Management Decisions in Nontraumatic Complaints of Arm, Neck, and Shoulder in General Practice

Study Population

In total, 31 GPs recruited eligible patients from September 2001 through December 2002. Included were patients aged 18 to 64 years who consulted their GP for a new complaint or a new episode of complaints of neck, upper back, shoulder, upper arm, elbow, forearm, wrist, or hand. The episode was considered new if patients had not consulted their GP for the same complaint during the preceding 6 months. Excluded were patients for whom the complaint could be explained by a trauma, fracture, malignancy, amputation, prosthesis, congenital defect, or previously diagnosed systemic or generalized neurological disorder.

Procedures

Before the GPs started to recruit patients, the GP characteristics were gathered during a structured interview.

During the first consultation patients received study information, an informed consent form, and a self-administered questionnaire from their GP. A fax was then sent to the research team with the initial diagnosis, expected prognosis at 6 months, and whether the complaint was recurrent. After the research team had received the completed informed consent and questionnaire (within 8 weeks), inclusion criteria were verified in the electronic medical file. Data on treatment management at the time of the consultation and patient and complaint characteristics were extracted from the self-administered questionnaire.

Outcome

Data on the following 6 common management options (yes/no) were gathered: request additional diagnostic tests (in addition to history taking and physical examination), give a corticosteroid injection, prescribe a medication (injections excluded), refer to a physiotherapist, refer to a medical specialist, and watchful waiting.

Variables Possibly Associated With Management

Possible associated variables were assigned to 1 of 3 domains: (1) diagnosis, (2) patient and complaint characteristics, and (3) GP characteristics. In accordance with the Dutch guidelines for shoulder complaints and epicondylitis, the diagnoses were categorized into 4 groups: specific shoulder diagnosis (impingement, frozen shoulder, biceps tendonitis), epicondylitis, other specific diagnosis, and nonspecific diagnosis.^6,7,14

Patient and complaint characteristics included age and sex, having employment, complaint severity,¹⁵ functional limitations,¹⁶ perceived general health,¹⁷ musculoskeletal comorbidity,^1,3 recurrent vs incident complaint, multiple-region complaint, social support,¹⁸ and somatization and distress.¹⁹ GP characteristics included sex, years in practice, special interest in musculoskeletal complaints, continuing education in musculoskeletal complaints, whether the physician was in group practice, and practice location (rural vs urban).²⁰

In addition, at the first consultation the treating GP scored expected nonrecovery at 6 months on a 4-point Likert scale ranging from very likely (1) to very unlikely (4). These scores were dichotomized into likely or unlikely.

Statistical Analyses

We evaluated the independent relationship of the variables of the 3 domains (diagnosis, patient and complaint characteristics, and GP characteristics) with the 6 management options by means of logistic regression for repeated measurements (generalized estimating equations),^21,22 with physicians as units of analysis and compound symmetry as variance/covariance structure. This model takes into account the correlation between patients consulting the same physician (PROC GENMOD, SAS version 8.2, SAS Institute, Cary, North Carolina).

Separate models were constructed for the 6 management options. For possible associated factors with clinically relevant classifications or validated classes, the existing cutoff points were used. If cutoff points were not available, ordinal scores were split based on the median score of the total population.

Analyses were performed both in univariate and multivariate models, resulting in the assessment of odds ratios (ORs) and corresponding 95% confidence intervals (CIs). First, variables with P <.20 in the univariate model were selected for multivariate analysis (enter procedure) within the domains of patient and complaint characteristics and GP characteristics separately. Second, variables with P<.20 in the multivariate model per domain were selected for the final analyses. Finally, variables with a significant odds ratio (P<.05) were kept in the multivariate model and presented in the results. The diagnostic group variable entered the multivariate model and was kept in the multivariate model independent of the level of significance. The largest diagnostic group is defined as the reference group because doing so is statistically the most efficient.

Response

In total, 798 patients fulfilled the entry criteria, and 682 (86%) entered the cohort after they returned the completed questionnaire and informed consent. The mean time between consultation and filling in the questionnaire was 2 weeks.

Patient/Complaint Characteristics

Table 1⇓ shows the distribution of the patient and complaint characteristics of the 682 patients in the current study. Of the total study population 47% was male and the median age was 45 years. One-half (51%) of the participants consulted their GP within 6 weeks after developing the complaint. After consultation, 229 patients (34%) received a specific shoulder diagnosis (impingement, biceps tendonitis, frozen shoulder), 95 (14%) received a diagnosis of epicondylitis, and 78 (11%) received another specific diagnosis (eg, carpal tunnel syndrome, de Quervain tenosynovitis, etc); in 280 patients (41%) no specific diagnosis was registered (Supplemental Appendix 1, available online at http://annfammed.org/cgi/content/full/7/5/446/DC1). Complaint severity in the previous week showed a median score of 6 (range 1–10) and functional limitations of arm, neck, and shoulder showed a median score of 35.4 (range 2.9–99.1).

