Truace
Subscribe

Depressive Symptom Predictors in Older Mexican Adults: Interaction Structures and Nonlinear Effects from Machine Learning Explainability

Background Depression among older adults remains a critical public health issue in Latin America, where aging populations face compounded social and health vulnerabilities. Objective To characterize psychosocial and clinical predictors of depressive symptoms among older Mexican adults using LASSO-guided logistic regression, Random Forest classification, and post-hoc explainability methods. Methods A cross-sectional analysis was conducted using baseline data from a cohort (n = 1,252) of adults aged ≥60 affiliated…

By the Truvace desk1 min read1 sourcePeer-reviewedTRV-2026-1759· cite & verify
Depressive Symptom Predictors in Older Mexican Adults: Interaction Structures and Nonlinear Effects from Machine Learning Explainability
A bill to amend title XVIII of the Social Security Act to provide Medicare beneficiaries with access to geriatric assessments and chronic care coordination services, and for other purposes by United States. Congress. Senate United States. Congress. Senate. Committee on Finance. Public domain

In brief

Researchers analyzed baseline data from 1,252 adults aged 60 and over affiliated to the Mexican social security institute IMSS in Mexico City. They combined LASSO-guided logistic regression with Random Forest classification, then used SHAP values, Friedman's H-statistic and ALE plots to interpret the models.

Low perceived social support, social isolation risk and male sex emerged as consistent predictors. The work matters because it suggests that interpretable machine learning can refine risk stratification in geriatric care, although the cross-sectional design and modest model discrimination limit how far the results can be applied.

Main points

  1. Low perceived social support, social isolation risk and male sex were consistently influential predictors of depressive symptoms.
  2. Logistic regression reached an AUC-ROC of 0.668, while the Random Forest model scored a significantly lower 0.588.
  3. Sex and age acted mostly through interactions with other predictors, with H-statistics of 0.867 and 0.604.
  4. The authors state that the LSNS-6 and MOS-SSS can serve as first-line screening tools in routine geriatric consultations.

The gain

Explainability methods applied to machine learning models revealed interaction structures and nonlinear effects, such as a U-shaped age pattern, that conventional regression could not detect in older Mexican adults.

The rundown

The study used a cross-sectional analysis of baseline cohort data. Depressive symptoms were measured with the 35-item CESD-R scale, and the modeling approach paired LASSO variable selection with a Random Forest classifier.

Post-hoc explainability showed a U-shaped association between age and depression risk. It also showed dose-response gradients for diabetes and hypertension complications, patterns not captured by conventional regression.

The authors conclude that risk stratification should draw on comorbidity burden, particularly type 2 diabetes or hypertension complications, rather than diagnosis alone.

Sources

  1. Peer-reviewedArchives of Medical Research2026-10-07

The debate