Risk Prediction for Pregnancy Induced Hypertension at University of Gondar Comprehensive Specialized Hospital, Northwest Ethiopia

Abstract

Background: Globally, up to ten percent of complications during pregnancy are related to pregnancy-induced hypertension that results in high maternal and perinatal morbidity and mortality. Early identifications of pregnant women who are at risk of developing pregnancy induced hypertension is important for better monitoring and reducing pregnancy induced hypertension and its complications. Objective: To develop and validate a risk prediction model for pregnancy induced hypertension based on the pregnant woman‟s profile at the first antenatal visit at University of Gondar Comprehensive Specialized Hospital, northwest Ethiopia. Methods: A retrospective study was conducted with a total of 973 pregnant women attending antenatal care enrolled from January 01/2017 to December 30/2021 at university of Gondar comprehensive specialized hospital. Data were collected from the women‟s medical records by Kobo collect digital data collection tools and exported to STATA version 17 for analysis. Least absolute shrinkage and selection operator method was applied to select predictors and entered to multivariable logistic regression. Significant predictors were used for nomogram development. Its performance was assessed using area under the curve and calibration plot. Internal validation was done by bootstrap technique. Decision curve analysis was performed Results: The incidence proportion of pregnancy induced hypertension was 15.7 % (95%CI 13.49, 18.16). Maternal age, residence, multiple pregnancy, gravidity and diabetes mellitus have statistically significant predictive power for pregnancy induced hypertension. The model has a discriminating power of area under the curve (AUC = 83.80% (95% CI; 79.9%, 87.64%)) and calibration (P-value = 0.677). Internal validation through bootstrap has an area under the curve of 83.80% (95% CI; 79.9%, 87.6%) with a calibration of (p-value = 0.436) with an optimism coefficient of zero. The decision curve analysis proved that the model has better net benefit than the treat all and treat none scenarios. Conclusion and recommendation: The model can be utilized as simple, inexpensive and implementable tool. We recommend the researchers to validate externally