Abstract Body (Do not enter title and authors here): Background: Hypertensive disorders of pregnancy (HDP) are a leading cause of maternal morbidity and contribute to cardiovascular disease later in life. Early identification of patients at highest risk of HDP may improve targeting of preventive therapies. Prior work within the Kaiser Permanente Northern California (KPNC) cohort demonstrated that early pregnancy blood pressure trajectories (BPT) improved HDP prediction compared to clinical factors alone. Our objective is to assess the performance of BPT models in a racially diverse patient population.

Methods: We identified patients within Penn Medicine with a delivery from 2012-2020 and ≥3 outpatient BP measurements between 0 to 20 weeks gestation. Patients with pre-existing hypertension (HTN) or serious medical conditions were excluded. Covariates were obtained from the electronic medical record (EMR). Each patient was assigned one of six BPT groups based on KPNC latent class growth models. The primary outcome of HDP diagnosis, including preeclampsia and gestational HTN, was identified using ICD codes. Multivariable logistic regression was used to model the relationship between patient factors and HDP diagnosis. Predictive models were compared using the C-statistic.

Results: Among 26,430 patients, the mean age was 30.8 years, 23% had Medicaid insurance, and 32% identified as Black. Overall, 22% of patients developed HDP and incidence varied by BPT group (Figure). After adjusting for age, race/ethnicity, parity, BMI, and diabetes, the moderate-stable group (aOR 5.42, 95% CI 3.7-8.3) and the elevated-stable group (aOR 10.2, 95% CI 6.9-15.8) had the strongest associations with HDP compared to the ultra-low-declining group. The prediction model including patient factors and BPT groups had better discrimination for HDP compared to patient factors alone (C-statistic 0.71 vs 0.66). Overall, models including BPT groups performed similarly to those from the KPNC cohort.

Conclusions: Compared to the KPNC cohort, the Penn cohort had higher prevalence of patients with HDP, who identified as Black and had Medicaid insurance. Despite these differences, early pregnancy BPT groups demonstrated a similar improvement in HDP prediction compared to patient risk factors alone.