Influence of discipline of provider and model of care on an arthritis educational intervention in primary care

Authors


Abstract

Objective

To identify both provider and organizational characteristics that predicted outcomes following an educational intervention (9-hour workshop and followup reinforcement activities) developed to improve the management of arthritis in primary care.

Methods

Providers completed a survey at baseline and at 6 months postworkshop, including a case scenario for early rheumatoid arthritis. Providers were asked how they would manage the case and their responses were coded to calculate a best practice score, ranging from 0–7. Two-level hierarchical linear modeling was used to determine which of the measured provider and organizational factors predicted best practice scores at followup.

Results

A total of 275 multidisciplinary providers from 131 organizations completed both baseline and followup surveys. Best practice scores increased by 17% (P < 0.01); however, the mean score at 6-month followup remained relatively low (2.68). Significant predictors of best practice scores at followup were discipline of provider and model of primary care in which they worked (P < 0.05), adjusting for baseline practice scores and clustering of providers within organizations. Physicians, nurse practitioners, and rehabilitation therapists scored higher than nurses, students, and other health care providers (P < 0.01). Physician networks scored significantly lower than providers from multidisciplinary-oriented models of care (P = 0.02).

Conclusion

These results have implications for the education of health professionals and the design of models of care to enhance arthritis care delivery.

INTRODUCTION

Arthritis and related conditions affect more than 4.5 million Canadians ages ≥15 years and in 2005–2006 accounted for close to 9 million physician visits (1). Most arthritis treatment occurs at the primary care level, yet many studies have documented the need for improved arthritis management in this environment (2–5). Challenges include physician delay in referring to specialists (4, 6, 7) and underutilization of nonpharmacologic interventions (6, 8, 9).

The dissemination of clinical practice guidelines (CPGs) has been suggested as one method for improving care delivery. In the past 10 years, multiple CPGs have been developed for osteoarthritis (OA) and rheumatoid arthritis (RA) (10–21). However, numerous reports have documented the failure of primary care providers to implement these guidelines due to lack of relevance in clinical practice and various patient, provider, and organizational characteristics (2, 9, 22, 23).

The Getting a Grip on Arthritis program is one of a handful of interventions designed to improve the uptake of arthritis CPGs in primary care (24). This interprofessional education program includes a 1-day workshop on OA and RA, followed by 6 months of reinforcement activities (25). In 2003, funding was received from Health Canada to implement the Getting a Grip on Arthritis program nationally. Thirty workshops were delivered in rural and urban communities across Canada. Six months after the workshop, providers reported increased confidence and satisfaction in managing arthritis and fewer perceived barriers to arthritis care. Recommendations of arthritis best practices, including referrals to rheumatology and nonpharmacologic interventions, significantly increased by >15% for early RA (26). This is particularly important given new guidelines recommending early referral for RA, preferably within 3 months of symptom onset (19, 27).

The purpose of this current study was to identify provider and organizational characteristics that predict the outcomes of this educational intervention designed to improve the management of arthritis in primary care.

Significance & Innovations

  • This study adds to the sparse literature on the implementation of arthritis best practices for early rheumatoid arthritis (RA) in primary care. It identified provider- and organizational-level characteristics that may influence arthritis care in the community, i.e., the management of arthritis may differ depending on the discipline of the provider and the model of care in which they work. Organizations that included multidisciplinary staff had higher best practice scores than physician-based models. Physicians, nurse practitioners, and rehabilitation therapists scored higher than nurses, students, and other health care providers. Nurse practitioners selected the broadest range of arthritis best practices, reflecting an expanded scope of practice.

  • These results point to the importance of the team in the management of the arthritis and the role that all primary care providers have in influencing the care provided to people with early arthritis.

  • These results could influence the design of future health educational interventions and serve as a model for the dissemination of clinical practice guidelines for other chronic diseases.

  • This study also confirms the need for ongoing educational efforts to improve arthritis care delivery in primary care, particularly given the need for early identification and treatment of RA.

MATERIALS AND METHODS

Intervention.

The design of the national Getting a Grip on Arthritis program has been described in detail elsewhere (25, 28) and is summarized briefly here. A total of 470 organizations known to treat patients with arthritis were identified. Two hundred fifty-four organizations (54%) agreed to participate. Health care providers from participating organizations were invited to a workshop in their community or region. All travel and workshop costs were covered by Health Canada funding.

