Unlocking Trauma Data’s Potential
Algorithms for Early Prediction
In Challenge Event 1, teams were provided real-world trauma center datasets containing thousands of cases with >500 variables. This data was collected from pre-hospital (medevac, ambulance) until 4 hours after hospital admission. Teams developed algorithms to predict the need for life-saving interventions during this time.
In Challenge Event 2, the data size will significantly increase in size. Additional data types will add to the training data’s complexity. To win a prize, teams’ performance must exceed the specified baseline algorithm as well as specificity, sensitivity and lead time thresholds
Life Saving Interventions (LSIs) are a medical care resource, activity, or procedure used to sustain life. LSIs have been grouped by type of injury and/or mechanism of treatment:
- Airway & respiration: Intubation, Cricothyroidotomy, Laryngoscopy, Mechanic Ventilation
- Vascular access & monitoring: Peripheral IV, Central Line Maintenance, Arterial Line Maintenance
- Bleeding control: Tourniquet, Pelvic Binder
- Vaso/cardioactive medications: Epinephrine, Dopamine, Phenylephrine, Vasopressin, Hydrocortisone
- Crystalloid products: Normal Saline (>500cc), Lactated Ringers (>500cc)
- Blood products: Packed Red Blood Cells, Plasma, Whole Blood Transfusion
- Cardiovascular procedures: CPR, Defibrillation/Cardioversion, Pericardiocentesis
- Chest decompression: Needle Decompression, Chest Tube Management
- Neurologic products & procedures: Craniotomy/Craniectomy, Hypertonic Saline, Antiepileptic Medications
- RSI sedation medications: Ketamine, Versed, Etomidate, Ativan, Propofol
- Limb salvage: Amputation, Fasciotomy
- Damage control procedures: Thoracotomy, Exploratory Laparotomy, IR Embolization
Performance measured for each case using two metrics:
Jaccard Index (JI): How accurate were the predictions?
Prediction Lead Time (PLT): How early were the correct predictions?
These innovative data teams developed cutting-edge algorithms to predict life-saving interventions from complex trauma data and ultimately improve medical decision-making.
- AI TEMPO | Alert for Intervention using Timeseries EMergency Physiological Observations | DARPA-funded
- ALICE | AI Life Saving Intervention Compute Engine | DARPA-funded
- AUSTERE | AI User Supporting Triage and Evacuation Recommendation
- CAMA | Center for Advanced Medical Analytics
- CNA | Center for Naval Analyses
- Coordinated Robotics
- CRITIC | Continuous Review and Intervention for Timely Care | DARPA-funded
- LENS | LSI Early Notification System | DARPA-funded
- MGB-Harvard
- MSAI
- Robotika
- TrueFit.AI
Leaderboard
Scoring criteria: Accuracy of LSI prediction (for any LSI and for more specific classifications, such as hemorrhage or airway interventions)
| Teams | Score | 1 | 2 |
|---|---|---|---|
| Coordinated Robotics* | 625 | 434 | 191 |
| MSAI* | 555 | 364 | 191 |
| CAMA* | 532 | 368 | 164 |
| LENS | 310 | 129 | 181 |
| AI TEMPO | 268 | 91 | 177 |
| CRITIC | 189 | 21 | 168 |
| TrueFit.AI* | 179 | 37 | 142 |
| Robotika* | 143 | 27 | 116 |
| ALICE | 109 | 18 | 91 |
| AUSTERE | 61 | 49 | 12 |
| MGB-Harvard* | 44 | 13 | 31 |
| CNA* | 35 | 8 | 27 |
* Self-funded
Overview
