Statistical Profiling of Drug-Resistant Tuberculosis Cases in Nigeria Using Logistic Regression and Discriminant Analysis

Drug-resistant tuberculosis (DR-TB) in Nigeria imposes substantial treatment burden and poor outcomes compared to drug-sensitive TB, and statistical profiling of DR-TB cases using multivariate methods can identify clinically actionable risk factors to guide testing prioritisation and treatment initiation. This study applies logistic regression and linear discriminant analysis to profile DR-TB cases among notified TB patients in Nigeria from 2020 to 2022. Case-level data from the National Tuberculosis and Leprosy Control Programme electronic register were used, covering 8,400 patients with known drug susceptibility testing results. The DR-TB case proportion was 14.2 percent. Logistic regression identified prior TB treatment (OR 6.3, p < 0.001), HIV co-infection (OR 2.4, p = 0.002), prison detention history (OR 3.8, p < 0.001), contact with known DR-TB case (OR 5.1, p < 0.001), and treatment facility in North West zone (OR 2.1, p = 0.008) as independent DR-TB predictors. Logistic model AUC was 0.84. Linear discriminant analysis with the same predictor set achieved 79 percent correct classification with a canonical correlation of 0.67. The two methods showed consistent predictor ranking, validating logistic findings. North West and North East zones showed the highest DR-TB prevalence by zone. The study provides a statistical risk profiling framework for DR-TB prioritisation and recommends its adaptation by the NTBLCP as a drug susceptibility testing triage tool to optimise limited laboratory resources across Nigerian TB facilities. Keywords: drug-resistant tuberculosis, logistic regression, discriminant analysis, NTBLCP Nigeria, risk profiling