Who wants to know if their care is rationed? Views of citizens and service informants

Abstract

Objective To explore the views of citizens and service informants about whether they would want to know about any rationing of their own health care.

Design In‐depth interviews using a semistructured schedule. Data were analysed using the methods of constant comparison.

Setting and participants Citizens and service informants.

Main variables studied Issues around health care rationing were explored within the context of the United Kingdom health care system.

Results The views of citizens and service informants were very similar in terms of whether they, personally, wanted to know about any rationing of their own care, with the vast majority wanting to be given this information. Informants were also similar in terms of their reasons for wanting to know about rationing: to be given a ‘good explanation’ to enable them to judge whether the decision made had been correct; and to enable them to change the decision if necessary, either through protest or payment. Many informants suggested that they would indeed be likely to react either by challenging the decision or by paying for care.

Conclusions The findings suggest that policies to be open with people about the rationing of care would be welcomed, but also indicate that if protest follows such openness, it may be difficult for the health service to cope with greater explicitness. Further research is needed among patient groups actually facing this situation.

Introduction

The issue of whether to maintain the implicit rationing that is a feature of most health care systems, or to move towards more overt denials of care, is at the forefront of policy discussion with regard to health care rationing. In the United Kingdom (UK), as in most other nations, rationing is mainly implicit. ^{1 , 2} Rationing is managed by doctors who are aware of limits on budgets, but who restrict care by telling the patient that they are unable to do anything to help them, rather than stating openly that there are no resources available with which to provide treatment. ² , ³ , ⁴ , ⁵ Even in the latest round of UK policy changes, the principle of access based entirely on need has been reiterated. ⁶

Implicit rationing is, however, becoming increasingly unacceptable because of its associations with unfairness and inefficiency. As Mullen points out ‘…it is as if a new value system has arisen which equates explicit rationing with hard, and thus strong, good decision making, while implicit rationing is characterized as soft, and thereby associated with weak, poor decision making’ (Mullen ⁷ , p. 53). The alternative is explicit rationing, a term which has been given a variety of meanings by different commentators and, in its most narrow sense, has been used to refer to the making of lists of conditions which will not be treated. ⁸ More generally, however, explicit rationing is concerned with making clear the decisions that have been made about resource allocation and the basis upon which these decisions have been made. ⁹ When commentators talk about explicit rationing, they do not usually concern themselves with openness at the level of the individual, but greater explicitness at the system level may inevitably lead to greater explicitness at the level of the individual. ¹⁰

Changes made by the two most recent administrations in the UK have increased the pressure for more explicit rationing. Both governments have altered the visibility with which decisions are made: in relation to the purchasing of services through a contracting process by the Conservative administration, ¹¹ and in relation to the commissioning of services through long‐term agreements by the current Labour government, as well as the development of institutions such as the National Institute for Clinical Excellence. ⁶ Both have placed increasing emphasis on the rights of patients through their development of charters ^{12 , 13} and their focus on public and patient participation. ^{6 , 14}

Many have welcomed the opportunity for more explicit decision making, ^{4 , 15 , 16} which, it is stated, can: reduce the burden of responsibility upon those who previously made decisions alone; ^{17 , 18} increase the consistency with which decisions are made; ^{17 , 18} increase equity; ^{17 , 18} increase efficiency; ¹⁶ increase trust (as opposed to deceiving the public); ^{15 , 17} reduce the influence of ‘shroud waving’ and vested interests; ^{15 , 17} and allow the public to judge and influence the values behind the decisions that are made. ^{15 , 17} Interestingly, practical experience has suggested that explicit rationing tends to result in an increase, rather than a decrease, in the basic healthcare package, ¹⁹ and thus the desire to increase the funding available may also provide a motive for increased explicitness.

