This research explores the relationship between support worker utilisation and patient- and team-level characteristics using a prospective longitudinal study of 20 older people's community rehabilitation teams. Between January and September 2009, 462 Whole Time Equivalent Staff and 1913 patients participated in the study. Patient dependency, health status (Therapy Outcome Measures, EQ-5D and Levels of Care tool) and demographic data were collected alongside detailed staff activity data for patients during the recruitment period. Multivariate analyses were used to determine the relationship between the proportion of care delivered by support workers team and patient variables. Support workers delivered up to 36% of direct patient care and spent less time per patient contact (36.0minutes, SD 37.7, range 0-600) than qualified professionals (mean time per contact 75.8minutes, SD 32.9, range 0-334). Less-dependent and female patients had a greater proportion of support worker input, with a 4.9% increase in face-to-face support worker time for every 0.1 unit increase in EQ-5D (95% CI 0.3-9.6, P=0.038), while females had an additional 5.1% of their total contact time with support workers compared with males (95% CI 1.9-8.4, P=0.002). In an analysis without EQ-5D, older patients had a greater proportion of support worker time (P=0.006). Other factors associated with support worker input include the referral source, intensity of care, usual living arrangements and proportion of support staff in a team. Results indicate that patient- and team-level factors are a determinant in the proportion of support worker care delivered to older people in the community, suggesting that it may be appropriate to introduce a more explicit targeted deployment of support workers based on patient-level characteristics, which may facilitate a more effective use of qualified practitioner time for more complex, dependent patients.