Development and validation of a nomogram for the prediction of late culture conversion among multi-drug resistant tuberculosis patients in North West Ethiopia: An application of prediction modelling

Introduction: Multi-drug resistant tuberculosis has impeded tuberculosis prevention and control due to its low treatment efficiency and prolonged infectious periods. Early culture conversion status has long been used as a predictor of good treatment outcomes and an important infection control metric, as culture-negative patients are less likely to spread tuberculosis. There is also evidence that suggests that delayed sputum conversion is linked to negative outcomes. Therefore, this study was aimed at developing a nomogram to predict the risk of late culture conversion in patients with multi-drug resistant tuberculosis using readily available predictors. Objective: The objective of this study was to develop and validate a risk prediction nomogram for the prediction of late culture conversion among multi-drug resistant tuberculosis patients in North-West Ethiopia. Methods: Multi-drug resistant tuberculosis data from the University of Gondar and the Debre Markos referral hospitals have been used and a total of 316 patients were involved. The analysis was carried out using STATA version 16 and R version 4.0.5 statistical software. Based on the binomial logistic regression model, a validated simplified risk prediction model (nomogram) was built, and its performance was evaluated by assessing its discriminatory power and calibration. Finally, decision curve analysis (DCA) was used to assess the generated model’s clinical and public health impact. Results: Registration group, HIV co-infection, baseline BMI, baseline sputum smear grade, and radiological abnormalities were prognostic determinants used in the construction of the nomogram. The model has a discriminating power of 0.725 (95% CI: 0.669, 0.781) and a P-value of 0.665 in the calibration test. It was internally validated using the bootstrapping method, and it was found to perform similarly to the model developed on the entire dataset. The decision curve analysis revealed that the model has better clinical and public health impact than other strategies specified. Conclusion: The developed nomogram, which has a satisfactory level of accuracy and good calibration, can be utilized to predict late culture conversion in MDR-TB patients. The model has been found to be useful in clinical practice and is clinically interpretable.