Background: Severity scoring systems are useful for assessing patient risk and predicting prognosis.
Methods: We developed a scoring system using data from a phase III study comparing antibiotics in hospitalized patients with complicated skin and soft tissue infections (study A), and used logistic regression analysis to identify factors contributing to treatment failure. We tested this system using data from a similar study (study B).
Results: In study A (n = 682), cure rates were lower in patients with at least 1 comorbid condition (P <0.05) and in the highest risk class (P = 0.05); elevated blood urea nitrogen, hyponatremia, anemia, lesion size, and surgical wound infection were independent predictors of failure (P <0.05). In study B (n = 166), findings were similar and significant for risk class (P <0.05). In the combined analysis (n = 848), cure rates were higher for linezolid than for the comparator in all patients (85% versus 77%; P <0.01) and in subanalyses by comorbid conditions, median score, and risk class (P <0.05).
Conclusions: We developed and validated a scoring system in which baseline variables predicted outcome. Patients with higher severity scores generally had poorer outcomes regardless of treatment group. Our finding that linezolid was an independent predictor of cure merits further evaluation.