To determine the probable effect of the Vaccines for Children (VFC) program on immunization coverage. Preintervention and postintervention study design, with data collected before and after enrollment in the VFC program. Twenty-three inner-city neighborhood physicians' offices in New York City. In 1993, 30 physicians were randomly selected from 8 neighborhoods with the highest proportions of Medicaid-eligible individuals in New York City. In 1995-1996, the 30 physicians were contacted again. Twenty-three agreed to an interview and medical record review. Within each office, the medical records of children aged 3 to 35 months, with at least 3 visits in a 3-month or longer period, were randomly selected. Medical record reviews were conducted for 173 eligible children in 1993 and 528 in 1995-1996. The VFC program was implemented in October 1994. The administration fee increased from $2 to $17.85; physicians received vaccines free. Up-to-date status for immunizations and lead and tuberculosis screening; percentage of visits that are missed opportunities to immunize; and percentage of visits that were well-child visits. Up-to-date status, missed opportunities to immunize, and well-child visits were compared across time using chi 2 analysis, corrected for the use of cluster sampling. Up-to-date status changed significantly before and alter enrollment in the VFC program (P < .05) for all immunizations and for lead and tuberculosis screening. For the diphtheria toxoid, tetanus toxoid, and pertussis vaccine, oral poliovirus vaccine, and measles, mumps, and rubella vaccine combined, coverage increased from 17.9% to 42.2%, up by 24.3 percentage points (P < .05). Missed opportunities to immunize did not change, but well-child visits increased from 15.0% to 21.6% (P < .05). Physicians generally attributed performance improvements to the VFC program and not to other competing hypotheses. The VFC program seems to be responsible for an increase in immunization rates among these physicians.