Related Experiment Video
Updated: Nov 15, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
7.1K
Developing Machine-Learning Prediction Algorithm for Bacteremia in Admitted Patients
Ebrahim Mahmoud1, Mohammed Al Dhoayan2,3, Mohammad Bosaeed1,4,5
1Department of Infectious Disease, Department of Medicine, King Abdulaziz Medical City, Riyadh, Saudi Arabia.
Infection and Drug Resistance
|March 4, 2021
Summary
Machine learning models can predict bacteremia in hospitalized patients with high specificity, though sensitivity is low. This approach shows promise for improving blood culture diagnostics despite limitations of traditional scoring systems.
Area of Science:
- Medical Informatics
- Infectious Diseases
- Machine Learning in Healthcare
Background:
- Bloodstream infections (BSI) lead to severe outcomes in hospitalized patients.
- Blood cultures are standard for suspected infections but often yield negative results (90%).
- Predictive models for bacteremia are limited.
Purpose of the Study:
- To evaluate machine learning models for predicting bacteremia.
- To identify factors associated with positive blood cultures.
Main Methods:
- Retrospective analysis of 36,405 blood cultures from 7,157 patients (2017-2019).
- Comparison of various machine learning models, including neural networks (NN).
- Logistic regression to identify predictive factors for bacteremia.
Main Results:
- 6.62% of blood cultures were positive.
- Neural networks achieved 88% specificity but low sensitivity for predicting bacteremia.
- Factors predicting bacteremia included longer admission duration, central line presence, and elevated lactic acid (>2 mmol/L).
Conclusions:
- Machine learning offers high specificity for predicting positive blood cultures, despite low sensitivity.
- Traditional scores like SIRS and qSOFA were not effective predictors.
- Further development of machine learning models may enhance bacteremia prediction accuracy.
