Patients may be admitted across administrative boundaries for a number of reasons. We compared cross-boundary flows into a teaching district for individual operations and diagnoses with flows into a group of non-teaching districts. We have attempted to distinguish cross-boundary care that is the result of natural geographic factors from that which reflects specialized supra-district hospital practice. Hospital activity data were analysed for patients treated in six districts in the Oxford Regional Health Authority area in the calendar years 1979-1986 and the financial year 1990-1991. In 1979-1986, 21.1 per cent of patients admitted in the teaching district, and 9.2 per cent in the non-teaching districts, were from other districts. Cross-boundary flows varied greatly between specialties. The non-resident workload in the teaching district increased a little over time from 21.1 per cent (1979-1986) to 24.5 per cent (1990-1991); no appreciable change was seen in the non-teaching districts. For many common conditions, the proportion of non-residents admitted in the teaching district was similar to that in the non-teaching districts, at about 10 per cent or less. For other conditions, it was higher than this but similar in both the teaching and non-teaching districts. Conditions were also identified for which the proportion of non-residents treated was clearly much higher in the teaching district than elsewhere. Geographical convenience accounted for about half the non-resident workload in the teaching district. Specialized supra-district care constituted a small but important part of the total workload.(ABSTRACT TRUNCATED AT 250 WORDS)