Richard Y Calvo1, C Beth Sise2, Michael J Sise3, Vishal Bansal4. 1. Scripps Mercy Hospital, Trauma Services, 4077 Fifth Avenue, San Diego, CA 92103, USA; SDSU/UCSD Joint Doctoral Program in Public Health (Epidemiology), 5500 Campanile Drive, San Diego, CA, 92182, USA. Electronic address: calvo.richard@scrippshealth.org. 2. Scripps Mercy Hospital, Trauma Services, 4077 Fifth Avenue, San Diego, CA 92103, USA. Electronic address: sise.beth@scrippshealth.org. 3. Scripps Mercy Hospital, Trauma Services, 4077 Fifth Avenue, San Diego, CA 92103, USA. Electronic address: sise.mike@scrippshealth.org. 4. Scripps Mercy Hospital, Trauma Services, 4077 Fifth Avenue, San Diego, CA 92103, USA. Electronic address: bansal.vishal@scrippshealth.org.
Abstract
INTRODUCTION: Pre-existing medical conditions (PEC) represent a unique domain of risk among older trauma patients. The study objective was to develop a metric to quantify PEC burden for trauma patients. METHODS: A cohort of 4526 non-severe blunt-injured trauma patients aged 55 years and older admitted to a Level I trauma center between January 2006 and December 2012 were divided into development (80%) and test (20%) sets. Cox regression was used to develop the model based on in-hospital and 90-day mortality. Regression coefficients were converted into a point-based PEC Risk Score. Performance of the PEC Risk Score was compared in the test set with two other PEC-based metrics and three injury-based metrics. An external cohort of 2284 trauma patients admitted in 2013 was used to evaluate combined metric performance. RESULTS: Total mortality was 9.4% and 9.1% in the development and test set, respectively. The final model included 12 PEC. In the test set, the PEC Risk Score (c-statistic: 79.7) was superior for predicting in-hospital and 90-day mortality compared with all other metrics. For in-hospital mortality alone, the PEC Risk Score similarly outperformed all other metrics. Combination of the PEC Risk Score and any injury-based metric significantly improved prediction compared with any injury-based metric alone. CONCLUSION: Our 12-item PEC Risk Score performed well compared with other metrics, suggesting that the classification of trauma-related mortality risk may be improved through its use. Among non-severely injured older trauma patients, the utility of prognostic metrics may be enhanced through the incorporation of comorbidities.
INTRODUCTION: Pre-existing medical conditions (PEC) represent a unique domain of risk among older traumapatients. The study objective was to develop a metric to quantify PEC burden for traumapatients. METHODS: A cohort of 4526 non-severe blunt-injured traumapatients aged 55 years and older admitted to a Level I trauma center between January 2006 and December 2012 were divided into development (80%) and test (20%) sets. Cox regression was used to develop the model based on in-hospital and 90-day mortality. Regression coefficients were converted into a point-based PEC Risk Score. Performance of the PEC Risk Score was compared in the test set with two other PEC-based metrics and three injury-based metrics. An external cohort of 2284 traumapatients admitted in 2013 was used to evaluate combined metric performance. RESULTS: Total mortality was 9.4% and 9.1% in the development and test set, respectively. The final model included 12 PEC. In the test set, the PEC Risk Score (c-statistic: 79.7) was superior for predicting in-hospital and 90-day mortality compared with all other metrics. For in-hospital mortality alone, the PEC Risk Score similarly outperformed all other metrics. Combination of the PEC Risk Score and any injury-based metric significantly improved prediction compared with any injury-based metric alone. CONCLUSION: Our 12-item PEC Risk Score performed well compared with other metrics, suggesting that the classification of trauma-related mortality risk may be improved through its use. Among non-severely injured older traumapatients, the utility of prognostic metrics may be enhanced through the incorporation of comorbidities.
Authors: Sascha Halvachizadeh; Lea Gröbli; Till Berk; Kai Oliver Jensen; Christian Hierholzer; Heike A Bischoff-Ferrari; Roman Pfeifer; Hans-Christoph Pape Journal: PLoS One Date: 2021-01-11 Impact factor: 3.240