Use of physician services in a population-based cohort of patients with polymyalgia rheumatica over the course of their disease

Authors


Abstract

Objective

To describe the use of generalist and rheumatologist services in a population-based cohort of patients with polymyalgia rheumatica (PMR) and to identify predictors of rheumatology care.

Methods

We identified all incident cases of PMR among residents of Olmsted County, Minnesota between 1970 and 1999. Patients were followed for a maximum of 5 years after their incidence date. Logistic regression and zero-inflated Poisson regression models were used to assess the association between rheumatology care and age, sex, giant cell arteritis (GCA), PMR relapses, corticosteroid complications, comorbidity, and various laboratory findings, adjusting for the total number of visits.

Results

Of the 364 incident cases of PMR eligible for this analysis, 67% were women and the mean age at incidence was 73 years. Over a mean followup of 4.1 years, individuals in this cohort utilized a total of 5,108 physician office visits and 2,015 telephone calls. The mean number of generalist and rheumatologist visits per person-years of followup during the first year of PMR was 7.02 and 2.15, respectively. Thereafter, there was a steady decline in both generalist and rheumatologist visits. One hundred forty-four (40%) patients had no rheumatologist visits and 102 (28%) had only 1 rheumatologist visit, mostly for diagnostic confirmation. Men and patients with several comorbid conditions were significantly more likely to be seen by rheumatologists (P < 0.001). However, once referred, women, older patients, and those with GCA, PMR relapses, and corticosteroid complications had significantly more rheumatologist visits (P < 0.001).

Conclusion

The use of physician services in PMR is considerable. Generalists provide the large majority of care. Rheumatologist involvement is generally limited to diagnostic confirmation and management of complications. The relative paucity of rheumatology care following the period of diagnosis may represent an opportunity for improving the care of patients with PMR.

INTRODUCTION

Polymyalgia rheumatica (PMR) is a common, nonfatal illness of middle-aged and older individuals characterized by aching and morning stiffness in the proximal regions of the extremities and torso, and elevated markers of inflammation, including erythrocyte sedimentation rate (ESR). The etiology of this inflammatory condition is still unknown (1). The prevalence of PMR among persons older than 50 years of age has been estimated at 6 per 1,000 (2). The reported incidence rates from different populations are quite variable, with a clear tendency for higher rates in studies of northern hemisphere populations. The overall age- and sex-adjusted annual incidence per 100,000 population aged ≥50 years in Olmsted County was estimated at 58.7 (95% confidence interval [95% CI] 52.8–64.7) (3).

The management of PMR is challenging because it involves careful and repeated titration of the corticosteroid dose according to the patient's symptoms and ESR, such that the patient is maintained on the lowest possible dose that controls PMR symptoms and minimizes complications (1, 4, 5). Most patients require 1 or 2 years of treatment. However, some patients may have a relapsing course and remain on low doses of corticosteroids for longer periods. Thus, patients with PMR can consume considerable amounts of physician services due to the chronicity of the disease and the difficulties with corticosteroid titration. However, no studies to date examined the practice patterns for PMR as well as the burden of the disease on the health care system. In addition, the role of the rheumatologist in the total care of patients with PMR has not been defined. The objective of this study was to describe the use of generalist and rheumatologist services in a population-based cohort of patients with PMR over the course of the disease and to identify predictors of rheumatology care.

PATIENTS AND METHODS

The study was conducted among the population of Olmsted County, Minnesota. The county is relatively isolated from other urban centers and almost all medical care is provided to the residents by a small number of providers. The characteristics of the local population are similar to US whites, except for the fact that a higher proportion of the working population is employed in the health care industry. All health care providers, including the Mayo Medical Center and the Olmsted Medical Group and its affiliated hospital, utilize a unit medical record system whereby all medical information on each individual is accumulated within a single record. Detailed information on all inpatient, outpatient, and emergency room visits; all laboratory results; pathology reports; correspondence (including telephone calls); and physician visits to nursing homes or private homes are kept in one record. These records are easily accessible because the Mayo Clinic has maintained, since 1909, extensive indices based on clinical and pathologic diagnoses and surgical procedures. These indices were extended, through the Rochester Epidemiology Project, to the records of other health care providers to county residents, resulting in the linkage of all medical records from all sources of care through a centralized system. The potential of this data system for population-based studies has been described previously (6, 7). This centralized system provides an ideal setting for an investigation of health services utilization in PMR because all aspects of health care services could be ascertained directly from the original medical records.

