“NEXUS Chest” Decision Instrument

Low-yield radiography in the setting of trauma is pervasive and costly, but, unfortunately guidance regarding appropriateness is poor.  The NEXUS group previously derived a chest imaging decision instrument, and this newly published article describes the validation study.

The good:  98.8% (CI 98.1-99.3%) sensitivity for any thoracic injury on imaging, and 99.7% (CI 98.2-100%) sensitivity for injuries of major clinical significance.

The really, really bad:  13.3% (CI 12.6-14.1%) specificity for thoracic injury or 12.0% (11.3-12.6%) specificity for major significance.

And, these numbers are probably subject to some limitations, considering about half the patients only received chest x-ray, rather than chest CT.  That said, the injuries missed by x-ray are not likely of major clinical significance – and the patients selected for x-ray alone in the run of standard practice were likely selected for a low pretest probability of serious injury, regardless.

The authors suggest their instrument, despite it’s terrible specificity, still represents a valuable rule-out option, theorizing that even the small reduction in imaging this rule represents is beneficial.  However, as we’ve covered before, one-way decision instruments are subject to cognitive bias and use as two-way rules, which may paradoxically increase imaging – although, in trauma, it’s hard to imagine a way to order more.  Careful adoption of this instrument will be required – perhaps only after clinical evaluation as a screening decision-support question in the CPOE, asking one last time if the patient possibly meets this very-low-risk criteria prior to ordering.

The exclusion from very-low-risk criteria, by the by:

  • Older than 60 years
  • Rapid deceleration mechanism (fall >20 ft, MVC >40mph)
  • Chest pain
  • Intoxication
  • Abnormal mental status
  • Distracting painful injury
  • Tenderness to chest wall palpation

“NEXUS Chest – Validation of a Decision Instrument for Selective Chest Imaging in Blunt Trauma”
http://www.ncbi.nlm.nih.gov/pubmed/23925583