Many countries where health care is funded from the public purse, face severe problems with their systems of care delivery. Public funding is often accompanied by public delivery; that is, by hospital and other medical facilities owned and operated by the state. Although, on occasion, such institutions can work successfully, in many other cases they do not. In fact, typically, they provide low-quality care, are inefficient in their use of resources, and are unresponsive to the needs and wants of their patients. In addition, they are usually close to monopolies with patients having relatively little alternative sources of treatment – especially if they are poor and cannot afford whatever private facilities may be available. They are also often directly funded from government, with budgets that are determined historically and that bear little relationship to their performance or activities.

Critics of the public delivery form of health service have linked the fact that institutions owned and operated by the state perform poorly with their monopoly status and the budgetary system. They have argued that if patients had more choice regarding where they could go for treatment, and if the money followed the choice, so that medical facilities would only successfully obtain resources if they successfully attracted patients, then the resultant competition would provide a powerful incentive for these facilities to improve almost all aspects of the service they provide: their quality, their responsiveness and their efficiency. Such a ‘quasi-market’ system would also be more equitable, with choices that are currently reserved only for those who can afford private care being extended to the less well off, and with the resultant rise in standards benefiting everyone.

However, many would not accept these arguments. They would point to the problems that quasi-market systems of this kind face, including the lack of genuine competition in the real world, the difficulty of providing information of a good enough quality to patients to enable them to make sensible choices, and the danger of cream-skimming (the selection, by providers, of easier or cheaper patients to treat). All this, they would argue, would vitiate the alleged advantages of choice and competition, and instead create a system encouraging exploitation and inequity.

This paper addresses some of these issues. It is a compressed version of the arguments in my recent book on the topic (Le Grand, Reference Le Grand2007), to which readers are referred if they wish to pursue the arguments in more depth. The paper begins with a discussion of alternative models for health care delivery, arguing that all have their merits and demerits, but pointing out that there are good theoretical arguments for preferring systems with a strong element of choice and competition in many situations. It then discusses some of the empirical evidence to see whether theory is born out in practice. Finally, it draws together practice and theory to discuss how choice and competition policies in health care can be designed so as to benefit from their advantages without incurring too many of their disadvantages.

Models of public service delivery

The basic reason why many publicly funded health care systems have experimented with quasi-markets in health care is because other models of public service delivery are perceived to have failed in this area. Essentially, there are three other such models, all of which have been tried in one form or another within most health care systems. First, there is the ‘Trust Model’, where professionals and managers are trusted to know what is best for their users, and to deliver high quality services without interference from government or any other source. Then there is the ‘Command and Control Model’, where central management sets targets for providers, rewards them if they succeed in meeting those targets, and penalises them if they fail. And third is the ‘Voice Model’, where users express their dissatisfaction (or satisfaction) directly to the providers. This can be through face-to-face conversations with the front-line providers themselves, through complaints to higher managers, or even through elected representatives.

All of these models have their advantages and disadvantages. The advantages of the trust model are that it requires little by way of monitoring and regulation (after all, an absence of these is in the nature of trust), and that, unsurprisingly, professionals and others working in the system like it, which in turn contributes to higher morale, and, perhaps, therefore, to higher productivity. The principal difficulty with the trust model concerns the assumption concerning provider motivation that is implicit within it. This is that providers are solely motivated by the desire to provide exactly the services that patients want and need, and that they have no more self-interested concerns. In terms of a metaphor I have used elsewhere, they are assumed to be “knights” not “knaves” (Le Grand, Reference Le Grand2006); that is, they are public-spirited altruists (knights), not self-centred egoists (knaves). In fact of course, like everyone else, professionals and managers who work in health systems are actually a mixture of knights and knaves, and inevitably therefore, at times, the service concerned is likely to be organised more in the interests of the provider than of the user.

