Abstract
Objective. The Research on Arthritis in Canadian Children Emphasizing Outcomes (ReACCh Out) cohort is a prospective inception cohort of patients with newly diagnosed juvenile idiopathic arthritis (JIA) seen in 16 Canadian pediatric rheumatology (PR) centers. We used data from this cohort to explore factors associated with longer time from symptom onset to the first visit to (PR), and with longer time from first visit to a diagnosis of JIA.
Methods. We included children enrolled in ReACCh Out within 6 months of JIA diagnosis, for whom the dates of symptom onset and first PR visit were recorded. We used Cox proportional hazard modeling to investigate the effects of history, physical examination, and laboratory evaluation on the interval from JIA symptom onset to first PR assessment.
Results. In total, 319 children from the cohort were included. Having a fever (hazard ratio 1.80, 95% CI 1.10, 2.93), any part South Asian ethnicity (HR 1.75, 95% CI 1.04, 2.95), highly educated parents (HR 1.69, 95% CI 1.18, 2.44), and limp (HR 1.55, 95% 1.16, 2.06) were significantly associated with shorter time from symptom onset to first PR assessment, while a history of heel pain or enthesitis (HR 0.61, 95% 0.38, 0.97) was significantly associated with a longer time to first PR visit.
Conclusion. Children with a history of a fever, limp, any part South Asian ethnicity, or highly educated parents were more likely to see PR sooner than patients without these features, while children with a history of enthesitis received PR care later than those without enthesitis.
Chronic childhood inflammatory arthritis is a common pediatric rheumatic disease that affects from 16 to 150 per 100,000 children less than 16 years of age1, yet children who develop this condition often are not recognized by primary and community healthcare providers, and frequently take a long time to be seen by a pediatric rheumatologist. Recent studies have shown median time from arthritis symptom onset to first pediatric rheumatology (PR) assessment was 13 weeks in Tübingen, Germany2, 20 weeks in the United Kingdom (UK)3, 24.4 weeks in British Columbia, Canada4, and a mean of 39.9 weeks in Australia5. Although definitive guidelines for referral times do not exist in Canada, the Alliance for the Canadian Arthritis Program (an alliance of arthritis patients, caregivers, volunteer associations, healthcare providers, researchers, government, and industry) recommends that new patients with inflammatory arthritis should be identified and treated within 4 weeks of seeing a healthcare professional6. In the UK, guidelines recommend all patients with suspected inflammatory arthritis should be seen by a rheumatologist within 6 weeks of referral from a general practitioner7. Factors associated with a longer time to assessment by a pediatric rheumatologist have been described in the UK8, and comparable data are emerging for children in Canada9.
The International League of Associations for Rheumatology (ILAR) classification system for juvenile idiopathic arthritis (JIA) is the most recent classification system for chronic childhood inflammatory arthritis10. We investigated factors associated with a longer time from arthritis symptom onset to first visit to a pediatric rheumatologist in children diagnosed with JIA, as well as those associated with longer time from first PR visit to diagnosis of JIA. We used data from a national prospective cohort of Canadian children with newly diagnosed JIA, the Research on Arthritis in Canadian Children Emphasizing Outcomes cohort (ReACCh Out). The process leading to a diagnosis and enrolment into the study cohort was divided into 2 models: the first used information available prior to the first PR assessment to examine factors associated with a longer time to the initial PR visit, and the second used information available at this initial assessment to examine factors associated with a longer time from first PR visit to a definitive diagnosis of JIA. We included only variables that we felt could conceptually influence the time intervals being examined in each particular model, as these variables could potentially be the focus of efforts to reduce the time from arthritis symptom onset to the first PR visit.
