Machine learning pipeline for blood culture outcome prediction using Sysmex XN-2000 blood sample results in Western Australia

Abstract

Background Bloodstream infections (BSIs) are a significant burden on the global population and represent a key area of focus in the hospital environment. Blood culture (BC) testing is the standard diagnostic test utilised to confirm the presence of a BSI. However, current BC testing practices result in low positive yields and overuse of the diagnostic test. Diagnostic stewardship research regarding BC testing is increasing, and becoming more important to reduce unnecessary resource expenditure and antimicrobial use, especially as antimicrobial resistance continues to rise. This study aims to establish a machine learning (ML) pipeline for BC outcome prediction using data obtained from routinely analysed blood samples, including complete blood count (CBC), white blood cell differential (DIFF), and cell population data (CPD) produced by Sysmex XN-2000 analysers. Methods ML models were trained using retrospective data produced between 2018 and 2019, from patients at Sir Charles Gairdner hospital, Nedlands, Western Australia, and processed at Pathwest Laboratory Medicine, Nedlands. Trained ML models were evaluated using stratified 10-fold cross validation. Results Two ML models, an XGBoost model using CBC/DIFF/CPD features with boruta feature selection (BFS) , and a random forest model trained using CBC/DIFF features with BFS were selected for further validation after obtaining AUC scores of $$0.76 \pm 0.04$$ 0.76 ± 0.04 and $$0.75 \pm 0.04$$ 0.75 ± 0.04 respectively using stratified 10-fold cross validation. The XGBoost model obtained an AUC score of 0.76 on a internal validation set. The random forest model obtained AUC scores of 0.82 and 0.76 on internal and external validation datasets respectively. Conclusions We have demonstrated the utility of using an ML pipeline combined with CBC/DIFF, and CBC/DIFF/CPD feature spaces for BC outcome prediction. This builds on the growing body of research in the area of BC outcome prediction, and provides opportunity for further research.

Recent Advancements in Machine Learning-Based Bloodstream Infection Prediction: A Systematic Review and Meta-analysis of Diagnostic Test Accuracy

“…Within this context, three studies utilized the publicly available Medical Information Mart for Intensive Care (MIMIC) database [41, 44, 48], while four studies included Complete Blood Count/Differential Count (CBC/DC) data [24, 25, 34, 40]. Furthermore, two of these studies also incorporated Cell Population Data (CPD) [24, 40], highlighting the integration of detailed hematologic parameters. Unstructured data utilization was observed in one study that utilized textual data [23].…”

“…In terms of data accessibility, most studies employed proprietary hospital data. Data sharing policies varied: one study explicitly stated that their data would not be shared [47], one study offered openly available data [34], and ten studies indicated that deidentified data could be provided upon reasonable request [24, 25, 27, 28, 29, 38, 39, 41, 42, 49], thus contributing to transparency and reproducibility. Notably, six articles did not specify the number of patients included in their analysis [22, 24, 25, 28, 31, 48].…”

“…The studies predominantly took place in inpatient settings (56%, n=17), with the emergency department (ED) (23%, n=7) and intensive care units (ICU) (20%, n=6) also represented. Within inpatient settings, nine studies examined general populations [20][21][22][23][24][25][26][27][28], and others focused on specific patient cohorts including those with central venous catheters (CVC) [30,31], systemic inflammatory response syndrome (SIRS) [29], and various other conditions. In ED settings, seven studies were identified [37][38][39][40][41][42][43], two of which addressed specific patient groups.…”

Evaluation of machine learning pipeline for blood culture outcome prediction on prospectively collected data in Western Australian emergency department

“…The two models evaluated in this study were trained in our previous study 19 . The XG model was trained using CBC, DIFF, and CPD features with boruta feature selection (BFS) and 1.5 class weights in favour of the positive BC result; and the RF model was trained using CBC and DIFF features, BFS, and balanced class weights.…”

“…In our previous work, we developed a ML pipeline for BC outcome prediction which achieved promising results using models trained with complete blood count (CBC), white blood cell differential (DIFF), and cell population data (CPD). 19 . In this study, we further validated this pipeline using a prospectively collected ED dataset.…”