The chief merit of the command and control model is that it can work – at least in the short term. In the early years of this century, the English National Health Service (NHS) adopted a wide variety of targets coupled with heavy management from the top; in consequence, some key aspects of service delivery (notably patient waiting times) have sharply improved. For instance, in 2002 the target was set that 98% of accident and emergency attendees should be treated, discharged or admitted to hospital within four hours of their arrival (at that time it was less than 20%). By 2005, this target had been achieved – and this despite an increase of over a quarter in the number of people attending accident and emergency admissions in that period (Department of Health, 2005, Statistical Supplement). Another example concerns elective surgery. In 1999, more than a quarter of the relevant patients in England were waiting longer than six months for surgery, and over 4% for more than a year. Key targets were set and by 2005, there was no one waiting longer than a year and only 5% waiting longer than six months (Bevan and Hood, Reference Bevan and Hood2006: 420, Table 1).

But, however well it might work in the short run, command and control creates difficulties in the longer run. Of these, perhaps the most significant is the demoralisation and demotivation of those on the front line of service delivery – especially if they are professionals who are not used to receiving orders and have been trained to believe that they will have substantial autonomy and independence in their work. Other problems include a possible distortion of priorities, and the incentive for ‘gaming’ behaviour of various kinds. The latter can range from a straightforward fiddling of the figures through to more subtle ways of meeting the target by changing behaviour in undesirable ways.

The voice model has its advantages as a means of public service delivery. It takes direct account of users’ needs and wants, at least as they themselves perceive them. Moreover, individual voice mechanisms, especially, can be rich in useful information. Telling providers what is wrong with the service they provide (and indeed what is right with it) can be very helpful to providers who desire to improve – indeed much more so than simply not turning up for appointments or, as with the choice model, just switching to another provider.

However, it too has its difficulties. In a no-choice world of publicly funded health care, patients who are dissatisfied with the quality of the treatment they are getting, or the responsiveness of the medical professionals or managers with whom they are dealing, have only a limited range of options open to them. If there is a private health care sector running in parallel to the public one, they can use that – or, at least, the wealthier among them can do so. Those who cannot afford this option can only complain, either directly to the professional or manager concerned, or to their superiors. In each case, the individual has to depend for a response on the goodwill – indeed the knightliness – of the person to whom they are complaining. In addition to being demanding to undertake, this is a fragile mechanism for improving quality. It offers little or no direct incentives for improvement to the knavish or self-interested professional or manager; and even the knightly, more altruistic, ones do not respond well to being challenged by pushy patients.

Moreover, in so far as complaining works at all, it favours the self-confident and articulate middle classes, thus tending to steer services in their direction at the expense of those for the less well off. The middle classes thus have a double advantage over the less well off. They are better placed to persuade the key decision-makers in the public service to meet their needs. And, if that fails, they can use the private sector. In neither case are equity and efficiency being served.

All of these models of service delivery thus have significant problems associated with them. So what about the model that is our principal concern here – the quasi-market model? It has some clear advantages over the others. Unlike the trust model, it channels both self-interest and altruism to serve the public good. If the money follows patient choices, then the hospital or practice that provides the better service will gain resources; that which provides the inferior service will lose. Whether the unsatisfactory providers are knights or knaves, they will wish to continue in business; the knaves because it is in their self-interest to do so, the knights because they want to continue to provide a good service to needy patients. But, to continue in business, they will have to improve the quality and responsiveness of the service they provide in order to attract patients, as well as the efficiency with which the service is delivered. Unlike the command and control model, it gives freedom and autonomy to professionals and managers, encouraging them to engage in innovation and creativity, and with no outside authority continuously telling them what to do. Unlike voice, in a quasi-market world where patient choice and provider competition is the norm, patients dissatisfied with the general quality of the service they can get from one provider – a hospital or a general practitioner (GP) practice – have the opportunity to go to another who can provide them with a better service. This gives considerable leverage to anyone who does want to voice their dissatisfaction: if the listeners to a complaint know that as the last resort the complainant can go elsewhere, they are much more likely to respond positively to the issues being raised. Choice gives power to voice. Moreover, the facts that now both poor and rich can exit if necessary, and that the less well off are no longer dependant on their ability to persuade professionals to get the service they want, can improve the equity of service delivery.