Using this data resource, an inception cohort of all cases of PMR first diagnosed between January 1, 1970 and December 31, 1999 among Olmsted County, Minnesota residents was identified as previously described (3). A trained nurse abstractor screened the medical records of individuals who had received 1 or more diagnoses of PMR, giant cell arteritis (GCA), or temporal arteritis. Individuals were included as PMR cases if they fulfilled the following 3 criteria: 1) age 50 years or older; 2) bilateral aching and morning stiffness (lasting 30 minutes or more) persisting for at least 1 month involving 2 of the following areas: neck or torso, shoulders or proximal regions of the arms, and hips or proximal aspects of the thighs; and 3) ESR elevated to more than 40 mm/hour (Westergren). Patients with suggestive clinical findings who fulfilled the first 2 of the 3 criteria, and who had a prompt response to low-dose corticosteroid therapy were also considered. Except for GCA, the presence of another disease that could explain the symptoms at the time the patient fulfilled the above criteria, such as active rheumatoid arthritis, was considered an exclusion criterion. If the patient was diagnosed with another rheumatologic disease during the course of followup, then followup for that patient was discontinued, but the patient was not excluded from the study. If the diagnosis was questionable, 3 rheumatologists reviewed all the clinical information and reached consensus on the diagnosis. Patients were assessed starting 30 days prior to incidence date until permanent remission of PMR, migration from Olmsted County, or a maximum of 5 years after the incidence date. Patients were considered in permanent remission if they had no PMR symptoms or relapse within 5 years of incidence date while no longer taking corticosteroids or were receiving a stable dosage of ≤5 mg/day, and had normal ESR. Information regarding all inpatient and outpatient physician encounters (including the telephone calls) was abstracted from the patients' complete medical records. Information on the occurrence of GCA and disease relapses was collected for all PMR patients who developed these disorders at any time during the followup period. The diagnosis of GCA was determined according to published American College of Rheumatology diagnostic criteria (8).

Disease relapses were defined as exacerbation of PMR symptoms (as assessed by both the physician and the patient) requiring an adjustment of treatment occurring at least 30 days after the incidence date. In addition, adverse events associated with corticosteroid or nonsteroidal antiinflammatory drug (NSAID) therapy were defined (9). These adverse events included diabetes mellitus (2 readings of fasting plasma glucose levels ≥140 mg/dl, or glucose tolerance test levels ≥200 mg/dl, excluding readings obtained during emergency room or inpatient care), symptomatic vertebral fractures, Colles' fracture of the wrist, proximal femur fracture, aseptic necrosis of the femoral head (the latter 4 clinical findings required radiographic confirmation), posterior subcapsular cataract (diagnosed or confirmed by an ophthalmologist), infections diagnosed by a physician or confirmed by culture (including bacteremia, sepsis, pneumonitis, and other infections excluding urinary tract and viral upper respiratory tract infections), upper gastrointestinal bleeding (endoscopic or clinical diagnosis supported by a drop in the hemoglobin concentration of ≥1 gm/dl), essential hypertension (2 consecutive blood pressure readings of at least 140/90 mm Hg, excluding readings obtained during emergency room or inpatient care), and myopathy (physician's diagnosis supported by documented proximal muscle weakness). Only new diagnoses or incident adverse events occurring after the diagnosis of PMR were included in the analysis. We computed a baseline comorbidity score for each patient as described by Charlson et al (10). For baseline laboratory values of hemoglobin and liver function tests, aspartate aminotransferase (AST), and alkaline phosphatase (AP), we considered the first recorded value in a 4-month window (1 month prior and 3 months following PMR incidence date).

Physician visits, telephone calls, and hospitalizations were summarized using a person-years of observation approach over the initial 5 years of followup for each patient (or until the date of last followup if <5 years). A logistic regression model was used to assess the association between rheumatology care in the first 6 months postdiagnosis and age, sex, GCA, any PMR relapses, corticosteroid complications, baseline ESR, hemoglobin, AP, AST, and comorbidity. A generalized linear model (Poisson distribution with log link) was used to assess the association between the number of rheumatologist visits and the same covariates for the entire (up to 5 years) course of the disease in each patient. An adjustment for the total number of physician visits (due to wide variation between subjects) and the substantial number of patients with no rheumatologist visits over their followup period was accommodated using a zero-inflated Poisson regression model modification for this latter analysis (11). The zero-inflated Poisson regression model was chosen to deal with the overdispersion of the data due to excess number of zero counts, namely individuals with no rheumatologist visits. The measure of overdispersion (deviance/degrees of freedom) was reduced from 3.7 using a usual Poisson regression model to 1.9 using the zero-inflated Poisson regression model.

