Machine learning-driven pathogen cluster analysis identifies high-risk subtypes of infected pancreatic necrosis in a multi-center cohort

Table 3 Adjusted associations between pathogen-defined clusters and 90-day mortality (Cox models)

Characteristics	Hazard ratio	Lower 95%CI	Upper 95%CI	P value
Age	0.997	0.979	1.016	0.785
APACHE II score	0.992	0.954	1.031	0.68
BISAP score	1.295	0.991	1.693	0.059
Cluster
Cluster α
Cluster β	0.337	0.118	0.965	0.043
Cluster γ	1.397	0.782	2.498	0.259
Cluster δ	2.35	1.136	4.863	0.021
Complications
No
Yes	0.793	0.476	1.323	0.375
Critical acute pancreatitis
No
Yes	0.857	0.451	1.631	0.639
Gender
Female
Male	0.809	0.414	1.582	0.536
Smoking or drinking
No
Yes	1.36	0.785	2.356	0.273
Type of pancreatitis
Biliary
Hypertriglyceridemia	1.101	0.613	1.977	0.747
Alcoholic	1.013	0.222	4.63	0.986
Other	0.875	0.405	1.893	0.735

Models adjusted for age, gender, Acute Physiology and Chronic Health Evaluation II score, etiology of pancreatitis (biliary, hypertriglyceridaemia, alcohol, other), complications, smoking or drinking and critical acute pancreatitis. Proportional-hazards assumptions were checked using Schoenfeld residuals and no material violations were observed. CI: Confidence interval; APACHE II score: Acute Physiology and Chronic Health Examination II score; BISAP score: Bedside Index of Severity in Acute Pancreatitis score; α: Enterococcus faecium/Enterobacter cloacae–predominant; β: Escherichia coli–dominant; γ: Multidrug-resistant organism-enriched; δ: Acinetobacter baumannii–Candida glabrata.