Triage in major incidents: development and external validation of novel machine learning-derived primary and secondary triage tools.

BACKGROUND: Major incidents (MIs) are an important cause of death and disability. Triage tools are crucial to identifying priority 1 (P1) patients-those needing time-critical, life-saving interventions. Existing expert opinion-derived tools have limited evidence supporting their use. This study employs machine learning (ML) to develop and validate models for novel primary and secondary triage tools. METHODS: Adults (16+ years) from the UK Trauma Audit and Research Network (TARN) registry (January 2008-December 2017) served as surrogates for MI victims, with P1 patients identified using predefined criteria. The TARN database was split chronologically into model training and testing (70:30) datasets. Input variables included physiological parameters, age, mechanism and anatomical location of injury. Random forest, extreme gradient boosted tree, logistic regression and decision tree models were trained to predict P1 status, and compared with existing tools (Battlefield Casualty Drills (BCD) Triage Sieve, CareFlight, Modified Physiological Triage Tool, MPTT-24, MSTART, National Ambulance Resilience Unit Triage Sieve and RAMP). Primary and secondary candidate models were selected; the latter was externally validated on patients from the UK military's Joint Theatre Trauma Registry (JTTR). RESULTS: Models were internally tested in 57 979 TARN patients. The best existing tool was the BCD Triage Sieve (sensitivity 68.2%, area under the receiver operating curve (AUC) 0.688). Inability to breathe spontaneously, presence of chest injury and mental status were most predictive of P1 status. A decision tree model including these three variables exhibited the best test characteristics (sensitivity 73.0%, AUC 0.782), forming the candidate primary tool. The proposed secondary tool (sensitivity 77.9%, AUC 0.817), applicable via a portable device, includes a fourth variable (injury mechanism). This performed favourably on external validation (sensitivity of 97.6%, AUC 0.778) in 5956 JTTR patients. CONCLUSION: Novel triage tools developed using ML outperform existing tools in a nationally representative trauma population. The proposed primary tool requires external validation prior to consideration for practical use. The secondary tool demonstrates good external validity and may be used to support decision-making by healthcare workers responding to MIs.

Artificial intelligence to enhance clinical value across the spectrum of cardiovascular healthcare.

Artificial intelligence (AI) is increasingly being utilized in healthcare. This article provides clinicians and researchers with a step-wise foundation for high-value AI that can be applied to a variety of different data modalities. The aim is to improve the transparency and application of AI methods, with the potential to benefit patients in routine cardiovascular care. Following a clear research hypothesis, an AI-based workflow begins with data selection and pre-processing prior to analysis, with the type of data (structured, semi-structured, or unstructured) determining what type of pre-processing steps and machine-learning algorithms are required. Algorithmic and data validation should be performed to ensure the robustness of the chosen methodology, followed by an objective evaluation of performance. Seven case studies are provided to highlight the wide variety of data modalities and clinical questions that can benefit from modern AI techniques, with a focus on applying them to cardiovascular disease management. Despite the growing use of AI, further education for healthcare workers, researchers, and the public are needed to aid understanding of how AI works and to close the existing gap in knowledge. In addition, issues regarding data access, sharing, and security must be addressed to ensure full engagement by patients and the public. The application of AI within healthcare provides an opportunity for clinicians to deliver a more personalized approach to medical care by accounting for confounders, interactions, and the rising prevalence of multi-morbidity.

Metabotypes of flavan-3-ol colonic metabolites after cranberry intake: elucidation and statistical approaches.

PURPOSE: Extensive inter-individual variability exists in the production of flavan-3-ol metabolites. Preliminary metabolic phenotypes (metabotypes) have been defined, but there is no consensus on the existence of metabotypes associated with the catabolism of catechins and proanthocyanidins. This study aims at elucidating the presence of different metabotypes in the urinary excretion of main flavan-3-ol colonic metabolites after consumption of cranberry products and at assessing the impact of the statistical technique used for metabotyping. METHODS: Data on urinary concentrations of phenyl-γ-valerolactones and 3-(hydroxyphenyl)propanoic acid derivatives from two human interventions has been used. Different multivariate statistics, principal component analysis (PCA), cluster analysis, and partial least square-discriminant analysis (PLS-DA), have been considered. RESULTS: Data pre-treatment plays a major role on resulting PCA models. Cluster analysis based on k-means and a final consensus algorithm lead to quantitative-based models, while the expectation-maximization algorithm and clustering according to principal component scores yield metabotypes characterized by quali-quantitative differences in the excretion of colonic metabolites. PLS-DA, together with univariate analyses, has served to validate the urinary metabotypes in the production of flavan-3-ol metabolites and to confirm the robustness of the methodological approach. CONCLUSIONS: This work proposes a methodological workflow for metabotype definition and highlights the importance of data pre-treatment and clustering methods on the final outcomes for a given dataset. It represents an additional step toward the understanding of the inter-individual variability in flavan-3-ol metabolism. TRIAL REGISTRATION: The acute study was registered at clinicaltrials.gov as NCT02517775, August 7, 2015; the chronic study was registered at clinicaltrials.gov as NCT02764749, May 6, 2016.

