Accuracy of risk prediction scores in pregnant women with congenital heart disease

Objective: To assess performance of risk stratification schemes in predicting adverse cardiac outcomes in pregnant women with congenital heart disease (CHD) and to compare these schemes to clinical factors alone.

Design: Single-center retrospective study.

Setting: Tertiary care academic hospital.

Patients: Women ≥18 years with International Classification of Diseases, Ninth Revision, Clinical Modification codes indicating CHD who delivered between 1998 and 2014. CARPREG I and ZAHARA risk scores and modified World Health Organization (WHO) criteria were applied to each woman.

Outcome measures: The primary outcome was defined by ≥1 of the following: arrhythmia, heart failure/pulmonary edema, transient ischemic attack, stroke, dissection, myocardial infarction, cardiac arrest, death during gestation and up to 6 months postpartum.

Results: Of 178 women, the most common CHD lesions were congenital aortic stenosis (15.2%), ventricular septal defect (13.5%), atrial septal defect (12.9%), and tetralogy of Fallot (12.9%). Thirty-five women (19.7%) sustained 39 cardiac events. Observed vs expected event rates were 9.9% vs 5% (P = .02) for CARPREG I score 0 and 26.1% vs 7.5% (P < .001) for ZAHARA scores 0.51-1.5. ZAHARA outperformed CARPREG I at predicting adverse cardiovascular outcomes (AUC 0.80 vs 0.72, P = .03) but was not significantly better than modified WHO. Clinical predictors of adverse cardiac event were symptoms (P = .002), systemic ventricular dysfunction (P < .001), and subpulmonary ventricular dysfunction (P = .03) with an AUC 0.83 comparable to ZAHARA (P = .66).

Conclusions: CARPREG I and ZAHARA scores underestimate cardiac risk for lower risk pregnancies in these women. Of the three risk schemes, CARPREG I performed least well in predictive capacity. Clinical factors specific to the population studied are comparable to stratification schemes.