Objective: To compare the concordance of family physicians' billing for evaluation and management services with medical record documentation. Design: Multimethod, cross-sectional observation study. Setting: Eighty-four family practices in northeast Ohio. Participants: Four thousand fifty-four outpatients visiting 138 family physicians. Main Outcome Measure: The degree of concordance between evaluation and management Current Procedural Terminology codes billed by physicians, with those codes assigned by trained research nurses using American Medical Association criteria to code medical records for the same visits. Results: Discrepancies between the multifactorial nature of family practice outpatient visits and the Current Procedural Terminology coding criteria, which dictate overcoding for depth rather than breadth, made coding difficult (multiple-rater κ statistic between research nurses = 0.36). Among 4137 outpatient visits with complete billing information, 57% of the Current Procedural Terminology codes generated by medical record review were concordant with the actual billing code assigned by physicians. Undercoding and overcoding occurred at a similar frequency (21% and 19%, respectively) and differed by more than 1 code in fewer than 4% of visits. Visits by new patients were more likely to be inaccurately coded than visits by established patients. Conclusions: Record documentation by community family physicians largely reflects the level of services billed using evaluation and management codes. Undercoding is as common as overcoding. Efforts from regulatory agencies should be redirected from penalizing physicians for overcoding to focusing on the development of coding criteria that reflect the multifactorial nature of outpatient primary care practice.