RESULTS

A total of 378 incident cases of PMR who were Olmsted County residents aged ≥50 years and who first fulfilled the diagnostic criteria for PMR between 1970 and 2000 have previously been identified (3). Of the 378 incident cases of PMR, 364 were eligible for this analysis (14 were excluded because they either subsequently denied research authorization or they had a serious comorbid disease, e.g., cancer, that dominated their medical care). Two hundred forty-four (67%) of these individuals were women and the mean age at incidence of PMR was 73.4 years (Table 1). Median pretreatment ESR was 52.2 mm/hour (range 2–136 mm/hour). Patients were followed for a median of 5.1 years (range 0.12–5.1 years) during which 178 (48.9%) patients experienced one or more disease relapses and 53 (14.6%) patients presented with GCA either at the time of PMR incidence or any time during the observation period. Overall, 310 (85%) patients received corticosteroids with or without NSAIDs over the course of the disease. Initial median corticosteroid dosage was 15 mg/day (range 1–100 mg/day). Fifty-five patients (15.1%) experienced at least 1 corticosteroid-related complication during the first 6 months and 216 patients (59.3%) had at least one corticosteroid-related complication during the entire followup period (Table 1). Overall, <25% of the corticosteroid complications occurred during the first 6 months.

Table 1. Demographic and clinical characteristics of 364 PMR patients*
 Overall (n = 364)
  • *

    PMR = polymyalgia rheumatica; ESR = erythrocyte sedimentation rate; GCA = giant cell arteritis; NSAIDs = nonsteroidal antiinflammatory drugs.

  • At least one corticosteroid complication following PMR diagnosis. Includes diabetes mellitus, vertebral fractures, Colles' fracture of the wrist, proximal femur fracture, aseptic necrosis of the femoral head, posterior subcapsular cataract, bacteremia, sepsis, pneumonitis, other infections, upper gastrointestinal bleeding, essential hypertension, and myopathy.

Sex, no. (%) 
 Females244 (67.0)
 Males120 (33.0)
Age at PMR diagnosis, years 
 Mean, median73.4, 74.1
 Range50.8–95.5
Length of followup, years 
 Mean, median4.2, 5.1
 Range0.12–5.1
ESR at diagnosis, mm/hour 
 Mean, median55.2, 52.0
 Range2–136
Clinical features, no. (%) 
 Aching/morning stiffness363 (99.7)
 Tenderness204 (56.0)
 Anorexia53 (20.1)
 Malaise164 (65.9)
 Weight loss49 (19.1)
Relapses, no. (%)178 (48.9)
GCA, no. (%)53 (14.6)
Corticosteroid complications, no. (%) 
 Within first 6 months55 (15.1)
 Entire followup period216 (59.3)
Treatment, no. (%) 
 NSAIDs alone54 (14.6)
 Corticosteroid alone110 (30.3)
 NSAIDs and corticosteroids200 (55.1)
Initial corticosteroid dosage, mg/day (n = 310) 
 Mean, median19.2, 15.0
 Range1–100

During a median 5.1 years of followup, this PMR cohort utilized 5,108 physician office visits and 2,015 telephone calls. Median percentage of generalist visits per person was 67% (25% and 75% percentiles: 0%, 83%). More than two-thirds of all physician visits (70%) and 61% of rheumatologist visits occurred during the first year of the disease (Figure 1). Thereafter, there was a steady decline in office visits. The mean number of generalist visits per person-year of followup was 7.0 during the first year, 3.8 during the second year, and ≤3 visits after the second year. The decline in the number of rheumatologist visits was much more pronounced, from 2.2 visits per person-year during the first year to 0.8 visits during the second year and ≤0.6 thereafter. Notably, in 144 (39.6%) patients, the disease was managed exclusively by generalists. One hundred two (28%) patients had only 1 rheumatologist visit and most of these visits occurred during the early months following diagnosis. Telephone calls with the provider were common, especially during the first year of diagnosis (Figure 1). Only 8% of telephone calls were with rheumatologists. Two-thirds of patients (244 [67%]) were hospitalized at least once during the followup period. Most of these hospitalizations were for reasons other than PMR and included disorders of the circulatory system (e.g., myocardial infarction, congestive heart failure, hypertension, stroke, transient ischemic attacks), respiratory system (pneumonia, chronic obstructive pulmonary disease, bronchitis), gastrointestinal system, genitourinary system, neoplasms, and fractures.

