TONG Chunyang

Professor of Tongji SEM, Doctoral Supervisor

Many healthcare systems use a referral process to determine if a patient should see a specialist. Indeed, China is planning to launch a referral-based healthcare reform. Patients typically visit a general practitioner who may either treat them or refer them to a specialist, depending on the complexity of their ailment. Specialists are often more expensive and their supply is limited in some regions and referral systems are designed to effectively utilize resources of varying expertise and costs. Patients who are referred to specialists tend to incur greater health care spending compared with those who remain within primary care, even after adjusting for health status. Borrowing from existing experience from other countries, we can see similar phenomenon. For example, there is concern about increasing referral rates in the US (Barnett et al. (2012)) and they are estimated to be twice as high as in Great Britain. The excessive referral to specialists is often decried as wasteful spending and Epstein (2015) states that excessive referral to expensive specialist care is one of the six major reasons driving up healthcare cost in the U.S. On the other hand, some patients with complex ailments need to be treated by a specialist rather than a general practitioner. Moreover, as is implicit in the anecdote above, generalists are concerned about potential risks, including lawsuits, of mistreatment claims if they do not refer patients. Song et al. (2014) argue that patient referral decisions in an outpatient setting are fundamental to collaboration among physicians and other health care professionals and “have been largely ignored in the payment reform debate”.

While referral rates are important, there is an important link between the referral rate of the primary care provider and the service provision delivered at the specialist. If the primary care providers refer too many patients, then the specialists may get overwhelmed and the wait time to get an appointment with a specialist could be longer. This is often the case because specialist resources are scarce in many regions of the world, including parts of the US. Higher referral rates combined with limited specialist capacity results in greater congestion and long wait times. According to a Fraser Institute survey (Barua et al. (2016)); the median waiting time between referral from a general practitioner to receipt of treatment from a specialist is 18.3 weeks in Canada. Similar statistics abound in many countries including in the US.

While congestion and waiting to get an appointment with a generalist may be an issue in some cases, access to and long waits for specialist service appears to be much more problematic based on a Commonwealth study of healthcare in 11 rich countries (see Exhibit 5 in Davis et al. (2014)). Stainkey et al. (2010) describe a study in Queensland, Australia, where times wait for specialists was over 2 years in some cases and how they worked with general practitioners to improve the referral process to reduce such waits. The long wait times can result in deterioration in patient health and poor health outcomes (Emberson (2014)). Other than longer waits, increased congestion may compel the specialist to spend less time with patients. The link between high congestion and shorter treatment time which results in lower care quality has been documented (Batt and Terweisch (2017), Chan et al. (2017)). Thus, the general practitioner’s referral decision has an impact on the specialist’s performance when specialist capacity is limited. Therefore, it is important that referral rates and specialist resource availability are carefully coordinated.

Payment schemes play a very important role in all aspects of healthcare. In our context, models used to compensate physicians play an important role in influencing both providers’ decisions and could either alleviate or exacerbate the problems with referral rates, wait time for specialists and care quality. Recently, there has been growing debate in the US about the relative efficacy of existing payment models in controlling healthcare costs and new models are being explored. Miller (2010) provides a detailed discussion of various payment models used in the US and their implications for providers and patients. As indicated in Miller (2010), a key objective of payment reform is to link payment with service outcome (quality) and incentivize physicians to save cost while achieving high quality—referred to as “affordable and accountable” healthcare. In particular, bundled payment, a new way of compensating healthcare service, has been widely touted as one promising payment scheme. In sharp contrast to traditional payment schemes under which service provided by different physicians are paid separately using either a flat fee per patient or fee-for-service, the payer under the bundled scheme pays a lump-sum to “a group of service providers” for each episode of care. The presumption is that providers then have an incentive to coordinate among themselves to reduce the cost and achieve high quality. Such coordination however requires accurate attribution of cost and quality to providers’ decisions. One of the challenges in designing good payment schemes in healthcare is the attribution of costs and pain suffered by a patient to a provider’s actions. A specialist’s decisions on treatment effort impacts the quality of care and costs incurred by patients due to suffering, ailments, relapse, etc. This in turn leads to unhappy patients, reputation, lawsuits, etc. for the specialist. In an ideal world, we would be able to accurately attribute patient outcomes to the provider’s actions. However, in reality, this is impossible since patient outcomes and corresponding costs depend on many factors, including patient characteristics and type of ailment, many of which may be outside the provider’s control. As a recent article (Beck (2015)) points out, it is not possible to hold a specialist fully accountable for healthcare outcomes. Moreover, there are many confounding factors leading to mixed results— Krumholz et al. (2013) and Brotman et al.(2016) show that measures such as readmission and mortality may be inversely correlated in the presence of partial accountability of providers. We explore how the partial accountability of specialists for patient outcomes impacts the performance of the payment schemes.

There has been considerable debate and analysis of how payment systems may result in excessive or inadequate services and specialist care (Barnett et al. (2012)) and how certain payment schemes have led to hospitals discharging patients more quickly than medically appropriate, for financial reasons (Neuet al. (2016)). However, there has been little analysis of how payment models may influence referral rates and the coordination of decisions by the general practitioner and specialist. In particular, to the best of our knowledge, there has been no analysis of how payment mechanisms may influence the linkage between referral rates and specialist resource limitations and the resulting impact on quality of care, wait times and healthcare costs. This work aims to fill this gap.

The objective of this work is to understand how different payment models in healthcare, particularly in the presence of partial accountability, can help address care coordination between the general practitioner and the specialist, which impact referral decisions, specialist workload and quality of care at the specialist. We address these issues by considering a stylized model comprised of three entities: a general practitioner (called generalist henceforth for simplicity), a specialist and a single payer (say a public agency). The payer contracts with and compensates the generalist and specialist on behalf of the patients. In this setting, we explore the performance of different payment systems, including the bundled systems that have been introduced in the US as well as systems where the payer may contract with the providers separately using different payment schemes (which we refer to as unbundled payment). We compare and contrast the different payment schemes and also compare them with the first best scheme, wherein a single central planner would make all the decisions and thus achieve the best outcome possible. However, the focus of the paper is on evaluating existing payment schemes and not on deriving the first-best scheme. In particular, since a bundled scheme has attracted attention recently and is deemed to be promising, our primary focus in this paper is to evaluate the effectiveness of the bundled scheme and also explore how the bundled scheme performs on some important operational metrics relative to traditional unbundled payment schemes.

We find that the bundled system results in higher referral rates, lower treatment times and lower service quality at the specialist stage as compared to the unbundled system. Moreover, while the total system costs are higher in the bundled scheme, the payer cannot fully extract the surplus in the unbundled scheme unlike in the bundled scheme. The unbundled scheme, featuring more controls over the physicians, can thus achieve a higher efficiency as compared to the bundled scheme. However, this higher efficiency comes with a concomitantly higher cost to the payer. This is because the payer aims to induce the ideal referral rate in the unbundled system by making a higher elective payment to the generalist, who therefore earns a strictly positive profit unlike in the bundled scheme. The bundled scheme achieves lower mistreatment cost at the generalist but higher quality costs at the specialist. Thus, the ability to contract separately with the two providers helps the unbundled scheme get closer to the first-best scheme in terms of referral rates, treatment time and quality costs at the specialist and total system costs. Referral rates are highest in the bundled scheme resulting in the lowest mistreatment costs. When specialist accountability is partial, the specialist spends less time with the patient in both the bundled and unbundled schemes as compared to the first-best scheme and quality costs are higher. When the specialist can be held almost fully accountable for patient outcomes, both the bundled and unbundled schemes do very well and achieve results similar to those in the first-best case.