Primary Care Team

Incentivising prevention by general practitioners: some research possibilities

Mark Dusheiko, Hugh Gravelle, Arne Hole, Peter Sivey

Abstract

The aim of the paper is to consider the possibilities for research on the effect of the incentive structure facing GPs on the preventive activities they undertake. We consider activities preventing conditions which generate negative externalities (eg flu vaccination, childhood immunisation) and those preventing conditions which do not (eg cervical screening).

Design of appropriate incentive schemes requires both theoretical analysis of their general properties and empirical analysis to produce estimates of the magnitudes of effects produced by changes in incentive schemes. We briefly describe the history of NHS incentive schemes with regards to GP prevention activity. We then use the example of the comparison of fee for service and target payments to consider what lessons can be drawn from the general literature on incentives and regulation and whether further theoretical modelling is required.

We discuss how far existing data can be used to throw light on (a) the effects of GP remuneration schemes on preventive activities; (b) the practice and other factors which influence prevention activities; (c) whether there is inequity in the prevention and (d) whether the degree of inequity varies with the type of incentive scheme.

1Introduction

The aim of the paper is to consider the possibilities for research on the effect of the incentive structure facing GPs on the preventive activities they undertake.

Design of appropriate incentive schemes requires both theoretical analysis of their general properties and empirical analysis to produce estimates of the magnitudes of effects produced by incentive schemes. As background, in section 2 we review the history of GP incentives directed at prevention. We consider in section 3 what lessons can be drawn from the general literature on incentives and regulation and whether further theoretical modelling is required, using the example of a comparison of fee for service and target payments.

In section 4 we examine how far existing data can be used to throw light on the effects of GP remuneration schemes on preventive activities. We focus on the impact of changes in the GP contract on the provision of preventive activities which have been undertaken for some time and where several years of data are available: childhood immunisations, flu vaccines and cervical screening. The first two are associated with obvious externalities whereas cervical screening is of benefit mainly to screened women and their families. We discuss how far existing data can be used to throw light on (a) the effects of GP remuneration schemes on preventive activities; (b) the practice and other factors which influence prevention activities; (c) whether there is inequity in the prevention and (d) whether the degree of inequity varies with the type of incentive scheme.

2History of GP incentives for prevention

In the past 40 years there have been four main GP remuneration regimes:

Pre 1990 contract. From 1966 to 1990 GPs income were paid by a mixture of capitation fees, lump sum payments (allowances) including those related to seniority, vocational training, and providing out of hours cover, and fee per item of service for night visits, cervical cytology, and for some vaccinations and immunisations. GPs could also claim back a fixed percentage of certain costs including practice staff, dispensing costs, locums, and study leave.

Even in the absence of direct financial incentives GPs have some motivation to undertake prevention. First, they bear some of the costs of treating sick patients. Second, they may have altruistic concerns for their health of their patients. Third, the capitation element provided an incentive to compete for patients via prevention services. But this may have been dampened under the pre 1990 contract by the requirement that patients who wished to change practices had first to obtain the consent of the practice they wished to leave. Moreover, since prevention activities are only a part of the services GPs provide to patients improvement in the quality of prevention would have had a diluted effect on patients’ decisions about the choice of practice.

1990 contract. The 1990 General Medical Services (GMS) contract was intended to increase the incentives for GPs to provide higher quality care. The capitation element was increased and patients did not have to get permission from practices they wished to leave. A number of direct prevention incentives were introduced including target payments for vaccinations and immunisations for children, a change from fee per item to target payments for cervical cytology, and session payments for health promotion clinics.

Primary medical service contract (PMS) From 1998 onwards primary care trusts have been able to negotiate an alternative PMS contract with their practices. The PMS contract required the practice to provide services in addition to those they would have provided under the standard GMS contract. The additional services might be related to particular conditions or to particular types of patients (for example rough sleepers). In return the practice received more income than it would have had under GMS. PMS contracts were locally negotiated so that their terms vary across and within PCTs. Although PMS practices were meant to provide GMS services, they did not have to make returns about those services which had previously been explicitly rewarded via target payments or fee per item payments. Consequently we know rather little about what PMS practices do. We do know however that GPs in PMS practices have markedly higher incomes than GPs in GMS practices. By 2004 about 40% of practices had switched to the PMS contract.

New GMS contract(nGMS). The nGMS contract introduced in April 2004 retained a reformed version of the capitation payment and other lump sum allowances, but also linked practice income to the number of points achieved in the Quality and Outcomes Framework (QOF) of 146 quality indicators, many relating to prevention (Roland 2004). The QOF payments resulted in a substantial increase in practice income compared with the old GMS contract and 2004/5 saw very few practices switching to PMS. PMS practices were required to take part in the QOF. The payments they received under QOF were subject to a lump penalty equivalent to loss of around 150 points (out of a maximum of 1050) but their marginal gain from improved QOF performance was the same as for nGMS practices.

