Development and Validation of a Renal Replacement after Trauma Scoring Tool.

Background : Stress on the healthcare system requires careful allocation of resources such as renal replacement therapy (RRT). The COVID-19 pandemic generated difficulty securing access to RRT for trauma patients. We sought to develop a renal replacement after trauma (RAT) scoring tool to help identify trauma patients who may require RRT during their hospitalization. Study Design : The 2017-2020 Trauma Quality Improvement Program (TQIP) database was divided into a derivation (2017-2018 data) and validation (2019-2020 data) set. A three-step methodology was used. Adult trauma patients admitted from the emergency department (ED) to the operating room or intensive care unit were included. Patients with chronic kidney disease, transfers from another hospital, and ED deaths were excluded. Multiple logistic regression models were created to determine the risk for RRT in trauma patients. The weighted average and relative impact of each independent predictor was used to derive a RAT score, which was validated using area under receiver-operating characteristic curve (AUROC). RESULTS : From 398,873 patients in the derivation and 409,037 patients in the validation set, 11 independent predictors of RRT were included in the RAT score derived with scores ranging from 0-11. The AUROC for the derivation set was 0.85. The rate of RRT increased to 1.1%, 3.3%, and 20% at scores of 6, 8, and 10, respectively. The validation set AUROC was 0.83. CONCLUSION : RAT is a novel and validated scoring tool to help predict the need for RRT in trauma patients. With future improvements including baseline renal function and other variables, the RAT tool may help prepare for the allocation of RRT machines/staff during times of limited resources.

Background: Stress on the healthcare system requires careful allocation of resources such as renal replacement therapy (RRT).The COVID-19 pandemic generated difficulty securing access to RRT for trauma patients.We sought to develop a renal replacement after trauma (RAT) scoring tool to help identify trauma patients who may require RRT during their hospitalization.

Study Design:
The 2017-2020 Trauma Quality Improvement Program (TQIP) database was divided into a derivation (2017-2018 data) and validation (2019-2020 data) set.A three-step methodology was used.Adult trauma patients admitted from the emergency department (ED) to the operating room or intensive care unit were included.Patients with chronic kidney disease, transfers from another hospital, and ED deaths were excluded.Multiple logistic regression models were created to determine the risk for RRT in trauma patients.The weighted average and relative impact of each independent predictor was used to derive a RAT score, which was validated using area under receiver-operating characteristic curve (AUROC).

RESULTS:
From 398,873 patients in the derivation and 409,037 patients in the validation set, 11 independent predictors of RRT were included in the RAT score derived with scores ranging from 0-11.The AUROC for the derivation set was 0.85.The rate of RRT increased to 1.1%, 3.3%, and 20% at scores of 6, 8, and 10, respectively.The validation set AUROC was 0.83.

CONCLUSION:
RAT is a novel and validated scoring tool to help predict the need for RRT in trauma patients.With future improvements including baseline renal function and other variables, the RAT tool may help prepare for the allocation of RRT machines/staff during times of limited resources.

Introduction
Renal replacement therapy (RRT) is a life-saving treatment for patients with insufficient renal function.[3] Despite its critical importance, resources for RRT have been severely limited and this shortage was exacerbated by the unprecedented needs of the recent COVID-19 pandemic. 1,2Trauma patients are at particularly high-risk for suffering acute kidney injury, with rates as high as 67% amongst the most severely injured. 35][6][7] Furthermore, the ability to identify trauma patients early who eventually require RRT would allow providers and hospital administrators the opportunity to triage resources to provide optimal care.Currently, there are several prediction models to evaluate patients with chronic kidney disease who will progress to need RRT. 8-14However, no tool has been developed to predict the need for RRT in trauma patients.The purpose of this study is to develop and validate a novel Renal replacement After Trauma (RAT) scoring tool to help identify trauma patients who require RRT during their index hospitalization.An accurate and validated scoring tool could serve to identify high-risk patients allowing appropriate counseling and ensuring the availability of RRT, especially in a time of resource limitation.prospective trauma data by trained professionals.This study was deemed exempt by our institutional review board, and a waiver of informed consent granted.

