Implementing pay-for-performance in Australian primary care: lessons from the United Kingdom and the United States

The Australian Government’s agenda for the reform of the health system is now under consideration.1,2 The National Primary Health Care Strategy recommends “pay-for-performance” as a key building block:

It has already been announced that the federal government is to take responsibility for all primary care funding and policy. Patients with diabetes will be encouraged to voluntarily enrol with a general practice for care of a chronic condition; capitation payments (set amounts paid to doctors for enrolled patients) will be made, and pay will be linked to performance. Given Australia’s past experience with pay-for-performance through the Practice Incentives Program (PIP), many of the conditions for the extension of a pay-for-performance scheme in Australian primary care seem to be in place. However, pay-for-performance can be expensive,3 and rigorous evidence that it leads to improvements in quality of care is lacking.

The aim of our article is to consider the international experience of pay-for-performance in primary care, identify the key lessons learned, and use these to examine what questions Australia may need to consider before such a scheme is introduced.

The Quality and Outcomes Framework (QOF) is a national primary care pay-for-performance scheme in the United Kingdom,4 described as “world-leading” at its introduction in 2004. The UK Government invested up to 20% of the primary care budget on the scheme, 90% of which was money additional to the existing health budget. Since 2004, there has been an increasing focus on clinical indicators (eg, the percentage of people with heart disease or with blood pressure below a defined target) (Box 1), which represented 52% of the framework in 2004, and increased to 66% in 2006 and 70% in 2009. Although the QOF is a voluntary system, over 99% of UK practices participate in it.5 Points are awarded for individual indicators in relation to the level of achievement of that indicator. There is a graduated scale of payments, from a minimum (25% initially, rising to 40% in 2006) to a maximum (usually 90%) once maximum points are reached.

The United States has the most diversity in pay-for-performance schemes in health care.6,7 There are hundreds of such schemes within US health maintenance organisations (HMOs), and some of these are large schemes involving coordination across a number of private health insurers.8 US national schemes are currently being developed and evaluated within the US Medicare system.9

There are few well designed empirical evaluations of the impact of pay-for-performance schemes.10,11 We reviewed the evidence from the UK, US and Australia and identified several important factors that should be considered when designing such schemes.12

It is crucial to account for intrinsic and contextual factors which may accompany the introduction of pay-for-performance. Schemes evolve over time and usually involve several complex interventions, including accreditation, education, quality improvement programs (eg, guidelines and feedback), investment in information technology (IT) and data collection systems, and professional and regional support structures.12 These are all necessary conditions for linking financial incentives to quality of care. In the UK, there was a long history of structural reform, performance management, clinical guideline development and development of IT infrastructure before the QOF was introduced. Critically, it was possible to collect good quality data on current baselines for quality of care from some general practices.13 Performance pay in the UK relies on data collection at the practice level. The QOF uses electronic data entry and extraction processes using agreed clinical terms (Read codes). There are no comparable nationally computerised clinical information systems in Australia, although some progress has been made through the Australian Primary Care Collaboratives. In the US, and in the Australian PIP, administrative data based on fee claims are used to calculate performance measures. This approach does not lend itself to a pay-for-performance scheme that covers multiple disease areas, because fee claims are indicators of process of care, not clinical outcomes.

The first rule of the QOF in the UK is “first, create a register”. This is possible because patients enrol with only one general practice. These registers focus predominantly on single conditions (eg, diabetes), but as the population ages and multiple morbidity is increasingly the norm, there is a need for measures that take into account the number and severity of conditions at an individual level. In Australia, disease registers are kept by practices in the PIP and by accredited practices, so enrolment is not strictly necessary for the maintenance of registers. However, enrolment would strengthen a pay-for-performance system, as it makes it easier to attribute changes in quality and performance to a specific provider responsible for coordinating the health care of patients.7 The incentives may be weaker if a patient visits multiple providers, since providers may be less likely to feel “responsible” or feel they have less control over patient outcomes. Enrolment for all patients could also prevent undesirable “double dipping” (patients appearing on the disease registers of more than one practice).

In the UK, negotiations to implement the QOF were simplified because a single funder (the UK Department of Health) could negotiate with doctors’ representatives and agree on a new national general practitioner contract. With multiple funders, the effects of incentives introduced by any one funder will be less effective. Experience from the US suggests that when a doctor contracts with multiple insurers or HMOs, a single HMO has less influence on changing GP behaviour. The scheme coordinated by the Integrated Healthcare Association in California has greater influence because it is a collaboration of several insurers in California that use pay-for-performance schemes covering 17% of the population (6.2 million people). The insurers pay 60% of the capitation payments to medical groups.7,14 In Australia, the federal government currently funds most GP services, but non-GP primary care services are often funded by state governments. Planned reform includes the federal government becoming the sole government funder of primary care services. Patients will continue to contribute through copayments, which could dilute the effect of a pay-for-performance scheme.

The UK QOF policy documents do not mention rewarding improvement in quality of care, just “providing” it. None of the incentives reward quality improvements, only achieving targets for “intermediate” outcome measures.15 No scheme directly measures or rewards improvements in health, as patients’ health-related quality of life before and after interventions is not routinely measured. The Californian Integrated Healthcare Association scheme had clear objectives for rewarding improvements in quality, but the incentives actually rewarded attainment of target levels of performance or position in relative rankings, not change or improvement. The US Medicare Physician Group Practice Demonstration scheme includes both threshold targets to reward achievement and an improvement target.9

There are several reasons why rewarding the achievement of a specified level of quality may not be as effective as rewarding improvements in quality. First, for those already providing high-quality care, the reward represents a windfall financial gain with no incentive to change behaviour. The QOF could be regarded as another way of paying GPs, rather than encouraging improvements in quality of care through changed behaviour. Second, rewarding the achievement of a threshold of quality does not encourage further improvements in quality beyond the threshold. Third, for doctors with very low baseline levels of quality, the costs of achieving a relatively high threshold may seem prohibitive and they may not participate. These are the very groups where substantial gains in quality of care could be made at little additional cost.

