Clinical assessment of infection in nonhealing ulcers analyzed by latent class analysis.

The distinction between bacterial colonization and infection relies on clinical judgement. Determining sensitivity and specificity of this judgement are problematic as no gold standard exists. Six specialists in wound management independently assessed 120 nonhealing chronic wounds. Sixty-five (54.2%) patients had venous ulcer, 18 (15%) arterial ulcers, 15 (12.5%) ulcerative pyoderma gangraenosum, 12 (10%) neuropathic or pressure ulcers, six (5%) vasculitis ulcers, and four patients had ulcers caused by a primary or metastatic cancer disease. Unrestricted latent class analysis was used for determining sensitivity and specificity in the observer's assessment of hypergranulation, redness, and overall impression of infection. Interrater agreement among observers was determined by restricted latent class analysis. The observers used the diagnoses (redness, hypergranulation, and overall impression of infection with different frequencies (p<0.001, Cochrane's Q test). A two latent class model fitted data. Sensitivity for hypergranulation ranged from 3 to 82%, for redness from 34 to 91% and for overall impression of infection from 37 to 90%. None of the observers were interchangeable. These results indicate that clinical assessment of chronic wounds for the presence of infection are difficult tasks accompanied by great variability and low reliability.