A trauma mortality prediction model based on the ICD-10-CM lexicon: TMPM-ICD10

Area of Science:

• Trauma research
• Medical informatics
• Public health

Background:

• Outcome prediction models are crucial for risk adjustment in trauma research.
• The transition to ICD-10-CM codes has outdated previous risk adjustment models.
• No established risk adjustment model based on ICD-10-CM codes currently exists.

Purpose of the Study:

• To develop and validate a novel outcome prediction model using ICD-10-CM codes.
• To compare the performance of the new model against the existing Injury Severity Score (ISS).

Main Methods:

• Utilized data from the National Trauma Data Bank (2016) including 773,388 patients.
• Constructed a trauma outcome prediction model (TMPM-ICD-10) solely from ICD-10-CM acute injury codes.
• Compared TMPM-ICD-10 performance against the ISS using discrimination and calibration metrics.

Main Results:

• TMPM-ICD-10 demonstrated superior discrimination (ROC 0.861) compared to ISS (ROC 0.830).
• The new model showed better calibration (HL 49.01) than ISS (HL 788.79).
• TMPM-ICD-10 achieved a lower Akaike Information Criteria (AIC 30579.49) than ISS (AIC 31802.18).

Conclusions:

• TMPM-ICD-10 offers enhanced discrimination and calibration over the ISS.
• The model can be computed without Abbreviated Injury Scale coding.
• TMPM-ICD-10 should supersede the ISS for injury severity assessment in ICD-10-CM coded data.