Key Factors and AI-Based Risk Prediction of Malnutrition in Hospitalized Older Women

Abstract

The numerous consequences caused by malnutrition in hospitalized patients can worsen their quality of life. The aim of this study was to evaluate the prevalence of malnutrition on the elderly population, especially focusing on women, identify key factors and develop a malnutrition risk predictive model. The study group consisted of 493 older women admitted to the Asunción Klinika Hospital in the Basque Region (Spain). For this purpose, demographic, clinical, laboratory, and admission information was gathered. Correlations and multivariate analyses and the MNA-SF screening test-based risk of malnutrition were performed. Additionally, different predictive models designed using this information were compared. The estimated frequency of malnutrition among this population in the Basque Region (Spain) is 13.8%, while 41.8% is considered at risk of malnutrition, which is increased in women, with up to 16.4% with malnutrition and 47.5% at risk of malnutrition. Sixteen variables were used to develop a predictive model obtaining Area Under the Curve (AUC) values of 0.76. Elderly women assisted at home and with high scores of dependency were identified as a risk group, as well as patients admitted in internal medicine units, and in admissions with high severity.

DigiHEALTH: Suite of Digital Solutions for Long-Term Healthy and Active Aging

“…This functionality provides health professionals and caregivers with a nutritional decision support system that considers not only the different nutritional needs (e.g., malnutrition risk, intake level, or the need for texture-adapted meals) but also the whole environment of an older patient, such as sociodemographic factors (e.g., sex, age…), psychosocial factors (e.g., psychosocial disorders), and morbidity factors (diseases). In addition, the personalized nutritional guidance [ 50 ] that offers this solution is divided into two types of recommendations, i) dietary and nutrition guidelines and ii) menu suggestions. In the first type, diet, fortification, food and liquid adaptation, supplementation, enteral nutrition, and follow-up areas are managed, whereas in the second type, different examples or suggestions of menu types and their ingredients are offered.…”

“…After identifying the key factors that contribute to malnutrition in older adults in a previous study carried out by our research group [ 50 ], those variables were used to develop an AI-based model for predicting the risk of malnutrition for this population. A total number of fifteen variables was introduced in the prediction system, and a three-optional response was obtained: low, medium, and high risk of malnutrition.…”

“…After identifying the key factors that contribute to malnutrition in older adults in a previous study carried out by our research group [50], those variables were used to de-…”

The geriatric nutritional risk index as a strong predictor of adverse outcomes following total shoulder arthroplasty

“… 14 Another study of hospitalized elderly females found that 63.9% were at risk of malnutrition or were experiencing malnutrition, proposing the negative influence of socioeconomic differences and increased depressive states. 28 Outside of disease, greater age in general exerts gradual yet dramatic effects on nutrition. Poor physical functionality hinders the ability for elderly individuals to maintain muscle mass, leading to progressive muscle catabolism, a common process in aging.…”

Assessing Calibration and Bias of a Deployed Machine Learning Malnutrition Prediction Model within a Large Healthcare System

“…This aligns with other single-center studies that nd that females are associated with higher risks of malnutrition compared to males, although the mechanisms behind this observation are still poorly understood. [24][25][26] This discrepancy highlights the need for deeper investigation as to the source of gender miscalibration or consideration or retraining of gender-speci c malnutrition risk prediction models. Furthermore, in sensitivity analyses, we discovered that calibration intercepts were signi cantly different between patients on commercial insurance, Medicaid, or Medicare.…”