Artificial Intelligence for Malnutrition Prediction: Integrating Clinical, Socioeconomic, and Environmental Determinants for Sustainable Development
TL;DR
This review critically examines the application of AI approaches — including supervised ML algorithms, deep learning architectures, natural language processing (NLP), and geospatial analytics — to malnutrition prediction across diverse populations and contexts and evaluates the evidence for multi-domain data integration within predictive frameworks.
Abstract
Malnutrition in all its forms continues to represent one of the most severe and preventable threats to human health and development, affecting an estimated 2.5 billion people globally and imposing profound burdens on low- and middle-income countries. Conventional nutrition surveillance tools remain episodic, resource-intensive, and inadequate for capturing the complex multidimensional determinants — clinical, socioeconomic, and environmental — that collectively drive malnutrition outcomes. The rapid maturation of artificial intelligence (AI) and machine learning (ML) offers transformative possibilities for malnutrition prediction by enabling the integration of heterogeneous data streams at scales and resolutions previously beyond reach. This review critically examines the application of AI approaches — including supervised ML algorithms, deep learning architectures, natural language processing (NLP), and geospatial analytics — to malnutrition prediction across diverse populations and contexts. It evaluates the evidence for multi-domain data integration within predictive frameworks, situates these developments within the agenda of the United Nations Sustainable Development Goals (SDGs), and identifies critical gaps concerning algorithmic bias, model interpretability, data privacy, and equitable deployment in resource-constrained settings. Evidence indicates that ensemble ML models incorporating remotely sensed environmental data, household socioeconomic indicators, and clinical measurements achieve substantially improved predictive performance relative to single-domain approaches. Significant barriers persist, however, particularly with respect to data governance, infrastructure deficits, and ethical accountability. This review calls for a transdisciplinary research agenda uniting nutrition science, data science, environmental epidemiology, and development policy to harness AI responsibly for malnutrition prevention and management, in alignment with the ambitions of sustainable development.