The models reported here have a common financial reporting system, and common faculty. The hospital-based practice demonstrates that a financially self-sufficient academic practice is possible, while the others show that care delivery to disadvantaged populations can be conducted at costs similar to those of a teaching practice. Nonetheless, the results may not be easily generalizable. Setting up and operating these new offices and clinics required a high degree of cross-institutional cooperation and trust. Crafting contracts that included academic departments and the practice plan, as well as hospitals, public schools, county social services, the city housing authority and a federally qualified community health center, required a level of institutional oversight and assistance which may be difficult to find in other settings. Yet, by challenging conventional wisdom, the models may encourage others to try their own experiments.
Having an appropriate avenue for billing is particularly important. In the models described here, operations and billing occurred under the aegis of a federally qualified community health center, an academic hospital, or an academic practice plan, as appropriate. Given the wide range of payers in the local community, and the range of reimbursement for identical visits, it was essential to identify and work with the operational unit that most closely matched the needs of the population served. For the family medicine center, the change from a faculty practice plan to a hospital-based clinic with both professional and facility fees allowed an already efficient practice to break even or better. The successful transition, however, was predicated on the willingness of the major payers to reimburse the facility fee charges.
Space cost was another variable, often ignored, that became a critical success factor. Costs for space varied from $0 to $55 per net square foot, which led to a difference in cost per visit of $22 to $101. For a practice close to breaking even, the cost of space can easily determine financial sustainability. Lower cost office space, or no space at all, can make a large difference.
Practice efficiency may be one of the greatest challenges. Overall, academic family practices see fewer patients, and generate fewer relative value units per provider, than their peers in private practice [8
]. In part, this may reflect an inherent inefficiency involved in the education of students and residents. However, the resident problem can be exaggerated. In the settings described here, faculty work units per full-time-equivalent (4217) were on par with the national mean for private practice (4244, MGMA) [8
]. Paradoxically, part of the higher productivity stems from the educational setting: residents are scheduled in sets of four per half day, mixed across the classes, so that each preceptor is responsible for more patients, and more work units, than would be possible if she were personally seeing patients. In addition, the family practice has a goal of being like a private practice in operation – of being "a practice that teaches" rather than a "teaching practice." With about 70% of visits with faculty and less than 30% with residents, and resident time clustered to maximize precepting efficiency, achieving private practice efficiencies is possible. Similarly, residents have been included in the community-based sites, but only after the practices were operating efficiently and community acceptance was achieved.
Finally, the centrality of midlevel practitioners has become clear. Highly trained physician assistants (PAs) and nurse practitioners (NPs), with physician support for the more complicated patients, provide excellent care at a cost per visit lower than that of a physician working alone. While it is not clear what the optimal mix is of physicians to midlevel providers, our experience has been that the financial performance of the team increases with the number of PAs and NPs, but that the need for interaction and supervision will restrict a single physician from working with more than 2–4 midlevel providers. It has not yet been possible to test the model beyond a configuration of two midlevel providers per one physician, however, due to a shortage of qualified PAs and NPs.