Comparison of the accuracy of hematological parameters in the diagnosis of neonatal sepsis: a network meta-analysis.

Abstract

Background: Currently, there are hundreds of hematological parameters used for rapid diagnosis of neonatal sepsis, but there is no network meta-analysis to compare the diagnostic efficacy of these parameters.

Methods: We searched for literature on the diagnostic neonatal sepsis and selected 20 of the most common parameters to compare their diagnostic efficacy. We used Bayesian network meta-analysis, Frequentist network meta-analysis, and individual traditional diagnostic meta-analysis to analyze the data and verify the stability of the results. Based on the above analysis, we ranked the diagnostic efficacy of 20 parameters and searched for the optimal indicator. We also conducted subgroup analysis based on different designs. GRADE was used to evaluate the quality of evidence.

Results: 311 articles were included in the analysis, of which 206 articles were included in the network meta-analysis. Bayesian models fond the top three of the advantage index were P-SEP, SAA, and CD64. In Individual model, P-SEP, SAA, and CD64 had the best sensitivity; ABC, SAA, and P-SEP had the best specificity. Frequentist model showed that CD64, P-SEP, and IL-10 ranked in the top three for sensitivity, while P-SEP, ABC, and I/M in specificity. Overall, P-SEP, SAA, CD64, and PCT have good sensitivity and specificity among all the three methods. The results of subgroup analysis were consistent with the overall analysis. All evidence was mostly of moderate or low quality.

Conclusions: P-SEP, SAA, CD64, and PCT have good diagnostic efficacy for neonatal sepsis. However, further studies are required to confirm these findings.