MATERIALS AND METHODS
Study population
The ReACCh Out cohort is a multicenter inception cohort of newly diagnosed children with JIA being followed by pediatric rheumatologists in 16 Canadian centers. Data collection for this cohort is described in detail elsewhere11. Briefly, the cohort includes patients diagnosed with JIA according to ILAR criteria10 within 12 months prior to enrolment11. Standardized data collection occurs at enrolment, every 6 months for 2 years, and then yearly. Information collected includes demographics, family and patient histories, medication histories, as well as laboratory data available at enrolment and followup. Additionally, the juvenile rheumatoid arthritis (JRA) core set measures12, the Juvenile Arthritis Quality of Life Questionnaire13, and the Quality of My Life Questionnaire14 are completed at study visits. This study included only those subjects enrolled into ReACCh Out within 6 months of JIA diagnosis between January 1, 2005, and June 30, 2007. Additional criteria for inclusion into this study: (1) the month and year of symptom onset and (2) the date of the first PR assessment. JIA subtype diagnoses were checked against criteria data submitted by site investigators for each patient as described11.
Statistical modeling
We used multivariable regression with Cox proportional hazard modeling to model the primary and secondary outcomes of interest: days from symptom onset to first PR assessment (model 1, primary outcome), and days from first PR assessment to definitive diagnosis of JIA (model 2, secondary outcome). As only the month and year of symptom onset were collected, the 15th day of the month was imputed as the day of symptom onset. We chose the 15th of the month and not the 1st of the month in order to avoid artificially increasing subsequent time intervals, which would increase potential bias. If the imputation resulted in a negative number of days from symptom onset to first PR assessment, then the first day of the month was imputed as the day of symptom onset.
All variables that had a conceptual rationale for inclusion were forced into the regression model. Rationale and categorization of these variables are listed in Table 1. Independent variables included in model 1 were clinical history (joint pain, joint swelling, limp, morning stiffness, enthesitis, fever, and systemic rash), family history of inflammatory arthritis, demographic information (age, sex, ethnicity), socioeconomic information (highest level of parental education), and distance to the PR center (based on the first 3 digits of the postal code). Independent variables in model 2 included clinical history (enthesitis, fever, systemic rash, psoriasis), physical examination findings at baseline (enthesitis, psoriasis, systemic rash, number of active joints), and laboratory features at enrolment [HLA-B27 status and rheumatoid factor (RF) status], as well as family history (psoriasis, inflammatory bowel disease, ankylosing spondylitis, and uveitis). Antinuclear antigen was not included as it does not influence the clinical diagnosis of arthritis, nor subtype classification.
We used the hazards ratios (HR) as measures of strength of association between dependent and independent variables. HR were considered statistically significant if p values were less than 0.05. The suitability of the assumption of proportional hazards was verified using the Schoenfield test. Although each participating province was initially part of model 1, no HR for the provinces were significant, and province was excluded from the final model to maintain an adequate number of events per variable.
Statistical analyses were performed using Stata version 10.1 (Stata Corp., College Station, TX, USA).
RESULTS
A total of 437 patients were enrolled during the period of time covered by this report, and 356 met the criterion for enrolment within 6 months after diagnosis. Of the latter a further 35 patients were excluded because of missing dates of symptom onset or first PR visit. Patients excluded were similar to included patients in age at symptom onset, sex and subtype distributions, interval from symptom onset to diagnosis, and distance from PR center (data not shown). Thus a total of 319 patients were included in this study, and patient characteristics are listed in Table 2. The JIA subtype distribution was as follows: 41.1% oligoarticular JIA, 21.3% RF-negative polyarticular JIA, 3.4% RF-positive polyarticular JIA, 6.3% psoriatic arthritis, 10.0% enthesitis related arthritis, 7.5% systemic arthritis, and 10.3% undifferentiated JIA.
The distribution of the time from symptom onset to the first PR assessment is shown in Figure 1. Overall, the median time from symptom onset to first PR visit was 115 days [interquartile range (IQR) 45, 219]. There were 52 patients for whom time from symptom onset to first PR visit was > 365 days (1 year). The median time from first PR visit to diagnosis of JIA was 0 days (IQR 0, 20), indicating that JIA was frequently diagnosed on the day of the first rheumatology assessment (Table 3). The median distance travelled to a PR center was 38.2 km (IQR 17.9, 117.3). Although the majority (59.2%) of patients lived within 50 km, 130 (40.8%) lived > 50 km, 68 (21.3%) > 150 km, and 49 (15.4%) > 300 km from a PR center.