Paediatric major incident triage: UK military tool offers best performance in predicting the need for time-critical major surgical and resuscitative intervention.

BACKGROUND: Children are frequently injured during major incidents (MI), including terrorist attacks, conflict and natural disasters. Triage facilitates healthcare resource allocation in order to maximise overall survival. A critical function of MI triage tools is to identify patients needing time-critical major resuscitative and surgical intervention (Priority 1 (P1) status). This study compares the performance of 11 MI triage tools in predicting P1 status in children from the UK Trauma Audit and Research Network (TARN) registry. METHODS: Patients aged <16 years within TARN (January 2008-December 2017) were included. 11 triage tools were applied to patients' first recorded pre-hospital physiology. Patients were retrospectively assigned triage categories (P1, P2, P3, Expectant or Dead) using predefined intervention-based criteria. Tools' performance in <16s were evaluated within four-yearly age subgroups, comparing tool-predicted and intervention-based priority status. FINDINGS: Amongst 4962 patients, mortality was 1.1% (n = 53); median Injury Severity Score (ISS) was 9 (IQR 9-16). Blunt injuries predominated (94.4%). 1343 (27.1%) met intervention-based criteria for P1, exhibiting greater intensive care requirement (60.2% vs. 8.5%, p < 0.01) and ISS (median 17 vs 9, p < 0.01) compared with P2 patients. The Battlefield Casualty Drills (BCD) Triage Sieve had greatest sensitivity (75.7%) in predicting P1 status in children <16 years, demonstrating a 38.4-49.8% improvement across all subgroups of children <12 years compared with the UK's current Paediatric Triage Tape (PTT). JumpSTART demonstrated low sensitivity in predicting P1 status in 4 to 8 year olds (35.5%) and 0 to 4 year olds (28.5%), and was outperformed by its adult counterpart START (60.6% and 59.6%). INTERPRETATION: The BCD Triage Sieve had greatest sensitivity in predicting P1 status in this paediatric trauma registry population: we recommend it replaces the PTT in UK practice. Users of JumpSTART may consider alternative tools. We recommend Lerner's triage category definitions when conducting MI evaluations.

Comparative analysis of major incident triage tools in children: a UK population-based analysis.

INTRODUCTION: Triage is a key principle in the effective management of major incidents. There is currently a paucity of evidence to guide the triage of children. The aim of this study was to perform a comparative analysis of nine adult and paediatric triage tools, including the novel 'Sheffield Paediatric Triage Tool' (SPTT), assessing their ability in identifying patients needing life-saving interventions (LSIs). METHODS: A 10-year (2008-2017) retrospective database review of the Trauma Audit Research Network (TARN) Database for paediatric patients (<16 years) was performed. Primary outcome was identification of patients receiving one or more LSIs from a previously defined list. Secondary outcomes included mortality and prediction of Injury Severity Score (ISS) >15. Primary analysis was conducted on patients with complete prehospital physiological data with planned secondary analyses using first recorded data. Performance characteristics were evaluated using sensitivity, specificity, undertriage and overtriage. RESULTS: 15 133 patients met TARN inclusion criteria. 4962 (32.8%) had complete prehospital physiological data and 8255 (54.5%) had complete first recorded physiological data. The majority of patients were male (69.5%), with a median age of 11.9 years. The overwhelming majority of patients (95.4%) sustained blunt trauma, yielding a median ISS of 9 and overall, 875 patients (17.6%) received at least one LSI. The SPTT demonstrated the greatest sensitivity of all triage tools at identifying need for LSI (92.2%) but was associated with the highest rate of overtriage (75.0%). Both the Paediatric Triage Tape (sensitivity 34.1%) and JumpSTART (sensitivity 45.0%) performed less well at identifying LSI. By contrast, the adult Modified Physiological Triage Tool-24 (MPTT-24) triage tool had the second highest sensitivity (80.8%) with tolerable rates of overtriage (70.2%). CONCLUSION: The SPTT and MPTT-24 outperform existing paediatric triage tools at identifying those patients requiring LSIs. This may necessitate a change in recommended practice. Further work is needed to determine the optimum method of paediatric major incident triage, but consideration should be given to simplifying major incident triage by the use of one generic tool (the MPTT-24) for adults and children.

