Comparative evaluation of the aggregate index of systemic inflammation (AISI) and its modified version for early detection of central line-associated bloodstream infection: a pilot study using machine learning techniques

Background: Central line-associated bloodstream infections (CLABSI) are a common and serious problem in critically ill patients; their early detecting is challenging. This study evaluated the predictive ability of the aggregate index of systemic inflammation (AISI) and its modified form for early identification of CLABSI within two calendar days following central line insertion, using a machine learning approach. Method: We conducted an analysis of patients who received central line insertion. Inflammatory indices were calculated using laboratory parameters obtained on second day post-insertion. Four machine learning algorithms were applied to evaluate their predictive performance for early CLABSI detection. Results: Among 234 patients who met the inclusion criteria, 39 were confirmed CLABSI cases. We found both indices significantly elevated in the CLABSI group. Modified AISI demonstrated the strongest performance using XGBoost, with the highest area under the ROC curve (0.99), 97% sensitivity and 98% specificity, indicating its potential as the better early screening marker for CLABSI than AISI. Conclusion: Both AISI and modified AISI demonstrated strong predictive value for early CLABSI detection, being both accessible and cost-effective. Modified AISI outperformed AISI in predictive performance. These findings support the need for the prospective validation of the modified AISI before clinical implementation.