Derivation and Validation of Hypertension Risk Prediction Model for US Black Women in US: A Population-Based Study Using Florida Birth Records (2005-2023)

Abstract

Background: Pre-pregnancy chronic hypertension is an important maternal health concern because it increases the risk of adverse pregnancy and cardiovascular outcomes. Black women experience a disproportionate burden of hypertension and pregnancy-related complications, yet less is known about long-term trends, associated risk factors, and risk prediction for pre-pregnancy chronic hypertension among Black women of reproductive age.Methods: This retrospective population based study used de-identified Florida birth certificate records from 2005 through 2023. The analytic population included Black women with live birth records and classifiable pre-pregnancy chronic hypertension status. The outcome was pre-pregnancy chronic hypertension, defined as hypertension present before pregnancy. Annual prevalence of pre-pregnancy chronic hypertension was estimated overall and by selected subgroups. Temporal trends were assessed using linear and segmented regression. Logistic regression models were used to examine demographic, socioeconomic, clinical, reproductive, and behavioral factors associated with pre-pregnancy chronic hypertension. A multivariable risk prediction model was developed using a 70% development sample and evaluated in a 30% validation sample. Model performance was assessed using discrimination, calibration, sensitivity, specificity, accuracy, positive predictive value, and negative predictive value at the Youden threshold. Prediction error was assessed using the Brier score. Additional evaluation included temporal validation, subgroup AUC analyses, and decision curve analysis. Results: The final analytic sample included 944,398 records, of which 27,377 were classified as having pre-pregnancy chronic hypertension, corresponding to an overall prevalence of 2.9%. Pre-pregnancy chronic hypertension increased from 2.15% in 2005 to 4.31% in 2023, representing an absolute increase of approximately 2.2 percentage points. Linear trend analysis showed an average annual increase of 0.108 percentage points per year (95% CI: 0.082–0.133; p<0.001). Segmented regression identified a change point around 2016, with a slower increase before 2016 of 0.046 percentage points per year (95% CI: 0.013–0.080) and a steeper increase after 2016 of 0.225 percentage points per year (95% CI: 0.170–0.279). In fully adjusted models, older maternal age, obesity, diabetes, and tobacco use were strongly associated with pre-pregnancy chronic hypertension. Compared with women aged 15–24 years, women aged 25–34 years had more than twice the odds of pre-pregnancy chronic hypertension (aOR=2.35; 95% CI: 2.27–2.44), while women aged 35 years or older had more than five times the odds (AOR=5.46; 95% CI: 5.23–5.71). Obesity was associated with over four times higher odds (aOR=4.11; 95% CI: 3.96–4.27), and diabetes was associated with nearly five times higher odds (aOR=4.85; 95% CI: 4.60–5.11). Tobacco use was also associated with higher odds (aOR=1.41; 95% CI: 1.32–1.50). In contrast, foreign-born nativity was inversely associated with pre-pregnancy chronic hypertension (aOR=0.61; 95% CI: 0.59–0.63). The final prediction model demonstrated acceptable discrimination, with an AUC of 0.774 in the development sample and 0.773 in the validation sample. At the Youden threshold of 0.0274, the validation model had 72.1% sensitivity, 68.8% specificity, 68.9% accuracy, 6.4% positive predictive value, and 98.8% negative predictive value. The Brier score was 0.027, indicating low overall prediction error. Conclusion: The prevalence of pre-pregnancy chronic hypertension among Black women in Florida increased from 2.15% in 2005 to 4.31% in 2023, representing an absolute increase of approximately 2.2 percentage points over the study period. Findings highlight the importance of earlier identification, prevention, and management of hypertension before pregnancy. The prediction model represented stable internal performance using routinely collected birth record variables, which may support population-level risk stratification, although external validation is needed before broader application. Keywords: pre-pregnancy chronic hypertension; Black women; maternal health; birth records; Florida; risk prediction; health disparities.