Care in rheumatoid arthritis (RA) is optimized by involvement of rheumatologists. We wished to determine whether patients suspected of having new-onset RA in Québec consulted with a rheumatologist, to document any delay in these consultations, and to determine factors associated with prompt consultation.
Physician reimbursement administrative data were obtained for all adults in Québec. Suspected new-onset cases of RA in the year 2000 were defined operationally as a physician visit for RA (based on the International Classification of Diseases, Ninth Revision diagnostic codes), where there had been no prior visit code to any physician for RA in the preceding 3 years. For those patients who were first diagnosed by a nonrheumatologist, Cox regression modeling was used to identify patient and physician characteristics associated with time to consultation with a rheumatologist.
Of the 10,001 persons coded as incident RA by a nonrheumatologist, only 27.3% consulted a rheumatologist within the next 2.5–3.5 years. Of those who consulted, the median time from initial visit to a physician for RA to consultation with a rheumatologist was 79 days. The strongest predictors of shorter time to consultation were female sex, younger age, being in a higher socioeconomic class, and having greater comorbidity.
Our data suggest that the vast majority of patients suspected of having new-onset RA do not receive rheumatology care. Further action should focus on this issue so that outcomes in RA may be optimized.
Although rheumatoid arthritis (RA) is the most common inflammatory arthropathy seen by primary care physicians, the diagnosis and treatment of this disorder may be problematic (1–3). Early referral of patients with RA to a rheumatologist is a key part of optimal clinical care (3, 4), and recommendation guidelines have been issued in the US and Canada (5, 6) because prompt use of disease-modifying antirheumatic drugs (DMARDs) can prevent joint destruction. Current opinion is such that RA is considered a reversible disabling condition if appropriate therapy is provided early (7, 8). Rheumatologists are experienced in the use of DMARDs, whereas family practitioners are often less comfortable prescribing these medications (1–3, 9). Despite the demonstrated benefits, many patients do not receive proper treatment early enough to benefit from rheumatologic care (1).
Several other studies have examined consultation with rheumatologists for patients with RA (10, 11), but none have focused exclusively on new suspected cases of RA and described factors associated with consultation. Since access to rheumatologists may be problematic (12) and patients with RA may consult other relevant specialists (e.g., physiatrists, immunologists, and internists) who have experience in managing RA, we also explored consultation with such relevant specialists.
The objectives of this study were to describe whether patients suspected of having new-onset RA consulted with a rheumatologist, to document these patients' delay to consultation with a rheumatologist, and to determine factors associated with shorter time to consultation with a rheumatologist. We examined variations in consultations with respect to geographic differences, the characteristics of the diagnosing physicians, and patient-related differences. We also evaluated rate of consultation and factors associated with consultation with relevant specialists.
MATERIALS AND METHODS
Data were obtained from a physician claims administrative database that includes all residents of the province of Québec. Our study population consisted of all adults age ≥18 years who had a visit to a physician for RA (International Classification of Diseases, Ninth Revision [ICD-9] code 714) in the year 2000 (n = 31,378). We then obtained records of all medical services provided to these persons for the period between January 1997 and June 2003.
All persons who had at least 1 ICD-9 code visit of RA in the year 2000 and no prior visit to a physician for RA in the preceding 3 years (n = 13,237) were termed incident suspected RA cases. Within this group, 10,001 (75.6%) persons had their incident RA diagnosed by a nonrheumatologist (the potential referring physician). The presumption is that a physician who records an RA visit believes that the person has RA, and on that basis may (or may not) recommend consultation with a rheumatologist. Although actual confirmation of RA diagnosis was not considered necessary for our purposes, we also used the validated criteria of MacLean et al (13) in identifying cases of incident arthritis: those with 2 coded visits at least 2 months apart but within a 2-year span, after verifying that there had been no claim in the previous 3 years for this condition.
