Margit Kooijman

42 asked to specify the reason for referring, if any. This way, thoughts on appropriate treatment between general practitioners and physiotherapists could be compared in more detail. Both groups of clinicians were requested to indicate the likelihood (very likely to very unlikely) of recovery taking place within two weeks, six weeks, one to two years or whether the patient would have a recurrence within two years, have permanent difficulties with activities of daily life or would require surgery. Finally, information was gathered on demographic details, years of experience, specific training on musculoskeletal or shoulder complaints, type of practice and if the practice had musculoskeletal ultrasound at their disposal. A letter with online login to the questionnaire was sent to all participants by post. After two weeks, as a reminder all non-responders received the questionnaire by post. Data analysis Descriptive statistics were used to summarize characteristics of general practitioners and physiotherapists. Non-response analyses were performed using t-tests and chi- square tests (a = 0.05). For each vignette and for both groups of clinicians, the proportion of responses was calculated. The response options ‘likely’ and ‘very likely’ were added together for each prognosis item. Whenever the same treatment option was available to both groups of clinicians, Chi-square tests were used to test differences in categorical data between them. All analyses were performed using Stata 15.0. Results Of the 500 questionnaires distributed to general practitioners, 29 were returned by post, 55 digitally and four were returned because of incorrect addressing. The net response rate was 16.9%. Of the 500 questionnaires distributed to physiotherapists, 55 were returned by post, 55 digitally and 42 were returned because of incorrect addressing. The net response rate was 24.0%. Non-response analysis showed that groups were comparable for age and gender except that responding general practitioners were somewhat younger (mean 46 years) than non- responding ones (mean 48 years, p = 0.0325). Table 1 presents demographic details of the responding clinicians.