Explainable Machine Learning to Understand the Social and Biological Determinants of Type 2 Diabetes in Mexico
Revista de Investigación e Innovación en Ciencias de la Salud · 16.04.2026 · 8 (2) · 1-15
DOI: 10.46634/riics.535 ↗Kurz gefasst
Fachartikel in der Zeitschrift Revista de Investigación e Innovación en Ciencias de la Salud, online erschienen am 16.04.2026. Verfasst von Mario Daniel Cervantes-Guerrero und 5 weiteren Autor:innen. Die Arbeit nennt 24 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Themen: Künstliche Intelligenz in der Pflege, Diabetes mellitus
Abstract (Originalsprache)
Introduction
Type 2 diabetes mellitus (T2DM) is a multifactorial disease associated not only with clinical factors but also with social determinants such as income, education, and access to healthcare. Its global prevalence continues to rise, particularly in low- and middle-income countries, making early detection and equitable management a public health priority.
Objective
To evaluate and interpret the performance of machine learning models that incorporate both clinical and socioeconomic data to predict the presence of type 2 diabetes in a Mexican population.
Methods
A dataset including 898 diabetic patients and 889 non-diabetic individuals randomly selected from the Hospital de Especialidades Siglo XXI (IMSS, Mexico City) was analyzed. Ten supervised algorithms (logistic regression, SVM, decision tree, random forest, KNN, naive bayes, AdaBoost, LightGBM, CatBoost, and XGBoost) were trained and validated using stratified 10-fold cross-validation. Data preprocessing included multiple imputation by chained equations (MICE), normalization, and encoding of categorical variables. Model interpretability was evaluated using SHapley Additive exPlanations (SHAP) to identify the most influential predictors.
Results
Ensemble methods, particularly XGBoost and LightGBM, achieved the best performance (accuracy = 0.94, AUC > 0.95). Glucose, diastolic blood pressure, and age were the strongest predictors, while socioeconomic variables such as income and education contributed additional predictive value. The integration of contextual variables improved model interpretability without reducing accuracy.
Conclusions
Explainable machine learning models integrating both clinical and socioeconomic data can enhance diagnostic precision and promote equity in diabetes detection. These tools offer potential applications in public health programs, facilitating the early identification of at-risk populations and supporting more equitable prevention and care strategies.
Zitieren
APA
Cervantes-Guerrero, M. D., Galván-Tejada, C. E., Cruz, M., Galván-Tejada, J. I., Barros, R. C., & Kupssinskü, L. (2026). Explainable Machine Learning to Understand the Social and Biological Determinants of Type 2 Diabetes in Mexico. Revista de Investigación e Innovación en Ciencias de la Salud, 8(2), 1-15. https://doi.org/10.46634/riics.535
Vancouver
Cervantes-Guerrero MD, Galván-Tejada CE, Cruz M, Galván-Tejada JI, Barros RC, Kupssinskü L. Explainable Machine Learning to Understand the Social and Biological Determinants of Type 2 Diabetes in Mexico. Revista de Investigación e Innovación en Ciencias de la Salud. 2026;8(2):1-15. doi:10.46634/riics.535
BibTeX
@article{CervantesGuerrero2026,
author = {Cervantes-Guerrero, Mario Daniel and Galván-Tejada, Carlos E. and Cruz, Miguel and Galván-Tejada, Jorge I. and Barros, Rodrigo C. and Kupssinskü, Lucas},
title = {Explainable Machine Learning to Understand the Social and Biological Determinants of Type 2 Diabetes in Mexico},
journal = {Revista de Investigación e Innovación en Ciencias de la Salud},
year = {2026},
volume = {8},
number = {2},
pages = {1--15},
doi = {10.46634/riics.535}
}
EUVIMED