Development and Validation of a Prediction Model for Gestational Hypertension in Mezam Division
Author(s): Nkem Ernest NJUKANG, Thomas Obinchemti EGBE, Nicolas TENDONGFOR, Tah Aldof YOAH, Kah Emmanuel NJI, Martyn SAMA, Fidelis Atabon AKO, Joseph KAMGNO
Objective: Our study aimed to develop and validate a prediction model for identifying women at increased risk of developing gestational hypertension (GH) in Mezam division, Northwest Region (NWR) of Cameroon.
Method: A retrospective cohort design was employed. Data for a cohort of 1183 participants were randomly divided into derivation (n = 578) and validation (n = 585) datasets. Inclusion criterion was women without chronic hypertension. Primary outcome was Gestational hypertension. A questionnaire and data abstraction form were used for data collection. Chi square (χ2) test, independent sample t-test and multivariate logistic regression (to derive the prediction model) were used for data analysis. For each significant variable, a score was calculated by multiplying coefficient (β) by 100 and rounding to the nearest integer. Discrimination was estimated by used of the c-statistic.
Results: DBP, SBP, hypertension in previous pregnancy, stress and smoking (scores 10, 6, 210, 56 and 18, respectively) were predictors of incident GH. The model accuracy was assessed by the area under the receiver operating characteristic curve (AUC), with optimal cut-off value 936. With the derivation dataset, sensitivity, specificity and AUC of the model were 75.9%, 80.8% and 0.828 (95% CI 0.772–0.884) respectively. The model was validated by dividing the aggregated scores into three ranges (low, moderate and high) and their cumulative incidence calculated which were; 3.5%, 6.1% and 39.4%, respectively, in the derivation dataset and 4.7%, 6.2% and 30.2%, respectively, in the validation dataset. Calibration was good in both cohorts. The negative predictive value of women in the development cohort at high risk of GH was 92.0% compared to 94.0% in the validation cohort.
Conclusions: The prediction model revealed adequate performance after validation in an independent cohort a