Application of a recursive partitioning decision tree algorithm for the prediction of massive transfusion in civilian trauma: the MTPitt prediction tool

Machine learning in the prediction of massive transfusion in trauma: a retrospective analysis as a proof-of-concept

PURPOSE

Early administration and protocolization of massive hemorrhage protocols (MHP) has been associated with decreases in mortality, multiorgan system failure, and number of blood products used. Various prediction tools have been developed for the initiation of MHP, but no single tool has demonstrated strong prediction with early clinical data. We sought to develop a massive transfusion prediction model using machine learning and early clinical data.

METHODS

Using the National Trauma Data Bank from 2013 to 2018, we included severely injured trauma patients and extracted clinical features available from the pre-hospital and emergency department. We subsequently balanced our dataset and used the Boruta algorithm to determine feature selection. Massive transfusion was defined as five units at 4 h and ten units at 24 h. Six machine learning models were trained on the balanced dataset and tested on the original.

RESULTS

A total of 326,758 patients met our inclusion with 18,871 (5.8%) requiring massive transfusion. Emergency department models demonstrated strong performance characteristics with mean areas under the receiver-operating characteristic curve of 0.83. Extreme gradient boost modeling slightly outperformed and demonstrated adequate predictive performance with pre-hospital data only, as well as 4-h transfusion thresholds.

CONCLUSIONS

We demonstrate the use of machine learning in developing an accurate prediction model for massive transfusion in trauma patients using early clinical data. This research demonstrates the potential utility of artificial intelligence as a clinical decision support tool.

Predictors of transfusion in trauma and their utility in the prehospital environment: a scoping review

Background: Hemorrhage is a leading cause of preventable mortality from trauma, necessitating resuscitation through blood product transfusions. Early and accurate identification of patients requiring transfusions in the prehospital setting may reduce delays in time to transfusion upon arrival to hospital, reducing mortality. The purpose of this study is to characterize existing literature on predictors of transfusion and analyze their utility in the prehospital context.Objectives: The objectives of this study are to characterize the existing quantity and quality of literature regarding predictor scores for transfusion in injured patients, and to analyse the utility of predictor scores for massive transfusions in the prehospital setting and identify prehospital predictor scores for future research.Methods: A search strategy was developed in consultation with information specialists. A literature search of OVID MEDLINE from 1946 to present was conducted for primary studies evaluating the predictive ability of scoring systems or single variables in predicting transfusion in all trauma settings.Results: Of the 5824 studies were identified, 5784 studies underwent title and abstract screening, 94 studies underwent full text review, and 72 studies were included in the final review. We identified 16 single variables and 52 scoring systems for predicting transfusion. Amongst single predictor variables, fluids administered and systolic blood pressure had the highest reported sensitivity (100%) and specificity (89%) for massive transfusion protocol (MTP) activation respectively. Amongst scoring systems for transfusion, the Shock Index and Modified Shock Index had the highest reported sensitivity (96%), while the Pre-arrival Model had the highest reported specificity (95%) for MTP activation. Overall, 20 scores were identified as being applicable to the prehospital setting, 25 scores were identified as being potentially applicable, and seven scores were identified as being not applicable.Conclusions: We identified an extensive list of predictive single variables, validated scoring systems, and derived models for massive transfusion, presented their properties, and identified those with potential utility in the prehospital setting. By further validating applicable scoring tools in the prehospital setting, we may begin to administer more timely transfusions in the trauma population.

Intelligent prediction of RBC demand in trauma patients using decision tree methods

BACKGROUND

The vital signs of trauma patients are complex and changeable, and the prediction of blood transfusion demand mainly depends on doctors' experience and trauma scoring system; therefore, it cannot be accurately predicted. In this study, a machine learning decision tree algorithm [classification and regression tree (CRT) and eXtreme gradient boosting (XGBoost)] was proposed for the demand prediction of traumatic blood transfusion to provide technical support for doctors.

METHODS

A total of 1371 trauma patients who were diverted to the Emergency Department of the First Medical Center of Chinese PLA General Hospital from January 2014 to January 2018 were collected from an emergency trauma database. The vital signs, laboratory examination parameters and blood transfusion volume were used as variables, and the non-invasive parameters and all (non-invasive + invasive) parameters were used to construct an intelligent prediction model for red blood cell (RBC) demand by logistic regression (LR), CRT and XGBoost. The prediction accuracy of the model was compared with the area under the curve (AUC).

RESULTS

For non-invasive parameters, the LR method was the best, with an AUC of 0.72 [95% confidence interval (CI) 0.657-0.775], which was higher than the CRT (AUC 0.69, 95% CI 0.633-0.751) and the XGBoost (AUC 0.71, 95% CI 0.654-0.756, P < 0.05). The trauma location and shock index are important prediction parameters. For all the prediction parameters, XGBoost was the best, with an AUC of 0.94 (95% CI 0.893-0.981), which was higher than the LR (AUC 0.80, 95% CI 0.744-0.850) and the CRT (AUC 0.82, 95% CI 0.779-0.853, P < 0.05). Haematocrit (Hct) is an important prediction parameter.

CONCLUSIONS

The classification performance of the intelligent prediction model of red blood cell transfusion in trauma patients constructed by the decision tree algorithm is not inferior to that of the traditional LR method. It can be used as a technical support to assist doctors to make rapid and accurate blood transfusion decisions in emergency rescue environment, so as to improve the success rate of patient treatment.

Development of an Accurate Bedside Swallowing Evaluation Decision Tree Algorithm for Detecting Aspiration in Acute Respiratory Failure Survivors

BACKGROUND

The bedside swallowing evaluation (BSE) is an assessment of swallowing function and airway safety during swallowing. After extubation, the BSE often is used to identify the risk of aspiration in acute respiratory failure (ARF) survivors.

RESEARCH QUESTION

We conducted a multicenter prospective study of ARF survivors to determine the accuracy of the BSE and to develop a decision tree algorithm to identify aspiration risk.

STUDY DESIGN AND METHODS

Patients extubated after ≥ 48 hours of mechanical ventilation were eligible. Study procedures included the BSE followed by a gold standard evaluation, the flexible endoscopic evaluation of swallowing (FEES).

RESULTS

Overall, 213 patients were included in the final analysis. Median time from extubation to BSE was 25 hours (interquartile range, 21-45 hours). The FEES was completed 1 hour after the BSE (interquartile range, 0.5-2 hours). A total of 33% (70/213; 95% CI, 26.6%-39.2%) of patients aspirated on at least one FEES bolus consistency test. Thin liquids were the most commonly aspirated consistency: 27% (54/197; 95% CI, 21%-34%). The BSE detected any aspiration with an accuracy of 52% (95% CI, 45%-58%), a sensitivity of 83% (95% CI, 74%-92%), and negative predictive value (NPV) of 81% (95% CI, 72%-91%). Using recursive partitioning analyses, a five-variable BSE-based decision tree algorithm was developed that improved the detection of aspiration with an accuracy of 81% (95% CI, 75%-87%), sensitivity of 95% (95% CI, 90%-98%), and NPV of 97% (95% CI, 95%-99%).

INTERPRETATION

The BSE demonstrates variable accuracy to identify patients at high risk for aspiration. Our decision tree algorithm may enhance the BSE and may be used to identify patients at high risk for aspiration, yet requires further validation.

TRIAL REGISTRY

ClinicalTrials.gov; No.: NCT02363686; URL: www.clinicaltrials.gov.