There are, however, also arguments for maintaining implicit rationing in health care. These have concentrated upon difficulties in ‘counterposing individual against collective wellbeing’ (Hunter, ²⁰ p. 30), the possible disutility from explicit rationing ¹⁰ and the potential impact upon the stability of the health care system. ²¹ Mechanic poses the latter argument, suggesting that explicit rules do not allow for individuals’ strength of preference to be considered, that individuals with the strongest preferences are least likely to accept easily explicit rationing which is not in their favour and that challenges by such individuals will weaken the resolve of health authorities to continue with explicit rationing of health care. Instead they will be forced to return to more flexible, implicit means of rationing care. ²¹ The work of Redmayne et al., ⁸ which shows that UK purchasing authorities who attempted to rule out certain procedures in the early 1990s have since relaxed such exclusions, illustrates this problem, ²¹ as does the case of Child B, where there was a strong family preference for further care and an unwillingness to accept the decision not to pursue further treatment options. ²² The difficulty in devising explicit rationing schemes which take account of the heterogeneity of patients and the need for flexible clinical responses has been noted, ²³ and one empirical study has suggested that explicit policies make little difference to the decisions made and that most rationing decisions could be left to clinical judgement. ²⁴

In practice, explicit rationing of health care is considered mainly from the perspective of the health system, although there is undoubtedly an ideological sentiment that regards greater openness in rationing as a good thing from the point of view of the population generally. What have seldom, if ever, been considered are the views of the public and those working within health care about the extent to which they want rationing to be explicit at the level of the individual patient: whether they themselves want to know if their care is being rationed. This paper reports some early exploratory research which has begun to tackle this subject. The issue was considered as a subsidiary topic in a research project whose main focus was the development of the economic theoretical model of the citizen–agent relationship in health care, ²⁵ this latter project being itself linked to a broader study of citizen participation in health care.

Methods

The methods, particularly the sampling strategies, for this early exploratory work were driven by the two main research projects. The broader study of citizen participation in health care primarily used focus groups as a means of exploring with informants their perceptions of the degree to which the public should participate in different types of rationing decision. Focus groups were held with both citizen and service informants and sampling strategies were used to select informants from the four different types of representation identified by Eyles and Litva. ²⁶ These four types of representation are: descriptive representation, those who represent the characteristics of the overall population; ascriptive representation, those who represent because of their particular knowledge; political representation, those holding political office who are given power to take decisions; and substantive representation, those with a professional role. Citizens were defined as members of the public or their representatives and were recruited to the project using random sampling from the electoral roll (descriptive representation), selection of informants from interest groups (ascriptive representation), and selection of elected representatives from the same localities as used in random sampling from the electoral roll (political representation). Service informants were defined as those sampled on the basis of their health service role and included both health professionals and those working in a voluntary capacity as non‐executive members of decision‐making boards and members of Community Health Councils (substantive representation). Service informants were selected using health commissioners as key informants. They either directly identified individuals who were asked to participate in the research (as in the case of general practitioners, hospital consultants, Community Health Council members, board members) or identified individuals working in health care trusts who were asked to identify further individuals to participate in the research (as in the case of allied health professionals, managers).

The research project examining the theoretical economic citizen–agent relationship in health care used in‐depth interviews, supplemented by the focus group material, to examine issues surrounding decision making in health care. Interviewees were sampled purposively from the groups of citizens (focusing on the descriptive and ascriptive representations of the public) and service informants with the aim of including the range of views about willingness to participate in health care decision making, informants with particular knowledge (including a range of health service roles among service informants) and informants expressing strong beliefs. The issue of wanting to know about rationing, the subject of this paper, was explored as a subsidiary issue with these interviewees.

During interviews topics were discussed as they arose and covered: whether the informant perceived health care as being rationed and how such rationing should take place; the role of the informant and others in health care rationing (in the case of service informants); important aspects of health care; who should make choices in health care and why; whether the informant would want to participate in health care decision‐making; and whether the informant would want to know about any rationing of his/her own care. Questions about whether the informant would want to know about any rationing of their own care tended to be asked towards the end of the interview, by which time the informant had already had the opportunity to discuss issues relating to rationing in some depth (both during the interview and their previous participation in a focus group).

All interviews were tape recorded, fully transcribed using a defined standard ²⁷ and each transcript was checked for accuracy. Analysis used the method of constant comparison. ^{28 , 29} Categories and subcategories were developed, modified and extended on the basis of emerging themes as the analysis was conducted; the technique of ‘questioning’ ²⁸ was used to enhance theoretical sensitivity. Detailed descriptive accounts were formed for each informant and matrices were used to facilitate comparison and contrast between informants. ³⁰ Separate matrices were developed for citizen and service informants and major themes were described across the two groups.