The program consisted of an accredited interprofessional provider workshop followed by 6 months of activities to both reinforce workshop learning and support the delivery of arthritis care in the community. Multiple approaches to support behavior change were used, including opportunities for skill development, multidisciplinary team learning, networking with local arthritis specialists, use of credible local role models as workshop faculty, personal goal setting, and reinforcement following the workshops (28). The content of the workshop focused on published guidelines for OA and RA (18, 25) adapted for primary care (for the 7 recommended best practices for early RA, see Supplementary Appendix A, available in the online version of this article at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)2151-4658). The workshop consisted of small group hands-on activities with many opportunities for providers to interact with local arthritis specialists. Sessions covered the differentiation between OA and inflammatory arthritis, nonpharmacologic and pharmacologic interventions, and a review of the musculoskeletal examination with trained patient educators (29). A communication module provided an interactive opportunity to discuss the challenges of interprofessional communications and referral (30). At the end of the workshop, providers met in their teams to discuss implementation of best practices within their organizations and communities, returning to their practices with personal goals to improve arthritis care. During the 6 months after the workshop, participants received a number of reinforcements, including newsletters, community resource lists, and educational materials for providers and their patients, and they were also invited to request additional support and resources to help them implement arthritis best practices in their community.

Study sample.

For this study, we used a convenience sample with data collected from the 553 providers from 189 organizations who completed the baseline assessment prior to attending a workshop. Of these, 275 providers from 131 organizations also completed the 6-month followup assessment and therefore were included in this study.

This research is in compliance with the World Medical Association's Declaration of Helsinki (Ethical Principles for Medical Research Involving Human Subjects) and received ethics approval from the University of Toronto, Health Canada, the University of Waterloo, and relevant provincial, regional, university, and hospital ethics boards across Canada.

Measures.

Study outcomes were measured in 2 ways. First, providers' recommendations of arthritis best practices for early RA were measured at baseline (before the workshop) and at 6 months following the workshop using a modified version of the Arthritis Community Research and Evaluation Unit (ACREU) Primary Care Survey (6, 8). Providers were asked to describe the investigations, interventions, and referrals that they would choose for the management of an early RA case scenario (26). Using a coding manual, 2 trained reviewers independently coded their open-ended responses. These coded responses were used to calculate the total best practice score, ranging from 0–7. One point was given for each of the 7 best practices recommended by the provider (see Supplementary Appendix A, available in the online version of this article at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)2151-4658). Second, in this study it was also possible to measure actual behavior related to 1 best practice, i.e., patient education. During the reinforcement phase of the project, staff tracked the number of requests made by an organization for educational materials or for assistance in donating arthritis books and videos to a public library. For each organization, a binary “patient education” variable was derived, where 0 = no requests and 1 = ≥1 request.

The ACREU survey also asked providers to rate both satisfaction and confidence with their ability to manage arthritis on a 10-point scale (where 1 = not satisfied/confident and 10 = extremely satisfied/confident) and to rate presence or absence of barriers to obtaining physical therapy (PT), occupational therapy (OT), social work, or rheumatology care. Barriers were defined as the following: available but waiting time was unacceptable, available but travel time was unacceptable, available but no confidence in the service, available but funding was a barrier, not available at all, or the provider was not sure if the service was available (0 = no barriers and 1 = ≥1 barrier). Age, sex, language, urban status (defined as population density >400 people per square kilometer with a population of at least 1,000), rural status (defined as population density ≤400 people per square kilometer or with a population <1,000), and discipline were also considered as potentially important provider characteristics. Population density was verified through the Statistics Canada web site for community profiles (http://www12.statcan.ca/english/profil01/PlaceSearchForm1.cfm).

Each organization was asked to complete a practice profile that requested information on urban/rural status, region, language, number of physicians, practice size, and model of care (Table 1). Practice size was defined as the total number of full-time–equivalent clinical staff. Additional organizational characteristics included the following: 1) team learning, defined as workshop attendance of ≥2 providers of different disciplines from the same organization (0 = no team learning or 1 = team learning), 2) the percentage of total clinical staff that attended the workshop, and 3) the percentage of total clinical staff who were physicians. Provider ratings of perceived barriers, satisfaction, and confidence for each organization were aggregated to derive an organizational measure of perceived barriers (percentage of respondents within an organization who perceived barriers to arthritis care), as well as to estimate mean baseline satisfaction and confidence of the organization's providers.

