OBJECTIVE To define risk factors for infections from dog- bite wounds and to model the probability of wound infection in patients presenting without infection who are treated as outpatients. METHODS A prospective survey of 769 consecutive dog-bite victims presenting over a two-year period to a community hospital emergency department (ED) with an emergency medicine residency program. A standardized wound-cleaning protocol was used, which included debridement and wound closure when indicated. Wounds were examined for infection at follow-up. Variables analyzed included demographic data (patient age, gender, race); wound information (wound age, type, number, location, depth); and treatment (prior to hospital, ED debridement, suturing, tetanus or rabies shots, antibiotics). RESULTS There were 734 patients with complete records. These patients had a mean age of 13.4 +/- 13.2 years (range, 4 months to 71 years). Infection was evident in 2.5% of the wounds upon presentation. There were 704 patients (765 wounds) managed as outpatients and without wound infection upon presentation. Wounds were distributed as follows: 26.7% head/neck, 20.4% hand, 15.7% arm, 10.1% trunk, 9.5% thigh, 15.9% leg, and 1.7% foot. There were 32.9% puncture, 39.9% full-thickness, and 60.1% partial-thickness wounds. Wound infections were diagnosed in 2.1% Of these wounds at follow-up. Wounds requiring surgical debridement had a sevenfold higher infection rate (p = 0.01). Patients more than 50 years of age had a sixfold higher infection rate than younger patients (p = 0.05). Stepwise logistic regression found the following variables to be the best predictors for wound infection: full-thickness [p = 0.006, odds ratio (OR) = 6.23], female gender (p = 0.048, OR = 2.88), and wound debridement (p = 0.024, OR = 5.01). Combinations of these three variables predict infection rates from 0.35% to 23.9%. CONCLUSIONS A low wound infection rate was seen in this cohort of dog-bite victims who were treated on an outpatient basis. Wound depth, patient gender, and wound debridement were the clinical variables that best predicted the likelihood of developing infection. Future interventional studies should concentrate on wounds with high probabilities of infection.