Physician Characteristics

Of the 31 recruiting GPs, 5 (16%) were female. The reported years of practice as a GP ranged from 1 to 35 years (median 13) years; 53% of the physicians reported to have a special interest in musculoskeletal complaints. About one-half of the practices were solo practices, and the remainder were group practices.

Management

Table 2⇓ displays the treatment options applied in both the total population and by diagnostic group. Within a maximum of 8 weeks, 9% of the total population received additional diagnostic testing, in 11% corticosteroid injections were given, and in 35% medication was prescribed. Of all patients, 23% was referred for physiotherapy treatment and 5% for medical specialist consultation. In 41% of the patients none of the specified options was applied.

Management Decisions

The complaint characteristics of long duration of complaints, high complaint severity, high functional limitations, and having a recurrent complaint were all negatively associated with watchful waiting (Table 3⇓).

Variables positively associated with the application of additional diagnostic tests were longer duration of complaints, a multiple-region complaint, and the GP having less than 13 years of practice (Table 3⇑). Negatively associated variables were having a diagnosis of epicondylitis (compared with a nonspecific diagnosis) and the GP working in a group practice.

Variables positively associated with a prescription of a medication were having a specific shoulder diagnosis, high functional limitations, high somatization, short duration of complaints, and the GP being male (Table 3⇑).

Variables positively associated with application of corticosteroid injections were having a specific shoulder diagnosis or other specific diagnosis, consulting a GP with special interest in musculoskeletal complaints, longer duration of complaints, high functional limitations, and being older (Table 4⇓).

Variables positively associated with referral for physiotherapy were longer duration of complaints, having a recurrent complaint, having more distress, and consulting a GP in a rural area (Table 4⇑).

Having a specific shoulder diagnosis or another specific diagnosis was negatively associated with referral for physiotherapy, but it was strongly positively associated with referral to a medical specialist. Furthermore, patients who reported higher functional limitations or consulted a GP with fewer years of experience were more frequently referred to a medical specialist.

In addition to the diagnostic category, a longer duration of complaints and higher functional limitations proved to be the patient and complaint variables most frequently associated with the 6 evaluated options in initial management of patients with nontraumatic arm, neck, and shoulder complaints (Table 5⇓).

In watchful waiting, only long duration of complaints, high complaint severity, many functional limitations, and a recurrent complaint played a role. In patients with these complaint variables, watchful waiting was opted for less frequently.

The results of the univariate analysis are presented in Supplemental Appendix 2 (available online at http://annfammed.org/cgi/content/full/7/5/446/DC1).

Influence of GP Characteristics

As expected, GP characteristics were not associated with watchful waiting; only complaint characteristics played an independent role. Several GP characteristics did play an independent role in the other 5 management options, however. Working in a solo practice and less practice experience were associated with additional diagnostic tests. The latter variable was also associated with referral to a medical specialist. These GP characteristics may imply some uncertainty regarding the diagnosis. More referral to physical therapists made by GPs in rural areas may reflect a closer cooperation between caregivers in such areas. Furthermore, female GPs prescribed relatively less medication compared with their male colleagues, although they did not use any other option more frequently.

Corticosteroid injections were most frequently given by those GPs with a special interest in musculoskeletal complaints. Their interest might result in more knowledge and training regarding these complaints (including giving a corticosteroid injection) and in more confidence in and more frequent use of corticosteroids.

In a study on confidence in primary care management, Glazier et al showed that GPs reported low scores on performing a joint injection or aspiration²⁴; the authors proposed that in addition to lack of interest, lack of training and experience is likely to explain the lower scores on confidence. On the other hand, the evidence related to corticosteroid injections was mainly restricted to short-term relief of symptoms,^25–28 which may be why some GPs selected this option less frequently.

Influence of Diagnosis

In the subgroups of shoulder complaints and epicondylitis, the Dutch College of General Practitioners has issued management guidelines.^6,7 The recommended management for shoulder complaints consists of a stepwise approach: watchful waiting (information and wait-and-see), analgesics (ideally, paracetamol; nonsteroidal anti-inflammatories as second-line choice, taken intermittently if no contraindications exist), corticosteroid injections, and referral for exercise therapy if functional limitations are still present after 6 weeks.

For epicondylitis a similar approach is recommended: watchful waiting, analgesics, or corticosteroid injections if pain hinders function. The initial management, as shown in Table 2⇑, seems mainly in line with the guidelines, taking into account the effect of hindrance and duration of the complaint. One difference seems to be the referral for physiotherapy in patients with epicondylitis, an option not recommended in the guideline.^6,7

Influence of Patient and Complaint Characteristics

The role complaint variables have in management indicates that, when a patient reports considerable hindrance, the GP will more frequently consider an active treatment option. Additionally, the results show that although medication is prescribed frequently, when patients have a long duration of nontraumatic arm, neck, or shoulder complaints, other options are explored.^6,7

No guidelines are available for the largest subgroup of nonspecific complaints. Patients with these complaints are relatively frequently referred for physiotherapy, whereas patients in the small subgroup of having other specific diagnosis are relatively often referred to a medical specialist. Although we report only on initial treatment decisions, these findings seem to be in line with the findings of the 2nd Dutch National Survey of General Practice. There, patients with International Classification of Primary Care codes that included complaints located at wrist and hand or fingers were most frequently referred to secondary care, with the highest referral rate for carpal tunnel syndrome.²⁹ For patients in this relatively small group whose complaints were of long duration and who had high functional limitations, special treatment or confirmation from a medical specialist seems to have been preferred. Additionally, high somatization scores were positively associated with prescription of medication. An explanation for this association may be that these bodily symptoms are interpreted as being more of a hindrance. Furthermore, in contrast to our expectation, being employed was not an indicator of treatment management.