HR of 1 indicates equal likelihood of first PR visit in the presence of the variable in question compared to its absence, while HR > 1 indicates increased likelihood, and < 1 indicates a reduced likelihood. Cox proportional hazard modeling of time from symptom onset to first PR visit in the 272 patients with complete data for all variables showed significant HR > 1 for a history of fever (1.80, 95% CI 1.10, 2.93), any part South Asian ethnicity (HR 1.75, 95% CI 1.04, 2.95), highest parental education at a university or postgraduate level (HR 1.69, 95% CI 1.18, 2.44), and history of a limp (HR 1.55, 95% CI 1.16, 2.06); while the HR was < 1 for a history of heel pain or enthesitis (HR 0.61, 95% CI 0.38, 0.97), as shown in Table 4. The likelihood ratio test for this model was 53.37, p < 0.0001, 18 df, with 15 events per variable.
The second Cox proportional hazard model examined the time from first PR assessment to a final diagnosis of JIA. Only patients whose diagnoses were not made at the first PR assessment were included in this model. Rheumatoid factor and HLA-B27 status were excluded from the model because these tests were not available for each patient. None of the HR in the reduced model were significant, although there was a trend toward a shorter time from first PR visit to a diagnosis of JIA with a family history of inflammatory bowel disease (HR 0.62, CI 0.38, 1.03; p = 0.064).
DISCUSSION
Children with JIA take a long time to see a pediatric rheumatologist in a variety of healthcare systems, with median or mean times from symptom onset varying from 13 to 40 weeks2,3,4,5,8. Identifying risk factors associated with a longer time to first PR visit will help focus efforts to minimize this time interval for children with JIA. The median interval of 115 days in our study (16.5 weeks) is within the variation previously reported2,3,4,5,8. The time from symptom onset to the first assessment by a pediatric rheumatologist can be divided into several components: the interval from symptom onset to presentation to a primary healthcare provider; time from first visit to a primary healthcare provider to referral to a pediatric rheumatologist; and time between referral and first assessment by a pediatric rheumatologist. In our study the variables assessed were primarily historical and could be related to both initial phases. For example, systemic symptoms, pain, functional limitations, cultural, and socioeconomic status are variables that may affect both a patient’s presentation to a primary healthcare provider and the time to referral. In our model we used ethnicity as a proxy for cultural influences and parental education as proxy for socioeconomic status.
The observation that heel pain increased time to first rheumatology assessment suggests that heel pain may be underrecognized both by families and by primary healthcare providers as a possible manifestation of enthesitis related arthritis. Other diagnoses such as Sever’s syndrome or Osgood Schlatter disease have findings that are often indistinguishable from enthesitis due to enthesitis related arthritis, especially in the absence of arthritis. This similarity may be a cause for the delay.
Our analysis showed patients with some South Asian ethnicity were seen sooner by a pediatric rheumatologist. This was true when holding factors such as education (our proxy for socioeconomic status) and distance from PR center constant. There are several possible explanations for this trend. First, as hypothesized (Table 1), it is possible that families with South Asian ethnicity have cultural expectations that lead them to seek earlier care. Second, it is possible that children with South Asian ethnicity have an overrepresentation of more severe types of arthritis, as differences in relative frequencies of subtypes have been reported15. We did not perform a statistical comparison of subtypes of JIA diagnosed in this ethnic group compared to the rest of the study population due to the small number of patients with some South Asian ethnicity (n = 21). We did not have data about the number of joints affected prior to the initial PR visit, and we cannot exclude an association between ethnicity and the number of joints involved between symptom onset and initial PR assessment15. Third, it is also possible that patients with South Asian ethnicity tend to reside in urban or suburban areas, thus facilitating access to subspecialty care. Although our model included distance to the PR center as a proxy for access, it was not able to distinguish reliably between urban and rural areas. The effect of ethnicity on accessing healthcare for children with arthritis requires further study, especially in a culturally diverse country such as Canada. The observation that highest parental education at a university level also decreased the delay is in keeping with reports of earlier referral of children with higher socioeconomic status9 and adults with higher education16 to rheumatologists.