Table 1. Description of models of primary health care*
Model of careStaffingFundingPopulation servedIntegration with community
  • *

    CHC = community health center; CSSS = centre de santé et de services sociaux; NP = nurse practitioner; PT = physical therapist; OT = occupational therapist; SW = social worker.

  • Networks are groupes de medecine de famille, family health networks, family health groups, and primary care networks.

  • In Canada, federal jurisdiction for health care is limited to the First Nations and military populations.

CHC/CSSSMultidisciplinary teams providing a broad range of medical and social services, most have physicians and nurses or NPs, some have PTs, OTs, SWs, health promoters, pharmacists, podiatristsProvincial global funding, salaried employees, or capitationDefined geographic location or target group, often socially disadvantaged or hard to serve populationsCooperate with the community to provide range of services, volunteer community board governance
NetworksPhysicians in solo or group practices along with nurses, NPs, and occasionally other providers, team members may be located in different physical locationsProvincial funding, payment to physician through mixed methods (per capita, session fees or fee for service), other staff salaried or contractRostered patientsInformal arrangements with community
FederalAboriginal services: physicians, nurses, NPs, community health representatives; often solo providers Military services: multidisciplinary providers including PT, OT, SWFederal funding, salariedSpecial populations (aboriginal, military)Aboriginal band council, military has varied arrangements with community for secondary and tertiary care services
RegionalMultidisciplinary providers plus fee-for-service physiciansRegional health authority funding, payment to physician mostly fee for service, other staff salaried or contractDefined geographic areaInformal arrangements with community

Statistical analyses.

Descriptive statistics were used to describe provider and organization characteristics at baseline, and nonparametric and parametric tests (Wilcoxon's 2-sample, chi-square test) were used to compare characteristics of study participants and nonparticipants. Statistical analyses were performed using SAS software, version 9.1.

Two-level hierarchical linear regression was used to identify the provider and organizational characteristics that predicted arthritis best practice scores at 6 months postworkshop, controlling for baseline scores and adjusting for potential clustering of providers within organizations. A partial F test was used to examine the statistical significance of the contribution of each predictor variable entered into the model. Backward elimination was used to eliminate variables with a P value less than 0.10. This more liberal P value was used to ensure that no potentially clinically relevant outcomes were missed.

A logistic regression model was used to identify the organizational characteristics predicting an organization's request for patient educational materials. Least squares means analyses were used to estimate the probabilities of requesting educational materials for the predictor variables included in the model.

Sample size.

This study involved secondary analysis of an existing data set and it was not possible to determine sample size a priori. A post hoc power calculation was done to determine whether the sample size was adequate to detect a 15% difference in scores for the early RA case scenario between models of care, adjusting for clustering around sites. The sample size required was calculated as 30 sites per model of care.

RESULTS

Sample characteristics.

This study included 275 providers who completed both the baseline and 6-month followup surveys (responders). As shown in Table 2, the mean age of responders was 42.7 years and the vast majority were women and English speakers. Responders were equally distributed in terms of rural/urban status and represented multiple disciplines as follows: 85 (30.9%) for nurses/licensed practical nurses, 62 (22.6%) for rehabilitation therapists (occupational or physical therapists), 62 (22.6%) for physicians, 30 (10.9%) for nurse practitioners (NPs), and 36 (13.0%) for other health care providers, including pharmacists, social workers, dietitians, and students. Baseline satisfaction and confidence in managing arthritis were moderate (4.9 and 6.2, respectively) and 97.5% of providers reported at least 1 barrier to arthritis care. In general, there were no differences between providers who responded at 6 months and those who failed to respond; however, the percentage of responders in Ontario was higher than in other regions (P < 0.01).

Table 2. Baseline provider characteristics of 6-month responders and nonresponders
Provider characteristics6-month responders (n = 275)Nonresponders (n = 278)P
  • *

    By Wilcoxon's 2-sample test.

  • Measured on a scale from 1–10, where 10 indicates being extremely confident/satisfied.

  • By chi-square test.

  • §

    Five missing.

  • Twenty-two missing.

  • #

    Rehabilitation therapists = physical or occupational therapist. Other = health care providers (e.g., pharmacists, dietitians, social workers), nonclinical staff, and students.

  • **

    One missing.

  • ††

    By Fisher's exact test.

