Predictive model for estimating risk of crush syndrome: A data mining approach

Noriaki Aoki, Janez Demsar, Blaz Zupan, Martin Mozina, Ernesto A. Pretto, Jun Oda, Hiroshi Tanaka, Katsuhiko Sugimoto, Toshiharu Yoshioka, Tsuguya Fukui

Research output: Contribution to journalArticlepeer-review

16 Scopus citations


BACKGROUND: There is no standard triage method for earthquake victims with crush injuries because of a scarcity of epidemiologic and quantitative data. We conducted a retrospective cohort study to develop predictive models based on clinical data for crush injury in the Kobe earthquake. METHODS: The medical records of 372 patients with crush injuries from the Kobe earthquake were retrospectively analyzed. Twenty-one risk factors were assessed with logistic regression analysis for three outcomes relating to crush syndrome. Two types of predictive triage models-initial evaluation in the field and secondary assessment at the hospital-were developed using logistic regression analysis. Classification accuracy, Brier score and area under the receiver operating characteristic curve (AUC) were used to evaluate the model. RESULTS: The initial triage model, which includes pulse rate, delayed rescue, and abnormal urine color, has an AUC of 0.73. The secondary model, which includes WBC, tachycardia, abnormal urine color, and hyperkalemia, shows an AUC of 0.76. CONCLUSIONS: These triage models may be especially useful to nondisaster experts for distinguishing earthquake victims at high risk of severe crush syndrome from those at lower risk. Application of the model may allow relief workers to better utilize limited medical and transportation resources in the aftermath of a disaster.

Original languageEnglish (US)
Pages (from-to)940-945
Number of pages6
JournalJournal of Trauma - Injury, Infection and Critical Care
Issue number4
StatePublished - Apr 1 2007


  • Crush injury
  • Crush syndrome
  • Data mining
  • Disaster
  • Earthquake
  • Prognostic model
  • Risk factors

ASJC Scopus subject areas

  • Surgery


Dive into the research topics of 'Predictive model for estimating risk of crush syndrome: A data mining approach'. Together they form a unique fingerprint.

Cite this