Factors Associated with Nutritional Status in Grassroots Recyclers in Ecuador: A Machine Learning Approach

TL;DR

This study uses machine learning to identify factors affecting the nutritional health of grassroots recyclers in Ecuador, a vulnerable group facing health and social challenges.

Contribution

Applies machine learning to uncover sociodemographic and health factors linked to nutritional status in an under-researched informal worker population.

Findings

71% of grassroots recyclers in Ecuador showed nutritional alterations, indicating a significant public health issue.

Factors like sex, age, residence, and healthcare access were significantly associated with nutritional status.

The CatBoost model with SMOTE achieved the best performance in predicting nutritional alterations.

Abstract

Public health relevance—How does this work relate to a public health issue?
Grassroots recyclers are a vulnerable occupational group who are exposed to adverse environmental conditions, food insecurity, nutritional risks, social inequity and health inequity, constituting relevant public health concerns.This study links nutritional status with sociodemographic, health, and work-related factors, which are determinants of health in an under-researched informal working population in Ecuador.

Grassroots recyclers are a vulnerable occupational group who are exposed to adverse environmental conditions, food insecurity, nutritional risks, social inequity and health inequity, constituting relevant public health concerns.

This study links nutritional status with sociodemographic, health, and work-related factors, which are determinants of health in an under-researched informal working population in Ecuador.

Public health significance—Why is this work of significance to public health?
The study provides empirical evidence on factors associated with nutritional status among grassroots recyclers, helping to address knowledge gaps related to health and nutrition in a vulnerable group.The application of machine learning models allows the identification of complex relationships among sociodemographic, territorial, occupational, and health-related variables relevant to public health research.

The study provides empirical evidence on factors associated with nutritional status among grassroots recyclers, helping to address knowledge gaps related to health and nutrition in a vulnerable group.

The application of machine learning models allows the identification of complex relationships among sociodemographic, territorial, occupational, and health-related variables relevant to public health research.

Public health implications—What are the key implications or messages for practitioners, policy makers and/or researchers in public health?
These results support and strengthen, with empirical evidence, the design of public policies and targeted interventions, strategies that incorporate periodic nutrition assessment and monitoring of the grassroots recyclers.By integrating approaches grounded in equity, territory, the life course and occupational health approaches, these strategies provide actionable guidance for public health practitioners, policymakers and studies to reduce nutritional inequities and improve quality of life in this vulnerable population.

These results support and strengthen, with empirical evidence, the design of public policies and targeted interventions, strategies that incorporate periodic nutrition assessment and monitoring of the grassroots recyclers.

By integrating approaches grounded in equity, territory, the life course and occupational health approaches, these strategies provide actionable guidance for public health practitioners, policymakers and studies to reduce nutritional inequities and improve quality of life in this vulnerable population.

Grassroots recyclers play a fundamental role in solid waste management in Ecuador; however, they often work under precarious conditions that may compromise their health. This study aimed to identify factors associated with nutritional status, operationalized as the presence or absence of nutritional alterations, among grassroots recyclers through supervised machine learning approaches. Data from 303 recyclers from three Ecuadorian cities (Cuenca, Macas, and La Libertad) were analyzed, incorporating sociodemographic, occupational, and health-related variables. Nutritional alterations were defined based on anthropometric and biochemical indicators, specifically, excess body weight and/or elevated total lipid levels. The results showed that 71% presented nutritional alterations, evidencing an important public health problem in this vulnerable population. Significant associations were observed with sex, age, canton of residence, ability to ride a bicycle, bicycle use for work, and attendance at medical check-ups. Among the evaluated models, CatBoost trained with SMOTE achieved the highest ROC-AUC value and the most balanced performance between classes, although sensitivity for individuals without nutritional alterations remained limited. Feature importance analysis highlighted sociodemographic, occupational, economic, and healthcare access factors, underscoring the multidimensional nature of nutritional risk and supporting the use of machine learning as a support tool for public health planning and targeted interventions.

Figures