What is the impact of primary care model type on specialist referral rates? A cross-sectional study

A key component of a primary care physician’s role involves acting as a gatekeeper to medical specialists and other health resources to ensure that patients receive appropriate and timely care. Studying referrals is of importance to policy makers and healthcare professionals, due to their significant effects on healthcare costs, quality of care, patient safety, and access to care [1, 2]. As such, having a clear understanding of how system level reforms impact specialist referrals is important.

Over the past two decades, many nations worldwide have initiated reforms to improve their delivery of primary health care. In Canada, the healthcare system is funded through public funds and is freely available to the population with the administrative responsibility of providing the healthcare services devolved to the individual provinces. Several provinces have recently redesigned their primary health care systems with the goal of improving patient access, preventive care, chronic disease care, and coordination with other health services [3].

In Ontario, Canada’s most populous province, the primary focus of reforms has been on the development of interdisciplinary health care teams and on shifting away from the traditional fee-for-service (FFS) physician remuneration model to a capitated payment approach [4]. Through capitation, providers receive a sex and age adjusted fixed payment per patient, which is independent of the number of services performed. Two types of capitation models have emerged with one being associated with funded interdisciplinary teams, and the second one without these teams. The primary care physicians were able to choose and apply to participate in either model, although limitations on the team funding restricted the total number of capitation interdisciplinary teams approved each year in Ontario. These models developed historically and now involve three quarters of the Ontario population. This is a natural experimental environment to understand the impact of primary care reforms on the health system as a whole. Previous studies comparing the delivery of care across the different models in Ontario have demonstrated differences in areas such as health promotion, chronic disease care, and health equity [5‐7]. Few studies have examined the impact of different models on other health care sectors such as specialty care.

The aim of this study is to examine differences in specialist referral rates amongst three main primary care models in Ontario, Canada: 1) Enhanced FFS, 2) Capitation- Non-Interdisciplinary (CAP-NI), and 3) Capitation – Interdisciplinary (CAP-I) . We hypothesized that 1) the two CAP models would have higher referrals than the FFS because of the reduced financial compensation for services delivered, and that 2) CAP-I practices would have lower referral rates than the CAP-NI group, as providers in this model have greater onsite resources and opportunity for collaboration than non-interdisciplinary practices

Six thousand three hundred and seventy providers caring for slightly over 10 million individuals were included in the analysis, with 53% cared for in the FFS practices, 26% in CAP-NI and 20% in CAP-I practices Over the two year examination timeframe, there were a total of 16,286,998 referrals across all medical specialties. The majority of specialist referrals were for diagnostic radiology (40.0%), followed by cardiology (8.7%), internal medicine (4.1%), and general surgery (4.1%).

Table 1 presents provider and patient characteristics for each model. In general, there were some differences in the year since graduation across models, and a higher percentage of foreign trained physicians in FFS practices. The patient population in FFS practices also tended to be sicker and a greater proportion resided in urban areas.

The results of the multivariable regression model, adjusting for both patient and provider characteristics are presented in Table 2. Few changes were observed in regression coefficients for patient level characteristics when adding provider characteristics to the model; results are therefore presented for the full multivariable model only. All patient and provider level characteristics (including model type) were significantly related to the referral rate. Referral rates were higher for providers who were female, had more years in practice and for those who were trained in Canada. In terms of patient characteristics, those that were older, sicker, female, and resided in urban areas had higher referral rates.

Figure 1 presents the least square mean estimates of the referral rates in the three model types. The CAP-I model had a lower unadjusted referral rate (755, 95% CI: 741-770 ) than both the CAP-NI (814, 95% CI: 799-827) and the FFS (827, 95% CI: 814-841 CAP-I) models. However, after adjusting for relevant patient and provider characteristics, FFS practices had the lowest referral rate (676, 95% CI: 666-687, p < 0.0001) when compared to the other two models (CAP-NI: 719, 95% CI: 705-734, p < 0.0001; CAP-I: 694, 95% CI: 681-707, p < 0.0001).

The two CAP models were also significantly different (p = 0.002). Pairwise comparisons of the adjusted referral rates [Table 3] revealed that the interdisciplinary CAP-I group had a 3.5% lower referral rate than the CAP-NI group (RR = 0.965, 95% CI: 0.943-0.987, p = 0.002).