We described rate of consultation with a rheumatologist and explored factors associated with time to consultation for these persons. Patient-related factors were age, sex, comorbidity, socioeconomic status, and proximity to available services. Comorbidity was characterized by Deyo et al's adaptation of the Charlson comorbidity index (14, 15). Socioeconomic status was based on a validated indicator that utilizes postal code to estimate neighborhood socioeconomic status and provide an ecologic index of material and social deprivation (16). We dichotomized socioeconomic status at the top 2 quintiles versus the lower 3. Proximity to available services was determined according to the density of primary and secondary care establishments and classified as high, moderate, or low depending on the availability of primary and secondary services (both, one, or neither available). Physician-related factors were the physician's sex, years since graduation, and type of specialty (specialist versus general practitioner).
First, we described the characteristics of all persons who had incident diagnostic visits in 2000 (n = 13,237), categorized according to who made the initial diagnosis. Next, we analyzed the data on those patients whose initial diagnosis was made by a nonrheumatologist (n = 10,001). Patients who consulted with a rheumatologist at any time during followup were compared with those who did not consult. We conducted bivariate analyses to determine associations between individual factors and the outcome (consultation with a rheumatologist). The incident RA visit was defined as time zero, and the time until the first visit to the rheumatologist was defined as the event time. Subjects who had not visited a rheumatologist by the end of their followup (i.e., by the end of the study on June 30, 2003, or by the subject's death) were censored at that time.
A multivariable Cox proportional hazards regression model (17) was used to estimate independent effects of the patients' characteristics (sex, age, comorbidity, socioeconomic status, and service availability), as well as the characteristics of the physicians who made the initial (incident) RA diagnosis (sex, years since graduation, and general practitioner versus specialist). We checked the proportional hazards assumption using the method proposed by Grambsch and Therneau (18). However, the conventional Cox model assumes independence of the observations, which may be incorrect if referral time depends on the particular practice style of the physician who made the first diagnosis (the potential referring physician). This would induce correlations between individual patients of the same physician, and ignoring this would affect statistical inference (19, 20). To account for such correlations, we used a novel bootstrap resampling procedure (21, 22) that ensures approximately correct standard errors, confidence intervals (CIs), and Type I error rates in the Cox regression analyses of clustered data (23, 24). Specifically, we used a 2-stage bootstrap procedure in which both physicians and their individual patients were resampled (25).
To investigate possible effect modifications, we used an exploratory approach of forward selection with P < 0.05 entry criterion to select those among the prespecified interactions that improved the model significantly. The following sets of interactions were considered: 1) patient age with patient sex; 2) specialty of physician who made the incident diagnosis (general practitioner versus specialist) with each of the following: patient sex, patient age, service availability, comorbidity, and socioeconomic status; 3) socioeconomic status with each of the following: service availability, patient sex, and patient age; and 4) comorbidity with patient sex.
We also sought to determine whether patients with RA consulted with other relevant specialists, defined as specialists in internal medicine, immunology, or physiatry. We believe that these relevant specialists may also have the skills to manage patients with RA, especially when there is limited access to rheumatologists. Also, in Québec, some internists and immunologists work with rheumatologists in hospital-based rheumatology clinics. We repeated the analyses with the outcome of referral to relevant specialists including rheumatologists, but this time the study population was limited to 9,057 patients whose first diagnosis was made by a physician who was not a rheumatologist or a relevant specialist. Because patients may consult a relevant specialist for a problem unrelated to arthritis, and our interest in this study was consultation for RA, in this analysis we restricted the events to those consultations where the relevant specialist listed an arthritis-related diagnosis (i.e., ICD-9 codes 710–720). We then explored factors associated with prompt consultation to a rheumatologist or relevant specialist using the approach described above.
