Objective  The random forest method was used to analyze the relationship between disease perception, doctor-patient communication, cognitive load, and the compliance of high-risk patients with early digestive tract cancer regarding endoscopic re-examination.

Methods A total of 303 patients at high risk for early digestive tract cancer from January 2022 to December 2022 were selected as research subjects. Questionnaire surveys were conducted using general information questionnaires, a simplified disease perception questionnaire, a doctor-patient communication evaluation scale, and various cognitive load scales. Spearman analysis was used to examine the relationships among disease perception, doctor-patient communication, cognitive load, and compliance with endoscopic reexamination in patients at high risk for early digestive tract cancer. Multivariate logistic analysis was performed to identify factors influencing patients' compliance with endoscopic reexamination. A random forest model was used to rank the importance of characteristic variables for compliance. The area under the receiver operating characteristic (ROC) curve (AUC) was used to evaluate model efficacy, and five-fold cross-validation was applied for validation.

Results Spearman analysis showed that poor compliance with endoscopic re-examination among patients at high risk for early digestive tract cancer was negatively correlated with disease perception and doctor-patient communication (r_s = －0.788, －0.815, P < 0.05), and positively correlated with cognitive load (r_s = 0.619, P < 0.05). Multivariate logistic analysis identified age (odds ratio OR = 1.159, 95% confidence interval CI: 1.052-1.276), family history of digestive tract tumors (OR = 0.081, 95% CI: 0.008-0.865), average monthly family income (OR = 0.994, 95% CI: 0.991-0.997), disease perception (OR = 0.745, 95% CI: 0.639-0.870), doctor-patient communication (OR = 0.694, 95% CI: 0.602-0.801), and cognitive load (OR = 1.351, 95% CI: 1.136-1.608) as factors influencing compliance with endoscopic re-examination in these patients (P < 0.05). The variable importance ranking in the random forest model was doctor-patient communication > disease perception > cognitive load > average monthly family income > age > family history of digestive tract tumors. The model's prediction AUC was 0.899 (95% CI: 0.843-0.921), and the average accuracy after five-fold cross-validation was 0.893.

Conclusion Disease perception, doctor-patient communication, cognitive load, average monthly household income, age, and family history of digestive tract tumors are key factors influencing compliance with endoscopic re-examination in patients with early digestive tract cancer.