The role of artificial intelligence in mortality prediction of patients aged 65 and over presenting to the emergency department

Aim: The aim of this study is to investigate the effectiveness of artificial intelligence algorithms in predicting short-term mortality risk in patients aged 65 and over presenting to the emergency department, compare the accuracy of models developed in the literature, and evaluate the potential contributions of these technologies to the clinical decision-making process. Material and Method: The study was approved by the Non-Interventional Clinical Research Ethics Committee of Uşak University with the approval number 330-330-08 dated 15.03.2024. The study is a prospective cross-sectional descriptive study. Results: A total of 2000 patients were included in the study. 49% of the patients (n=986) were female, and 51% (n=1014) were male, with a mean age of 77.98 ± 7.72 years. The comorbidity rate was 93.8%, and the medication use rate was 90.5%. The mortality rate was found to be 22.6% (n=452). In the logistic regression analysis performed formortality, high age (p=0.001), low glaskow coma scale (GCS) (p=0.001), low systolic (p=0.001) and diastolic blood pressure (p=0.001), low heart rate (p=0.001), low SPO2 (p=0.001), high fever (p=0.001), sinus bradycardia (p=0.001) and other bradycardias (p=0.001), hospital admission (p=0.001), chronic kidney disease (CKD) (p=0.001), asthma/COPD (p=0.036), heart failure (p=0.010), Parkinson's disease (p=0.001), arrhythmia (p=0.001), history of stroke (CVE) (p=0.001), history of cancer (p=0.001), history of bypass surgery (p=0.001), high WBC (p=0.001), high lymphocyte count (p=0.026), low eosinophils and hemoglobin (p=0.001), high glucose (p=0.001), creatinine (p=0.001), BUN (p=0.001), CK (p=0.001), bilirubin (p=0.001), AST (p=0.001), ALT (p=0.001), LDH (p=0.001), sodium (p=0.021), potassium (p=0.001), CRP (p=0.001), high lactate (p=0.001), low albumin (p=0.001), low calcium (p=0.001), and low pH (p=0.001) were found to be statistically significant. In the mortality analysis performed using artificial intelligence algorithms, the artificial intelligence score (p=0.001) and the artificial intelligence risk status (p=0.001) were found to be statistically significant. Conclusion: Artificial Intelligence-Based Risk Scores can be effectively used for early mortality prediction in older patients in emergency department triage; these models should be integrated into routine clinical practices. Extended, multi-center, and prospective studies are needed to evaluate the generalizability of these findings and the effects of artificial intelligence algorithms in real-world settings. vii Keywords: Emergency department, Elderly patient, Artificial intelligence