Design and Development of an Automated System for Early Detection of Elderly Frailty

de Enciso García, J. (2024). Design and Development of an Automated System for Early Detection of Elderly Frailty. Trabajo Fin de Titulación (TFG). Universidad Politécnica de Madrid.

Abstract:
The ageing of the population and the resulting increase in life expectancy have a significant impact on countries’ healthcare systems, especially with regard to mental frailty of older people. This increase in longevity leads to higher costs of medical care, as well as an increase demand for mental health care problems related to ageing, putting additional pressure on available health resources. The main objective of this study is to develop a machine learning system that can assist clinicians in the early detection and prediction of mental frailty. Initially, in order to achieve this objective, an extraction of the most relevant features was performed, which allowed for a reduction in the number of assessments a patient has to undergo. This not only alleviates the burden on clinicians by eliminating irrelevant data collection, but also enables the detection of frailty without the necessity of so many tests. Conversely, the models developed demonstrate satisfactory capabilities in the prediction of mental frailty, as they permit the extraction of relationships between relevant features and the development of models using various machine learning alternatives, such as Multilayer Perceptron (MLP) and convolutional networks (CNN). After the exploration, the models developed were evaluated according to their F1-score, with the aim of identifying and improving system performance and supporting prevention strategies, in conjuction with other diagnostic tests, such as FFP. Furthermore, particular emphasis was placed on the incorporation of temporal dependencies between data collected at different times. This approach allows a better understanding of the evolution of mental frailty over time, contributing to a more personalised and effective treatment. In light of the outcomes observed, a number of algorithms were evaluated in comparison to those derived from the study conducted by Leghissa [1]. This analysis has identified those algorithms and models that have demonstrated superior performance in the detection and prediction of mental frailty in its early stages. These findings provide a robust foundation for future research and the implementation of clinical decision support systems in this area.