This study, a prospective, randomized trial comparing two inpatient staffing models, was undertaken to compare clinical and financial outcomes for general medicine inpatients assigned to resident (teaching) or staff (nonteaching) service. Key outcome measures included: 1) length of stay; 2) total charges; 3) laboratory, radiology, pharmacy, and supplies charges; 4) in-hospital mortality and mortality within 6 months of admission; and 5) 15-day readmission rate. The study took place at Henry Ford Hospital, a 937-bed urban teaching hospital in Detroit, Michigan; the subjects included all general internal medicine patients admitted to a single nursing unit of Henry Ford Hospital between October 1, 1987 and September 30, 1988. When the unit was fully staffed and operational, patients admitted to the Staff Service had a 1.7-day lower average length of stay than patients admitted to the Resident Service (P greater than 0.005), lower average total charges of $1,681 (P greater than 0.01), and significantly lower laboratory and pharmacy charges. No statistically significant differences in mortality rates or readmission rates were found. Even though personnel costs are invariably higher on an attending service, this staffing arrangement can be financially viable because of more efficient patterns of care. Shorter length of stay may be translated either into cost savings or increased revenues in order to offset higher salary costs. Teaching hospitals may wish to consider an attending service as one way to reduce house officer work loads, offer more opportunities for training in ambulatory settings, and adjust to a smaller pool of applicants for residency positions.