This paper presents findings relating to informants’ perceptions about whether they would want to know about any rationing of their own care. These findings explore issues relating to why informants would or would not want to know about any rationing of their own care; they do not seek to be statistically representative and cannot be generalized quantitatively. Verbatim quotes from informants illustrate the findings. Ellipses are used to denote missing speech —‘umm’, ‘err’, ‘you know’, ‘I mean’— and repeats of words which do not add to meaning are removed without use of ellipsis.

Ethics approval for the entire project was granted by two of the relevant three local ethics committees. The third stated that the nature of the project was such that ethics approval was not required. Written consent was obtained from all informants.

Results

Semi‐structured interviews were conducted with 24 informants (13 citizens, 11 service informants) between July 1997 and October 1998. The 13 citizens were split between those originally selected from the electoral roll (descriptive representation, n=7) and from interest groups (ascriptive representation, n=6). Interviews lasted between 30 minutes and 1 hour 30 minutes. Informants were evenly spread across both gender and age distributions: there were 12 female informants; informants were aged between 29 and 71, with nine aged between 25 and 44, eight aged between 45 and 64 and six aged over 65 (age data missing for one informant). Most informants (17) had visited the general practitioner in the previous year, fewer than half (7) had had an outpatient appointment and none had stayed in hospital as an inpatient.

Findings are discussed separately for citizens and service informants in relation to three themes. The first is whether informants wanted to know about any rationing of their own care, the second relates to the reasons they gave for their views, and the third describes their anticipated reaction. A final section in the findings considers the associated issue that emerged for informants from the service setting: whether they should tell others about health care rationing.

Citizens’ preferences for knowing about health care rationing

There was an overwhelming preference among citizen informants that they should know if their own care was being rationed, with the vast majority being quite clear about their views:

‘They can come out with all this sort of thing when at the bottom line it is all about money. Now I would much sooner be told.’ (Citizen, female, aged 62)

Two informants had responses that differed depending upon the type of decision that was being discussed. One felt that he would want to know about life and death decisions but would look the other way for other types of decisions. In contrast, the other felt that she would not want to know about decisions regarding potentially terminal illnesses but would be able to cope with knowing for other types of decision. One further informant, whose views were less clear, talked about the death of his son in such a way that it appeared that it was important to him to believe that everything possible had been done:

‘We lost our son…, I still accept that there was no other way…I think you’ve got – at the time you was believing that he was being given the best judgement and the best of services that were available…’ (Citizen, male, aged 63)

Despite the overwhelming preference among citizen informants to know about any rationing of their own care, some informants suggested that others might not want to know about the rationing of care and that doctors should use their instincts to decide how much information individuals want to be given in relation to rationing:

‘I still suspect that doctors use their gut feeling, instinct, whatever, as to how much people really do want to know and how much they want it spelt out, and I think that’s right.’ (Citizen, female, aged 50)

Reasons for their views

Almost all informants were able to provide reasons for wanting to know about any rationing of their own care. The reason most often stated was that they would want a ‘good explanation’ as to why the decision had been made. This ‘good explanation’ was in part an expression of rights and in part to enable the informant to form a view about the adequacy of the decision:

‘The more knowledge I think you can get on it, I think the better…I mean it’s your own body. So you’ve got a perfect right to know, what’s happening to it.’ (Citizen, male, aged 68)

‘If there is a decision made I’d like to maybe have some understanding why that decision was made, in terms of the technical side of it, it might be just on a technical basis – you know there’s no hope, ‘you can have this expensive drug, but I can assure you it’s not going to make any difference’…and the risks involved, and then so maybe I could weigh it up and decide whether they were right.’ (Citizen, male, aged 31)

Other reasons were also suggested by informants. Some informants indicated that they would want to know about the rationing of care so that they could do something about the decision if necessary:

‘Well if I know about it perhaps I could kick up a fuss and get it. If I don’t know, then there’s no chance at all, if you do know there might be some chance.’ (Citizen, male, aged 65)

This possibility of changing a decision was also linked with the desire for a good explanation, the implication being that if the explanation was not perceived to justify the rationing then the informant would seek to change the decision.

Other reasons that were suggested less frequently included the need to inform people because exclusions of care would soon become obvious, that doctors have too much power in decision making if they do not have to be open, and that it was patronizing not to be told about rationing decisions.

