The impact of missing trauma data on predicting massive transfusion.

Area of Science:

• Clinical research methodology
• Biostatistics
• Trauma medicine

Background:

• Missing data is a common challenge in clinical research, particularly impacting trauma studies.
• Assessing the impact of missing data on clinical risk prediction models is crucial for reliable outcomes.

Purpose of the Study:

• To describe a sensitivity analysis method for evaluating the impact of missing data on clinical risk prediction algorithms.
• To assess the performance of three blood transfusion prediction models using a trauma dataset with inherent missing data.

Main Methods:

• Utilized data from the PRospective Observational Multicenter Major Trauma Transfusion (PROMMTT) study.
• Evaluated three massive transfusion (MT) prediction models using complete case analysis and multiple imputation.
• Conducted a sensitivity analysis to determine the upper and lower bounds for correct classification percentages.

Main Results:

• The PROMMTT study included 1,245 subjects, with missing data ranging from 2.2% to 45%.
• Complete case analysis and multiple imputation yielded similar correct classification percentages for the MT prediction models.
• Sensitivity analysis revealed upper-lower bound ranges of 4%, 10%, and 12% for correct classification across models.

Conclusions:

• Evaluating clinical prediction models with missing data can be misleading, especially with numerous variables and moderate missingness.
• The proposed sensitivity analysis effectively illustrates the influence of missing data on risk prediction algorithms.
• Reporting upper-lower bounds for correct classification may provide more informative insights than multiple imputation alone.