Influence of Prognosis

We expected indicators of poor prognosis also to be indicators of management. The associations of long duration of complaints, having a recurrent complaint, and a high somatization score were in line with our expectation. Little social support and having musculoskeletal comorbidity, however, were not associated with any of the 6 management options.

Perhaps the GP’s prognosis directly influences management rather than the included indicators of poor prognosis; however, there may be some overlap. Because we had collected the 6-month prognosis according to the GP, we also explored its univariate relationship with the 6 management options. Poor prognosis showed a negative association with watchful waiting (OR = 0.6; 95% CI, 0.5–0.9). Poor prognosis showed a positive association with additional diagnostic tests (OR = 2.7; 95% CI, 1.7–4.6) and with referral for physiotherapy (OR = 2.1; 95% CI, 1.5–3.0). The association with referral for medical specialist care (OR = 1.6; 95% CI, 0.7–3.5) was not significant (besides the low OR, the prevalence of the outcome was also low). Prescription of medication (OR = 1.1; 95% CI, 0.9–1.5) and application of corticosteroid injection (OR = 0.9; 95% CI, 0.6–1.1), however, did not show a relation with the expected prognosis. This is in line with the short-term relief of symptoms as a treatment goal with these 2 options while waiting for a positive natural course.

Treatment Options and Evidence

Although there is evidence of effect for some of the treatment options in nontraumatic arm, neck, and shoulder complaints, these effects are mainly reported as a short-term effect (mainly short-term pain relief) or a limited effect.^{25–28,30,31} The lack of solid evidence in favor of the treatment options explored in our population may explain variability in management and may leave more room for the personal preferences of both the GP and the patient. Aiming at short-term relief of symptoms may lead to favoring a corticosteroid injection, whereas contraindications or fear of side effects of nonsteroidal anti-inflammatory medications may lead to favoring any alternative.

Strengths and Limitations of the Study

In the present study we studied only initial management in nontraumatic complaints of arm, neck, and shoulder; thus, additional management options may be applied if the complaints persist. Also, because of the 8-week response time, 21% of the patients consulted their GP more than once before filling in the study questionnaire. The number of consultations variable that we added to the final model was associated with all 6 outcomes. Thus, the more consultations, the more frequently a management option other than watchful waiting was applied. Adding the number of consultations to the models hardly changed the odds ratios of the other variables, however. Furthermore, because the different management options were not mutually exclusive, we also checked for co-occurrence of management options. In the initial phase of management, multiple options were applied in 114 (17%) of the patients, mainly analgesics and another option (n = 94; 14%). The most frequent combination was with referral for physiotherapy (n = 57; 8%). Because our heterogeneous population mainly consists of patients with complaints at neck, shoulder, or elbow, these results mainly apply to these particular subgroups. For patients with certain specific diagnoses (eg, carpal tunnel syndrome; osteoarthritis of elbow, wrist, and/or hand; or hernia of the neck), these results need to be interpreted carefully because of the small numbers. The number of GPs included in the study (31) is not large, but it seems sufficient based on the confidence intervals of the estimates of the GP characteristics.

In the present study we focused on the diagnosis, the patient and complaint, and the GP characteristics and their association with management options. Other variables may also play a role, however. In a prescriptive model for evidence-based clinical decisions, Haynes et al describe a combination of patients’ circumstances, patients’ preferences and actions, and best evidence research, with an overlap area representing clinical expertise.³² Although we included many variables, we did not include patient preference. Furthermore, we did not include such economic variables as insurance and costs, which may also contribute to management choices.³³ A study on medical practice variation in general practice, however, concluded that a clinical rather than an economic model is a more plausible explanation of interpractitioner variation.³⁴

Consequences for Practice and Future

Overall, the diagnostic categories of having a long duration of complaints and reporting many functional limitations were most frequently associated with the choice for a management option. In watchful waiting, the complaint variables (long duration of complaints, high complaint severity, many functional limitations, and a recurrent complaint) played a role in deciding for this management option less frequently. When deciding for 1 of the 5 other management options, however, several GP characteristics played a role as well. That less experience was related to more additional diagnostic tests and more referral for medial specialist care may imply uncertainty regarding the diagnosis. Regarding treatment options, lack of solid evidence in favor of 1 certain option seems to leave room for personal preferences. Further evidence on the effectiveness of management (especially in the large group of nonspecific diagnoses) may reduce variability resulting from GP characteristics.