Margit Kooijman

General discussion | 121 long pain duration at baseline, somatisation, higher age and presence of previous pain episodes were mentioned in two systematic reviews on this topic 22,25 . Limitations Our meta-analysis provides us with the scientifically most objective evidence for a clear and thus workable set of prognostic factors. Our survey provides us with less robust yet detailed information on the opinions and experiences of clinicians. A disadvantage of both studies is that they focus on the part of treatment by the clinician. Patients’ thoughts and beliefs on a switch to prognostic instead of diagnostic based treatment and use of musculoskeletal ultrasound, would have added to a more complete image. So where do we go from here? It is very interesting that our findings in patients with shoulder pain match the brief set of generic prognostic factors already acknowledged. For patients with shoulder pain without a clear biomedical diagnosis (e.g. suspicion of full rotator cuff tear), this similarity supports targeting treatment based on prognostic factors rather than pain site and diagnosis alone. That being said, reservations from existing literature mentioned above alert us not to replace diagnosis for its own sake by classification on future outcome for its own sake 29 . With musculoskeletal ultrasound, only the category ‘referral to secondary care’ showed higher agreement between clinicians and this is precisely the group of patients in which least uncertainties exist and traditional diagnostic labels suffice. Considering this and the costly investment of equipment and education, current prevalence of musculoskeletal ultrasound is already quite extensive. For patients presenting with a more complex complaint (e.g. those with a longer duration of complaints, a gradual onset, concomitant pain in other areas and/or persisting pain), a more comprehensive approach is required. Still, the question is whether the changing nature of the constellation of symptoms in clinical practice can be caught in the prognostic approach 23 . Also a prognostic model needs new input, by monitoring patients over time, to reflect the situation of the patient as well as possible and modify treatment correspondingly. All things considered, I think that for a small group of patients (e.g. suspicion of full rotator cuff tear) we shouldn’t complicate things while for the other, large group of patients, we should acknowledge that management isn’t simple. Whether it concerns usage of musculoskeletal ultrasound or a prognostic model, input by the patient is essential. Recommendations for clinical practice and research For patients with a traumatic cause of shoulder pain (e.g. suspicion of full rotator cuff tear), musculoskeletal ultrasound may be used to verify tissue damage to facilitate appropriate referral to secondary care. For those patients without a clear medical