This
6][17] The 2017-2018 TQIP dataset was queried for trauma patients who were 18 years of age or older and admitted from the emergency department (ED) to the operating room (OR) or intensive care unit (ICU).Also, any trauma patient admitted initially to the surgical floor but subsequently was upgraded to the ICU during their hospitalization was included, as these patients may potentially require RRT.Patients with chronic kidney disease, transferred from another hospital, or who died in the ED were excluded.This 2017-2018 data served as the derivation set to develop the RAT scoring tool.The dataset was separated into two groups.All trauma patients who required RRT (either hemodialysis or continuous RRT) during the index hospitalization were included in the (+) RRT group.All others comprised the (-) RRT group.A univariate analysis was used to compare the two groups based on demographics (e.g., age and sex) and comorbidities (e.g., cirrhosis, diabetes, hypertension, and congestive heart failure).In addition, injury profile data including specific organ injuries were recorded.Finally, operations (based on International Classification of Diseases codes) and complications including packed red blood cell transfusions, sepsis, unplanned intubation, ICU admission or return to the OR, ventilator days, length of stay and mortality were collected.Variables with a p-value < 0.2 were included in stepwise multivariate logistic regression models to identify independent risk factors for RRT.9][20][21] The weighted average and odds ratio of each independent factor were used to inform multiple iterations of the scoring tool, which was simplified to provide its ease of use.The area under the receiveroperating curve (AUROC) was calculated after each iteration to verify its continued accuracy.
After deriving the RAT scoring tool, validation was performed using the 2019-2020 TQIP dataset, using the same inclusion/exclusion criteria.An AUROC was performed based on the 2019-2020 dataset and directly compared to the AUROC of the derivation 2017-2018 dataset for validation of the RAT Scoring Tool.All analyses were performed with IBM SPSS Statistics for Windows (version 24, IBM Corp, Armonk, NY).
Multiple logistic regression models identified 11 independent predictors of RRT which were male sex, mechanical ventilation, comorbidities such as cirrhosis, diabetes, hypertension or congestive heart failure, hypotension on arrival, packed red blood cell transfusion within 4 hours of presentation, operation involving the respiratory, gastrointestinal, hepatobiliary, and urinary system within 24 hours of presentation, renal injury, and lower extremity fracture.Each predictor had a similar effect on the risk for RRT and the RAT score was derived with scores ranging from 0-11 with each variable carrying equal weight (Table 5).The AUROC for the derivation set was 0.85.The rate of RRT increased steadily from 1.1%, 3.3%, and 20% at scores of 6, 8, and 10, respectively (Figure 1).Few patients achieved a RAT score of 11 and therefore the incidence of RRT could not be calculated for the maximum RAT score.
The TQIP 2019-2020 validation dataset was comprised of 409,037 patients who met inclusion/exclusion criteria for the RAT Scoring Tool validation analysis.The AUROC curve for the validation set was 0.83, similar to the derivation set (Figure 2).

Discussion
RRT is an expensive and labor-intensive, albeit life-saving resource. 22In order to maximize the benefit from such a limited resource, it is prudent for health care providers and administrators to identify potential patients who will require RRT at an early stage in their hospitalization.11][12][13][14]23,24 This large national analysis spanning 4 years of data identified risk factors for RRT in adult trauma patients who are most susceptible to develop acute renal failure after injury (i.e., admission to the OR from ED or admission to the ICU at some point during their hospitalization).These risk factors include male sex, mechanical ventilation, comorbidities (i.e., cirrhosis, diabetes, hypertension, or congestive heart failure), hypotension on arrival, packed red blood cell transfusion within 4 hours of presentation, operation involving the respiratory, gastrointestinal, hepatobiliary, and urinary system within 24 hours of presentation, renal injury, and lower extremity fracture.As a result, an easy-to-use integer-based RAT scoring tool was developed and validated using contemporary nationwide data from trauma patients.
Most of the risk factors identified in this study are known to have an impact on renal failure following injury which provides further justification for their inclusion in the novel RAT scoring tool. 4,6,18,25In addition, a major advantage of the RAT scoring tool is its simplicity and the availability of most variables shortly after presentation which allows for a timely identification of trauma patients who may need RRT.0][21] These variables were also more common in patients receiving RRT in this study.However, the addition of rhabdomyolysis and obesity lowered the quality of the RAT scoring tool and thus were excluded from the RAT score.Also, having all variables readily available provided further support for excluding rhabdomyolysis from the RAT scoring tool as well.Additionally, sickle-cell disease/trait has been shown to be a risk factor for exertional rhabdomyolysis but its association with post-traumatic RRT could not be evaluated as no patients within the RRT group carried the diagnosis. 26Lastly, fasciotomy was evaluated for inclusion in the RAT Scoring Tool but did not improve the model and was excluded to maintain ease-of-use.Future studies are needed to evaluate these findings and determine if a more narrow subset of these variables may prove helpful in further honing the RAT scoring tool.
This study has numerous limitations, including the inherent potential for misclassification and missing variables within a large national dataset.There are also institutional variations in criteria for initiation of RRT which are not accounted for in this analysis as this data is not available within TQIP.Additionally, TQIP does not provide information regarding the time to initiation of RRT and the use of intravenous contrast for imaging so its effects on renal failure cannot be evaluated.Although, recent studies have shown intravenous contrast does not affect renal complications. 27,28In terms of methodology, there are multiple techniques to develop risk scoring tools such as machine learning or decision tree analysis that may prove more helpful.