Anticipated reactions

Not all informants spoke about how they might react to the knowledge that their care was being rationed but a number of informants suggested that they might protest against the decision:

‘You might protest and you go against your otherwise maybe ideas or beliefs for self‐preservation, I don’t know, or preservation of your family.’ (Citizen, male, aged 31)

Some of those who spoke about protesting suggested that this would be against their general views or inclinations, but that if put in that situation they would do whatever was in their means to safeguard themselves and their families:

‘Because we’re all selfish when it comes to that one. When it’s you and your family, that’s really all you’re interested in, at that moment…we are selfish because you read and hear of cases where you disagree with what they’re doing.’ (Citizen, male, aged 65)

Informants also suggested the possibility of paying for care:

I would want to know why and if necessary get the extra care I needed or treatment I needed. But I would be fortunate enough that I could finance it if I wanted to. (Citizen, female, aged 37)

Around a third of the interviewees suggested that they would accept the decision to ration treatment, although all qualified this view:

‘…this is the way that surely that most people would react, you’d kick and scream but at the end of the day you would have to come to terms with the situation.’ (Citizen, female, aged 68)

‘I’d certainly want a good explanation as to why. And if it was reasonable enough I would probably accept it. But…that’s me as an individual.’ (Citizen, male, aged 71)

Two interviewees perceived that they had already experienced explicit constraints on care. One had experienced the rationing of a screening test in early pregnancy, whilst another had experienced the rationing of tuberculosis vaccination for her children. In both cases the informants were aware that equivalent care was available in other locations and both had made attempts to obtain the rationed treatment themselves, with the first purchasing the screening test and the second protesting:

‘I paid over a hundred pounds for it because it was available and it was suggested. Unless you can’t afford it I suppose you do try and have it…we came to the conclusion that we wanted to have the early screening done.’ (Citizen, female, aged 29)

‘I did have a word with my doctor about it…and I had a word with…the hospital as well because I was absolutely horrified when I found out that they don’t get the choice here…’ (Citizen, female, aged 37)

The first of these two informants also spoke about how she would have felt if she had been unable to afford to pay for the screening test:

‘If I…couldn’t afford it then I would have just dismissed that at that stage I suppose. You would feel quite frustrated I think, especially if you did, come long‐term, have problems.’ (Citizen, female, aged 29)

Service informants’ preferences for knowing about health care rationing

As with citizens, most service informants wanted to know about any rationing of their own care:

‘I think I would. I would want to know in great detail about what the options were for my care and my children’s care.’ (Service informant, female, aged 47)

Personally as a member of the public, yes…(Service informant, male, aged 62)

There were, however, two exceptions among this group. For one, his preferences varied dependent upon the type of denial that was being envisaged. He suggested that he would not want to know about the denial of emergency care, such as treatment for severe burns, but would want to know about the denial of elective treatments. One further informant had very different views from all other informants, both citizen and service. He stated clearly that he would not want to know about any rationing of care for himself or his family.

Despite their own general desire to know about the rationing of their own care a number of informants suggested, like citizens, that not every individual would want this information.

Reasons for their views

Reasons for wanting to know about the rationing of care varied. As with citizens, service informants wanted to be able to make their own judgements about the decision that had been made, as well as potentially be able to do something about the decision:

‘Why would I want to know? [pause] It is interesting that, I think that one of the reasons I would want to know is what value is being placed on yourself as a person, what value is being placed on you by a doctor in terms of your condition and what his true opinion might be of your prognosis, which he may not have told you.’ (Service informant, male, aged 67)

‘I’d want to know why. And I would want…to know how I could challenge the decision. Whether I challenged it or not is another matter. But I would want to know that I could. And I would want the opportunity to think about it.’ (Service informant, male, aged 39)

One informant spoke at length about why he felt he would not want to know about the rationing of care. He felt that it would be cruel and inhumane for individuals to know that their care was being rationed:

‘Because if there’s nothing you can do about it and on a lot of these decisions, there isn’t anything you can do about it…it’s rather like saying “Oh well, you could go to the moon, but actually you’ll never be the astronaut that’s chosen”. I really would rather not know about that.’ (Service informant, male, aged 44)

The informant whose preferences for knowing about the rationing of care were dependent upon the type of care analysed his reasons for providing different answers. He suggested that for the more stressful decisions there would not be time to deal with the issue:

‘The only reason I can think of for the difference between those two answers is, is that it’s to do with emotion. That a Cochlear implant type situation is likely to be less immediate, less acutely stressful than a severe burn, and there’s probably more time to deal with the answer and come to terms with it. And to think it through.’ (Service informant, male, aged 39)