The characteristics of the 13,237 persons with incident RA visits in 2000 are described in Table 1. The mean ± SD age for the entire group was 57.1 ± 17.1 years. The database included persons who attended their first diagnostic visit with a general practitioner (7,945 [60%]), a rheumatologist (3,236 [24.4%]), a relevant specialist (944 [7.1%]), or another specialist (1,112 [8.4%]). We compared patients who were first diagnosed by a rheumatologist with those first diagnosed by a relevant specialist. The 2 groups were similar in terms of age, but a greater proportion of those first diagnosed by a rheumatologist were in the higher socioeconomic group (P < 0.001) and living in a high service availability area (P < 0.001).
Table 1. Characteristics of 13,237 patients with incident rheumatoid arthritis with a first coded diagnostic visit in 2000
Of the 13,237 total persons, 10,001 were first assigned an RA diagnosis by a nonrheumatologist (with 7,945 [79.4%] of these diagnosed by a general practitioner), and only 2,735 (27.3%) of these 10,001 saw a rheumatologist within the subsequent 2.5–3.5 years (Figure 1). Of the 2,735 patients who consulted with a rheumatologist, 468 (17.1%) received confirmation of the RA diagnosis by the rheumatologist, 396 (14.5%) were diagnosed with osteoarthritis by the rheumatologist, 996 (36.4%) had other arthritis-related diagnoses (e.g., gout, connective tissue disease, ankylosing spondylitis, etc.), and the remaining 875 (32%) had other, nonarthritis-related diagnoses. Mean ± SD time between first diagnostic visit and consultation with the rheumatologist was 189.59 ± 253.12 days (median 79 days, interquartile range [IQR] 28–228). Of the 2,735 patients who were referred, 1,452 (53.1%) were seen by the rheumatologist in the first 3 months, a total of 1,933 (70.7%) were seen within the first 6 months, and a total of 2,265 (82.8%) were seen within 12 months. The consultation patterns to rheumatologists and to relevant specialists are depicted in Figure 2.
Using the 2-coded visit algorithm (i.e., 2 coded RA visits at least 2 months apart within the span of 2 years), 2,675 patients fit the criteria, of whom 2,237 were diagnosed by a nonrheumatologist. Of those 2,237, 692 (30.9%) subsequently consulted a rheumatologist.
Bivariate analyses indicated that patients who consulted with rheumatologists were younger (mean age 55.1 years versus 57.4 years; P < 0.0001) and had a higher comorbidity index (P = 0.02). Consultation rates were higher among women (29.9% versus 22.8%; P < 0.0001), patients in the higher socioeconomic category (30.0% versus 26.2%; P < 0.0001), patients who resided in areas where services were highly available (28.0% versus 24.7%; P = 0.004), and patients who were first diagnosed by a female physician as opposed to a male physician (29.4% versus 26.7%; P = 0.02).
Results of the multivariable Cox regression model of time to consultation with a rheumatologist are shown in Table 2. For each patient and physician characteristic, Table 2 shows the adjusted hazard ratios (HRs) with bootstrap-based 95% CIs. An HR <1 indicates that a given characteristic is associated with longer time to consultation, and an HR >1 indicates a shorter time, i.e., a higher probability of having a consultation during the followup period. Younger patients, women, patients in the higher socioeconomic class level, and patients with greater comorbidity had significantly shorter times to rheumatologic consultation.
Table 2. Cox regression analysis of factors associated with time until consultation with a rheumatologist
Hazard ratio (HR) <1 indicates that a given characteristic was associated with a longer time to consultation, and HR >1 indicates a shorter time. To account for the clustering of patients within physicians' practices, 95% confidence interval (95% CI) was obtained by the bootstrap technique.
P < 0.05, i.e., bootstrap-based 95% CI excludes 1.
Only 1 of the interactions was statistically significant: patient's age and sex (adjusted HR 0.993, 95% CI 0.988–0.998). For example, the HR for consultation with a rheumatologist for a 40-year-old woman compared with a 40-year-old man is 1.538. The HR decreases for 57-year-olds to 1.365, and to 1.211 for 74-year-olds. This implies that although women have a shorter time to consultation across all ages, as patients get older the mean time to consultation for women and men becomes closer.