2.1Cervical cytology

Pre 1990 contract. GPs were paid a fee per cervical cytology test performed (£8.9[1] in 1989/90). See Figure 1. The GP indifference curves are U shaped. GPs always prefer a higher payment. Additional testing is costly to them in terms of effort or expenditure but: it is assumed that they are partially altruistic. Hence, at least for low levels of testing (and hence effort and costs), they may prefer more testing to less, other things (payment) equal.

1990-2004 contract. GPs were paid for cervical cytology by means of target payments. The targets were based on the percentage of eligible women on the GP’s list who received a smear test in the previous 5 and a half years. The main eligibility criterion were age: between 25 and 64 (Scotland had slightly different age ranges). There were two target rates, 50% and 80%. GPs meeting a target would receive a payment based on number of eligible patients. In 2003/4 the low target payment was approximately £1000 and the high payment approximately £3000 for practices with an average number (around 430 in 1997) of eligible women on their lists.

See Figure 2. If the marginal utility of testing is negative (because the costs outweigh any altruistic motivation) then GPs will cluster at zero, 50% or 80%. If they are altruistic any testing rate is observable.

nGMS contract. In the QOF GPs are paid according to the number of points they achieve, with no points for less than 25% of eligible patients tested, and 11 points for 80% or more tested, with points increasing linearly between 25% and 80%. A point was worth £120 in 2005/2006 for an average size practice. (A 11 further points available for other aspects of the cervical screening service such as the practice having a system for following up inadequate or abnormal smears.).

The incentive structure is illustrated in Figure 3.

2.2Child Immunisation

Pre 1990 contract. GPs were paid on a fee-for-service basis for administering immunisations to children.

1990-2004 contract. GPs were paid according to target payments based on how the number of immunisations administered relates to that required (on average) to achieve full immunisation of children aged under two. The targets were percentages, but the target population was not as simply defined as with cervical cytology. The number of immunisations necessary in a given year to achieve full coverage of a GP’s patients depended on the number of children in different age groups. A rate of 90% achieved the high target and a rate of 70% achieved a low target. The high rate was approximately £3000 and the low rate approximately £1000 in 2003/4.

nGMS contract. In the new contract the payment system remains the same. However, it appears there has been a significant increase in value of the target payments. The high rate became £7,965 and the low rate, £2,655 (when calculated for 2003/4 levels).

2.3Influenza Vaccination

Pre 1990 and GMS contract to 2000. GPs were not directly paid for influenza vaccinations but they were reimbursed at a fixed rate intended to cover the costs of the vaccine. Since they could buy the vaccine at a price less than the reimbursement rate they could boost their income via flu vaccinations. Their payment function (reimbursement less payment for vaccine) may have yielded an increasing marginal payment if the unit price they paid fell with a higher volume purchased. The incentives were broadly similar to an explicit fee for service regime.

In 1998 the DH recommended flu vaccination for the over-75’s (Breeze et al, BMC Family Practice 2004) but with no change in the reimbursement scheme.

In 2000 the recommendation was extended to all over-65s and an an item of service fee of £6.45 per vaccine was introduced. Health authorities (not individual practices) were set a target of 60% of all over 65s to be vaccinated but there was no financial incentive for HAs or practices associated with this target(Chief Medical Officer, 2000).

In October 2003 the recommendation was extended to include high risk groups under-65 including chronic respiratory and circulatory diseases and diabetes (Amendment to GMS SFA October 2003). The item of service fee was also extended for vaccination of these extra patient groups.

nGMS contract. The flu vaccine incentive payments remained unchanged in the new GP contract.

3Some theory: targets versus fee for service

Patients have private benefits and costs of prevention provided by GPs. These determine their demand for preventive care from GPs and their compliance with GP prescriptions of preventive care. The private benefits and costs may differ from the social benefits and costs for contagious diseases, if there are caring externalities and because patients do not bear all the costs of ill health. The GP incurs a cost of prevention and may have an altruistic concern about private or social benefit. She will also bear some of the costs of future ill-health.

The private and social benefits and costs vary across patients. It is often argued that GP cost and social benefit are positively correlated so GPs give low priority to the most deserving (socially beneficial) cases.

The regulator (DH or PCT) will have some information about patients (age, gender, post-code and so can condition the payments on patient type. The regulator will also have information about the practice characteristics and could condition payments on, for example, the size of the practice. But benefits and costs will also depend on factors unobservable to the regulator. The regulator does not know the realised benefits and costs of particular patients (otherwise the incentive design problem is trivial), just their joint distribution conditional on the factors observed by the regulator. GPs better information about benefits and costs since they know more about their practices and their patients.

