Prediction of mortality among severely injured trauma patients A comparison between TRISS and machine learning-based predictive models.

BACKGROUND: Given the huge impact of trauma on hospital systems around the world, several attempts have been made to develop predictive models for the outcomes of trauma victims. The most used, and in many studies most accurate predictive model, is the "Trauma Score and Injury Severity Score" (TRISS). Although it has proven to be fairly accurate and is widely used, it has faced criticism for its inability to classify more complex cases. In this study, we aimed to develop machine learning models that better than TRISS could predict mortality among severely injured trauma patients, something that has not been studied using data from a nationwide register before. METHODS: Patient data was collected from the national trauma register in Sweden, SweTrau. The studied period was from the 1st of January 2015 to 31st of December 2019. After feature selection and multiple imputation of missing data three machine learning (ML) methods (Random Forest, eXtreme Gradient Boosting, and a Generalized Linear Model) were used to create predictive models. The ML models and TRISS were then tested on predictive ability for 30-day mortality. RESULTS: The ML models were well-calibrated and outperformed TRISS in all the tested measurements. Among the ML models, the eXtreme Gradient Boosting model performed best with an AUC of 0.91 (0.88-0.93). CONCLUSION: This study showed that all the developed ML-based prediction models were superior to TRISS for the prediction of trauma mortality.

Development and Internal Validation of a Multivariable Prediction Model for Mortality After Hip Fracture with Machine Learning Techniques.

In order to estimate the likelihood of 1, 3, 6 and 12 month mortality in patients with hip fractures, we applied a variety of machine learning methods using readily available, preoperative data. We used prospectively collected data from a single university hospital in Copenhagen, Denmark for consecutive patients with hip fractures, aged 60 years and older, treated between September 2008 to September 2010 (n = 1186). Preoperative biochemical and anamnestic data were used as predictors and outcome was survival at 1, 3, 6 and 12 months after the fracture. After feature selection for each timepoint a stratified split was done (70/30) before training and validating Random Forest models, extreme gradient boosting (XGB) and Generalized Linear Models. We evaluated and compared each model using receiver operator characteristic (ROC), calibration slope and intercept, Spiegelhalter's z- test and Decision Curve Analysis. Using combinations of between 10 and 13 anamnestic and biochemical parameters we were able to successfully estimate the likelihood of mortality with an area under the curve on ROC curves of 0.79, 0.80, 0.79 and 0.81 for 1, 3, 6 and 12 month, respectively. The XGB was the overall best calibrated and most promising model. The XGB model most successfully estimated the likelihood of mortality postoperatively. An easy-to-use model could be helpful in perioperative decisions concerning level of care, focused research and information to patients. External validation is necessary before widespread use and is currently underway, an online tool has been developed for educational/experimental purposes ( https://hipfx.shinyapps.io/hipfx/ ).

Centenarian hip fracture patients: a nationwide population-based cohort study of 507 patients.

Background and purpose - Several studies suggest a global increase of centenarians during the 21st century. We describe temporal trends of hip fracture incidence and mortality in this group and compare these patients with a group of younger hip fracture patients with regards to comorbidities and mortality. Patients and methods - The full study population included all hip fractures that occurred in Denmark (n = 154,047) between 1996 and 2012. Patients aged 100 or above were identified (n = 507) and hip fracture patients between the ages of 70 to 99 years (n = 124,007) were used for comparison. Data were accessed from national registries. Trends in incidence over time were analyzed using a log-linear regression model, mortality was analyzed using the Kaplan-Meier estimator and trends in mortality over time were analyzed using a log-binomial regression model to obtain relative risk estimates. Results and interpretation - The centenarian patients had fewer comorbidities than the younger comparison group, but mortality was higher at all timepoints. There was no statistically significant change in mortality over time but the incidence of hip fracture among centenarians decreased during the same time period. Our findings describe the characteristics of an emerging group of hip fracture patients and could be of use in the planning of healthcare in the years to come.

Metaanalysis of risk factors for mortality in patients with hip fracture.

