Development and validation of a lung cancer risk prediction model in Ethiopia

An advanced stage of presentation is common among lung cancer patients in low-resource settings, leading to late diagnosis and poor prognosis. Early identification of high-risk individuals is essential to enable targeted prevention and timely diagnosis. We conducted a case–control study including 1,255 adults (251 lung cancer cases and 1,004 controls) to develop a lung cancer risk prediction model. A multivariable logistic regression model was fitted, and 10-fold cross-validation and 2,000 bootstrap resamples were performed to check internal validation. A simplified point-based risk score was derived to enhance practical applicability. The final model included age, sex, household solid fuel use, asbestos exposure, years of smoking, years of exposure to secondhand smoke, prior tuberculosis, and family history of cancer as predictors. The model demonstrated good discrimination (AUC = 0.775), with consistent performance across bootstrap resampling and cross-validation. At the optimal cutoff, the model achieved 72% sensitivity and 72% specificity. The simplified risk score showed comparable discriminatory performance and classification accuracy. This study presents a locally relevant lung cancer risk prediction model and a simplified risk score based on readily obtainable predictors. These tools may support the identification of high-risk individuals in resource-limited settings and inform targeted screening and prevention strategies. However, external validation is required to establish the model’s generalizability. Recalibration to population-level incidence would be required if the model is used to estimate absolute risk or define population-based clinical or public health thresholds.