Other independent variables that may affect the length of time to referral to PR originate from the primary healthcare provider’s assessment, the provider’s knowledge about the condition being considered, and accessibility of specialist services. Direct information about primary healthcare provider knowledge was not collected in this study; but studies have suggested that the source of referral may affect referral time3,8. In our study, distance to PR center had no effect on time to first PR assessment, but due to our sample size a moderate effect of distance cannot be definitively excluded (hazard ratio 0.82, 95% CI 0.47 to 1.41 for distances more than 150 km). This contrasts with the experience of one center in Germany, where greater distance to the regional PR center increased delay to referral2. In another study, local availability of rheumatologists in Ontario was also an important factor associated with utilization of specialist services for musculoskeletal complaints in adults17. Our findings are remarkable since a considerable proportion of patients, 21.3%, lived more than 150 km from a PR center, and the median distance to a PR center was 38.2 km (IQR 17.9, 117.3), which is similar to the median of 38.8 km in the German study2. Although some of our patients are followed in outreach clinics in British Columbia and Saskatchewan, none of the patients in this study were seen for their initial assessment in traveling clinics. This observation suggests that accessibility to the network of PR clinics in Canada is not impeded by travel for patients who are referred to subspecialty care. However, patients who are not referred to a pediatric rheumatologist could not be included in this study, and therefore a possible bias cannot be excluded.
The final interval prior to PR assessment is the time from referral to the first visit with a pediatric rheumatologist. The duration of this interval may be affected by the information received during the triage process and subspecialist availability; but this information was not identified in our study.
Our analysis differs from that reported by the Childhood Arthritis Prospective Study (CAPS), a multicenter cohort of children with new onset inflammatory arthritis in the UK8. In the CAPS study, disease activity and demographic characteristics at enrolment were assessed, but only erythrocyte sedimentation rate (ESR) had a significant effect on duration of symptoms prior to the first PR assessment in their regression model of the entire cohort, and both lower ESR and older age at symptom onset had significant effects when analysis was limited to children with oligoarticular JIA8. Our first model focused specifically on features available prior to the first PR assessment as these were variables that may have initiated a referral and were available for analysis.
Our second model, in which we included historical features and physical examination findings at enrolment to model factors associated with a longer time from PR assessment to a definitive diagnosis of JIA, failed to show any significant hazard ratios. A high number of patients received their JIA diagnosis at the first PR visit and were therefore excluded from the model. We did not have enough patients with HLA-B27 and RF status in the remaining patients to include these in our model as these tests were not part of our study protocol, but rather were performed selectively to assign JIA subtype diagnoses. Consequently they were not available for the entire study population.
Our distribution of JIA subtypes is similar to that in studies that used the ILAR classification criteria for JIA, although our proportion of patients diagnosed with oligoarticular JIA is closer to the lower end of the reported range of 46%–51%3,8. Although unlikely, we cannot exclude the possibility of a referral bias affecting the proportion of our study patients with oligoarticular JIA. Since only 67% of our patient population were of European background, it is also possible that the multiethnic distribution of our patients affected the proportions of ILAR subtypes15.
We used data collected for an inception cohort of patients with JIA that focuses on outcomes and was not originally designed to track referral processes. Accordingly, some data that might have permitted an examination of causes for delayed referral were not collected. For example, date of referral and data relating to the primary healthcare provider’s assessment and investigations were not identified in our study. Our study included only patients ultimately seen at a PR center. Although Canada has a universal healthcare system and PR centers are available in all provinces, we cannot rule out that selection bias may be present in our results. We did not collect data about the number and types of healthcare providers seen prior to the initial PR visit, and therefore cannot comment on the pathways to care followed by these patients. Despite these shortcomings our study identified previously unreported variables that affected time from symptom onset to first PR assessment. These included history of limping and high parental education, which decreased time to assessment, and a history of heel pain, which lengthened this time interval.