Anticipated reactions

Around half of the informants spoke about how they felt they would react to knowing about the rationing of care. Some spoke in terms of their anticipated feelings:

‘If I was told “All this treatment might help you, but you’re not having it,” I’d resent it…’(Service informant, male, aged 44),

‘And I’d quite possibly get very angry about it. I don’t think I’d say “Well, that’s OK. Thank you very much for explaining it so coherently. I’m very pleased that the money is being used to treat, you know, burns patients.” I’d be saying “What about a fair deal for people with severe hearing impairment?” And I would find it very difficult to accept it. Whether I would come to accept it, or whether…I would come to resent it I don’t know…’(Service informant, male, aged 39)

Most informants indicated that they would want to do something about the decision, either by protesting against the decision or by somehow acquiring the funds to pay for care:

‘If a doctor said to me “Oh well this is the drug, it is very expensive it might give you a few months”, I would probably want those few months…I would probably do exactly as this particular patient did [protest].’ (Service informant, female, aged 50)

…make more people aware that it’s being rationed, get more people involved, raise funds, the more people that know about it the more something’s probably going to be done about it so it won’t become rationed. (Service informant, female, aged 30)

Service informants’ preferences for telling others about health care rationing

Interestingly, for service informants, the question of whether they would want to know about rationing of their own care brought up the parallel issue of whether they, in turn, should be telling others about health care rationing. There was a range of views about whether they should be providing this information to others:

…I think very few of us will actually tell the patient “Well, this test, this treatment might help you, but you’ll never get to the top of the priority list, so forget it.” That’s an incredibly cruel thing to do to someone. (Service informant, male, aged 44)

‘It depends on what context it was being talked about, I couldn’t turn round to somebody…on a hospital bed and say “Well sorry we’re not looking after you today because we’ve only got three nurses on so you’ll need to fend for yourself for the day”…’(Service informant, female, aged 30)

‘I think they have a right to a certain level of knowledge, but I also feel that it is our job to make the balance as well, so that you don’t want to say “Oh patient × in two years time we are going to reduce this, we are going to cut this down, oh by the way you will hardly notice it but we are going to tell you anyway”.’ (Service informant, female, aged 50)

…on the whole I think that people should be told the truth and yes they will cope with it…they might have to do a bit of adjustment, and it might shatter a few illusions, but yes, I think they should be told. (Service informant, female, aged 35)

Interestingly, informants who were reluctant to inform patients indicated that it would be less problematic to inform those who, in one way or another, would have the resources to affect the situation:

‘…unless they happen to be an MP and might be able to do something about it…If you knew someone had private insurance or was well‐to‐do…you might mention it.’ (Service informant, male, aged 44)

Two informants had been involved in the explicit rationing of one particular form of care (the provision of aids and appliances), and both suggested that reactions among the public varied, but that the majority had been accepting:

…it varies between acceptance and I guess, although it didn’t feel like it at the time…it felt like “Oh my God I’m being overwhelmed with complaints,” the majority of people were very accepting, and it was the vociferous minority who argued back and…threw the complaints in, of which some were…accepting once an explanation as to “Why is this is happening, where has this come from?” had been given…and then there were people who wouldn’t let go until they had pushed it right to the highest levels they could and arguably who probably ended up clawing back some of the service they’d lost, people that had powerful allies like MPs, like the CHC, like MENCAP…(Service informant, female, age not reported)

Discussion

This paper has reported early exploratory work looking at the views of citizens and service informants in terms of whether they would want to know about any rationing of their own care. Despite the small numbers sampled, it is interesting to note the high degree of agreement among informants. The vast majority of both citizens and service informants did want to know about such rationing, perhaps unsurprisingly given similar findings from research looking at individuals’ information needs concerning terminal illness. ³¹ There was also a strong tendency to suggest various responses which would ensure that the rationed care was forthcoming. The major exception to this pattern was from one service informant who was quite adamant that he would not want to know about health care rationing. Interestingly, this informant was the only one directly involved in the immediate treatment of severe life‐threatening conditions and thus he may have been more aware than other informants of the level of anxiety facing hospital patients.