Of course, there are limitations to the applicability of these kinds of arguments to all forms of health care. Patients who have suffered an accident or are seriously ill are unlikely to be able to make any kind of choice of provider, and may have to rely upon others (attending doctors or ambulance crews) to make the choice for them. Some forms of medical treatment are one-off (your appendix can only be taken out once); in such cases, information gained about the quality of treatment may be of little use in deciding where to go for other forms of treatment. Some people – perhaps an elderly person or one suffering from a debilitating long-term condition – may prefer not to have to make the necessary decisions.

However, the number of conditions where choice is impossible or unwanted should not be exaggerated. Experience of one form of treatment at a hospital can give insights into the quality of care provided for other treatments at the same facility. And the recent British Social Attitudes Survey showed that most people do want choice of medical facility – with, interestingly, larger majorities in favour of choice among the less well off than the middle classes (Appleby and Alvarez, Reference Appleby and Alvarez2005).

So, in theory at least, elements of quasi-markets, especially patient choice and provider competition, can be used to promote responsiveness, quality, efficiency and equity across wide areas of health care. But do things really work out that way in practice?

Choice and competition in practice

Carol Propper and colleagues have recently produced a comprehensive survey of the US and other international evidence concerning the effects of competition between providers, especially hospitals (Burgess et al., Reference Burgess, Propper and Wilson2005). They found evidence that, as the model would predict, competition in the US both reduced costs and increased quality, so long as prices were fixed. But the US also provides examples of where problems can arise. Information provided to patients on quality was often too complex to be used effectively by them or, indeed, by institutional buyers of health care. In fact, it was most widely used by the providers of that care itself, sometimes in ways that may have harmed patients. Providers concentrated on improving what was measured, which was not necessarily that which contributed to health. The fixed price system may have induced ‘cream-skimming’, whereby hospitals try to attract patients whose treatment costs they expect to be below the fixed price they are being offered, and to ‘dump’ patients whose costs they expect to be above that price.

There are also useful lessons to be derived from the British experience with the Conservative Government’s NHS internal market, 1991 to 1997. NHS activity rose faster than resources (and rose at a relatively faster rate than before the reforms). This suggests that overall, despite some well publicised increases in transaction costs, there was an increase in efficiency in the NHS that was attributable to those reforms (Le Grand et al., Reference Le Grand, Mays and Mulligan1998). Moreover, following the partial rollback of the internal market that was carried out by the new Government in 1997, efficiency fell (Le Grand, Reference Le Grand2002). Although many analysts predicted that cream-skimming would cause equity problems, no cream-skimming was observed in practice. The principal equity concern arose from the differences between the two types of purchasers, one of which, GP fund-holders (where GPs held the budget of the hospital care of their patients), was more successful in getting a better deal for their patients. In particular, GP fund-holders were effective in bringing down waiting times, reducing hospital referrals and holding down prescription costs. They were also able to generate improvements in the responsiveness of providers.

There was no evidence of any increase in choice for patients, and although there were changes over the period in indicators of quality such as waiting lists and patient-satisfaction surveys, it was difficult to attribute these to the reforms. One study, however, found an increase in mortality from heart attacks in hospitals that were under greater competitive pressure (Propper et al., Reference Propper, Burgess and Green2004).

Overall, despite some changes in culture, measurable changes were relatively small, and perhaps not as great as was predicted by the reforms’ advocates, or as was feared by their critics. This appears to have been because competition within the market was limited, and this in turn may have been because some of the essential conditions for the market to operate were not fulfilled. More specifically, the incentives for the market players were too weak and the constraints imposed by central government were too strong. This interpretation is reinforced by the fact that the area where there were the greatest changes, GP fund-holding, was the one where the incentives were greatest and the constraints the weakest.

