Comparing measures of injury severity for use with large databases

J Trauma. 2002 Aug;53(2):326-32. doi: 10.1097/00005373-200208000-00023.


Background: After recent debate about the best measure of anatomic injury severity, this study aimed to compare four measures based on Abbreviated Injury Scale scores derived using ICDMAP-90-the Modified Anatomic Profile (ICD/mAP), Anatomic Profile Score (ICD/APS), Injury Severity Score (ICD/ISS), and New Injury Severity Score (ICD/NISS)-with the International Classification of Diseases-based Injury Severity Score (ICISS).

Methods: Data were selected from New Zealand public hospital discharges from 1989 to 1998. There were 349,409 patients in the dataset, of whom 3,871 had died. Models were compared in terms of their discrimination and calibration using logistic regression. Age was included as a covariate.

Results: The ICISS and ICD/mAP were the best performing measures. Adding age significantly improved the discrimination and calibration of almost all the models.

Conclusion: The ICISS is a viable alternative to ICDMAP-based measures for coding anatomic injury severity on large datasets.

Publication types

  • Comparative Study
  • Research Support, Non-U.S. Gov't

MeSH terms

  • Calibration
  • Databases as Topic*
  • Humans
  • Logistic Models
  • Medical Records Systems, Computerized*
  • New Zealand / epidemiology
  • Sensitivity and Specificity
  • Trauma Severity Indices*
  • Wounds and Injuries / diagnosis*
  • Wounds and Injuries / mortality