A C C E P T E D
2023 by the American College of Surgeons.Published by Wolters Kluwer Health, Inc.All rights reserved 17 and decision tree analysis favors the most common variables rather than focusing on the most predictive variables.As these most predictive variables for RRT are not the most common, decision tree analysis may not prove as helpful as our current methodology, but merits future study.
Also, there are several other known predictors of acute renal failure including, but not limited to, urine output, baseline serum creatinine, creatine phosphokinase, intravenous fluid balance, serum lactate, mean arterial pressure, hemorrhagic shock duration, and exposure to nephrotoxic medications which are not available within TQIP. 6,19,21,25,29Acknowledging this, we believe this current study serves as a scaffold for future prospective studies to build upon and incorporate more granular data (i.e., serum lactate/creatinine, urine output, and continuous physiologic data) to further hone the RAT Scoring Tool.Finally, it is important to acknowledge the limitations of clinical application for the RAT score in its current form given that the highest predicted risk of post-traumatic RRT is 20% with a RAT score of 10.Although this predicted risk of 20% is somewhat low, this is over 400 times the national incidence for trauma patients.
Furthermore, the RAT Scoring Tool is the first trauma specific risk tool to predict the need for RRT with good predictive capability and has the potential to be an invaluable tool for resource allocation with further improvements.

Conclusion
The RAT score is a novel and validated scoring tool to predict the need for RRT in trauma patients.This tool may be able to help hospital systems better prepare for the allocation of precious resources including RRT machines and trained staff to safely manage patients with acute renal failure, especially during periods when resources are limited.However, prior to

:
renal replacement therapy, dialysis, trauma, kidney failure, scoring tool, TQIP American College of Surgeons.Published by Wolters Kluwer Health, Inc.All rights reserved study was conducted through a retrospective review of the Trauma Quality Improvement Program (TQIP) database, a multicenter database that systematically collects American College of Surgeons.Published by Wolters Kluwer Health, Inc.All rights reserved American College of Surgeons.Published by Wolters Kluwer Health, Inc.All rights reserved American College of Surgeons.Published by Wolters Kluwer Health, Inc.All rights reserved American College of Surgeons.Published by Wolters Kluwer Health, Inc.All rights reserved American College of Surgeons.Published by Wolters Kluwer Health, Inc.All rights reserved both of these variables are known within 24-hours of presentation and thus still occur early enough in hospitalization to aid with prognostication and resource allocation.

Figure 2 :
Figure 2: Area under the receiver operating characteristic curve (AUROC) for development of the American College of Surgeons.Published by Wolters Kluwer Health, Inc.All rights reserved Figure 1 Figure 2

Table 1 .
Demographics of Patients Included in The Derivation Dataset Based on Renal Replacement Therapy Requirement

Table 2 .
Injuries and Procedures of Patients Included in the Derivation Dataset Based on Renal Replacement Therapy Requirement

2023 by the American College of Surgeons. Published by Wolters Kluwer Health, Inc. All rights reserved
2023 by the American College of Surgeons.Published by Wolters Kluwer Health, Inc.All rights reserved

Table 3 .
Complications of Patients Included in the Derivation Dataset Based on Renal Replacement Therapy Requirement

Table 5 .
Development of the Renal Replacement After Trauma Scoring Tool under receiver-operating characteristic curve; CHF, congestive heart failure; DM, diabetes mellitus; HTN, hypertension; PRBC, packed red blood cell