Accepted for/Published in: Journal of Medical Internet Research
Date Submitted: Sep 4, 2022
Date Accepted: Jan 10, 2023
Development and assessment of assisted diagnoses models using machine learning for identifying elderly patients with malnutrition: a cohort study
ABSTRACT
Background:
Older patients are at increased risk of malnutrition due to many factors, relating to poor clinical outcomes.
Objective:
This study aimed to develop an assisted diagnoses model using machine learning (ML) for identifying elderly patients with malnutrition.
Methods:
We reanalyzed a multicenter, observational cohort study including 2734 elderly patients. Baseline malnutrition was defined using the GLIM (Global Leadership Initiative on Malnutrition) criteria, and the study population was randomly divided into a derivation group (n = 2128) and a validation group (n = 532). Five ML algorithms were used and we further explored the relationship between features and the risk of malnutrition by using the Shapley Additive exPlanations (SHAP) visualization method. BMI over 21 was associated with a higher risk of GLIM.
Results:
2,734 inpatients were included and GLIM criteria were determined retrospectively. In the external validation cohort, the top three models by the area under the receiver operating characteristic curve (AUROC) were LightGBM (92.1%), the XGBoost (91.9%), and the RandomForest model (91.5%). Additionally, the analysis of the importance of features revealed that body mass index (BMI), weight loss, and calf circumference were the strongest predictors to affect GLIM.
Conclusions:
Using machine learning we could get an accurate and dynamic diagnosis of malnutrition based on the GLIM criteria. The cut-off values of laboratory tests generated by SHAP could provide references for the identification of malnutrition. Clinical Trial: ChiCTR-EPC-14005253
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.