Author	S.J. Gawler
Author	S. Iliffe
Author	K. Walters
Author	R. Morris
Abstract	Background: Therapist and dose effects may be important in falls prevention exercise service delivery because the evidence for falls prevention exercise has a very specific prescription, therefore, maximising therapist skills, minimising 'therapist drift' and encouraging compliance could enhance patient outcomes. Methods: The aims of this research were to study (a) the effect of the therapist (delivering the exercise programme), and (b) the effect of exercise dose, on falls outcomes within a group exercise intervention. The primary objective was to establish any difference in the number of falls for subjects participating in the intervention and the secondary objective was to establish any difference in falls risk factors (balance and lower limb power) for these subjects, according to their therapist and separately, their dose. Multilevel modelling, which is designed for clustered data, was used to investigate the magnitude of therapist and dose effects, and to explore whether specific therapist characteristics were individually associated with the falls outcomes. Results and Conclusion: Unconditional multilevel models showed some variance between patients grouped by therapists of up to 6% of the overall variance in falls outcomes. These effect sizes are small, but in a standardised exercise intervention, especially within the research setting, they would not be expected to be large due to quality assurance procedures reducing variability between therapists. The therapist characteristics investigated, however, did not explain this therapist-level variance, and it may be that the characteristics studied did not include those that make a difference to falls outcomes. Another explanation for the unconvincing evidence of therapists effects is that the main trial (within which my study was nested) was not set up to investigate therapist effects and therefore was not powered for this. The dose effect analysis showed that the dose of the exercise intervention was not an independent predictor of falls rate nor falls risk factors. It is possible that the dose investigation was affected by the high numbers of non-fallers within the recruited population. The use of our protocol and documents for the quality assurance of the intervention within research was effective at standardisation and ensuring fidelity, and this approach could be used as part of falls prevention exercise service delivery to reduce 'therapist drift'.