Figure 1.

Health services utilization over the entire course of disease of 364 Olmsted County, MN residents (≥50 years of age) who first fulfilled the diagnostic criteria for polymyalgia rheumatica between 1970 and 2000.

The majority (201 [92%]) of the total 220 first-time visits to rheumatologists occurred during the first 2 months following diagnosis (Figure 2). Men were referred earlier than women (P = 0.033). Almost 63% of male patients saw a rheumatologist within the first month of PMR diagnosis as compared with 49.2% of female patients. By the end of the sixth month of PMR diagnosis, 66.7% of male patients and 54.2% of female patients had visited a rheumatologist at least once.

Figure 2.

Time to first rheumatologist visit by sex.

We then examined demographic and clinical predictors of rheumatology care using logistic regression and zero-inflated Poisson regression models. Univariate analyses indicated that men (odds ratio [OR] 1.70, 95% CI 1.08–2.68) and patients with a Charlson comorbidity score of >3 (OR 1.83, 95% CI 1.18–2.83) were significantly more likely to see a rheumatologist within the first 6 months of the disease (Table 2). Although not statistically significant, these results also suggested a trend toward younger patients and those with near normal ESR values being more likely to see a rheumatologist. In multiple logistic regression analysis, only the Charlson comorbidity score (P = 0.016, OR 1.74, 95% CI 1.11–2.74) was significantly associated with rheumatology care during the first 6 months postdiagnosis. The association with sex was no longer significant (OR 1.44, 95% CI 0.89–2.32).

Table 2. Univariate predictors of seeing a rheumatologist during the first 6 months of PMR diagnosis and ever seeing a rheumatologist during the entire followup period*
 Seeing a rheumatologist during the first 6 months OR (95% CI)Seeing a rheumatologist during entire followup OR (95% CI)
  • *

    PMR = polymyalgia rheumatica; OR = odds ratio; 95% CI = 95% confidence interval; ESR = erythrocyte sedimentation rate; GCA = giant cell arteritis.

  • ORs and 95% CIs were computed from the estimated coefficients and standard errors in univariate logistic regression models, adjusted for total number of physician visits.

  • ORs and 95% CIs were computed from the estimated coefficients and standard errors in univariate zero-inflated Poisson regression models, adjusted for total number of physician visits.

  • §

    Laboratory values were classified as normal or abnormal according to the following ranges: normal hemoglobin, males 13.5–17.5 gm/dl; normal hemoglobin, females 12.0–15.5 gm/dl; normal aspartate aminotransferase 12–31 U/liter; normal alkaline phosphatase 100–250 U/liter. Values outside these ranges were categorized as abnormal.

Age, per 10 years0.83 (0.65–1.06)0.77 (0.60–0.99)
Male, female is the reference1.70 (1.08–2.68)2.29 (1.39–3.78)
ESR, per 10 mm/hour0.94 (0.87–1.01)0.92 (0.85–1.00)
GCA, yes/no1.24 (0.60–2.54)1.17 (0.61–2.23)
Relapse, yes/no1.09 (0.67–1.76)0.84 (0.53–1.31)
Corticosteroid complications, yes/no1.09 (0.61–1.96)1.02 (0.65–1.59)
Alkaline phosphatase (n = 214)§  
 Normal (n = 168)1.00 (reference)1.00 (reference)
 Abnormal (n = 46)0.97 (0.50–1.91)1.24 (0.60–2.56)
Aspartate aminotransferase (n = 235)§  
 Normal (n = 186)1.00 (reference)1.00 (reference)
 Abnormal (n = 49)1.62 (0.81–3.21)2.10 (0.97–4.51)
Hemoglobin (n = 307)§  
 Normal (n = 147)1.00 (reference)1.00 (reference)
 Abnormal (n = 160)0.83 (0.52–1.31)1.00 (0.62–1.61)
Charlson comorbidity index ≤ 31.00 (reference)1.00 (reference)
Charlson comorbidity index > 31.83 (1.18–2.83)1.90 (1.21–2.99)

In univariate zero-inflated Poisson regression analyses, men (OR 2.29, 95% CI 1.39–3.78) and patients with Charlson comorbidity scores of >3 (OR 1.90, 95% CI 1.21–2.99) were significantly more likely to see rheumatologists over the entire followup period (Table 2). In the multiple-predictor zero-inflated Poisson regression model, male sex (P = 0.018) and Charlson comorbidity score (P = 0.027) remained statistically significant predictors of ever seeing a rheumatologist (adjusting for total number of visits, age, ESR, GCA, relapses, and corticosteroid complications), whereas younger age (P = 0.068) and lower ESR (P = 0.071) were borderline significant.