Consultation with a relevant specialist.
Of the 9,057 incident RA cases in the year 2000 that were not diagnosed by a rheumatologist or relevant specialist, 3,145 (34.7%) consulted with rheumatologists or other relevant specialists over the next 2.5–3.5 years. Of the 3,145 patients who consulted, 1,970 (62.6%) were seen by the rheumatologist or relevant specialist in the first 3 months, 2,441 (77.6%) were seen within 6 months, and 2,734 (86.9%) were seen within 12 months. For those who were seen by a relevant specialist, the mean ± SD time between diagnosis and consultation was 197.68 ± 276.5 days (median 67 days, IQR 21–251).
Bivariate analyses indicated that patients who consulted with relevant specialists were older (P = 0.008), had a higher comorbidity index (P < 0.0001), and were first diagnosed by physicians who graduated more recently (P < 0.0001). Consultation with a relevant specialist was slightly higher among those in the lower socioeconomic level (8.1% versus 5.9%; P = 0.0002) and among those who resided in the low service availability areas (9.8% versus 7.0%; P = 0.03).
Time to consultation with either a rheumatologist or a relevant specialist was significantly shorter for women, younger patients, and those with a higher comorbidity index (Table 3). Once again, the interaction between patient age and sex was statistically significant, i.e., as age increases, the difference in time to consultation between men and women decreases.
Table 3. Cox regression analysis of factors associated with time until consultation to a rheumatologist or relevant specialist*
Relevant specialists are internists, immunologists, or physiatrists. HR = hazard ratio; 95% CI = 95% confidence interval.
HR <1 indicates that a given characteristic was associated with a longer time to consultation, and HR >1 indicates a shorter time. To account for the clustering of patients within physicians' practices, 95% CI was obtained by the bootstrap technique.
P < 0.05, i.e., bootstrap-based 95% CI excludes 1.
Only 27.3% of patients first diagnosed with RA by a nonrheumatologist consulted with a rheumatologist over the next 2.5–3.5 years. This proportion increased to 30.9% when using the more stringent criteria (2-coded diagnostic visits), and to 34.7% when including consultation to relevant specialists. For those who did consult, mean time until consultation was lengthy: approximately 6 months. Nevertheless, it is encouraging that ∼50% of these patients were seen within 3 months, which is considered a benchmark for referral in early arthritis (26–28).
Our results regarding referral to rheumatologists in Québec are similar to the results reported for Ontario (26.4%  and British Columbia (26–48% over 1–5 years ). It must be pointed out that in all of these studies, a case of RA is defined as that coded by a physician in a reimbursement database. We contend that, if a patient receives a coded diagnosis of RA by their physician, the physician probably believes that the patient has RA. Thus, we are evaluating the rate of consultation to rheumatology for those who were diagnosed with RA by their nonrheumatologist physician. One may argue that this rate should be higher if physicians are to follow recommendations of referral to rheumatology for patients with RA (5). On the other hand, this does not include those patients for whom a physician did not code a diagnosis of RA but may have suspected it and therefore referred the patient to a rheumatologist for confirmation of diagnosis or further investigation.
Among those who consulted during followup, median time from incident visit to rheumatology consultation was 79 days in our study, which is longer than the 66 days that was estimated for patients in British Columbia in the study by Lacaille et al (10). Based on information from the Canadian Rheumatology Association, there were 85 rheumatologists in Québec and 52 rheumatologists in British Columbia during the time of our study; the per capita ratio of rheumatologists is somewhat higher in British Columbia (1:75,149) than it is in Québec (1:85,147), which may explain this difference in part. In general, problems with access to rheumatology may be related to the declining number of rheumatologists available to serve the Canadian population (12). Two other studies found median lag times of 16 and 36 weeks from symptoms to referral for patients with confirmed diagnoses of RA (29, 30). A Norwegian study indicated a median lag time of 8 weeks from first physician encounter until rheumatology consultation, which is shorter than what we found, possibly due to the higher per capita ratio of rheumatologists in that region of Norway (1:48,000) (29). An earlier American study covering the time period between 1987 and 1990 (a decade before the benefits of early use of DMARDs and early referral started to be publicized [4, 9]) found a median lag time of 18 weeks for patients belonging to a health maintenance organization with reportedly excellent access to rheumatology care (30).
