A latent class model for defining severe hemorrhage: experience from the PROMMTT study

Area of Science:

• Trauma critical care
• Medical informatics
• Predictive modeling

Background:

• Existing models for severe hemorrhage (SH) often define it as 10+ red blood cell (RBC) units transfused within 24 hours (massive transfusion, MT).
• This definition may inaccurately exclude SH patients who die before receiving 10 RBC units, questioning model validity.
• A latent class model offers a more accurate method for identifying SH patients.

Purpose of the Study:

• To develop and validate a latent class model for improved identification of severe hemorrhage (SH) in trauma patients.
• To compare the accuracy of the latent class model against the traditional massive transfusion (MT) definition.

Main Methods:

• SH classification was modeled as a latent variable using emergency department admission data, 24-hour blood product ratios, and survival status.
• Patients with a posterior probability of SH ≥ 0.5 were classified using the latent class model.
• The latent class model's SH classification was compared with the traditional MT definition using data from the PROMMTT study.

Main Results:

• The latent class model classified 25.3% of 913 patients as SH, with 83.8% overall agreement with the MT definition.
• Crucially, 84% of patients who died before receiving 10 RBC units were correctly identified as SH by the new model.
• Among those not classified as SH, a significant proportion had head injuries.

Conclusions:

• The latent class model improves SH classification by including patients who die before reaching the traditional 10-unit MT threshold.
• This refined SH classification is advantageous for developing more accurate prediction algorithms.
• Further research is recommended to refine SH classification and potentially replace the traditional MT definition.