To illustrate some of the issues we consider with a simple model. We assume that patients are passive. Then the distinction between social and private benefit is irrelevant as patients make no decisions. The social benefit from prevention is b: the discounted sum of the value of health benefits and treatment costs. The only prevention cost, denoted k, is incurred by the GP. F(b,k) is be the joint distribution of benefits and prevention costs across patients where . GPs know the realised value of k for each patient. They also know the distribution function. GPs are selfish and care only about their income: revenue R from the incentive scheme minus their costs of prevention. (Since GPs are selfish we do not need to specify whether they observe b or not: it has no effect on their decisions.)

The regulator knows the distribution function but not the realised values of b and k for treated patients. The regulator can only observe the number x of patients treated and so the payment to the GP can only depend on x: R= R(x)

Since patients impose different prevention costs, the GP cost function is determined by the GP’s choice of patients to treat. The GP will always select x patients to minimise cost at any x. Thus the GP will select patients in increasing order of unit cost k until she has x patients. Let be the unit cost of the marginal patient, defined by

(1)

where fk, Fk are the marginal density and distribution function.

The total and marginal cost of treating x patients is

, (2)

and marginal cost increases or decreases depending on the sign of

Using (1) we see that

(3)

(since the cheapest patients are selected first), but we need to make assumptions about the conditional density function to sign .

The welfare function is the sum of patient benefits minus GP costs:

(4)

where

(5)

and marginal benefit from treating more patients is (using (3))

(6)

The marginal benefit is positive but whether it increases or decreases with x depends on the form of the distribution.

The GP objective function is R(x) – C(x). The regulator’s problem is to choose the function R(x) to maximise welfare. The second best optimum has x = x* where x* = argmax{B(x)-C(x)} and satisfies

(7)

The first order conditions are necessary but not sufficient in the absence of assumptions about the distribution function. Notice that optimum is restricted by the regulator’s poor information. The GP will be treating patients for whom bk and not treating patients for whom bk.

In Figure 4 first best prevention treats patients in areas  and  in part (i). The second best treatment regime is to treat x* patients but patients in  who should be treated are not and those in  are treated but should not be.

The second best can be achieved by an infinite number of forcing contracts which give the GP a higher net income at x* than elsewhere. In particular any target scheme with a target payment with TC(x*) paid if and only if xx*,or any fee for service scheme with R(x) = R0+ p*x C(x*) where p* = , will achieve the second best.

Thus with this specification there are no grounds for preferring fee for service to target payments. The latest GP contract predominantly uses targets to incentivise GPs. This suggests one immediate aim for theoretical modelling: determining under what, if any, circumstances targets yield higher welfare than fee for service.

This requires enriching the model sketched above to include some additional features of primary care to examine their implications for optimal fee for service and target payments (for example they may imply increasing the number of target steps or a multi-art fee for service regime) and which may imply that one type of contract is better than the other.

multiple types of practice. Practices have different distributions of benefits and costs. Does this imply that one incentive scheme is preferable to the other? Does this depend on whether the type is observable? Does it depend on whether there can be as many different contracts as types?

multi-tasking. GPs provide many different types of activity and information conditions differ across the activities. Holmstrom and Milgrom (1991) have shown that this can lead to incentive schemes which are low powered. Does it also imply that some of schemes should be linear and some targets?

sequential effort. GP treat patients sequentially during a period and may be uncertain about how many they will be able to treat in a given period. Holmstrom and Milgrom (1987) have shown this type of specification can lead to the optimal contract being linear ie fee for service.

multiple periods. GP actions in one period will provide information about her type to the regulator and, in the absence of commitment by the regulator, the GP take this into account when chosing action in any period (Laffont and Tirole, 1988). Does this imply that one regime is better than another?

exception reporting. With fee for service GPs have only one choice variable: the number of patients treated. But with targets defined in terms of the proportion of eligible patients treated they have an additional choice variable: the declared number of eligible patients. GPs can achieve the target proportion by reducing the declared number of eligible patients by removing patients from their lists or by falsely declaring them ineligible (exception reporting). Thus the target scheme may give rise to unproductive gaming by GPs.

practices as teams. Most GPs work in profit sharing partnerships and so the interaction of the contract and the internal profit sharing arrangements must be considered. Holmstrom (1982) shows that no profit sharing scheme with budget balance (sum of shares always sums to one) can induce the efficient amount of unobservable effort from partners. But if the regulator uses target payment mechanism this finesses the budget balance problem and can induce efficient effort with no shirking by partners.