Rewarding improvements in quality would encourage those with low baseline levels of quality to improve their care (ie, where the largest increases are likely to be made).6,16,17 This can be achieved either by setting successive thresholds close together (eg, at 5% intervals) or to pay only when a certain increase has been achieved (eg, a 5% increase). Additional costs for those achieving a given change in quality from a low baseline are likely to be lower than the costs for those already providing high-quality care, where the “easy” gains have already been made. The increasing costs of achieving each successive improvement should then be reflected in the payment schedule. Rewards for those who already provide high-quality care, but find it difficult to improve any further, should be part of a background payment scheme rather than a key element of the pay-for-performance scheme.

Possible unintended consequences, such as a focus on the remunerated areas at the cost of unremunerated areas (“gaming”), need to be thought through and managed carefully. Few empirical studies have examined potential adverse consequences of pay-for-performance.18 The QOF was intended to reward quality of care rather than numbers of registered patients, and also to improve data capture, care processes, patient outcomes and doctors’ working conditions.4 However, evidence suggests that over the lifetime of the QOF, patients in the UK now find it harder to obtain continuity of care.19 Other evidence from the UK suggests that there were positive spillovers to unremunerated disease areas.20,21 A practice that provides high-quality care in one disease area is also likely to provide high-quality care in other disease areas.

Gravelle and colleagues found that UK GPs who failed to score maximum QOF points in the first year “gamed” the system by increasing their exception-reported patients (patients who would not count in their QOF scores) in the following years.22 “Patient dumping” is therefore another risk in pay-for-performance schemes.18 There is also a risk that external incentives may crowd out intrinsic motivation (the desire to do a task well for its own sake), especially if the external incentives are not closely aligned to doctors’ views of their role.23 Current evidence suggests that financial incentives are most likely to be effective in influencing professional behaviour when performance measures and rewards are aligned to the values of the staff being rewarded.24

In the UK, practice nurses have absorbed a higher proportion of the clinical workload associated with the QOF.23 One unintended consequence is that nurses do not benefit directly from the incentives that drive their work.25 Payments to individual professionals may not encourage team work or coordination of care (and thus may not improve quality of care when team-based coordinated care is needed) if there are large differences in payments between team members.25

A balance needs to be achieved in deciding how much funding will be allocated to pay-for-performance schemes. The QOF substantially overspent initially, because GP performance was underestimated before the scheme was introduced.26 Additional government spending on health care was used to fund the scheme, including a substantial increase in doctors’ incomes of 35% over 2 years, when the scheme was introduced. The 20%–25% of doctor income that comes from the QOF contrasts sharply with the experience of using pay-for-performance in the US, where rewards were less than 10% of revenue, which may not have been large enough to cover the costs of changing behaviour.6-8 Incentives were funded from within existing budgets or savings, and so for some schemes no new resources were invested. Australian Medicare data from 2003 suggest that the PIP represented an average of 9% of GPs’ revenue from Medicare,27 though only a proportion of this would have been used for the pay-for-performance elements (service incentive payments [SIPs]28 and outcome payments). There is little evidence to suggest what proportion of revenue should be paid using pay-for-performance, but the 20%–25% of revenue in the QOF is likely to be too high.

There should be a mix of payments to the practice and to individual health professionals for quality improvements. There is little evidence to suggest that one or the other should be targeted.29 Paying teams raises issues of how a team is defined, and how the financial rewards are used by the team.3,30 These issues may mitigate the intended effect of the financial incentives, though it does give discretion to the team to use the rewards in a number of ways, which might include investing in further quality improvements, or directly rewarding team members, or a combination.17 Any payments to team members should be equitable with respect to their skill and effort, or those who feel unfairly treated may lose motivation to change behaviour.

Pay-for-performance in Australia should evolve. It is not possible to transplant a system such as the QOF to Australia, mainly because Australia does not currently have the IT infrastructure to support such a scheme. However, Australia already has a pay-for-performance scheme, with SIPs and outcome (threshold) payments for diabetes, cervical screening and Indigenous health. A practice-level data collection and reporting system is also being developed and used in the Australian Primary Care Collaboratives. These schemes could be built on and developed, while taking into account the key lessons learnt from overseas.

The PIP is already evolving. The new Indigenous health incentive requires voluntary patient enrolment as well as performance incentives. The recent announcement of voluntary enrolment for diabetes patients, along with capitation payments to practices and performance payments (yet to be defined), is an important step in this direction. Even without this, there is a strong argument for changing the existing SIP program, and re-investing the resources into revised outcome payments providing rewards for annual percentage improvements in patients receiving completed cycles of care. In the longer term, investment in quality improvement and reporting requires investment in an integrated IT infrastructure.

If pay-for-performance is to be introduced into Australia, key lessons should be learned from the experiences of other countries (Box 2). In particular, pay-for-performance should be used only as one part of a wider, multiple-systems-based strategy for quality improvement, and should not be seen as a panacea. Pay-for-performance should be used explicitly to drive quality improvement, and not simply to reward those who are already providing high-quality care. Before such a scheme is contemplated, the method of data entry and extraction should be considered a key priority, with IT specialists included early in discussions.