Artificial Intelligence based Early Identification of Bacteremia in Patients Admitted iIn Medical ICU Experience from A Tertiary Care Hospital in Southern India
Year: 2025
Type: article
Study design: Observational Study
Abstract: INTRODUCTION: Prediction of bacteremia is a clinically important but challenging task. An artificial intelligence (AI) model has the potential to facilitate early bacteremia prediction, aiding critical care physicians in making timely decisions and reducing unnecessary medical costs. In this study, we aim to develop and externally validate an Artificial Intelligence bacteremia prediction model (AI-BPM). We also aim to evaluate its impact on bacteremia prediction by comparing blood culture results and AI – BPM prediction results.(1) OBJECTIVES: Primary: To develop an Artificial Intelligence Bacteremia Prediction Model for the early identification of bacteremia in patients admitted in Medical ICU. Secondary: 1. To evaluate the efficacy of AI BPM by comparing blood culture results with AI BPM prediction results. 2. To assess the sensitivity and specificity of AI BPM. 3. To assess the positive predictive value and negative predictive value of AI BPM. MATERIALS AND METHODS: Setting: Department of Critical Care Medicine, GG Hospital. Design: Prospective Observational Study Population: All patients admitted in the Medical ICU with atleast 1 set of blood cultures taken. Duration: 1 year (December 2023–December 2024) METHODS: We created an AI BPM model using Naïve Bayes classification algorithm. The parameters used to define bacteremia in the prediction model were the presence of comorbidities, recent hospitalizations, presence of indwelling devices, recent history of antibiotic use, fever, hypotension, tachypnea, tachycardia, altered mental status, desaturation, leukocytosis/leukopenia, high neutrophil count, raised CRP, raised lactate, deranged liver and renal function tests, APACHE 4 score, presence of pus cells and bacteria in urine, recent surgical interventions, and use of immunosuppressive medications.(2) The development dataset is randomly split into 2, in which half of the data is used for AI BPM creation and the remaining data is used for validation of the AI BPM. RESULTS: This is an ongoing study CONCLUSION: This is an ongoing study.
Authors: Greeshma R, Anvar M Ahammed
Language: English
Subfield: Epidemiology
Field: Medicine
Domain: Health Sciences
Open Access status: diamond
OpenAlex ID: W4410920364