Development and validation of a risk prediction model for uncontrolled hypertension among adult hypertensive patients on anti-hypertensive treatment at the University of Gondar Comprehensive Specialized Hospital, Northwest Ethiopia

Abstract

Background: In Ethiopia, uncontrolled hypertension accounts nearly half of hypertensive patients on anti-hypertensive treatment. Uncontrolled hypertension increases the risk of heart failure, chronic renal failure, and other disorders. Predict the risk of uncontrolled hypertension among adult hypertensive patients on anti-hypertensive treatment is crucial to give more efficient intervention for high risk patients. However, evidence is lacking about risk prediction of uncontrolled hypertension among adult hypertensive patient on anti-hypertensive treatment in Ethiopia. Objective: To develop and validate a risk prediction model for uncontrolled hypertension among adult hypertensive patients on anti-hypertensive treatment at the University of Gondar Comprehensive Specialized Hospital, Gondar, Ethiopia, 2024. Methods: An institutional based retrospective follow up study were carried out from January 01/2019 to January 1/2023 among 849 adult hypertensive patients on anti-hypertensive treatment at University of Gondar Comprehensive Specialized Hospital. The participants were selected using computer generated simple random sampling technique. Data were collected using Kobo toolbox and analyzed with STATA and R version 4.3 software. Important predictors selected by Least Absolute Shrinkage and Selection Operator were entered to multivariable logistic regression. Statistically and clinically significant predictors after model reduction were used for the nomogram development. Overall performance of the model was assessed by the Brier score. Area under receiver operating characteristic curve was used to assess ability of the model to discriminate between those with and without the outcome. Calibration plot and Hosmer Lemeshow test were used to assess the agreement between observed outcomes and predictions. A bootstrap random generated sample was used for internal validation. Decision curve analysis is used to determine the usefulness of a prediction model for clinical decision-making. Results: The cumulative incidence of uncontrolled hypertension was 48.65% (95% CI: 45.29, 52.01). The nomogram was developed from eight statistically and clinically significant predictors: creatinine, total cholesterol, low density lipoprotein cholesterol, baseline systolic blood pressure, age, anti-hypertensive medication, comorbidities, and triglycerides. The Area Under Receiver Operating Characteristic curve of nomogram was 0.8052 (95% CI; 0.7760, 0.8345%) for original data set and 0.7925 (CI 0.76249, 0.82260) for the internal validated data set. The calibration plots of the nomogram exhibited optimal agreement between the predicted and observed values, with a P-value of 0.66. The optimal cutoff value for the predicted probability was 0.41, indicating sensitivity of 81% and specificity of 64.22%. Moreover, the decision curve analysis revealed that the nomogram would add net clinical benefits at the threshold probabilities less than 90%. Conclusion and recommendation: The developed nomogram demonstrated very good level of discrimination and well calibration, using predictors including creatine, total cholesterol, low density lipoprotein, systolic blood pressure, age, anti-hypertensive medication, comorbidities, and triglycerides. Healthcare providers should incorporate the nomogram into clinical practice to enhance the management of hypertensive patients after external validation. This allows for identification of hypertensive patients on antihypertensive treatment at a higher risk of having uncontrolled hypertension and give intervention based on individual risk, including prescribing anti-hypertensive treatments tailored to the characteristics of hypertensive patients, withhold or change antihypertensive treatment, scheduling follow-up appointments, and other services.