Age, mean ± SD years (range)42.7 ± 9.9 (21–74)43.1 ± 9.6 (22–66)0.58*
Satisfaction with ability to manage arthritis, mean ± SD4.9 ± 2.24.7 ± 2.40.27*
Confidence in managing arthritis, mean ± SD6.2 ± 2.55.8 ± 2.70.15*
Sex, no. (%)  0.29
 Female223 (82.6)§220 (85.9) 
 Male47 (17.4)36 (14.1) 
Language, no. (%)  0.09
 English231 (84.0)218 (78.4) 
 French44 (16.0)60 (21.6) 
Rural/urban status, no. (%)  0.25
 Rural135 (49.1)149 (53.6) 
 Urban140 (50.9)129 (46.4) 
Discipline, no. (%)#  0.53
 Family physicians62 (22.6)52 (18.8)** 
 Nurse practitioners30 (10.9)22 (7.9) 
 Nurses/licensed practical nurses85 (30.9)92 (33.2) 
 Rehabilitation therapists62 (22.6)70 (25.3) 
 Other36 (13.0)41 (14.8) 
Perception of barriers, no. (%)  0.34
 ≥1268 (97.5)270 (98.9)§ 
 None7 (2.5)3 (1.1) 
Region, no. (%)  < 0.01††
 Ontario22 (8.0)3 (1.1) 
 Atlantic65 (23.6)76 (27.3) 
 Prairies91 (33.1)54 (19.4) 
 British Columbia/Alberta63 (22.9)95 (34.2) 
 Quebec34 (12.4)50 (18.0) 

The 275 providers who responded at 6 months came from 131 organizations. Characteristics of these organizations are shown in Table 3. Four different models of primary health care were represented as follows: community health center/centre de santé et de services sociaux (CHC/CSSS) at 59 (45.0%), regional models at 47 (35.9%), networks at 15 (11.5%), and federal models at 10 (7.6%). The mean number of providers per organization was 2.1 (min-max 1–19). The majority (77.1%) of organizations ordered patient educational materials or organized the donation of books to a local library in the 6 months following the workshop.

Table 3. Baseline characteristics of organizations with providers that responded at 6 months and those with providers that did not respond*
VariableParticipating organizations (n = 131)Nonresponding organizations (n = 58)P
  • *

    FTEs = full-time equivalents; CHC = community health center; CSSS = centre de santé et de services sociaux.

  • The 131 organizations represented at first 27 workshops.

  • By Wilcoxon's 2-sample test.

  • §

    By chi-square test.

  • By Fisher's exact test.

Practice size (FTEs), mean ± SD (range)21.9 ± 35.8 (1–192)18.2 ± 31.2 (1–154)0.19
Organization satisfaction, mean ± SD (range)4.9 ± 1.7 (1–9)4.4 ± 2.3 (1–8)0.49
Organization confidence, mean ± SD (range)6.1 ± 1.9 (1–10)5.2 ± 2.8 (2–8.5)0.50
Language, no. (%)  0.34§
 English110 (84.0)38 (65.5) 
 French21 (16.0)20 (34.5) 
Rural/urban status, no. (%)  0.69§
 Rural73 (55.7)38 (65.5) 
 Urban58 (44.3)20 (34.5) 
Perceived at least 1 barrier to arthritis care, no. (%)  1.00§
 Yes130 (99.2)56 (96.6) 
 No1 (0.8)2 (3.4) 
Model of care, no. (%)  < 0.01
 Regional47 (35.9)20 (34.5) 
 CHC/CHSS59 (45.0)19 (32.8) 
 Networks15 (11.5)1 (1.7) 
 Federal10 (7.6)18 (31.0) 
Region, no. (%)  < 0.01
 Ontario14 (10.7)0 (0.0) 
 Atlantic24 (18.3)12 (20.7) 
 Prairie Provinces47 (35.9)5 (8.6) 
 British Columbia/Alberta36 (27.4)33 (56.9) 
 Quebec18 (13.7)8 (13.8) 
Ordered patient educational materials/donated books to library, no. (%)  < 0.01§
 Yes101 (77.1)30 (51.7) 
 No30 (22.9)28 (48.3) 
Team learning, no. (%)  < 0.01§
 Yes74 (56.5)57 (98.3) 
 No57 (43.5)1 (1.7) 

With the exception of networks, where 27% reported having physicians only, all organizations reported having teams and 74 (56.5%) sent providers of more than 1 discipline to the workshops (team learning). CHC/CSSS were more likely to have sent a team compared to the networks, federal organizations, and regional organizations (72.9% versus 13.3%, 50.0%, and 51.1%, respectively; P < 0.01). With one exception, all participating organizations had providers that perceived at least 1 barrier to arthritis care. On average, organization satisfaction and confidence in ability to manage arthritis at baseline were moderate (4.9 and 6.1, respectively).

