Predicting mortality from burns: the need for age-group specific models.

Area of Science:

• Trauma surgery
• Burn care research
• Biostatistics in medicine

Background:

• Traditional burn mortality prediction models often use aggregated age data.
• Existing models may not accurately reflect age-specific variations in burn injury outcomes.
• Hypothesis: Age significantly influences burn mortality, necessitating age-specific predictive models.

Purpose of the Study:

• To develop and compare age-specific burn mortality models (children, adults, seniors) against an all-ages model.
• To assess the predictive accuracy of these models using data from the National Burn Repository.
• To identify how age, burn size (TBSA), and inhalation injury impact mortality across different age groups.

Main Methods:

• Utilized data from the American Burn Association National Burn Repository (2000-2009).
• Employed mixed-effect logistic regression to build mortality models for all ages and specific age groups (<18, 18-60, >60).
• Evaluated model performance using the area under the receiver operating curve (AUC).

Main Results:

• Overall mortality was 4%, with significant variation by age (seniors 17%, children <1%).
• Age, TBSA burn, and inhalation injury were significant predictors across all models.
• Age-specific models showed distinct differences, particularly for children and seniors, with seniors exhibiting higher mortality risk per age and burn size increase.

Conclusions:

• A 'one-size-fits-all' approach to burn mortality prediction is inadequate.
• Age-specific models, especially for pediatric and geriatric populations, offer improved accuracy.
• Recommendations for age-tailored burn outcome prediction models are warranted.