. Cox proportional hazards models were used to predict CVD hospitalization, following individuals from 43 days after delivery discharge until CVD, censoring at in-hospital noncardiovascular death, subsequent pregnancy, or end of study. Risk scores and LRs for CVD hospitalization were generated using model coefficients. Model performance was characterized using C-statistics with 95% CIs. RESULTS: In this cohort of 1,992,972 individuals (median follow-up 3.7 years [Q1-Q3: 1.6-7.9 years]), the LRs for CVD within 1 year of delivery were 0.55 (95% CI: 0.49-0.61), 1.42 (95% CI: 1.28-1.57), and 7.81 (95% CI: 6.69-9.11) for standardized risk scores of <0, 0 to <3, and ≥3, respectively, which was similar at 5 and 10 years postpartum. Most individuals (59.3%) had a risk score of <0, 39.1% had a risk score between 0 and 3, and 1.5% had a risk score ≥3. The C-statistic for the multivariable Cox model was 0.669 (95% CI: 0.668-0.670). CONCLUSIONS: The lack of sociodemographic, clinical, and health behavior information may have contributed to the low predictive performance of this score for postpartum CVD hospitalizations. However, this score derived from a national administrative database, provides an initial framework for estimating future cardiac risk, which may provide guidance regarding the need and intensity of postpartum CVD surveillance.