Redefining ß-blocker response in heart failure patients with sinus rhythm and atrial fibrillation: a machine learning cluster analysis.

BACKGROUND: Mortality remains unacceptably high in patients with heart failure and reduced left ventricular ejection fraction (LVEF) despite advances in therapeutics. We hypothesised that a novel artificial intelligence approach could better assess multiple and higher-dimension interactions of comorbidities, and define clusters of ß-blocker efficacy in patients with sinus rhythm and atrial fibrillation. METHODS: Neural network-based variational autoencoders and hierarchical clustering were applied to pooled individual patient data from nine double-blind, randomised, placebo-controlled trials of ß blockers. All-cause mortality during median 1·3 years of follow-up was assessed by intention to treat, stratified by electrocardiographic heart rhythm. The number of clusters and dimensions was determined objectively, with results validated using a leave-one-trial-out approach. This study was prospectively registered with ClinicalTrials.gov (NCT00832442) and the PROSPERO database of systematic reviews (CRD42014010012). FINDINGS: 15 659 patients with heart failure and LVEF of less than 50% were included, with median age 65 years (IQR 56-72) and LVEF 27% (IQR 21-33). 3708 (24%) patients were women. In sinus rhythm (n=12 822), most clusters demonstrated a consistent overall mortality benefit from ß blockers, with odds ratios (ORs) ranging from 0·54 to 0·74. One cluster in sinus rhythm of older patients with less severe symptoms showed no significant efficacy (OR 0·86, 95% CI 0·67-1·10; p=0·22). In atrial fibrillation (n=2837), four of five clusters were consistent with the overall neutral effect of ß blockers versus placebo (OR 0·92, 0·77-1·10; p=0·37). One cluster of younger atrial fibrillation patients at lower mortality risk but similar LVEF to average had a statistically significant reduction in mortality with ß blockers (OR 0·57, 0·35-0·93; p=0·023). The robustness and consistency of clustering was confirmed for all models (p<0·0001 vs random), and cluster membership was externally validated across the nine independent trials. INTERPRETATION: An artificial intelligence-based clustering approach was able to distinguish prognostic response from ß blockers in patients with heart failure and reduced LVEF. This included patients in sinus rhythm with suboptimal efficacy, as well as a cluster of patients with atrial fibrillation where ß blockers did reduce mortality. FUNDING: Medical Research Council, UK, and EU/EFPIA Innovative Medicines Initiative BigData@Heart.

The BCD Triage Sieve outperforms all existing major incident triage tools: Comparative analysis using the UK national trauma registry population.

BACKGROUND: Natural disasters, conflict, and terrorism are major global causes of death and disability. Central to the healthcare response is triage, vital to ensure the right care is provided to the right patient at the right time. The ideal triage tool has high sensitivity for the highest priority (P1) patients with acceptably low over-triage. This study compared the performance of major incident triage tools in predicting P1 casualty status in adults in the prospective UK Trauma Audit and Research Network (TARN) registry. METHODS: TARN patients aged 16+ years (January 2008-December 2017) were included. Ten existing triage tools were applied using patients' first recorded pre-hospital physiology. Patients were subsequently assigned triage categories (P1, P2, P3, Expectant or Dead) based on pre-defined, intervention-based criteria. Tool performance was assessed by comparing tool-predicted and intervention-based priority status. FINDINGS: 195,709 patients were included; mortality was 7·0% (n=13,601); median Injury Severity Score (ISS) was 9 (IQR 9-17); 97·1% sustained blunt injuries. 22,144 (11·3%) patients fulfilled intervention-based criteria for P1 status, exhibiting higher mortality (12·8% vs. 5·0%, p<0.001), increased intensive care requirement (52·4% vs 5·0%, p<0.001), and more severe injuries (median ISS 21 vs 9, p<0.001) compared with P2 patients.In 16-64 year olds, the highest performing tool was the Battlefield Casualty Drills (BCD) Triage Sieve (Prediction of P1 status: 70·4% sensitivity, over-triage 70·9%, area under the receiver operating curve (AUC) 0·068 [95%CI 0·676-0·684]). The UK National Ambulance Resilience Unit (NARU) Triage Sieve had sensitivity of 44·9%; over-triage 56·4%; AUC 0·666 (95%CI 0·662-0·670). All tools performed poorly amongst the elderly (65+ years). INTERPRETATION: The BCD Triage Sieve performed best in this nationally representative population; we recommend it supersede the NARU Triage Sieve as the UK primary major incident triage tool. Validated triage category definitions are recommended for appraising future major incidents. FUNDING: This study is funded by the National Institute for Health Research (NIHR) Surgical Reconstruction and Microbiology Research Centre. GVG also acknowledges support from the MRC Heath Data Research UK (HDRUK/CFC/01). The views expressed are those of the authors and not necessarily those of the NIHR, the Department of Health and Social Care, or the Ministry of Defence.

