Network meta-analysis of the accuracy of early warning assessment tools for mortality among trauma patients

He Congcong; Bi Zhengqin; Zhang Yu; Chen Wuying; Lan Quanming; Zhou Limin

doi:10.3969/j.issn.1009-4237.2026.02.010

Abstract：Objective To explore and compare the accuracy of different warning assessment tools for early death prediction among trauma patients.Methods The databases of China Biology Medicine disc,CNKI,Wanfang data,VIP,Cochrane Library,PubMed,Web of Science,and Embase were searched for published Chinese or English diagnostic studies on the effectiveness of in-hospital mortality assessment tools for trauma patients,in May 2024.Two evaluators independently screened the literature,extracted information and evaluated the reference quality using the diagnostic research quality assessment tool(QUADAS-2).Meta-analysis was performed using the RevMan 5.3 soft-ware,and network meta-analysis was performed using the Stata 26.0 software.Results A total of 38 references in-volving 36 assessment tools were included.For further analysis,the assessment tools studied by＜2 references were excluded,and finally a total of 18 assessment tools were analyzed in the study.QUADAS-2 determined a low overall risk of bias.The surface under the cumulative ranking curve results showed that the top three tools for sensitivity were:TRISS(trauma and injury severity score,89.2%)＞RSI(reverse shock index,85.0%)＞CRAMS scale(cir-culation,respiration,abdomen/thorax,motor,and speech,68.4%).The top three tools for specificity were:A-PACHE Ⅱ(81.8%)＞SI(shock index,77.0%)＞RAPS(rapid acute physiology score,68.1%).Conclusion A phased strategy is recommended for trauma assessment.At the emergency scene and during initial hospital arrival,RSI and SI can be used for rapid triage.During the subsequent in-hospital monitoring phase,TRISS or APACHE Ⅱ can be employed for precise prognosis prediction,depending on the available medical resources.