It should be noted at this point that there are some methodological issues associated with this research, and that the findings must be interpreted in the light of these. The first is that, at times, the hypothetical nature of the question made it difficult for some informants to know how they would respond to the rationing of care:

‘Far enough removed hopefully from a situation like that, I would like to think that I could help to make a rational decision as well on it, but of course, that’s because I’m not in that situation, and…I can imagine that that probably wouldn’t be the case if you did find yourself in that situation, it’s very easy to talk theoretically, but very difficult in practice.’ (Citizen, female, aged 50)

I don’t know what I would do…I think it would very much depend on what was wrong with me and whether we are talking about quality of life of an extra few weeks. (Citizen, female, aged 62)

Although this issue is not problematic in the sense that it was the perceptions of citizens and service informants that were of interest here, and these questions will inevitably be hypothetical for many in these groups, it does suggest a strong need to supplement this research with additional perceptions from patients, for whom the issues will be less hypothetical.

No distinction was made by the researcher about whether care was being denied just to the informant (with other patients receiving this care) or was part of a more general exclusion (such as those envisaged under the remit of the National Institute for Clinical Excellence (NICE)). Informants themselves did not pursue this issue, concentrating instead on how they themselves would react to the denial of care. This may be because NICE was not in existence at the time that the fieldwork was conducted and informants tended to assume that rationing would be ‘by postcode.’ Whether individuals have different feelings about these two situations is an issue worthy of further exploration, given that the two situations, whilst both resulting in the denial of care, are very different in terms of equity.

It is impossible to know how representative the views found here are of the general population, but the high level of agreement among those from a variety of backgrounds and experiences that they would want to know about the rationing of their own health care is inevitably of interest. More research could be conducted on more representative groups to determine the extent to which this finding is replicated across a general population and whether similar findings result in other health care systems. More particularly, it may be important to determine the extent to which those in a service environment are willing to meet the wishes expressed here, by telling people about rationing of their care.

Notwithstanding these methodological issues, the research highlights an issue which is bound to be of interest to policy makers. It suggests that many people want to know about any rationing of their own care and, importantly, that some would anticipate acting on that information either through protest or by paying for care. If, however, more people were informed explicitly that they were not receiving care, there could be a real crisis of confidence in the health care system. Whilst individuals may wish to know about any denial of their own care, a sudden change in policy could potentially lead to an increase in numbers of apparent denials which would almost certainly be negatively reported by the media. People could therefore experience a sudden drop in their confidence that the health system would be able to provide for their own health care needs. Indeed such an outcome could be expected given the concern that was already apparent among citizen informants who, from the media, were aware of cases where care had been rationed and who showed increasing anxiety that the National Health Service might be ‘lost’.

Certainly in the UK, a change in policy towards the explicitness of rationing that had this effect would be at odds with one of the stated health care policies of this government, which is to rebuild confidence in the NHS, ⁶ although it may be that, by clarifying entitlements as well as by reducing expectations, in the long term the policy may be beneficial.

A further problem would concern the extent to which the service could cope with a much more open system. This may not just be in terms of the practical issues associated with instability, discussed below, nor those associated with greater numbers of complaints, but in relation to the emotional impact on staff. A number of informants suggested that they would be angry or resentful if their care was to be rationed, and increasingly open rationing at the patient level may be difficult for staff to deal with.

Given that many informants suggested that they would fight the rationing of their own care, Mechanic’s arguments that explicitness will cause instability in the system ²¹ would seem to be borne out by these findings. This may be particularly problematic if those whose care is explicitly rationed have significant public support and thus are able to apply political pressure. Increased explicitness could well be expected to lead to increased funding generally as the political costs of openness become apparent. ³²

Although exploratory, these findings are important in that they suggest a strong desire for people to know if their own care is being rationed, as well as the potential for a number of people to fight any rationing of which they are aware. Further work is now required to determine the extent to which these findings are replicated among patients.

Acknowledgements

I am grateful to the following: the informants who took part in this research; Stephen Frankel, who supervized the thesis from which this research is taken; Jenny Donovan, John Eyles, Andrea Litva, Kieran Morgan, Mike Shepherd, Julia Abelson, Jo Tacchi and Will Warin who were involved in the broader project from which informants were drawn for this particular research; South and West Research and Development Directorate who funded the broader project; and two anonymous referees for helpful comments on an earlier version. The Department of Social Medicine of the University of Bristol is the lead centre of the MRC Health Services Research Collaboration.