Background
Establishing a diagnosis for musculoskeletal conditions constitutes a major health care expenditure. The most common reason for a patient to consult a health care provider is musculoskeletal pain or disability,[1] and such complaints account for up to 28% of all problems seen in the family practice office.[2] Musculoskeletal disorders are also the leading cause of long-term physical disability and absence from the workforce.[1]

Shoulder pain is the third most common complaint addressed by primary care physicians.[3] A large percentage of these patients (up to 60%) present with pain secondary to a rotator cuff tear (RCT).[4] Other common pathologies include glenohumeral osteoarthritis (OA), traumatic instability (I), and adhesive capsulitis (AC). Diagnosis requires a detailed history and physical examination. However, given the complex nature of the shoulder joint and the possible presence of multiple pathological lesions, establishing a diagnosis may be challenging.

The evaluation of a patient with shoulder pain involves an initial assessment, which may be followed by imaging to verify the diagnosis. If necessary, the patient is then referred to an orthopaedic specialist for treatment.

It is assumed that advanced imaging modalities will confer added benefit and increase diagnostic accuracy. However, several studies refute this claim and demonstrate the use of additional advanced imaging may have minimal impact on the care patients ultimately receive.[5,6] Thus the increased demand for advanced imaging and subsequent specialist referral may serve little purpose and may increase wait times for these services. Believing that a stringent evaluation of advanced imaging and referral patterns in assessing shoulder pain would be of value, we proposed a study to investigate the assessment of common shoulder pathologies encountered in the primary care setting. The study also sought to analyze imaging relevance and diagnostic concordance between referring physicians and orthopaedic surgeons. Other outcomes proposed included the wait time between referral and orthopaedic assessment, and the percentage of surgical referrals. We hypothesized a discordance between referring and orthopaedic specialist diagnoses. To date, we are not aware of any previous studies that have investigated this topic.

Methods
We conducted a retrospective, observational cohort study utilizing a patient chart review. The study included all patients referred to the shoulder practice of this article’s senior author (DPG) between January 2011 and May 2015. The study was approved by the UBC Clinical Research and Ethics Board. Patient demographic characteristics were collected along with clinical findings. Averages, means, and standard deviations relevant to the outcome measures were calculated using Microsoft Excel (Redmond, WA).

Patient enrollment
Patients referred to the surgeon’s clinic during the study period were identified through an electronic medical record system (Plexia) using International Classification of Diseases codes for one or more of the following: rotator cuff tear, glenohumeral osteoarthritis, instability, and adhesive capsulitis. Patients with these diagnoses were enrolled sequentially until target numbers for each group were reached. Inclusion criteria required patients to be 18 years of age or older, exhibit skeletal maturity, and have insurance through the Medical Services Plan of British Columbia. Patients were excluded from the study if they were uninsured or were patients under WorkSafeBC or ICBC.

Outcome measures
Outcome measures included the referring diagnosis and imaging studies available at the time of specialist assessment. Data were also extracted from the referral database regarding wait times from primary care referral to specialist assessment, the final diagnosis as determined by the orthopaedic specialist, and the operative cases.

“Final” diagnosis was defined as the diagnosis determined by the orthopaedic specialist preoperatively and postoperatively.

“Indicated” imaging was defined according to standard orthopaedic management and literature. For example, X-ray is currently considered an indicated imaging modality for diagnosing adhesive capsulitis and ruling out an intrinsic cause for motion loss (e.g., glenohumeral arthritis). X-ray studies are also considered indicated for suspected instability to identify concentric reduction and the absence of fracture or bone loss. X-ray plus ultrasound or MRI is considered indicated for rotator cuff disease.

“Not medically indicated” imaging was defined as investigations beyond what is necessary for diagnosis or treatment.

“Underinvestigated” was used to describe the absence of imaging necessary to formulate a treatment plan.

An example of not medially indicated imaging included the utilization of CT or MRI in the case of adhesive capsulitis. An example of underinvestigation would be the absence of an X-ray for all pathologies investigated.

Results
A total of 150 patients were included in this study, and 41 of these (27%) were ultimately surgical referrals. The mean wait time from primary care referral to specialist assessment was 138 (99) days.

Table 1 [10] presents the diagnoses and demographic characteristics of all patients. As expected, more males than females were referred for instability and more females than males for adhesive capsulitis.

Table 2 [11] compares the specialist and referring diagnoses and classifies the imaging investigations as indicated, not medically indicated, or underinvestigated based on the underlying pathology. The referring diagnoses of common shoulder pathologies were concordant with the specialist diagnoses 21% to 97% of the time, and patients underwent indicated investigations 40% to 78% of the time.

Table 3 [12] provides a summary of discordant diagnoses (i.e., where specialist diagnoses differed from referring diagnoses). Unspecified shoulder pain was the most common referring diagnosis applied to all the pathologies studied. Adhesive capsulitis was the most common pathology referred as another diagnosis.

Table 4 [13] classifies the diagnostic imaging as not medically indicated or underinvestigated for each shoulder disorder. The most common example of not medically indicated imaging was the use of MRI for adhesive capsulitis. The most common example of underinvestigation was the failure to obtain X-ray images at the time of referral for all shoulder pathologies.

Table 5 [14] compares preoperative and postoperative diagnoses and shows that only one diagnosis was changed postoperatively.

