Margit Kooijman

110 yet, even been disregarded 19 . The findings of present study should therefore be seen as a stepping-stone towards more research, in which personality traits deserve to be taken into account. Study limitations Limitations of present study include the fact that the physiotherapist recorded patients’ opinion on outcome and that ‘severity of complaints’ is a broad construct. This is one disadvantage of using previously gathered data. However, the same method was used at the beginning and end of treatment, allowing comparison in time. The database from which data were derived for this study, now holds more specific patient reported outcome measures, which is to be preferred. Also, there was a relatively high number of patients with missing data on severity of complaints. Patients with missing data were younger, more often had a recurrent complaint which more frequently arose gradually. Although the existing body of evidence shows conflicting results regarding these characteristics, current study identified age and recurrent complaints as predictors of outcome, therefore, caution is needed in interpreting our findings. On the other hand, the percentage of missing data on the dependent variable between physiotherapists differed greatly and was not correlated to the mean difference in severity of shoulder complaints. This suggests that missing data is at least partly explained by general poor registration by physiotherapists instead of selectively omitting disappointing results. Lastly, it is difficult to put the BIG 5 scores into perspective because norm scores barely exist, let alone for a comparable sample of trained professionals. Lack of heterogeneity could be an explanation for not finding an association between personality traits and treatment outcome. This can only be verified through future studies. Nevertheless, one benefit of a cohort study like this is that is does allow variability in other aspects and thus reflects actual practice more accurately; physiotherapists of different practices participated and were free in choice of treatment. Another strength of current study is that data was collected on a large number of patients and we were able to investigate prognostic factors both at the physiotherapist and at the patient level in an attempt to increase our understanding of its interaction. Since neither personality traits nor psychological factors like fear avoidance beliefs were measured at the patient level, we recommend exploring these in future research. Conclusions In conclusion, a therapist effect exists in the rehabilitation of patients with shoulder complaints in primary care physiotherapy settings. The physiotherapist explains 12% of variance in change of severity of the shoulder complaint and further investigation of this effect showed that therapists with more extraverted personalities influenced treatment