Predictive Modeling of Cancer Information Overload and Screening Attitudes: A Cross-Sectional Study

To explore the association between cancer information overload and attitudes toward cancer screening among internal medicine patients.

To identify predictors of screening attitudes using statistical and machine learning models.

A cross-sectional study was conducted with 410 internal medicine outpatients. Data were collected using the Cancer Information Overload Scale and the Attitude Scale for Cancer Screening. Analyses included t-tests, ANOVA, and regression. Seven machine learning models (KNN, SVM, ANN, RF, XGBoost, CART, Elastic Net) were compared using 10-fold cross-validation (R², RMSE, MAE). Statistical significance was set at P < .05.

Participants' mean age was 38.25 ± 12.46 years; 69.8% were female. Mean information overload and attitude scores were 17.14 ± 4.91 and 100.13 ± 13.58, respectively. Regression analysis showed a significant negative association between information overload and screening attitudes (β = -0.321, P < .001; R² = 0.103; 95% CI [-1.144, -0.634]). Among machine learning models, Elastic Net Regression (α = 0.2, λ = 1) achieved the best performance (RMSE = 12.6, MAE = 10.8), confirming cancer information overload as the strongest predictor.

Cancer information overload is inversely associated with screening attitudes. Machine learning models enhance interpretability, emphasizing the importance of managing information burden to improve cancer screening engagement.