Finally, there is evidence from the choice experiments that were tried out in England under the Labour Government in the early years of this century (Dawson et al., Reference Dawson, Jacobs, Martin and Smith2004; Coulter et al., Reference Coulter, le Maistre and Henderson2005). These offered a choice of hospital to patients who had been waiting more than six months for certain kinds of surgery (chiefly cataracts and hearts). Help was provided with transport costs, and patients were given a ‘patient care adviser’ (PCA) to help them with the relevant choices; points to which I return below.

Take-up of choice was high – perhaps not surprisingly given that the patients concerned had already been waiting six months. Significantly, there was no difference in take-up between socio-economic groups defined in terms of income, class or education, though the unemployed took up the offer of choice less frequently that the employed. The pilots also had a significant impact on waiting times in the areas in which they operate. For instance, in all of England except London, in the six months ending in March 2003 compared with the same period in the previous year, there was a fall of 2% in ophthalmology referrals received and seen, and a 6% decline in the mean waiting time. However, in London, where the choice project was in operation, there was an increase in referrals of 5% to 6%; but this was accompanied by a decrease in waiting times of 17%.

So what are the lessons that can be learned from all this about choice and competition policy design? The principal ones may be summarised under the headings of competition, information and cream skimming.

Competition

For the quasi-market model to work – that is, for it to provide the necessary incentives for greater quality, efficiency and responsiveness – there have to be competitors, actual and/or potential. That is, there have to be alternative providers from which to choose; there have to be easy ways for new providers to enter the market, and, correspondingly, for failing providers to leave or exit from it; and there have to be ways of preventing existing providers engaging in anti-competitive behaviour, such as colluding with one another against the interests of users, or trying to create local (or even) national monopolies. In short, the competition must be real.

The condition concerning the ability to enter and exit the market is important. New providers face barriers to entry into any market. An obvious one in health care is the capital cost of setting up a new facility; this could be quite considerable, especially if high-technology equipment is required. A less obvious one concerns the habits of users; if people are used to being referred to their local hospital then it can be difficult to persuade them to use a new or a different provider. In such cases, it may be necessary to offer some kind of assistance to new providers, for instance, through guaranteeing them a higher price for their services, or guaranteeing a specific volume of business. However, such assistance should be strictly time limited.

Then there is the crucial question of ‘exit’ – or, more generally, how to deal with failing hospitals or other medical facilities. It is critical for the choice and competition model that there is some mechanism for dealing with failure that imposes costs on failing institutions. For if there is no cost to failure, then much of the incentives that are so important for generating the desired outcomes disappear.

Now dealing with inefficient or ineffective providers presents perennial difficulties for all systems for delivering public services. One advantage of quasi-markets in this respect is that failure under choice and competition is obvious. If a hospital or other medical facility is failing in terms of quality, and if it is recognised as such by potential users, then it will not be chosen. In consequence, its revenues will fall and its quality failure will be reflected in a financial failure. The failure will be clear; moreover, since few people will be choosing the facility, it will affect a relatively small number of people directly. Hence, it will not be necessary to have some additional mechanism for checking quality; and, if it becomes necessary to close the facility, there will be relatively few patients affected.

Both the decision to intervene and the intervention itself are undertaken by an agency independent of government. This could be an industry regulator, as in the privatised utilities in the UK where regulators have statutory powers to act to protect consumers in the event that the utility should face financial distress or failure. In those cases, the regulator does not have to wait until a firm becomes insolvent to act, and its powers enable it to approve or reject a financial restructuring.

Also, there is a danger in any market that the actors in the market will behave in ways that damage competition. Examples include agreements to drive up prices, arrangements to divide up the market and not to poach on each other’s territory, and attempts to try to take over competitors so as to create a monopoly. Again, to prevent unwarranted or unhelpful political interference, the answer here would seem to be to have a rule-driven system implemented by an independent regulator. It would, in fact, be sensible for this regulator to be the same as that for deciding upon the entry and exit of providers; for all the relevant decisions are aspects of competition and indeed all are part of the business of making competition real.

