Predicting Self-Reported Lung Cancer Diagnosis Using Statistical and Machine Learning Approaches: A 2024 NHIS Study

Lung cancer remains an important public health concern, and cigarette smoking is one of its most established associated factors. This study examined demographic, behavioral, socioeconomic, and health-related factors associated with self-reported lung cancer diagnosis among U.S. adults using the 2024 National Health Interview Survey (NHIS). Survey-weighted descriptive analyses and logistic regression were used to estimate population-level prevalence and adjusted associations. Logistic regression and random forest were also compared for respondent-level prediction using a stratified 80/20 training and test split. Among 32,573 adults with a valid lung cancer outcome, 145 reported a previous diagnosis. The weighted prevalence was 0.312%. Older age and smoking history showed the strongest adjusted associations. Each 10-year increase in age was associated with 2.13 times the odds of reported lung cancer, while former and current smokers had odds ratios of 4.05 and 5.63 compared with never smokers. In the test sample, logistic regression achieved an ROC-AUC of 0.899 and identified 24 of 28 cases, while random forest identified only one. Greater model complexity did not improve prediction in this rare-outcome survey dataset.