INTRODUCTION: The aim of this meta-analysis is to assess the association of three different clinical score systems with the mortality in hip fracture patients. METHODS: A literature search was conducted on November 13, 2011 using PubMed and Embase. The search yielded 315 publications which were reviewed on the basis of the inclusion criteria. RESULTS: Thirteen studies were included for further processing. The following clinical score systems were found to be of prognostic value for mortality in hip fracture patients: a high American Society of Anesthesiologists (ASA) score of three or above (odds ratio (OR): 3.07; 95% confidence interval (CI): 2.78-3.38; p < 0.00001, 15,625 study participants included), a Charlson Comorbidity Index (CCI) score of one or more (OR: 2.05; 95% CI: 1.79-2.34; p < 0.00001, 13,570 study participants included) and dementia (assessed with Mini Mental State Examination or obtained from journal extraction) (OR: 2.73; 95% CI: 1.64-4.57; p = 0.0001; 1,782 study participants included). CONCLUSION: The present meta-analysis showed that the ASA score, the CCI score and assessment of preexisting dementia are useful in predicting the mortality of hip fracture patients.

Routine blood tests as predictors of mortality in hip fracture patients.

OBJECTIVE: The aim of this study was to assess the evidence for the association between different biochemical markers at admission and mortality through a meta-analysis. DATA SOURCES: PubMed-, Embase-, Cochran Library and the Web of Knowledge were searched for cohort studies. STUDY SELECTION: Eligible studies were observational studies with a study population larger than 150 subjects, a mean age above 60 years and a study duration below 730 days. DATA EXTRACTION: Characteristics of studies and outcomes of all-cause mortality were extracted from the retrieved articles. Data were pooled across studies for the individual biomarker using random- or fixed-effect analysis. DATA SYNTHESIS: 15 eligible studies of 5 different markers on mortality were studied. The following markers were found to be of prognostic value on mortality in hip fracture patients: low haemoglobin (odds ratio, 2.78; 95% confidence interval, 2.17-3.55; P<0.00001, 3148 subjects included), low total lymphocyte count, TLC (odds ratio, 2.60; 95% confidence interval, 1.61-4.20; P<0.00001, 1689 subjects included), low albumin (odds ratio, 1.83; 95% confidence interval, 1.31-2.56; P=0.0004, 1680 subjects included), low albumin/low TLC (odds ratio, 3.00; 95% confidence interval, 1.81-4.99; P<0.0001, 704 subjects included), low albumin/high TLC (odds ratio, 3.39; 95% confidence interval, 1.83-6.29; P=0.0001, 704 subjects included), high creatinine (odds ratio, 1.58; 95% confidence interval, 1.25-1.99; P=0.0001, 3761 subjects included), and high PTH (odds ratio, 15.43; 95% confidence interval, 3.60-66.14; P=0.0002, 525 subjects included). CONCLUSION: Biochemical markers at admission are valid predictors of mortality in hip fracture patients.

Value of routine blood tests for prediction of mortality risk in hip fracture patients.

BACKGROUND: There is a 5- to 8-fold increased risk of mortality during the first 3 months after a hip fracture. Several risk factors are known. We studied the predictive value (for mortality) of routine blood tests taken on admission. METHODS: 792 hip fracture patients were included prospectively; blood tests were taken on admission. Follow-up data on mortality were obtained from the civil registration system. Patients were divided into 2 groups based on whether they had survived at least 90 days after the hip fracture. To estimate which laboratory tests could be used to predict outcome, we used receiver operation characteristic (ROC) curves. RESULTS: 3-month mortality could be predicted with 69% accuracy from the level of plasma creatinine in standard admission blood tests. The mortality in patients with elevated levels of creatinine was almost 3-fold that of the patients with normal creatinine. Mortality was also associated with age, low blood hemoglobin, high plasma potassium, and low plasma albumin levels. INTERPRETATION: Our findings could be of use in identifying patients who might benefit from increased attention perioperatively.