Refusing to admit a patient to intensive care is guaranteed to generate media hype about bed provision.But little research has been done on optimum ICU bed numbers. Klim McPherson and Alison Metcalfe report

How many intensive care beds should the NHS provide? If the need is unequivocal, numbers can be determined by relevant morbidity rates with some degree of built-in over-capacity for times of peak need.But if there is uncertainty about appropriate referrals and admissions, this can cause problems of unmet need, lack of referrals in times of high demand, or even overuse.

With the inevitable publicity about the most extreme examples, how can we determine appropriate bed provision for intensive care? At the moment any refused admission to intensive care is inevitably taken as evidence of a need for more beds, but healthcare planners should be aware of an apparent paradox.

The relationship between intensive care provision and refusal rates is central to understanding the debate.

Intensive care is highly technical, expensive and labourintensive care for acutely sick people and is the centre of much speculation about the possibilities of serious underprovision of both staff and beds in the NHS.

1 But little intensive care has been subjected to randomised trials to examine the effects of the extra provision, and there are large local geographic variations in use rate.

2 Policy makers like to believe that provision is determined by clinical need.But in circumstances of manifest clinical uncertainty, need is somewhat subjective and intensive care is no exception to this.

Some key workers in the field have singled out intensive care as being so self-evidently beneficial that randomised trials might be unethical.

3 But too often assumptions are made on grounds of intrinsic (biological) plausibility without the necessary empirical validation of efficacy.

In the US the expenditure on intensive care alone is more than 1 per cent of GDP, while in the UK the figure is more like 0.1 per cent.Since our GDP per capita is lower anyway, the expenditure on intensive care here is around 20 times lower - and clearly there remain serious questions of underprovision.

The basic assumption is that refusal to admit patients to intensive care will give rise to unnecessary deaths, and that the ethical policy solution is to provide more beds or staff to drive refusal rates down.We have reported the relationship between refusal to admit and deaths in some detail.

4Here we discuss the relationship between bed numbers and refusals.

Unmet need The more strategic relationship between intensive care bed provision and the refusal rate in England is of particular interest. In 1993, we carried out a survey of adult intensive care provision in England for the Department of Health.

5We surveyed all 234 intensive care units, receiving replies from around 75 per cent, and sought estimates of the number of patients who had been refused admission due to a shortage of staff or beds during 1992.

There is a strong correlation between refusals and ICU bed provision, relative to the estimated catchment area population, with more refusals in areas with lower bed provision (see figure 1).

The strongest determinant of refusals, as a proportion of referra ls, seems to be the number of beds per head of catchment population.Very roughly,1.5 intensive care beds per 100,000 population is associated with refusal rates of around 10 per cent, while an increase to around 5.3 beds per 100,000 is associated with 4 per cent refusal rates.Thus a four-fold increase in the number of beds might be expected to result in a halving of the refusal rate.

Such a halving, if achieved, would still be regarded as unacceptable, since any refusal to admit appropriately referred patients to intensive care is taken as evidence for underprovision.

So pressure for more beds would still not cease at this level if it was driven largely by refusal as a measure of unmet need.

The implicit political assumption is still that supply responds to need: medical need that depends on relevant morbidity rates and not the supply of services.The observed relationship in our study between bed supply and admissions is shown in figure 2.

Here we see another predictable relationship, where high bed provision is associated with more than three times as many patients admitted per head of population than low bed provision.So to increase beds four-fold might cut refusals by half, but will allow a more than three-fold increase in admissions.

For example, we can imagine two very similar communities of 500,000 people, where one has 7.4 ICU beds and another about four times as many - 26.2 beds. If the observed relationships hold, the first community would admit 500 patients each year to its ICU and the second 1,700, ignoring any cross-boundary flow.

Assuming that the above refusal rates applied - 10 per cent and 4 per cent of referrals, respectively - the number of patients refused admission each year would be 56 and 71 respectively (see box above).

