Intended for healthcare professionals

General Practice

The clinical course of shoulder pain: prospective cohort study in primary care

BMJ 1996; 313 doi: (Published 07 September 1996) Cite this as: BMJ 1996;313:601
  1. Peter Croft, professor of epidemiologya,
  2. Daniel Pope, research assistantb,
  3. Alan Silman, director, for the Primary Care Rheumatology Society Shoulder Study Groupb
  1. a Industrial and Community Health Research Centre, School of Postgraduate Medicine, University of Keele, Stoke on Trent ST4 7QB
  2. b ARC Epidemiology Research Unit, University of Manchester, Manchester M13 9PT
  1. Correspondence to: Professor Croft.
  • Accepted 16 May 1996

Current knowledge of the clinical course and efficacy of treatment for shoulder pain comes mainly from studies of hospital patients. However, only a few patients experiencing such pain require referral to a specialist. Although shoulder pain is common in the general population, the outcome of patients presenting in general practice is unknown.1 We conducted a prospective cohort study to determine the outcome of shoulder pain in primary care.

Methods and results

Twelve general practitioners recruited 166 patients who consulted with a new episode of shoulder pain during one year. They recorded demographic information, diagnosis, management, and an assessment of passive elevation of the shoulder; patients assessed the disability associated with their symptoms with a validated 22 item disability questionnaire.2 To assess outcome, identical disability questionnaires were sent to patients six and 18 months after consultation, together with a question measuring self assessed change in symptoms in the intervening period.

A total of 125 (75%) patients provided follow up information at six months and 95 (57%) at 18 months. There were no differences in markers of severity of disease at baseline in these patients and non-responders.

At consultation, patients reported considerable disability, typically scoring positive on half of the 22 disability items (table 1). Injection was a common treatment (58%), capsulitis was the most common diagnosis (39%), and almost one quarter (22%) of patients reported a previous episode of shoulder pain.

Table 1

Positive responses to 22 items of disability questionnaire at baseline and follow up. Values are numbers (percentages)

View this table:

After six months, self reported disability had improved (mean difference in disability scores, 4.1; 95% confidence interval 2.8 to 5.3). The most frequently reported disability items were the same as those at baseline (table 1). Only 21% of patients reported complete recovery.

A baseline disability score above the median value of 10, a duration of symptoms greater than a month, having received an injection at consultation, and having had shoulder pain in the past were significantly associated with poorer outcome at six months (P<0.05, Wilcoxon rank sum test for equality of medians). Patients who had severely restricted passive elevation at baseline (less than 101°) also had a poorer outcome at six months.

After 18 months the disability score had further decreased (mean difference 4.0; 2.5 to 5.4), although only 49% of patients reported complete recovery.


The assumption that shoulder problems are short lived, isolated episodes3 is not supported by data here, which show that a quarter of patients with a new episode recall a previous problem; such a history influences the outcome of the new episode; and only one in five of these new episodes resolved completely six months later and half had not resolved after 18 months. These findings in a cohort of general practice consulters with shoulder pain parallel the findings of Hazleman et al from an elderly community sample that most shoulder disorders were still causing problems three years later.4

The recruiting practitioners in our study were asked to include all patients with new episodes of shoulder pain. This was not verified independently, so our findings may not reflect the full range of episodes presented in primary care. The practices were therefore stratified into three groups according to the number of patients recruited. Baseline characteristics of patients in the three groups did not differ; this argues against selection bias.

The poorer outcome at six months in those who were given an injection may reflect the more severe baseline symptoms in this group (at consultation they reported having had their symptoms for longer and had a lower angle of passive elevation). However, this finding does question the effectiveness of injections in reducing longer term morbidity.

The finding that the duration of the symptom influences outcome may reflect an adverse consequence of delay in presentation. It is not known whether encouragement to attend a doctor and to receive treatment early in an episode might improve the longer term outcome and reduce the incidence of chronic shoulder problems.

Members of the group who participated in the study were Drs Penny Aeberhard, Martin Allen, Robert Curzon, Ahmet Fuat, Roger Gadsby, Iain Gilchrist, Sally Hull, the late Andrew Martynoga, Ian Robertson, Keith Slinger, Owen Thurtle, Andrew Watson, and David Wheeler. We are grateful to Rita Boswell and Lesley Jordan for their help with the study.


  • Funding Arthritis and Rheumatism Council.

  • Conflict of interest None.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.