Conclusions
This retrospective, observational cohort study evaluated assessment and imaging protocols and referral practices for shoulder pain in the primary care setting. Referring diagnoses for 150 patients were compared with the final diagnoses by an orthopaedic specialist. In addition, imaging available at the time of specialist assessment was classified as either indicated, not medically indicated, or underinvestigated. Wait times after referral were also recorded.

Advanced imaging often not medically indicated
The results confirmed our initial hypothesis: there was a limited correlation between referring and specialist diagnoses. Also, there was an increased preponderance toward not medically indicated investigations for shoulder pain prior to specialist referral.

While history taking, physical examination, and preliminary imaging provide the essential information required to accurately diagnose a shoulder disorder, more advanced imaging has been shown to be of limited value in the early diagnostic stages.[5,6] However, in our study, subjects underwent advanced imaging even in cases where it would not affect management. Several explanations may account for the use of advanced imaging.

Patients often present to the referring physician with shoulder pain early in the disease process. This can limit accurate examination as pain inhibition prevents assessment, and repeated visits for unresolved shoulder pain can lead to advanced imaging and subsequent specialist referral. For example, a patient presenting with frozen shoulder in the acute inflammatory phase may not tolerate a full examination or physiotherapy.

The delay to specialist referral can also coincide with the natural history of the disease and mean the patient presents for specialist evaluation in the later phases of disease. This allows for a more reliable examination and subsequent diagnosis. This is consistent with a low concordance rate of 21% between referring and specialist diagnoses for adhesive capsulitis and the highest rate of not medically indicated imaging.

Patients may also demand investigations from the referring physician because the Internet has misled them regarding the value of MRI for shoulder pathology. This becomes very challenging for front-line care providers. It may also be assumed that an orthopaedic specialist requires advanced imaging before a shoulder referral is accepted. This, in addition to the requirements of third-party insurers such as WorkSafeBC and ICBC, adds a level of complexity when patients present with nonspecific shoulder pain.

Referring diagnoses of rotator cuff tear and glenohumeral osteoarthritis had similar low rates of concordance with specialist diagnoses and an increased preponderance toward not medically indicated investigations. A possible explanation for this may be the diffuse nonspecific pain often noted in patients with rotator cuff pathology. Radiating symptoms associated with concurrent acromioclavicular arthritis and trapezial-related pain further complicate establishing a diagnosis.

Study strengths and limitations
This study has several strengths, including sample size and study design. The number of patients enrolled was large for an orthopaedic study and the use of a retrospective design eliminated any alterations in practice patterns that can occur during a prospective study. However, we note that a retrospective study may also be prone to bias.

Study limitations include the fact that all patients were enrolled as a result of referral to a single orthopaedic specialist; the referral base of primary care physicians may not be representative of the overall physician population; and the correct final diagnosis was defined as that of the specialist. This last limitation does not account for possible instances where the specialist was mistaken and the referring physician was correct. However, we did review follow-up appointments with the orthopaedic specialist for all patients enrolled in the study and found no change in diagnosis on subsequent visits. There was also only one change at the time of operative intervention from rotator cuff tear to rotator cuff syndrome (i.e., no rotator cuff tear was identified at the time of arthroscopy).

Another potential limitation may apply to patients with concomitant diagnoses, which is common with shoulder pathology. For example, a patient may have presented with both adhesive capsulitis and a rotator cuff tear. In this case, if the clinical picture was more in keeping with adhesive capsulitis, the case was labeled as such. Therefore, there may have been an overestimation of discordant diagnosis for adhesive capsulitis; however, it should be noted that no patient with a specialist diagnosis of adhesive capsulitis underwent surgery for rotator cuff tear. Also, some patients were referred simply for shoulder pain despite imaging reports confirming a rotator cuff tear or osteoarthritis. Thus, some referring physicians whose diagnoses were reviewed in this study may have known the actual diagnosis but chose to describe it simply as shoulder pain. As these cases were classified as being incorrectly diagnosed, it is possible that this lack of specificity led to an overestimation of diagnoses deemed discordant.

Study implications
The results of the present study have significant implications for both the individual patient and the health care system. As wait times to see specialists increase, triage by the orthopaedic specialist is contingent on the diagnosis of the referring physicians. For the patient, an incomplete or incorrect diagnosis may lead to imaging that is not medically indicated. This can delay referral and optimal treatment, whether it be for conservative care or operative intervention. For the health care system, such delays can lead to an increase in both direct and indirect costs, including the cost of imaging wait times across all specialties. For example, a patient may currently wait as long as a full year to receive a routine shoulder MRI at a local care centre.

The results of this study indicate that the assessment of shoulder pain is a difficult multifactorial problem. Looking at four common shoulder pathologies, there is a high rate of imaging deemed not medically indicated and discordance between the referring and specialist diagnoses. This translates into increased wait times for specialist assessment and a low rate of surgical cases extracted from the referrals. These study results highlight the need for increased collaboration between referring physicians and orthopaedic specialists.

Competing interests
None declared.

This article has been peer reviewed.

Dr Jarvie is a PGY-5 orthopaedic surgery resident at the University of British Columbia. Dr Pike is a clinical assistant professor in the Department of Orthopaedics at the University of British Columbia and a staff orthopaedic surgeon at St. Paul’s Hospital. Dr Goel is a clinical assistant professor in the Department of Orthopaedics at UBC and a staff orthopaedic surgeon at Burnaby General Hospital.