CEM-Trauma Expected Mortality Model

Intended Use

CEM-Trauma is an investigator-developed expected-mortality model for registry-based trauma quality review, observed-to-expected benchmarking, model audit, and research. It is intended for post hoc review after AIS/ISS injury coding is available. It is not an American College of Surgeons, TQP, or TQIP product, benchmark, or endorsed calculator.

Regulatory Status

This tool has not been cleared or approved by the U.S. Food and Drug Administration (FDA). It is provided for quality-review, educational, and research purposes. It is not intended for use as the sole basis for clinical decision-making.

Model Derivation & Limitations

• Population: Derived from 3.8 million adult trauma patients in ACS TQP/TQIP Participant Use File data (2017–2022) and temporally validated on 1.6 million patients (2023–2024). Performance in non-TQIP populations, pediatric patients (<16 years), or international registries has not been assessed.
• Missing data: Approximately 34% of TQIP patients were excluded for missing vital signs or ISS. Excluded patients are likely more severely injured, meaning model performance may be overestimated relative to the complete trauma population.
• Prediction scope: The model predicts in-hospital mortality only. It does not predict long-term survival, functional outcome, disability, or quality of life.
• Population estimates: Predictions reflect average outcomes across large patient groups. Individual patient outcomes will vary. A predicted mortality of 10% means that, on average, 10 of 100 similar patients die—not that this specific patient has a 10% chance of death.
• Temporal drift: Model coefficients were locked in April 2026. As trauma care practices evolve, recalibration may be necessary. No automated updating mechanism is included.
• Transfer patients: Physiology at ED presentation may be altered by pre-hospital interventions. The model does not adjust for transfer status.
• EMR deployment: Real-time EMR use would require reliable injury coding, local validation, recalibration, monitoring, and institutional governance. The current implementation is best suited to registry/QI review rather than live triage.

Data Privacy & Security

• All computation is local. No patient data is transmitted to any server. All calculations run entirely in your browser. No cookies are used, and no value you enter ever leaves the page. Page visits are counted in aggregate by a cookieless tool that cannot identify you or follow you across sites.
• No patient identifiers are collected. This tool requires only de-identified clinical parameters (injury severity, vitals, age, mechanism).
• Browser storage: This tool does not store any entered data. Closing or refreshing the page clears all inputs. If installed as a mobile app, no data persists between sessions.
• Content Security Policy: This page enforces a strict CSP that blocks external scripts, connections, and embedding. No data can be exfiltrated by injected code.

Liability & Disclaimer

This tool is provided "as is" without warranty of any kind, express or implied, including but not limited to the warranties of merchantability, fitness for a particular purpose, and noninfringement. The authors shall not be liable for any claim, damages, or other liability arising from the use of this tool. The treating clinician retains full responsibility for all clinical decisions.

Citation

Manuscript currently under double-blind peer review; the full citation is temporarily withheld and will be restored upon publication.

Contact

For questions about the model, validation data, or institutional deployment: Contact