Information

Information is crucial to the quasi-market model. More specifically, if patient choice is to act as an effective driver of quality, it is necessary to rely upon the user’s judgement about the quality and responsiveness of the service and necessary for providers to react to choices made on the basis of those judgements.

In health care this is clearly a key issue, since much of the relevant information is of a technical nature that most patients will have difficulty dealing with. And in fact there is little evidence that, when, for instance, presented with information about the quality of outcomes by individual surgeons, patients actually use that information to make the appropriate judgements (Marshall et al., Reference Marshall, Shekelle, Leatherman and Brook2000; Hibbard, Reference Hibbard and Peters2003; Burgess et al., Reference Burgess, Propper and Wilson2005).

However, all is not lost. Even if not driven by patient choice, there is evidence that providers do use published information to improve their performance – even if, as noted earlier, they may on occasion game the system. This may be because of professional pride (the naming and shaming phenomenon discussed in a previous chapter), or because they believe that, although patients do not directly use the information, it will eventually affect patient choice through the impact on their reputation and other less direct factors.

Moreover, there are ways of making information more accessible to patients. One method that has proved very successful in helping overcome the patient information problem and encouraging user choice in the UK was the PCA used in the choice pilot experiments mentioned above. PCAs were trained staff, sometimes with a clinical background, who advised on choice of provider; they also gave advice on other matters, including (from those who were clinically trained) clinical ones, and offered support and reassurance. They were very popular with patients.

Cream-skimming

Any quasi-market system carries within it the danger of cream-skimming: that, instead of users choosing providers, providers choose users and do so on the grounds of cost. So in health care there is the possibility that popular hospitals, perhaps with waiting lists or queues for treatment, will only choose to treat those patients who are easiest or the cheapest to treat.

One possibility for dealing with this is to introduce some kind of stop-loss insurance scheme, whereby hospitals faced with a patient whose treatment costs lie well outside the normal range get allocated extra resources once the cost has passed a certain threshold. These would have to be justified as catastrophic costs (not as the result of poor-quality care). This has the advantage of removing the incentive to discriminate against high-cost patients, but carries with it the problem that the hospitals concerned have no incentive to economise on treatment once the threshold has been passed.

A second possibility is to take the admission decisions completely away from hospitals. Hospitals and other treatment centres would be required to accept whoever was referred to them.

A third alternative is to risk-adjust the tariff system such that higher risk patients have a higher tariff associated with them. A certain amount of this is already going to happen under the payment-by-results national tariff system. If fully risk adjusted, this could eliminate the incentive to cream-skim completely. However, risk adjustment is a complex and difficult business; perfectly accurate risk adjustment is arguably an impossible one. But so long as risk adjustment is not perfect, there will remain an incentive to cream-skim. Risk adjusted payments also provide the incentive to up-code patients to more lucrative high-price categories.

Overall, we do not know whether risk selection or cream-skimming will turn out to be a problem associated with the new quasi-market. As the policy is implemented, an equity audit needs to be undertaken, partly to check on the progress of the supported choice package, but also to assess whether there are adverse equity consequences arising from cream-skimming. If this does emerge as a concern, policy options for dealing with it, such as these, need to be explored.

Conclusion

Publicly funded health care systems that incorporate quasi-market elements, such as patient choice and provider competition, can achieve the ends of health care policy. But they must be properly designed so as to meet the conditions for effectiveness. There must be mechanisms for ensuring: that the entrance for new providers is easy; that exit can take place, and that the relevant decisions are immune from political interference; that patients are given the relevant information and are helped in making choices, especially the less well off; that there is help with transport costs, preferably again targeted at the less well off. And the opportunities and incentives for cream-skimming should be eliminated, either through not allowing providers to determine their own admissions or through properly risk adjusting the fixed price system.