Development of biomarker combinations for postoperative acute kidney injury via Bayesian model selection in a multicenter cohort study

Table 3 Combinations selected by BMA methods and their estimated performance

	Maximum posterior model probability combination	Median probability combination
AUC (95% CI)
Sustained mild AKI	0.80 (0.78, 0.87)	0.81 (0.78, 0.87)
Severe AKI	0.81 (0.76, 0.90)	0.83 (0.76, 0.90)
Posterior model probability	0.20	0.20
Odds ratios (95% CI)^a
Log plasma IL-6^b		1.58 (1.12, 2.26)
Log plasma NT-proBNP	1.60 (1.28, 2.02)	1.58 (1.26, 1.99)
Log plasma h-FABP	2.00 (1.33, 3.02)	1.85 (1.22, 2.82)
Change in serum Cr^c	1.80 (1.55, 2.11)	1.79 (1.54, 2.10)
Posterior variable probability
Log plasma IL-6^b		0.57
Log plasma NT-proBNP	1.00	1.00
Log plasma h-FABP	0.76	0.76
Change in serum Cr	1.00	1.00

Abbreviations: BMA Bayesian model averaging, AUC area under the receiver operating characteristic curve, CI confidence interval, AKI acute kidney injury, IL-6 interleukin-6, NT-proBNP N-terminal-pro-B-type natriuretic peptide, h-FABP heart-type fatty acid binding protein, Cr creatinine
^aThe odds ratios and corresponding 95% CIs are based on logistic regression with sustained mild AKI as the outcome
^bThe results for plasma IL-6 are given only for the median probability combination as plasma IL-6 was not included in the maximum posterior model probability combination
^cper 0.1 mg/dL

ISSN: 2050-7771