P Ghorbani1, L Strömmer1. 1. Division of Surgery, Department of Clinical Science, Intervention and Technology (CLINTEC), Karolinska Institute, Stockholm, Sweden.
Abstract
BACKGROUND: The wide disparity in the methodology of preventable death analysis has created a lack of comparability among previous studies. The guidelines for the peer review (PR) procedure suggest the inclusion of risk-adjustment methods to identify patients to review, that is, exclude non-preventable deaths (probability of survival [Ps] < 25%) or focus on preventable deaths (Ps > 50%). We aimed to, through PR process, (1) identify preventable death and errors committed in a level-I trauma centre, and (2) explore the use of different risk-adjustment methods as a complement. METHODS: A multidisciplinary committee reviewed all trauma patients, which died a trauma-related death, within 30 days of admission to Karolinska University Hospital, Stockholm, in the period of 2012-2016. Ps was calculated according to TRISS and NORMIT and their accuracy where compared. RESULTS: Two hundred and ninety-eight deaths were identified and 252 were reviewed. The majority of deaths occurred between 1 and 7 days. Ten deaths (4.0%) were classified as preventable. Sixty-seven errors were identified in 53 (21.0%) deaths. The most common error was inappropriate treatment in all deaths (21 of 67) and in preventable deaths (5 of 13). Median Ps in non-preventable deaths was higher than the cut-off (<25%) and Ps-TRISS was almost twice as high as Ps-NORMIT (65% vs 33%, P < .001). Two clinically judged preventable deaths with Ps <25% would have been missed with both models. Median Ps in preventable deaths was above the cut-off (>50%) and higher with Ps-TRISS vs Ps-NORMIT (75% vs 58%, P < .001). Three and 4 clinically judged preventable deaths would have been missed, respectively, for TRISS and NORMIT, if using this cut-off. CONCLUSION: Preventable deaths were commonly caused by clinical judgment errors in the early phases but death occurred late. Ps calculated with NORMIT was more accurate than TRISS in predicting mortality, but both perform poorly in identifying preventable and non-preventable deaths when applying the cut-offs. PR of all trauma death is still the golden standard in preventability analysis.
BACKGROUND: The wide disparity in the methodology of preventable death analysis has created a lack of comparability among previous studies. The guidelines for the peer review (PR) procedure suggest the inclusion of risk-adjustment methods to identify patients to review, that is, exclude non-preventable deaths (probability of survival [Ps] < 25%) or focus on preventable deaths (Ps > 50%). We aimed to, through PR process, (1) identify preventable death and errors committed in a level-I trauma centre, and (2) explore the use of different risk-adjustment methods as a complement. METHODS: A multidisciplinary committee reviewed all traumapatients, which died a trauma-related death, within 30 days of admission to Karolinska University Hospital, Stockholm, in the period of 2012-2016. Ps was calculated according to TRISS and NORMIT and their accuracy where compared. RESULTS: Two hundred and ninety-eight deaths were identified and 252 were reviewed. The majority of deaths occurred between 1 and 7 days. Ten deaths (4.0%) were classified as preventable. Sixty-seven errors were identified in 53 (21.0%) deaths. The most common error was inappropriate treatment in all deaths (21 of 67) and in preventable deaths (5 of 13). Median Ps in non-preventable deaths was higher than the cut-off (<25%) and Ps-TRISS was almost twice as high as Ps-NORMIT (65% vs 33%, P < .001). Two clinically judged preventable deaths with Ps <25% would have been missed with both models. Median Ps in preventable deaths was above the cut-off (>50%) and higher with Ps-TRISS vs Ps-NORMIT (75% vs 58%, P < .001). Three and 4 clinically judged preventable deaths would have been missed, respectively, for TRISS and NORMIT, if using this cut-off. CONCLUSION: Preventable deaths were commonly caused by clinical judgment errors in the early phases but death occurred late. Ps calculated with NORMIT was more accurate than TRISS in predicting mortality, but both perform poorly in identifying preventable and non-preventable deaths when applying the cut-offs. PR of all trauma death is still the golden standard in preventability analysis.