There were few differences between responding and nonresponding organizations; however, participation rates were higher among those organizations that did not send teams to the workshop and among networks and were lower among federal organizations (P < 0.01). Participation rates also varied by region and were higher in the Prairie Provinces and Ontario (P < 0.01). Not surprisingly, nonresponding organizations requested fewer patient educational materials or the donation of books to a local library (P < 0.01).

Best practice scores for early RA.

At baseline, the total best practice score based on the case scenario was low, with a mean ± SD of 2.28 ± 1.57. At 6 months postworkshop, the mean ± SD score increased by 17% (P < 0.01) but remained low at 2.68 ± 1.74.

Predictors of best practice scores at 6 months postworkshop.

Baseline best practice scores were predicted by provider satisfaction with ability to manage arthritis and discipline, controlling for all other variables in the model (P < 0.10). These baseline differences were accounted for by adding them as additional predictor variables in the development of the final model. The final statistical model is summarized in Table 4. Mean best practice scores at 6 months postworkshop were predicted by providers' discipline and the model of care in which they worked, after controlling for clustering of providers within organizations (P < 0.05).

Table 4. Results of final regression models predicting early rheumatoid arthritis best practices at 6 months postintervention*
Predictor variableEstimateSEDFTPr > |t|
  • *

    CHC = community health center; CSSS = centre de santé et de services sociaux; NP = nurse practitioner.

  • Two-level hierarchical linear regression.

  • Physical and occupational therapists.

  • §

    Logistic regression.

Model 1. Outcome: providers' best practice scores at 6 months     
 Intercept2.250.311267.37< 0.0001
 Model of care     
  CHC/CSSS0
  Federal0.100.431360.230.8184
  Regional0.100.251360.400.6887
  Physician network−0.960.41136−2.330.0213
 Discipline     
  Physicians0
  Nursing−0.890.30136−2.970.0035
  Rehabilitation therapists−0.120.31136−0.370.7085
  NP0.510.371361.380.1698
  Other−1.590.36136−4.44< 0.0001
Model 2. Outcome: organizations' request for educational materials§     
Intercept2.570.441265.82< 0.0001
 Model of care     
  CHC/CSSS0
  Federal−2.841.02144−2.790.006
  Regional−0.480.66144−1.800.074
  Physician network0.911.151440.800.426
 Team learning     
  No−1.130.63144−1.800.074
  Yes0

Providers from the physician networks scored significantly lower than providers from other models of care (P = 0.02), after adjusting for discipline and baseline best practice scores. The network providers scored, on average, almost 1 point lower on the best practice score than the reference model (CHC/CSSS). This means that after the program, providers from the networks recommended 1 fewer best practice (intervention or referral) than providers from the CHC/CSSS.

Further, nursing staff and the “other” health providers group scored significantly lower than physicians, NPs, and rehabilitation therapists (P < 0.01), controlling for model of care. Nursing staff recommended almost 1 fewer best practice on average than did physicians, and the “other discipline” group recommended 1.5 fewer best practices on average than physicians adjusting for other variables. NPs recommended, on average, 0.5 more best practices than physicians (not significant), 1.4 more best practices than nurses, and 2 more best practices than other disciplines after the program.

Figure 1 shows the recommendations of individual best practices at 6 months postworkshop by model of care. In general, the most frequently recommended best practices were PT/exercise, OT interventions, patient education, and referral to rheumatology. Recommendations for nonsteroidal antiinflammatory drugs (NSAIDs) and disease-modifying antirheumatic drugs (DMARDs) were generally low. Providers from the regional models were more likely to recommend OT interventions than other models of care (P < 0.05 by chi-square test). Providers from the networks were more likely to recommend NSAIDs and less likely to recommend exercise and OT interventions (P < 0.05).

Figure 1.

