Designing payments for GPs to improve the quality of diabetes care

Performance pay for doctors has been introduced in many countries, including the United Kingdom through the Quality and Outcomes Framework (QOF) and the United States through the patient-centred medical home model.1 The effectiveness of these models remains in question, although there is emerging evidence that these schemes can reduce hospital admissions.2-4 In Australia, the Coordinated Care for Diabetes Pilot (CCDP) begins in 2012.5 The key elements of the pilot (Box) are voluntary patient enrolment, a flexible payment for each diabetes patient to cover allied health services (among other things) and a pay-for-performance element. General practitioners will continue to be able to charge fee-for-service payments and claim diabetes-related payments from the Practice Incentives Program (PIP), but will no longer be able to claim the Chronic Disease Management Medicare Benefits Schedule (MBS) items for GP management plans or team care arrangements.5

Doubts about the effectiveness of financial incentives are raised, not only by poor design of evaluations, but also by poor design of funding models. Careful design, based on theory and empirical evidence, is essential in designing interventions to change professional behaviour.6,7 This helps to ensure that the intervention is likely to be effective, and that unintended and undesirable consequences of incentives are minimised.

Our aim in this article is to propose three essential features of the flexible payment-per-patient and the pay-for-performance elements of the new CCDP.8,9

An objective of the CCDP is that incentive payments will be paid for:

Payment should be made for improvements in the quality of care, which requires measuring changes in quality over time. Previous schemes have made payments for the achievement of a level of quality, often based on a relatively high threshold, rather than for an improvement in quality. Using thresholds may not lead to changes in behaviour because: (i) doctors in practices with already high levels of quality of care can claim the highest threshold payment without changing their behaviour; (ii) there are no incentives to go beyond the threshold, and (iii) doctors in practices with low levels of quality of care have little incentive to reach a high threshold, as they perceive the costs to be greater than the financial reward.10

One option is to reward improvements in quality between two time points, and set successive thresholds close together, say at 5% intervals.8 This would encourage those with low baseline levels of quality to improve the quality of their care. Furthermore, economic theory would predict that behaviour is only likely to change if the level of the payment is at least equal to the costs of improving the quality of care. These costs will vary across practices depending on their baseline level of quality and the complexity of the health problems of their patients.3 Allowing payments to vary in line with these cost variations would lead to a potentially more effective intervention. The costs of improving quality of care from 10% to 15% are likely to be lower than those of improving from 90% to 95%. Practices that already attain high standards of care find it difficult and costly to improve further and would not be able to claim an improvement payment. Thus, an element of performance pay should recognise the high performance levels already achieved. The “improvement” payment should be higher than the “achievement” payment, or there would be little incentive for practices to improve their quality of care.

A fixed payment per patient gives GPs an incentive to minimise costs, as any surplus from the budget can be kept as personal income or reinvested in patient care, and any deficits are borne by the provider.11 This is good for governments, who want to control costs and ensure costs are predictable, because it shifts the financial risk to GPs. However, if this risk is perceived to be too high, GPs may choose not to participate in the scheme or to enrol only “healthy” patients, so a careful balance needs to be struck between risk-sharing and the strength of incentives.

The incentives to minimise costs and “cream-skim” are already ameliorated to some extent in the proposed CCDP. GPs can still claim standard MBS items and PIP payments for their patients with diabetes. Total revenue per patient will vary according to the number of visits made per year, and the length of each visit, which partly reflects complexity. However, this may not cover intensity in terms of the use of other resources in the practice, such as practice nurses or the costs of employing allied health professionals.

One option is to “risk-adjust” fixed and performance payments, so that higher payments are made to practices with patients with higher needs or more complex health problems than the average. There is much international literature on this subject, including on the use of risk adjustment in the design of the patient-centred medical home pilots in the US.12 Risk adjustment requires data on the primary care costs of diabetes care for all patients and how this varies according to patients’ characteristics. Unfortunately, there are no routinely available data, and very little information on the use of allied health services by patients with diabetes and the associated costs.13,14 Primary care and hospital data would need to be linked to obtain adequate measures of patients’ severity of illness and diabetes complications.14 It is doubtful whether adequate risk adjustment can be developed and implemented in time for the pilot. However, given continued access to MBS and PIP payments, the additional level of risk for practices under the new scheme may not be high, although this will vary across practices.

A further way that cream-skimming can take place for enrolled patients is exception reporting.15,16 Paying for the proportion of patients who achieve a certain target means that payments are based on a ratio, with the numerator equal to the number of patients reaching the target and the denominator equal to the number of patients in the population. Exception reporting occurs when practices exclude patients from the denominator, thus increasing measured performance and earning income, while not improving quality. For example, in the first year of the UK Quality and Outcomes Framework, a median of 5.4% (range, 0–40%) of practices used exception reporting for patients with diabetes. This translated into median gains of between £1700 and £15 000 per practice.15

One way to avoid exception reporting is to make payments based on the numerator only — that is, on the number of patients whose quality of care improves from one period to the next. This would involve a payment for each patient achieving a desired change in performance. Payments can only be increased from one time period to the next if the number of patients achieving a target, or whose quality of care improves, increases.

The design of the new payment system in the CCDP should attempt to maximise the impact of the incentives on quality of care while also ensuring an “appropriate” sharing of financial risk with providers, in addition to minimising any unintended consequences, such as exception reporting and cream-skimming. This will partly depend on the validity and reliability of quality indicators, but experience from the Australian Primary Care Collaboratives Program suggests that such data can be extracted from practices. The balance of additional revenue between the flexible payments and performance-pay elements is also crucial, and can influence behaviour as well as costs. Careful design of incentive schemes is essential for their success.