Men with PMR were significantly more likely to see a rheumatologist compared with women (68% versus 57%; P = 0.001) (Table 3). However, the univariate zero-inflated Poisson regression analyses indicated that once referred to a rheumatologist, women visited their rheumatologists more often (P < 0.001), i.e., among the subset of 220 patients who had at least one rheumatologist visit, the mean ± SEM number of rheumatologist visits for women and men was 3.8 ± 0.4 and 2.7 ± 0.3, respectively (Table 3). Similarly, although GCA, disease relapses, and corticosteroid complications were not associated with ever seeing a rheumatologist (Table 2), patients with these conditions visited their rheumatologists more frequently once they were referred (Table 3). Patients with GCA had a mean of 4.7 rheumatologist visits as compared with 3.1 for those who did not have GCA (P < 0.001). The mean number of rheumatologist visits for patients with at least one disease relapse was 4.0 compared with 2.9 for those with no relapses (P < 0.001). Patients with corticosteroid complications had a mean of 3.8 rheumatologist visits compared with 2.7 among those with no complications (P < 0.001). When considering only the 220 patients with at least one rheumatologist visit in the multiple-predictor variable analysis, the total number of visits to rheumatologists were significantly increased among women (P = 0.015), patients with GCA (P < 0.001), patients with relapses (P < 0.001), and patients with large numbers of corticosteroid complications (P = 0.001).

Table 3. Percentage of patients with rheumatologist visits among 364 incident cases of PMR and mean ± SEM number of rheumatologist visits in those with such visits (n = 220) during the entire followup period*
 Percentage with rheumatologist visits (n = 364)Mean ± SEM number of rheumatologist visits among those with visits (n = 220)P
  • *

    Data were summarized by subgroups of the predictor variables considered in the zero-inflated Poisson regression analysis. PMR = polymyalgia rheumatica; SEM = standard error of the mean; ESR = erythrocyte sedimentation rate (52 mm/hour was observed median); GCA = giant cell arteritis.

  • P values based on univariate zero-inflated Poisson models containing just the corresponding (row) covariate and total number of visits as predictor variables. See text for multiple predictor model results.

  • These variables were used as continuous covariates (predictor variables) in the zero-inflated Poisson regression model analyses.

  • §

    Laboratory values were classified as normal or abnormal according to the following ranges: normal hemoglobin, males 13.5–17.5 gm/dl; normal hemoglobin, females 12.0–15.5 gm/dl; normal aspartate aminotransferase 12–31 U/liter; normal alkaline phosphate 100–250 U/liter. Values outside these ranges were categorized as abnormal.

Age, years   
 50–59683.3 ± 0.6 
 60–69653.0 ± 0.40.628
 ≥70583.5 ± 0.3 
Sex   
 Male682.7 ± 0.3< 0.001
 Female573.8 ± 0.4 
ESR, mm/hour   
 ≤52653.4 ± 0.40.194
 >52563.3 ± 0.3 
GCA   
 GCA at any time during followup684.7 ± 0.8< 0.001
 No GCA ever593.1 ± 0.2 
Relapse   
 Relapse at any time614.0 ± 0.4< 0.001
 No relapse602.9 ± 0.3 
Corticosteroid complications   
 At least 1623.8 ± 0.4< 0.001
 None572.7 ± 0.4 
Alkaline phosphatase§   
 Normal (n = 109)653.4 ± 0.40.941
 Abnormal (n = 32)703.5 ± 0.6 
Aspartate aminotransferase§   
 Normal (n = 118)633.5 ± 0.30.093
 Abnormal (n = 36)732.9 ± 0.6 
Hemoglobin§   
 Normal (n = 94)643.7 ± 0.40.069
 Abnormal (n = 100)623.2 ± 0.4 
Charlson comorbidity index   
 0–3553.6 ± 0.40.022
 ≥4683.1 ± 0.4 

