Designing incentives for good-quality hospital care

Public hospitals in Australia are in for a shake-up over the next few years, with boards being reintroduced in many states and activity-based funding (ABF) being rolled out nationally.1 ABF will replace global or historic budgets for hospitals in most states. National casemix classifications will be agreed, and the work of hospitals described and priced using these classifications. Australian refined diagnosis-related groups (AR-DRGs) will be used for inpatients, and other classifications will describe outpatient, emergency department, mental health and subacute activity.

The national introduction of ABF will immediately improve transparency of federal funding support for hospital activity. The rhetoric associated with ABF also emphasises efficiency of hospital care. Much of this will depend on how states, who will manage the system, pass on incentives to local hospital networks. However, the National Health Reform Agreement goes further and declares that the Independent Hospital Pricing Authority (IHPA) will have regard to, among other things, clinical safety and quality (clause B12a) in setting a “national efficient price” for hospital activity.2 It is probably a stretch to add “quality improvement” as a third objective (in addition to funding transparency and clinical safety), but certainly the National Health Reform Agreement signals a potential role for the IHPA in this area.

Although generally still in pilot stages, there has been increased attention to the extent to which financial incentives might encourage good-quality care (“pay for performance”) and to what degree financial disincentives can be built in to discourage poor-quality care. A hesitant toe has been dipped into the water in England, where the Payment by Results scheme has recently introduced “best-practice” pricing strategies in four areas: cholecystectomy, hip fractures, cataracts and stroke.3 In the United States, almost all contemporary implementations of funding reform involve a blend of efficiency and pay-for-performance aspects.4 However, the policy enthusiasm for pay-for-performance appears to be greater than the evidence, which suggests only moderate efficiency gains.5

What has moved from the stages of discussion and pilot testing to implementation is the obverse of pay for performance: non-pay for non-performance.

Routine hospital datasets used for ABF distinguish between patients’ comorbidities present on admission and hospital-acquired complications. In the US Medicare system, a limited list of hospital-acquired conditions has been excluded from being used in assigning cases to its casemix classification and thus affecting activity-based payment. The current non-pay for non-performance regime shifts only a small amount of funding around, but has been extensively debated, perhaps because of its potential to affect hospital reputation.

There are several options within the area of non-pay for non-performance, relating to which complications should not attract payment for their management, and what financial impact should be imposed.

In its most narrow implementation, non-payment can be targeted at those complications that are clearly preventable and should never occur, with wrong-site surgery being the best example. The US Medicare system goes further, focusing on complications that “could reasonably have been prevented through the application of evidence-based guidelines” — such as catheter-associated urinary tract infections.6 This would apply to conditions in the index hospital admission or to readmissions (or both). At the other extreme, one could determine that the treatment of all hospital-acquired conditions should not be funded.

When a list of non-fundable conditions has been decided, one can then consider what kind of impact the scheme should impose. The softest form of payment impact is simply to exclude hospital-acquired complications from consideration in DRG assignment, as currently happens in the US. Most DRGs make a distinction between whether or not comorbidities and complications are involved, with more comorbidities and complications being associated with greater expected costs and a higher payment weight. At present, the DRG algorithm treats hospital-acquired complications and diagnoses present on admission (comorbidities) equally. This is, in a sense, perverse, as a hospital could receive a higher payment for a complication acquired by a patient under its care. However, including complications in DRG assignment (the current approach in Australia) could be argued to be fair to the hospital, as complications require additional resources to treat, and some are not preventable with current medical knowledge.

A variant of this approach is to identify a risk-adjusted rate of complications for each hospital, and then not fund activities in hospitals with a rate of complications higher than the expected risk-adjusted rate.7

The toughest form of payment impact would be to apply penalties for excessive rates of poor performance. From 2013, the US Medicare program will impose penalties in the form of reductions in payment to hospitals (phased in, of up to 3% of payments in 2015) for excess readmission rates for acute myocardial infarction, heart failure and pneumonia.8

Without any adjustment to AR-DRGs for non-performance, the system now being implemented is such that hospitals will, in many cases, receive a higher level of payment for treating patients they have harmed.

Adjusting payment for hospital-acquired conditions would create a strong incentive for managers to work with clinicians to reduce hospital-acquired conditions. The greater the payment impact and the greater the adverse event coverage that a non-pay for non-performance policy has, the greater the incentive for hospitals to reduce hospital-acquired conditions or readmissions attributed to suboptimal care.

However, there are risks that must be taken into account in a non-pay for non-performance policy. The range of hospital adverse events will be measured by what is recorded and coded in the patient record. Coding guidelines for some conditions in the narrowly defined US list have been challenged,9 as has the degree of “preventability” of some of them.10 If a large range of hospital-acquired conditions is targeted, hospitals run the risk of being unfairly labelled and stigmatised if the financial impact is disclosed publicly. But a narrower range of identified “preventable” conditions, while minimising this risk, also limits the policy impact and may result in hospitals focusing their quality endeavours on the limited range of monitored indicators.

Strategies with tough financial impacts may lead to underrecording of diagnoses and hence require more coding audits. Strategies that exclude hospital-acquired conditions from consideration in DRG assignment will be neutral from a coding incentive perspective (although the “present on admission flag” will need to be audited nationally). However, strategies that involve risk adjustment of complications might actually encourage better coding.

So where to from here? First, Australia should not withhold payments for readmissions, partly because it is difficult to attribute responsibility for readmissions in the Australian system. Adopting the US list of hospital-acquired conditions could represent a safe and modest start to linking payment policy and health care quality issues in Australia, but would only shift a trivial proportion of payments.11 In addition, as the US list can be challenged, a better path might be to use the work of the Australian Commission on Safety and Quality in Health Care to create our own shortlist of conditions linked to quality of care. The list should only include conditions where responsibility for the adverse event can be clearly attributed to the hospital and its staff, perhaps by failure to implement the Commission’s standards, such as those relating to patient identification. Initially, the impact could also be limited by excluding relevant complications from DRG assignment (Box).

Alternatively, Australia could start more robustly and use a DRG classification system that does not take any hospital-acquired complications into account. Care for a patient with a hospital-acquired complication would be funded as if the complication had not occurred.12 Because of the way payment weights are calculated, the payment for such a case would effectively incorporate the costs of the average hospital-acquired complication rate. Hospitals with the average rate of hospital-acquired complications would thus not be affected by this policy, while hospitals with a lower rate would be rewarded and those with a higher rate adversely affected.

The national introduction of ABF allows us to reposition ABF in Australia. It is being introduced without budget cuts — a coincidence which dogged initial implementation in Victoria. Incorporating signals about the importance of quality, its impact on the health system and, more importantly, its impact on affected patients would be another step forward in ABF development and health care safety and quality in this country.