Prediction of Low Dietary Adherence in Patients with Type 2 Diabetes Mellitus: A Comparative Study of Nomogram and CART Decision Tree Models

Le résumé fourni par la source

Objective: This study aimed to develop and compare two visual prediction models—a nomogram and a classification and regression tree (CART) decision tree—for identifying the risk of low dietary adherence among patients with type 2 diabetes mellitus (T2DM). Methods: Clinical and psychosocial data were obtained from 356 inpatients with T2DM recruited from two tertiary general hospitals in Baoding, Hebei Province, China. Dietary adherence was assessed using a validated dietary behavior adherence scale for patients with T2DM. Based on the median adherence score, participants were classified into a low-adherence group (n = 190) and a high-adherence group (n = 166). Candidate predictors were screened using univariate analysis and logistic regression, after which a nomogram and a CART decision tree model were constructed. Internal validation was performed using 1,000 bootstrap resamples. Model discrimination was evaluated using the area under the receiver operating characteristic curve (AUC), and the predictive performance of the two models was compared. Results: The nomogram model incorporated age, perceived susceptibility, perceived severity, perceived barriers, self-efficacy, negative coping, openness, and conscientiousness as predictors of low dietary adherence. The CART decision tree model identified self-efficacy, openness, future consideration, perceived susceptibility, and conscientiousness as key classification variables. The AUC values of the nomogram and CART decision tree models were 0.864 and 0.814, respectively, indicating good discriminative ability for both models. The difference in AUC between the two models was statistically significant (z = −2.3562, P < 0.05). Conclusion: Both the nomogram and CART decision tree models showed favorable predictive performance for low dietary adherence in patients with T2DM. The nomogram demonstrated slightly superior overall discrimination, whereas the CART decision tree provided an intuitive classification structure. These visual prediction tools may assist healthcare professionals in early identification of patients at risk of poor dietary adherence and support individualized dietary management interventions.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.