So increasing the number of beds by nearly four-fold between the two communities might increase the number (and rate per 100,000 population) of patients refused admission to intensive care by some 30 per cent.This is presumably because the extra beds themselves create an extra demand, not all of which can eventually be met.

The Audit Commission is, therefore, quite right to postulate independent measures of quality as the arbiter of intensive care provision.

6Survey results Data from our study indicated that approximately 73,000 cases were referred to ICUs in England in 1992.We estimated that 80 per cent of these cases were appropriately referred and that the relative risk of dying among patients refused admission, compared with those admitted, was between one-and-a-half and two times higher.

Based on the results of our study the number of deaths attributable to ICU underprovision could lie between 2,100 and 2,500 a year for appropriately referred cases.

This is in the same order ofmagnitude as deaths from road traffic accidents (but far less than the 138,000 avoidable deaths due to smoking).

The important question is by how much these deaths might be avoided by providing more intensive care beds.

degreesThe changing relationship between mortality and refusal as bed provision increases is crucial for policy makers. It must be true that, on average, refusals at high levels of bed supply must be associated with lower relative risks of death.Therefore the perceived need for more beds, while lessened, still exists against a decreasing attributable consequence on mortality.

This data already indicates a lower apparent attributable mortality associated with refused admission to an ICU than many expected, and also some confusion, both about who should be appropriately referred and who should be admitted.Such uncertainty will be likely to give rise to overprovision both at patient level and at the macro provision level - unless and until independent measures of the effect of intensive care among particular sub-groups can be reliably collected.

Increasingly, the rational response is to find out more about the role of high-dependency beds, and about the attributable effects of important aspects of intensive care in increasing health gain and preventing mortality.

This work demonstrates that seeming ly plausible strategies - played out too often in intense and simplistic media discussion - to provide services that compensate for crudely measured unmet need may simply exacerbate the problem.

But need is generated by health professionals who, like everyone else, may often have concerns over and above the needs of the community they serve.

In a state of manifest uncertainty, it seems sensible to cover the most extreme possibility - which is precisely how overservicing may come about.This may be important in even the most financially constrained systems.

Practice styles have a momentum of their own, much of which may have surprisingly little to do with the objective of improving the population's health.

Key points

Little research has been done on the number of intensive care beds the UK needs, and provision varies widely across the country.

There is no indication of systematic underprovision of intensive care unit beds in the UK.

There is confusion about appropriate referrals and admissions.

Research suggests that with the current number of beds there are 2,100-2,500 avoidable deaths a year.

But it is unclear how many of these would be avoided by providing extra beds.


1 Rowan Ketal , Vessey M. Intensive Care Society's APACHE II study in Britain and Ireland-I: Variations in case-mix of adult admissions to general intensive care units and impact on outcome.Br Med J , 1993; 307: 972-977.

2 McPherson K, Wennberg J, Hovind O, Clifford P. Small area variations in the use of common surgical procedures: an international comparison of New England, England and Norway.NEJM 1982; 30: 13101314.

3 Black N.Why we need observational studies to evaluate the effectiveness of health care. Br Med J 1996; 312: 1215-1218.

4 Metcalfe M, McPherson K. Provision of Intensive care in the UK.The Lancet 1997; 350: 1399.

5 Metcalfe M, McPherson K. Study of Intensive Care in England 1993 .Department of Health, 1995.

6 Aud it Comm iss ion .Critical to Success .Audit Commission, 1999.

Two communities of 500,000 people in England

ICU beds per 100,000 population 1.5 5.3

Average number ICU beds 7.4 26.2

Annual cost @£1,000 per bed day£2.7m£9.6m

Admission rate per 100,000 population 100 340

Expected number of admissions 500 1,700

Refusal rate (as percentage of referrals) 10 4

Total referrals 556 1,771

Number of refusals 56 71

Refusals per million population 112 142

Klim McPherson is professor of public health epidemiology and Alison Metcalfe is research scientist, cancer and public health unit, London School of Hygiene and Tropical Medicine.