DISCUSSION

The use of physician services in PMR was substantial. The majority (67%) of physician services in this population-based cohort were provided by generalists. Most first-time referrals to rheumatologists occurred within the first weeks of diagnosis. Men and patients with several comorbid conditions were significantly more likely to be seen by rheumatologists. Although men were more likely to be seen by rheumatologists than women, once referred, women had significantly more rheumatologist visits. Interestingly, indicators of PMR disease severity, such as occurrence of GCA, PMR relapses, or corticosteroid complications, were not associated with first-time rheumatologist visits. However, the presence of these conditions was associated with an increased number of rheumatologist visits among those patients with at least one rheumatologist visit. To our knowledge, this is the first study to report the patterns of physician (especially rheumatologist) care in PMR. Our community-based study sets the background for future studies aiming to define the costs and consequences of care in PMR as well as opportunities for optimizing management of these patients.

Our results demonstrate that the majority of physician services for patients with PMR occur at or around the period of diagnosis. Most referrals to rheumatologists occurred during the first 2 months following diagnosis. These visits were likely for confirmation of PMR diagnosis and recommendations regarding treatment. In addition, there was a tendency to refer younger patients and those with low ESR values. Together, these findings suggest that referral to a rheumatologist is driven primarily by diagnostic uncertainty. Clearly, a young patient with near normal ESR poses more diagnostic challenges than an older patient with a high ESR and typical symptoms (12). Despite various diagnostic criteria, the diagnosis of PMR can be challenging, mainly due to atypical clinical features and a long list of differential diagnoses (8, 13–17). A retrospective chart review of 123 PMR patients referred to a tertiary care rheumatology clinic in Canada demonstrated that the accuracy of PMR diagnosis by nonrheumatologists was very low (24%), and that a large number of inappropriate tests were performed by nonrheumatologists in an effort to reach diagnosis (18). These diagnostic difficulties are common in many musculoskeletal diseases (19–21). Early access to a rheumatologist can substantially reduce unresolved diagnostic and therapeutic problems (22–24).

Disease management challenges in PMR are not limited to the diagnosis. Although PMR is generally regarded as a benign disease with no adverse impact on long-term survival (3, 25), management is particularly challenging due to lack of specific disease severity indicators and the relatively high likelihood of disease relapses and serious corticosteroid complications (9, 26). Consequently, there is considerable variation in the management of PMR (27). Although a substantial number of patients in our study (201 [55.2%] within first 2 months) saw a rheumatologist at or near diagnosis, a much smaller proportion (8 [2.2%]) were referred after the initial 6-month period. Almost 40% of the PMR patients were managed exclusively by generalists, and 28% of the patients had only 1 rheumatologist visit, mostly for diagnostic confirmation only. Clearly, for the majority of these patients, rheumatologist involvement throughout the disease course was minimal.

In addition, none of the disease severity-related characteristics, namely relapses, GCA, and corticosteroid complications were associated with rheumatology referrals. However, among the relatively few patients who had access to rheumatology care, those who experienced disease relapses, GCA, or corticosteroid complications visited the rheumatologists more frequently. These findings suggest that such patients required more intense rheumatologic care.

Previous studies of predictors of specialist referrals in a variety of conditions other than PMR revealed various factors associated with referrals. These include patient-related (age [28–30], sex [31–40], race [41], income [42], patient and family requests [43], and insurance coverage [44]), physician-related (age, sex, certification, and experience [45, 46]), and practice-related (size, location [47, 48], fund holding [49], and access to and type of specialty care [50, 51]) factors. In a recent study of musculoskeletal referrals in an academic hospital for knee and shoulder pain, duration of pain, insurance status, and sex were not associated with referral status (52). In our study, men were twice as likely to see a rheumatologist than women, both during the early and later course of the disease. However, once referred to a rheumatologist, female patients visited their rheumatologists more frequently. Studies to date have consistently shown the same pattern of sex differences in access to and use of health care services in various therapeutic areas (31–40), including musculoskeletal diseases (53–57). A study from the Netherlands reported an ∼1-month delayed referral of women with rheumatoid arthritis to rheumatologists, despite similar disease severity in both sexes (56). Women are also less likely to undergo arthroplasty for degenerative arthritis (53–55). However, similar to our findings, women tend to utilize more health care services (58–60). These findings may be due to a number of factors. Earlier referrals for men might be due to more atypical or more severe presenting symptoms of PMR in men than women (61). In fact, in our study, male patients were significantly younger, had less severe pain, less likely to have anorexia at presentation, and had higher hemoglobin values. Other possible explanations include sex differences in preferences (54), higher rates of comorbidity in women than men (59), or even sex differences in the likelihood of seeking help (62, 63).