This study demonstrates significant differences in specialist referral patterns between primary care models. After accounting for patient and provider characteristics, physicians practicing in the FFS model had a lower referral rate than the physicians in the two capitated models (i.e., CAP-NI and CAP-I), while physicians in the interdisciplinary CAP-I model had a lower referral rate than those in the non-interdisciplinary CAP-NI model.

On a study population level, the observed differences in referral rates equate to 111,059 and 37,391 fewer referrals by FFS versus CAP-NI and CAP-I, respectively – a total difference of $22.3 million in initial referral appointment costs (assuming that the average cost of a referral is $150 [12]). The observed referral cost difference between CAP-I and CAP-NI is $ 9.9 million. We expect this cost difference to increase when including other costs that are frequently incurred during or following a typical referral (e.g., patient travel costs, time off work, repeat diagnostic testing, follow up specialist visits etc.).

These results confirm our hypotheses which were premised on 1. Remuneration influencing physician behaviour, and 2. Team structure influencing capacity. Physicians that are paid primarily through a fee-for-service approach may have an incentive to bring patients in multiple times to bill for more services instead of referring out to a specialist. On the other hand, in a capitation-based payment model, physicians are given a fixed lump-sum payment for each patient regardless of the number of services they provide and may be more inclined to pass on the treatment of their patients to specialists. This is in line with findings from several European-based primary care studies, which found that FFS was associated with a 9-12% lower referral rate than capitation or salary payment models [13‐16].

Physicians practicing in the interprofessional capitation model had a lower referral rate than those working in the CAP-NI, potentially because the presence of allied health professionals such as a social worker, nurse educator and pharmacist gives greater capacity for care afforded to the doctor by the team structure that results in a lower pressure to offload complex, time consuming patients. For example a team comprised of a nurse and a pharmacist can support insulin initiation for a poorly controlled patient with diabetes thereby avoiding the referral to a specialist endocrinology clinic. A team with a social worker can help provide mental health support to patients thereby avoiding referral to a psychiatrist.Evidence examining the impact of interdisciplinary primary care practices on specialist referrals is limited and mixed, with certain studies showing no impact while others have shown a decrease in referral rates [17, 18]. Thus,there is a need for ongoing research in this area to understand the impact of team based care from both an access and an economic perspective. Patient characteristics had the largest impact on referral rates; inversing the observed associations between referral rate and model. Patients that were older, sicker, female and urban residents had higher referral rates, a finding which is consistent with studies that have been conducted internationally and in Canada [1, 10, 11, 19]. Females were 17.2% more likely to have a referral than males; however, this is likely due in most part to referrals made to obstetrics/gynecology. In the case of rurality, urban residents were 6.5% more likely to have a referral than those in rural areas. Primary care physicians in rural areas likely provide a select range of specialty services due to the shortage of local specialists and travel challenges for patients within these communities [10, 20].

In addition, gender of the physician had a significant impact on referral rates with female physicians more likely to refer patients compared to male doctors. The reason for this difference is unclear and likely involves multiple factors ranging from differences in time spent with the patient resulting in more indepth examinations leading to referral, or to differences in risk tolerance [21]. Our results are consistent with a previous study in Ontario [10] and a recent study from the Netherlands [21]. This study was limited due to low power with the total number of physicians being only 44 with 14 female physicians in the sample. Other studies have not found this same effect of physician gender on referral patterns and suggest it may be related to other organizational aspects of the practice that differ between men and women [22, 23]. Nonetheless, as there is increasing proportion of women in the physician workforce, differences in referral rates could have major implications for health system planning and resources. Future studies should include physician gender as a variable as earlier research on the variability in referral rates did not commonly include physician gender in the analysis.

These findings demonstrate that there is a significant association between primary care model type and referral rates. Whether a lower rate of referral is more appropriate or not is not known and requires further investigation. Our findings suggest that physician remuneration and practice team structure may account for the differences in models; however, further investigation is required to better understand whether other factors associated with primary care model also impact referral rates. This study is one of few to comprehensively examine the association between primary care model type and referral rates on a large population basis. This study provides data which will help policy makers understand the impact of recent primary care reforms on speciality care and will also provide support in planning and projecting future referral initiatives and their impact on health care costs.