Making sense of paying for performance in health care: short-term targets versus patient-relevant outcomes

There has been growing interest around the world in the use of pay-for-performance programs in health care. These programs are based on the seemingly strong economic logic that financial rewards can encourage behaviour in healthcare providers to promote outcomes, such as quality of care and efficiency (or, conversely, that penalties will discourage behaviour that is negative to such outcomes). A local example of pay-for-performance is the Medicare Practice Incentive Program that rewards general practices for targeted levels of care in areas such as diabetes, asthma, prescribing and cervical screening.¹ In relation to asthma, for instance, practices receive payments to sign on patients and general practitioners (GPs) are paid A$100 per year for each patient who completes a designated cycle of care.

The conclusion from systematic reviews in this area,^2–4 as well as recent reports,^5–8 is that despite many studies, the worldwide evidence of the effectiveness of pay-for-performance in health care is fairly weak. Certainly, there have been limited examples of success. The Australian Government Diabetes Care Project, through a quality improvement intervention with payment reform comprising risk-adjusted capitation payments alongside pay-for-performance incentives, saw some improvement in intermediate clinical indicators, adherence to recommendations and patient satisfaction.⁵ General practice incentives in the English National Health Service (NHS) have led to improvements in some quality of care indicators, but with significant increases in the cost to government.⁶ Despite these examples, the general conclusion from the literature is reflected in a recent European Observatory on Health System report: ‘(i)n common with many other authors, we too find that P4P [pay-for-performance] has not produced the direct significant change that many advocates hope for…’.⁷

Several reasons have been posited for the lack of effect of pay-for-performance schemes.

The first is that often incentives are misdirected; for example, they reward practices, whereas the achievement of targets requires change at the level of individual providers. This is a problem if provider remuneration is not linked to the practice-level incentives (e.g. when individual physicians or nurse practitioners are paid by fixed salaries and are therefore cut off from the bonus payments that are paid to the practice). In practice, although non-financial factors such as loyalty to the organisation or peer recognition may have an influence, there is nothing in the incentives as such to effect individual behaviour change.

Second, incentive levels may not be adequate to influence behaviour. Although we can always make them more effective by simply paying more, incentives also have to be sustainable. As such, cost-effectiveness considerations ought to be central to decisions about implementing pay-for-performance programs.

Third, money isn’t everything: a complex mix of social and professional norms drive behaviour and these are more influential than the monetary rewards and penalties to individual actors. An example of interventions that exploit this idea is the NPS Medicines Insights Program. This program provides GPs with data on their own prescribing relative to peers, thereby using social comparison as a means of ‘nudging’ positive behaviour.

By paying for activities that were previously undertaken for ‘free’, pay-for-performance programs potentially ‘crowd out’ the social norms and goodwill that may have otherwise held sway.^9,10 Furthermore, such cultural changes in motivation, once implemented, are difficult to reverse; once we start to pay individuals to undertake a task, we cannot expect them to undertake that task unpaid in future. A study of financial incentives to improve rates of screening for diabetic retinopathy and cervical cancer in medical facilities run by Kaiser Permanente in California indicated that although such incentives improved screening levels, the removal of such incentives was followed by falls in the levels of screening to levels below (albeit slightly) those observed at the pre-incentive baseline.¹¹

One flow-on effect from incentivising a certain set of tasks is that attention and effort can be drawn away from other tasks. Such incentives can also lead to overuse and duplication. An example is fee-for-service payment, which is not usually put forward as an example of an incentive scheme but is, in one important respect, ‘pay-for-performance’ because it involves providers being paid more the more services they deliver. Such a system in general practice in Australia has often been implicated in overservicing, waste and a lack of coordination in care.^12–14

In addition to these explanations as to why they may or may not work, there is the broader question of whether they are targeting the right outcomes. Applications of pay-for-performance have generally been based on payment for short-term targets that are not necessarily meaningful to patients. Typically, incentive programs in health care reward or penalise on the basis of short-term behaviour change, such as targets in relation to vaccination rates, blood pressure control and screening. As such, they involve carrots or sticks doled out separately on a year-by-year basis with no memory built into the system. In this regard they are predicated on the notion of a one-shot doctor or provider–patient relationship where achievements in one interaction count for nothing in how performance is assessed in the next.

However, provider–patient relationships are often long term, involving repeat interactions, and are built on social norms such as trust and reputation. In these instances, a policy maker may be better advised to align the incentives to long-term goals and shift behaviour designed to achieve targets, over say 5 years, such as the control of overall cardiovascular risk, the prevention of unplanned admissions to hospital, continuity of care and quality of life. Viewing interactions between providers and patients within the context of a relationship rather than as a series of isolated transactions will enable the design of incentives that are better aligned with their common goals. However, the achievement of these long-term outcomes may be influenced by a range of external factors, which inevitably complicates the task of attribution. This problem can, to some extent, be addressed through the triangulation with data on secondary indicators of process and activity.

Notwithstanding these measurement issues, there is an important reason for making providers accountable for outcomes that are subject to external factors: it encourages the tailoring of patient care to individuals’ personal circumstances, and the social and physical environments in which they live, rather than on the mechanistic delivery of individual services.

Despite their limited evidence of success, pay-for-performance programs in health care potentially have a role in Australia if incentives target outcomes that are meaningful to patients rather than short-term measures of activity. In designing such programs, decision makers need to recognise the common interest of patients and providers in achieving basic goals, such as long-term health, quality of life and continuity of care, and reward on the basis of these goals. For such measures to be effective, they also need to be designed so that rewards are in some form transmitted to the individual providers responsible for performance. Given that such incentives are costly and can have side effects, decisions to introduce pay-for-performance should also be subject to evaluation alongside other ways to motivate behaviour, such as through the influence of social and professional norms.

The author has no competing interests to declare.