Recommendation of individual best practices for early rheumatoid arthritis by model of care. DMARDs = disease-modifying antirheumatic drugs; NSAIDs = nonsteroidal antiinflammatory drugs; SW = social worker; OT = occupational therapy; PT = physiotherapy; CHCs = community health centers; CSSS = centre de santé et de services sociaux.

Figure 2 shows the recommendations of individual best practices at 6 months postworkshop by provider discipline. Rehabilitation therapists and NPs were more likely to recommend patient education and rehabilitation interventions than other disciplines (P < 0.05 by chi-square test). Nurses, NPs, and rehabilitation therapists were more likely to recommend psychosocial interventions, while NPs and physicians were more likely to refer to rheumatology and to recommend NSAIDs and DMARDs (P < 0.01).

Figure 2.

Recommendation of individual best practices for early rheumatoid arthritis by discipline. DMARDs = disease-modifying antirheumatic drugs; NSAIDs = nonsteroidal antiinflammatory drugs; SW = social worker; OT = occupational therapy; PT = physiotherapy; Other = other health professional/nonclinical staff/students; Rehab = physiotherapist or occupational therapist; NPs = nurse practitioners.

Predictors of organizations' requests for educational materials.

In the logistic regression model, requests for patient educational materials were regressed on baseline mean level of organization satisfaction and confidence, percentage of respondents perceiving barriers to care, percentage of physicians on staff, percentage of staff at the workshop, practice size, model of care, and team learning. Only model of care and team learning remained as predictors in the final model (P < 0.01). After adjusting for team learning, model of care was a significant predictor of the request for educational materials with the estimated probability of organizations requesting materials being significantly lower for federal organizations (30%) compared to the other 3 models of care (≥82%; P < 0.01).

After adjusting for model of care, team learning was a significant predictor, with organizations that sent a multidisciplinary team to the workshop being more likely to request educational materials than those that did not (P = 0.07).

DISCUSSION

The Getting a Grip on Arthritis program was successfully implemented in a variety of primary care settings across Canada and modestly improved providers' recommendations for arthritis best practices for early RA at 6 months following the workshops. This study involved analyses of data collected as part of a before-and-after study, which allowed us to identify provider- and organization-level factors that predicted recommendations of arthritis best practices following the program. Discipline of the provider and the model of care in which they worked predicted total best practice scores at 6 months. Physicians, NPs, and rehabilitation therapists scored higher than nurses, students, and other health care providers (P < 0.01). Providers from traditional physician networks had lower best practice scores than providers in the multidisciplinary CHC/CSSS. On average, providers from networks recommended 1 fewer best practice (intervention or referral) after the program than providers from the other models of care. In particular, they were less likely to recommend exercise and OT interventions. It is unclear to what extent this more traditional physician-centered model of primary care might explain the lower overall best practice scores and fewer recommendations for nonpharmacologic interventions. It might also be that physicians wait to see the response to medications before referring to other services. However, it should also be noted that due to the small number of networks participating, the study might have been underpowered to detect differences in scores between the networks and other models of care.

The Getting a Grip on Arthritis educational program may have been more relevant for CHC/CSSS because it was developed originally through a collaborative effort with the CHCs in Ontario (25) and the CHC/CSSS were represented on the Partners' Group that designed the national program. There is a paucity of research examining the relative benefits of different models of care; however, Abelson and Lomas (31) found that CHCs were more likely to deliver health promotion programming and to use nonphysicians to deliver care than fee-for-service practices. This might be due to provider characteristics, the mix of disciplines employed in the CHC model, or the fact that the providers are salaried, giving them more time to address a broader range of patient issues. This has implications for policy makers in terms of funding and staffing of primary care models for chronic disease management. For instance, the current family health teams in Ontario, some of which grew out of the physician networks involved in this project, now have designated funding for other disciplines on their teams.

Physicians, NPs, and rehabilitation therapists had higher scores than nurses and other disciplines. For NPs and physicians, these differences likely reflect differences in scope of practice, since physicians and NPs are able to directly address the pharmaceutical management of patients and access specialists. It is noteworthy that NPs scored modestly higher than physicians, likely reflecting their more frequent use of nonpharmacologic interventions.

The overall low recommendations for the use of NSAIDs and DMARDs may be problematic given that guidelines recommend the initiation of these medications within 3 months of symptom onset (19, 27). However, it may not be so problematic if rheumatology referrals are made early, services are locally available, and wait lists are reasonable. Additional educational interventions or alternate strategies may be required to support the potentially important roles of rehabilitation therapists and nurses in advocating for early referral to specialists.

