Factors associated with positive findings of deep infection on computed tomography among patients with extremity...

Cellulitis shares several clinical features with fulminant deep soft tissue infections, creating a diagnostic dilemma for which clinicians are increasingly using cross-sectional diagnostic imaging to resolve. However, the role of imaging in apparent cellulitis is poorly defined. In particular, the clinical utility of CT in extremity cellulitis is underexplored. The purpose of this study was to determine patient and clinical factors that increase likelihood for finding of deep infection on CT among patients with extremity cellulitis.

A retrospective observational study was performed of patients with cellulitis of the extremities who were examined with computed tomography from January 1, 2012 to December 31, 2021. Demographics, medical history, and clinical and laboratory characteristics were collected. The presence of positive findings on CT report was recorded. Repeated measures logistic regression was used to evaluate significant predictors of positive CT.

There were 496 eligible patient encounters, and CT was positive for deep infection in 39 cases (7.9 %). Median patient age was 56 years of age, 58.1 % were male, and most patients had a history of diabetes mellitus, history of tobacco use, and/or obesity. Significant predictors for positive CT on multivariable analysis included gas on pre-CT radiographs, febrile leukopenic state, injection drug use, purulence on exam, and white blood cell count extremes. Elevated CRP was also a significant positive predictor on bivariate analysis.

Finding a deep infection on CT in a patient with extremity cellulitis is uncommon and is predominantly associated with a high-risk clinical picture, and/or with high index of suspicion based on pre-CT plain films. In the absence of these patient factors or compelling findings on radiographs, CT does not typically provide clinically actionable information for extremity cellulitis and should not be used as part of standard evaluation.