Children with JIA are at risk for complications such as uveitis, erosions, growth abnormalities, and flexion contractures18. Although acceptable length of time of symptom duration prior to first PR assessment has not been universally established, the longterm effects of untreated disease can be devastating, and current standard practice includes early treatment of JIA18. Our preliminary analyses suggest that patients with longer delay to inital PR assessment have greater disability and poorer quality of life during their early disease course19. Further followup of this cohort will be critical to answer questions about the longterm effects of longer time to the first PR assessment.
In summary, in our study of patients from the ReACCh Out cohort, patients with traditional features associated with inflammatory arthritis, such as a limp and fever, were seen by a pediatric rheumatologist sooner than patients without these symptoms. The children of parents with a university degree or postgraduate training were also more likely to be seen sooner, suggesting that these parents may be either more aware of abnormalities or better able to navigate the healthcare system and advocate for their children. It is less clear why patients with South Asian ethnicity had shorter times from onset of symptoms to first PR assessment. Patients with enthesitis, on the other hand, were seen later than those without enthesitis. Future advocacy and education efforts that include promoting awareness of the presenting symptoms and signs of JIA will hopefully improve access to PR care for these patients. Our results suggest the possibility that such advocacy and awareness of JIA may be more important than physical barriers to accessibility, such as distance to a PR center, at least in Canada. Prospective cohort studies such as ReACCh Out and CAPS will continue to provide important data about the short- and longterm influence of delays in diagnosis on eventual outcomes.
Acknowledgment
We thank Narayan Sainaney for data analysis, Michele Gibbon for project coordination, and the following for substantial contributions to acquisition of data: Coordinating Center: McGill University, Montreal, Quebec: Sarah Campillo, Gaëlle Chédeville, Karen Duffy, Beth Hazel, Rosie Scuccimarri. Participating Institutions: Dalhousie University, Halifax, Nova Scotia: Adam Huber, Bianca Lang, Suzanne Ramsay, Elizabeth Stringer; McMaster University, Hamilton, Ontario: Maggie Larché, Peter Dent; Memorial University, St. John’s, Newfoundland: Paul Dancey; Université Laval, Ste. Foy, Quebec: Anne-Laure Chetaille, Jean Dorval; Université de Sherbrooke, Sherbrooke, Quebec: Gilles Boire, Allessandra Bruns; Université de Montreal, Montreal, Quebec: Claire St-Cyr; University of Alberta, Edmonton, Alberta: Janet Ellsworth, Claire LeBlanc; University of British Columbia, Vancouver, British Columbia: Roxana Bolaria, David Cabral, Katherine Gross, Kristin Houghton, Ross Petty, Stuart Turvey; University of Calgary, Calgary, Alberta: Nicole Johnson, Paivi Miettunen, Robert Rennebohm, Heinrike Schmeling; University of Ottawa, Ottawa, Ontario: Johannes Roth; University of Saskatchewan, Saskatoon, Saskatchewan: Alan Rosenberg; University of Toronto, Toronto, Ontario: Suzanne Benseler, Bonnie Cameron, Brian Feldman, Ron Laxer, Rayfel Schneider, Earl Silverman, Lyn Spiegel, Shirley Tse. The authors also thank the patients in the ReACCh Out study for their generous participation and the research assistants at each center.
Footnotes
-
Supported by grants from the Canadian Institutes of Health Research (69301QNT) and the Fast Foundation. Dr. Shiff is supported by the Clinician Investigator Program at the University of British Columbia, and CIHR Frederick Banting and Charles Best Canada Graduate Scholarships Master’s Award. Dr. Yeung is supported by an Investigator Award from The Arthritis Society. Dr. Duffy is supported by The Montreal General Hospital Foundation Scholarship Award.
- Accepted for publication June 16, 2010.