Machine learning can predict disease manifestations and outcomes in lymphangioleiomyomatosis.

BACKGROUND: Lymphangioleiomyomatosis (LAM) is a rare multisystem disease with variable clinical manifestations and differing rates of progression that make management decisions and giving prognostic advice difficult. We used machine learning to identify clusters of associated features which could be used to stratify patients and predict outcomes in individuals. PATIENTS AND METHODS: Using unsupervised machine learning we generated patient clusters using data from 173 women with LAM from the UK and 186 replication subjects from the US National Heart, Lung, and Blood Institute (NHLBI) LAM registry. Prospective outcomes were associated with cluster results. RESULTS: Two- and three-cluster models were developed. A three-cluster model separated a large group of subjects presenting with dyspnoea or pneumothorax from a second cluster with a high prevalence of angiomyolipoma symptoms (p=0.0001) and tuberous sclerosis complex (TSC) (p=0.041). Patients in the third cluster were older, never presented with dyspnoea or pneumothorax (p=0.0001) and had better lung function. Similar clusters were reproduced in the NHLBI cohort. Assigning patients to clusters predicted prospective outcomes: in a two-cluster model the future risk of pneumothorax was 3.3 (95% CI 1.7-5.6)-fold greater in cluster 1 than cluster 2 (p=0.0002). Using the three-cluster model, the need for intervention for angiomyolipoma was lower in clusters 2 and 3 than cluster 1 (p<0.00001). In the NHLBI cohort, the incidence of death or lung transplant was much lower in clusters 2 and 3 (p=0.0045). CONCLUSIONS: Machine learning has identified clinically relevant clusters associated with complications and outcome. Assigning individuals to clusters could improve decision making and prognostic information for patients.

Violence-related knife injuries in a UK city; epidemiology and impact on secondary care resources.

BACKGROUND: The incidence of knife-related injuries is rising across the UK. This study aimed to determine the spectrum of knife-related injuries in a major UK city, with regards to patient and injury characteristics. A secondary aim was to quantify their impact on secondary care resources. METHODS: Observational study of patients aged 16+ years admitted to a major trauma centre following knife-related injuries resulting from interpersonal violence (May 2015 to April 2018). Patients were identified using Emergency Department and discharge coding, blood bank and UK national Trauma Audit and Research prospective registries. Patient and injury characteristics, outcome and resource utilisation were collected from ambulance and hospital records. FINDINGS: 532 patients were identified; 93% male, median age 26 years (IQR 20-35). Median injury severity score was 9 (IQR 3-13). 346 (65%) underwent surgery; 133 (25%) required intensive care; 95 (17·9%) received blood transfusion. Median length of stay was 3·3 days (IQR 1·7-6·0). In-hospital mortality was 10/532 (1·9%). 98 patients (18·5%) had previous attendance with violence-related injuries. 24/37 females (64·9%) were injured in a domestic setting. Intoxication with alcohol (19·2%) and illicit drugs (17·6%) was common. Causative weapon was household knife in 9%, knife (other/unspecified) in 38·0%, machete in 13·9%, small folding blade (2·8%) and, unrecorded in 36·3%. INTERPRETATION: Knife injuries constitute 12·9% of trauma team workload. Violence recidivism and intoxication are common, and females are predominantly injured in a domestic setting, presenting opportunities for targeted violence reduction interventions. 13·9% of injuries involved machetes, with implications for law enforcement strategies.

Genetic algorithm-based classifiers fusion for multisensor activity recognition of elderly people.

Activity recognition of an elderly person can be used to provide information and intelligent services to health care professionals, carers, elderly people, and their families so that the elderly people can remain at homes independently. This study investigates the use and contribution of wrist-worn multisensors for activity recognition. We found that accelerometers are the most important sensors and heart rate data can be used to boost classification of activities with diverse heart rates. We propose a genetic algorithm-based fusion weight selection (GAFW) approach which utilizes GA to find fusion weights. For all possible classifier combinations and fusion methods, the study shows that 98% of times GAFW can achieve equal or higher accuracy than the best classifier within the group.