Our cohort of PMR patients also relied substantially on telephone calls for medical advice. These telephone calls were mainly for corticosteroid dose adjustments. This pattern of care is relatively common in diseases with an extended corticosteroid treatment course (e.g., transplantation, inflammatory bowel diseases), and as demonstrated in previous studies, it demonstrates overall satisfaction with telephone medicine (64, 65). Future studies on the economic burden of PMR should take this important aspect of care into account, because effective telephone communication can improve patient satisfaction and substantially reduce the need for more costly office consultations (65, 66).

In this study, we did not compare the quality of care or outcomes for PMR patients whose main physician was a rheumatologist with those who were cared for by a nonrheumatologist. Although the magnitude of the differences was small, some studies to date suggest that early specialist referrals and frequent visits can improve disease outcomes (67–76). In addition, recent studies have shown less frequent use of pharmacotherapy by nonrheumatologists (77, 78), and inappropriate use of corticosteroid therapy in patients with PMR and other rheumatic conditions (9, 79, 80). Therefore, rheumatologists can contribute significantly to successful management of PMR patients throughout their disease course during which corticosteroid therapy and disease relapses are the main challenges (9, 26).

The strengths of our study include use of a large population-based incidence cohort of PMR patients; diagnostic accuracy using a standardized systematic approach for case ascertainment; complete and detailed data collection through a complete review of medical records including all inpatient and outpatient encounters, procedures, and telephone calls by all health care providers; and the use of novel statistical analysis techniques to account for overdispersion of the data due to individuals with no rheumatologist visits and within-subject correlations. Olmsted County is a relatively isolated community in which nearly all medical care is provided by the Mayo Clinic and Olmsted Medical Center and their affiliated hospitals. Variability of clinical practice due to institutional characteristics is minimal and effective collaboration is the mainstay of the health care system. In Olmsted County, the local population has easy access to health care and, therefore, our results showing opportunities for improving care of PMR patients through increased rheumatology care are likely to be conservative.

Our results must also be viewed in light of the study limitations. First, as some racial and ethnic groups are underrepresented in Rochester, MN, where the population in 2000 was 90.3% white according to the US census data, the results of our population-based study are only generalizable to the US white population. Second, health services utilization in this study represents patterns of care in Olmsted County and may not necessarily be representative of national patterns. For example, most primary care physicians have ready access to specialists on the phone without necessarily generating new specialist office visits. Such informal consultations would not be captured in our study unless recorded in the medical records. The unique strength of our data source is the ability to examine the various determinants of rheumatologist referrals in a single population, from which all aspects of medical care can be optimally ascertained because of the availability of longitudinal data. Finally, we did not take into account various physician factors (i.e., age, sex, certification, experience) (45, 46) or practice-related factors (i.e., size, location, type of specialty care) (47–51) previously shown to impact access to specialty care. The impact of physician- and practice-related factors would be minimal in Olmsted County due to the limited number of providers and the integrated multidisciplinary team approach in clinical practice. Health insurance coverage (30, 44) and income (42) could be related to the association we observed for age. Younger and possibly working patients, with a high proportion employed in the health care industry, would have more flexible health insurance coverage and easier access to specialist care as compared with elderly retired Medicare beneficiaries (44).

In conclusion, we have shown that PMR patients consume considerable physician services and that most patients never see a rheumatologist. Virtually all referrals to rheumatologists are for diagnostic purposes early in the disease course. Although the available evidence suggests that rheumatologists can play a key role in minimizing disease morbidity and treatment-related complications, their involvement late in the disease course is limited only to a small minority of patients. Difficulties associated with the diagnosis of PMR as well as the importance of careful corticosteroid dose titration and the well-documented risks of late complications argue for increased rheumatologist involvement throughout the disease course (9, 18). Future research on the management of PMR should address the impact of rheumatology care on long-term outcomes and quality of life in PMR.

Acknowledgements

The authors wish to acknowledge Margaret Donohue, RN, for performing data abstraction, and Gregory Pond for his help with the statistical analyses.

Ancillary