Model of care and team learning predicted the likelihood of an organization requesting patient educational materials during the 6-month followup period. These results highlight the importance of the team in the management of arthritis since it may be unreasonable to expect physicians to deal with the broad range of issues associated with complex diseases like arthritis (31). Indeed, studies in arthritis have demonstrated that multidisciplinary team care can improve function, psychosocial outcomes, and overall health, as well as decrease pain and disability (32, 33).

Most providers (97.5%) perceived barriers at baseline; this lack of variability in the response may explain the failure of this variable to be a predictor. However, perceptions of barriers did decrease significantly following the workshop (26). Educational interventions may not be enough. Other issues may still need to be addressed such as the overall shortage of rheumatologists and the lack of publicly funded rehabilitation (34). To understand this better, qualitative studies might further explore discipline-specific barriers to the implementation of arthritis best practices.

This study had several strengths. It was the first study to evaluate the provider- and organizational-level factors that might influence the outcomes of an interprofessional educational program for the implementation of arthritis CPGs in primary care, building on the results of an Ontario pilot study that evaluated the program in CHCs only (25). The current study involved additional models of care, therefore extending the generalizability of the pilot study results and adding to the sparse literature on the evaluation of different funding and organizational models (35, 36). It was also the first study to examine the potential differences in outcome based on the model of care in which a provider works. The current study also examined differences that exist in the management of arthritis by discipline, which suggests areas for continuing educational interventions, e.g., the nurse's role as advocate for referral to specialists.

Future studies need to also consider other factors not available in this data set such as provider beliefs and attitudes (37), as well as system- (38) and patient-level factors that might also influence these outcomes. For instance, Nutting et al (39) found that 75% of identified barriers to depression care were related to patient factors such as patient resistance, noncompliance, and psychosocial factors. In other chronic diseases, factors such as patient beliefs and attitudes (40, 41) and patient preference (42) have been shown to influence providers' ability to implement best practices.

Interpretation of the results of this study must be done cautiously. Best practice scores were based on providers' responses to a written case scenario and therefore may not reflect actual practice, underestimating or overestimating provider performance. Jones et al (43) identified 11 articles where written case scenarios were assessed for criterion validity; they concluded that it was unclear whether written case scenarios could be used as proxies for actual performance. As indicated earlier, the Getting a Grip on Arthritis case scenario scores have been validated by assessing a subset of patients of providers who participated in the program (25). These patients reported receiving more information about arthritis and more referrals were made to rehabilitation services in the community compared to a control group. In this current study, there was evidence that providers from CHC/CSSS, regional organizations, and networks acted on at least 1 best practice by ordering more educational materials for dissemination to patients compared to federal organizations. Federal organizations included 2 special populations, aboriginal communities and military organizations, and were not specifically targeted in this project. Therefore, this finding might be due to the lack of relevance or cultural appropriateness of the materials. As well, the number of federal organizations that participated was small and the number of dropouts at 6 months was higher, making it possible that the study was underpowered to detect changes in this group. Further research is needed to better understand the specific learning needs of these special populations and to develop an intervention specific to their needs.

As seen in other studies evaluating educational interventions for the implementation of arthritis CPGs (24), a large number (46%) of sites did not volunteer to participate in the program and 50% of the providers were lost to followup at 6 months. It may be that this program is only effective for motivated providers who participate fully in the program. It has also been suggested that low burden placed on providers to collect data and financial incentives for completion of the data collection process might improve provider retention in studies (44). Future studies might focus on understanding organizational constraints to participation in research and test strategies to successfully engage organizations and to recruit and retain providers.

This study adds to the sparse literature on the implementation of CPGs for early RA in primary care and showed that the management of arthritis may differ depending on the discipline of the provider and the model of care in which they work. Results suggest that arthritis care might be improved if the model of care includes multidisciplinary providers and points to the importance of the team in the education of people with early arthritis. These results could influence the design of future continuing health educational interventions and serve as a model for the dissemination of CPGs for other chronic diseases. This study also confirms the need for ongoing educational efforts to improve arthritis care delivery in primary care, particularly given the need to identify and treat RA early.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Lineker had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Lineker.

Acquisition of data. Lineker.

Analysis and interpretation of data. Lineker, Husted, Brown.

Ancillary