BACKGROUND: The risk factors, microbial etiology, differentiation, and clinical features of purulent and non-purulent cellulitis are not well defined in Taiwan. METHODS: We conducted a retrospective cohort study of hospitalized adults with cellulitis in Taiwan in 2013. The demographic characteristics, underlying diseases, clinical manifestations, laboratory and microbiological findings, treatments, and outcomes were compared for patients with purulent and non-purulent cellulitis. RESULTS: Of the 465 patients, 369 had non-purulent cellulitis and 96 had purulent cellulitis. The non-purulent group was significantly older (p = 0.001) and was more likely to have lower limb involvement (p < 0.001), tinea pedis (p = 0.003), stasis dermatitis (p = 0.025), a higher Charlson comorbidity score (p = 0.03), and recurrence at 6 months post-infection (p = 0.001) than the purulent group. The purulent group was more likely to have a wound (p < 0.001) and a longer hospital stay (p = 0.001) and duration of antimicrobial therapy (p = 0.003) than the non-purulent group. The etiological agent was identified in 35.5 % of the non-purulent cases, with β-hemolytic streptococci the most frequent cause (70.2 %). The etiological agent was identified in 83.3 % of the purulent cases, with Staphylococcus aureus the predominant pathogen (60 %): 50 % of these were methicillin-resistant S. aureus (MRSA). In multivariable analysis, purulent group (odds ratio (OR), 5.188; 95 % confidence interval (CI), 1.995-13.493; p = 0.001) was a positive predictor of MRSA. The prescribed antimicrobial agents were significantly different between the purulent and non-purulent groups, with penicillin the most frequently used antimicrobial agent in the non-purulent group (35.2 %), and oxacillin the most frequent in the purulent group (39.6 %). The appropriate antimicrobial agent was more frequently prescribed in the non-purulent group than in the purulent group (83.2 % vs. 53.8 %, p < 0.001). CONCLUSIONS: The epidemiology, clinical features, and microbiology of purulent and non-purulent cellulitis were significantly different in hospitalized Taiwanese adults. Purulence was a positive predictor of MRSA as the causal agent of cellulitis. These findings provide added support for the adoption of the IDSA guidelines for empirical antimicrobial therapy of cellulitis in Taiwan.