Institutional members access full text with Ovid®

Share this article on:

TMPM–ICD9: A Trauma Mortality Prediction Model Based on ICD-9-CM Codes

Glance, Laurent G. MD*; Osler, Turner M. MD†; Mukamel, Dana B. PhD‡; Meredith, Wayne MD§; Wagner, Jacob MD, PhD¶; Dick, Andrew W. PhD∥

doi: 10.1097/SLA.0b013e3181a38f28
Original Articles

Objective: To develop and validate a new ICD-9 injury model that uses regression modeling, as opposed to a simple ratio measurement, to estimate empiric injury severities for each of the injuries in the ICD-9-CM lexicon.

Background: The American College of Surgeons now requires International Classification of diseases ninth Edition (ICD-9-CM) codes for injury coding in the National Trauma Databank. International Classification of diseases ninth Edition Injury Severity Score (ICISS) is the best-known risk-adjustment model when injuries are recorded using ICD-9-CM coding, and would likely be used to risk-adjust outcome measures for hospital trauma report cards. ICISS, however, has been criticized for its poor calibration.

Methods: We developed and validated a new ICD-9 injury model using data on 749,374 patients admitted to 359 hospitals in the National Trauma Databank (version 7.0). Empiric measures of injury severity for each of the trauma ICD-9-CM codes were estimated using a regression-based approach, and then used as the basis for a new Trauma Mortality Prediction Model (TMPM-ICD9). ICISS and the Single-Worst Injury (SWI) model were also re-estimated. The performance of each of these models was compared using the area under the receiver operating characteristic (ROC), the Hosmer-Lemeshow statistic, and the Akaike information criterion statistic.

Results: TMPM-ICD9 exhibits significantly better discrimination (ROCTMPM = 0.880 [0.876–0.883]; ROCICISS = 0.850 [0.846–0.855]; ROCSWI = 0.862 [0.858–0.867]) and calibration (HLTMPM = 29.3 [12.1–44.1]; HLICISS = 231 [176–279]; HLSWI = 462 [380–548]) compared with both ICISS and the Single Worst Injury model. All models were improved with the addition of age, gender, and mechanism of injury, but TMPM-ICD9 continued to demonstrate superior model performance.

Conclusions: Because TMPM-ICD9 uniformly out-performs ICISS and the SWI model, it should be used in preference to ICISS for risk-adjusting trauma outcomes when injuries are recorded using ICD9-CM codes.

We developed and validated a new ICD-9 Trauma Mortality Probability Model that uses regression modeling, as opposed to a simple ratio measurement, to estimate empiric injury severities for each of the injuries in the ICD-9-CM lexicon. Because Trauma Mortality Probability Model-ICD9 uniformly out-performs International Classification of diseases ninth Edition Injury Severity Score and the Single-Worst Injury model, it should be used in preference to International Classification of diseases ninth Edition Injury Severity Score for risk-adjusting trauma outcomes when injuries are recorded using ICD-9-CM codes.

From the *Department of Anesthesiology, University of Rochester School of Medicine, Rochester, New York; †Department of Surgery, University of Vermont Medical College; ‡Department of Medicine, Center for Health Policy Research, University of California, Irvine, California; §Department of Surgery, Wake Forest University School of Medicine; ¶Department of Surgery, University of Vermont College of Medicine; and ∥RAND.

Supported by the Agency for Healthcare and Quality Research by grant (RO1 HS 16737).

The views presented in this manuscript are those of the authors and may not reflect those of Agency for Healthcare and Quality Research or of the American College of Surgeons Committee on Trauma.

Reprints: Laurent G. Glance, MD, Department of Anesthesiology, University of Rochester Medical Center, 601 Elmwood Avenue, Box 604, Rochester, NY 14642. E-mail: Laurent_Glance@urmc.rochester.edu.

© 2009 Lippincott Williams & Wilkins, Inc.