We found that women tended to consult with rheumatologists more often and had shorter times from diagnosis to consultation. This concurs with other studies (10, 31) and may be indicative of the fact that RA occurs more frequently in women; perhaps physicians are more certain of their diagnosis of RA in women and consequently recommend consultation with specialists more often. Younger persons were also more likely to consult with rheumatologists, as were those with a higher socioeconomic status, which concurs with the report by Lacaille et al (10). Interestingly, our results indicated that age was an effect modifier for sex: the impact of older age on delaying the consultation was stronger for women than it was for men. In addition, persons with more comorbidity were more likely to consult with a rheumatologist or relevant specialist. Presumably these persons have increased contact with specialists, who may be more apt to refer them (32).
Data-based studies such as ours have inherent limitations. These include lack of quality control over diagnostic coding and no indication of the severity of the diagnosed disease. As discussed above, we did not insist on a high sensitivity of diagnosis by the diagnosing physicians. If the diagnosing physician coded a visit as RA, we assume that he or she believed that the patient had RA and thus should have been referred to a rheumatologist. Severity may have an impact on perceived need for consultation; those with more severe disease may be diagnosed sooner and specialized treatment may be considered more urgent (33). Data derived from the administrative databases describe actual consultations with specialists. We cannot know whether these constitute all referrals made, because some referrals may not be carried through by the patient. We were not able to address other issues related to accessibility, such as waiting times to get an appointment or difficulties on the part of patients to attend appointments (e.g., physical disabilities, lack of transportation, inclement weather, etc.). Lastly, we classified persons as having suspected new-onset RA if they did not have a visit coded for RA in the 3 previous years. However, some cases of RA may have been in remission for >3 years, and thus have been erroneously included as incident cases in our cohort. Patients who had been in remission may have been more likely to consult with a rheumatologist because they may have done so in the past. If that is the case, then the rate of consultation among truly incident RA cases in our study may actually be lower than that reported.
The comparison of those who were initially diagnosed by a rheumatologist with those who were diagnosed by a relevant specialist suggests that persons living in moderate to lower service availability areas were more likely to be initially diagnosed by a relevant specialist. Some lesser-populated regions of Québec do not have rheumatologists but may have internists.
There is a low rheumatology consultation rate for patients suspected of having new-onset RA. We believe that physicians should be encouraged to refer all patients suspected of having new-onset RA to rheumatologists in order to optimize care. One concerning possibility is that our results reflect a shortage of rheumatologists in the province of Québec. With increasing workloads for rheumatologists and long waiting lists for appointments, it may be important to adopt shared care models. Shared care of these new-onset patients may improve outcomes and limit disability caused by RA (34). Other potential solutions include both improved medical education and the implementation of good screening techniques to help primary care physicians identify probable cases of RA. In addition, triaging rheumatology consultations according to degree of concern on the part of the referring physician may help patients with inflammatory arthritis receive prompt, appropriate management, thus maximizing outcomes for people with this potentially disabling condition.
Study design. Ehrmann Feldman, Bernatsky, Leffondré, Roy, Zummer, Abrahamowicz.
Acquisition of data. Ehrmann Feldman, Haggerty, Roy.
Analysis and interpretation of data. Ehrmann Feldman, Bernatsky, Haggerty, Leffondré, Tousignant, Xiao, Abrahamowicz.