Toxic epidermal necrolysis (TEN) is a rare drug-related life-threatening acute condition. Sepsis is the main cause of mortality. Skin colonization on top of impaired barrier function promotes bloodstream infections (BSI). We conducted this study to describe the epidemiology, identify early predictors of BSI, and assess the predictive value for bacteremia of routine skin surface cultures. We retrospectively analyzed the charts of all patients with Stevens-Johnson syndrome (SJS) and TEN hospitalized over an 11-year period. Blood cultures and skin isolates were recovered from the microbiology laboratory database. Early predictors of BSI were identified using a Cox model. Sensitivity, specificity, and negative and positive predictive values of skin cultures for the etiology of BSI were assessed. The study included 179 patients, classified as having SJS (n = 54; 30.2%), SJS/TEN overlap (n = 59; 33.0%), and TEN (n = 66; 36.9%). Forty-eight episodes of BSI occurred, yielding a rate of 15.5/1000 patient days. In hospital mortality was 13.4% (24/179). Overall, 70 pathogens were recovered, mainly Staphylococcus aureus (n = 23/70; 32.8%), Pseudomonas aeruginosa (n = 15/70; 21.4%), and Enterobacteriaceae organisms (n = 17/70; 24.3%). Variables associated with BSI in multivariate analysis included age >40 years (hazard ratio [HR], 2.5; 95% confidence interval [CI], 1.35-4.63), white blood cell count >10,000/mm3 (HR, 1.9; 95% CI, 0.96-3.61), and percentage of detached body surface area >or=30% (HR, 2.5; 95% CI, 1.13-5.47). Skin cultures had an excellent negative predictive value for bacteremia due to S aureus (especially methicillin-resistant strains) and P aeruginosa, but not for those due to Enterobacteriaceae organisms. In contrast, the positive predictive value was low for all pathogens studied.To our knowledge, this is the largest study describing the epidemiology and risk factors of BSI in patients with SJS/TEN. The body surface area involved is the main predictor of BSI. Excellent negative predictive values of skin cultures for S aureus and P aeruginosa bacteremia should help clinicians consider targeted empirical antibiotic choices when appropriate.