Literature DB >> 12147111

Determining the most effective level of TRISS-derived probability of survival for use as an audit filter.

Anne-Maree Kelly1, Jon Nicholl, Janette Turner.   

Abstract

OBJECTIVE: To determine the most effective cut-off of TRISS-derived probability of survival (TRISS-PS) for the selection of trauma deaths for audit, using a large sample of trauma deaths from the United Kingdom (UK).
METHODS: TRISS-PS and avoidability of death (as judged by an independent peer review panel) were compared for a sample of 222 trauma deaths. Sensitivity, specificity and predictive values were calculated for the 0.5 screening cut-off. ROC curves were derived to assess the ability of different levels of TRISS-PS to identify avoidable deaths. Calculations were made for both the raw sample and the sample adjusted for the sampling method used.
RESULTS: For the weight-adjusted sample, the sensitivity of TRISS-PS greater than 0.5 for the detection of avoidable death is 80% (95% CI 61-91%), the specificity is 86% (95% CI 80-90%), PPV 42% (95% CI 29-56%) and NPV 97% (95% CI 93-99%). Twenty percent of avoidable deaths would have been 'missed' if the 0.5 level of audit filter had been used. Based on the same sample, the best cut-off is at TRISS-PS 0.33, with a sensitivity of 90% and specificity of 80%. It is estimated that this cut-off would have selected 62 deaths for audit and failed to identify 2 out of 25 avoidable deaths.
CONCLUSION: The previously accepted audit filter of TRISS-PS of greater than 0.5 fails to identify a significant proportion of avoidable deaths. This study suggests that the most effective level of audit filter cut-off of TRISS-PS for the trauma system studied is 0.33. This level would identify 90% of avoidable deaths with 80% specificity. Similar ROC curve analysis could be used to determine appropriate TRISS-PS cut-offs for institutions or other trauma systems.

Entities:  

Mesh:

Year:  2002        PMID: 12147111     DOI: 10.1046/j.1442-2026.2002.00309.x

Source DB:  PubMed          Journal:  Emerg Med (Fremantle)        ISSN: 1035-6851


  2 in total

1.  The role of trauma scoring in developing trauma clinical governance in the Defence Medical Services.

Authors:  R J Russell; T J Hodgetts; J McLeod; K Starkey; P Mahoney; K Harrison; E Bell
Journal:  Philos Trans R Soc Lond B Biol Sci       Date:  2011-01-27       Impact factor: 6.237

2.  Machine Learning Models of Survival Prediction in Trauma Patients.

Authors:  Cheng-Shyuan Rau; Shao-Chun Wu; Jung-Fang Chuang; Chun-Ying Huang; Hang-Tsung Liu; Peng-Chen Chien; Ching-Hua Hsieh
Journal:  J Clin Med       Date:  2019-06-05       Impact factor: 4.241

  2 in total

北京卡尤迪生物科技股份有限公司 © 2022-2023.