Evaluating the Efficacy of Artificial Intelligence Models in Predicting Intensive Care Unit Admission Needs

Overview

This study aims to evaluate the efficacy of two artificial intelligence (AI) models in predicting the need for ICU admissions. By comparing the AI models' predictions with actual clinical decisions, we aim to determine their accuracy and potential utility in clinical decision support.

Intensive care units (ICUs) are critical components of healthcare systems, providing life-saving care to patients with severe and life-threatening conditions. Timely and accurate prediction of ICU admission needs is essential for improving patient outcomes and optimizing hospital resource allocation. Delayed ICU admissions have been consistently associated with higher morbidity and mortality rates. With the advent of artificial intelligence (AI) in healthcare, there is an opportunity to enhance clinical decision-making by leveraging AI models to predict ICU needs accurately. AI models, such as ChatGPT and Gemini, can process vast amounts of complex data to identify patterns that might not be immediately evident to human clinicians, potentially improving the speed and accuracy of ICU admission decisions. This is an observational retrospective study. Data were collected from electronic health records (EHRs) from a hospital retrospectively. Data were extracted from EHRs and included: Demographic data: Age, gender, and basic patient characteristics. Clinical parameters: Medication information, consultation details, ECG findings, imaging results, comorbid conditions (e.g., diabetes mellitus, hypertension, heart failure, COPD, cerebrovascular events), and laboratory values (e.g., hemoglobin, hematocrit, platelet count, PT, INR, procalcitonin, ALT, AST, bilirubin, sodium, potassium, chloride, glucose, creatinine, urea, albumin, thyroid function tests). Prediction data: AI model predictions and actual ICU admission decisions.

Conditions

Interventions

Eligibility

Locations (1)

Outcomes