The NHS Confederation’s impressive policy salon series launched itself into 2010 with gusto: its January iteration looking at the hardy perennial issue of variations of supply and cost.

The main speaker of Variation of supply and cost – higher quality with fewer resources was a visiting US academic expert, involved in the production of the leading ‘blble’ on healthcare variation in the USA.

He was introduced by a UK policy expert, who touched on the relevance of variations to the current stage of UK health reform, with specific reference to the current health select committee inquiry into commissioning, and its focus on provider dominance. The recent Nuffield Trust report into variation between the four NHS ‘home nations’ differences was also highlighted. Drawing attention to variation is, warned the UK speaker, “unpopular and hard to establish”, but vital to meaningful healthcare improvement – especially the “key message that more is not always better”.

The findings of significant variation in US healthcare had, noted the UK speaker, been replicated in UK studies of the NHS – the first major study, by Newton and others in the Oxford region found huge variations of up to 40% in practice. McPherson’s 1996 work on the whole of England found 90% high-low variation, and Bevan’s 2004 work finding 50-70% high-low variation.

The main speaker described how work on the US reference book had begun in the late 1970s, to examine dilemmas of access, quality and costs, and draw lessons from unwarranted variations in effective care. Unwarranted variation is defined as that which cannot be explained by patient illness; the dictates of evidence-based medicine; and / or patient preference.

One would naturally expect some variation in medicine. Additionally, the sense of patient preference and choice driving decisions in the US remains a very strong notion. Patients value things differently to providers. This reference work is about what is termed ‘residual variation’, that we can lay at the feet of providers.

So these are nuisance factors, to control out so we can assess providers fairly.

The process started in 1973, when Jack Wennberg’s study of Vermont (published in Science) looked at small-area variation in health care delivery. It had very good data, including nursing homes, physician visits, hospital utilisation. Another critical aspect of this was the geography of Vermont - very conducive to studies, with its scattered small town and very small city and rural community population. Small hospitals were evenly scattererd, possibly due to the population being generally loyal to their local hospital, as are physician staff. Loyalty of this kind allowed the study to create hospital service areas – markets for acute medical care. An epidemiologist, Wennberg applied populations and denominators to this work, which was interested in events of medical care, and population-based measurements. The Wennberg study described tremendous variation across a relatively homogenous population. Medical journals would not publish it.

Variations were seen in physician use, bed and physician capacity. The immediate response was that Vermont’s local populations really differed significantly and it was all about what patients needed (and later, questions of what they wanted).

Today, the nationwide US study compares the biggest Los Angeles hospitals against the Mayo Clinic. But variation has remaine3d tremendously high in Vermont as recently as four years ago. (Wennberg surveyed the Vermont population with Fowler to assess their posited health variations in response to criticism of his ’73 findings, and found them quite slight).

In the early 1990s, the team behind the national study sought a way to project their findings into the national publlic policy world much more visibly, as the Dartmouth Atlas (www.dartmouthatlas.org). Wennberg realised that the real audience for this would be public policy thinkers; congressional staffers; and people in Congress. They needed peer influence, but knew the limitations of publishing for an academic saudience, all reading one another’s work in ‘very good research journals and having those we know read our paper’.

The Atlas is not peer-reviewed. Instead, it gives a display of variation, and supplies interpretation and suggests remedies. There have been 20 national editions, some condition-specific. In the first 6 months of 2009, the website had 118 million media impressions, and about 2,000 unique media markets.

Variations are found in per capita spending across hospital referral regions (n=306 for tertiary care). There is influence from state-funded Medicare – for those over-65. US has higher public spend per capita than the UK, but other half of it is not paid publicly.

What role for price and volume?
The US shows a lot of difference. How much is about price, and how much volume?

About 1/3 of differences are in price and 2/3 volume. Some places that are very high in price (New York) are low in volume – contrastingly, Miami’s volumes are remarkably high. A paper coming out in Health Affairs promises “to parse this out in exquiite detail”!

Variations are unwarranted regarding effective care – studies of acute myocardial infarction (AMI) patients receiving beta blockers back in 1995 demonstrated huge variations in technical quality of care. Graphs showed that all the highest-prescribing regions had big medical centres locally. For this type of case, the ‘right’ rate is near-100% prescription of beta-blockers; but the study found most are middle-spread. Information has driven change in this area.

Domains of effective care
The domains of effective care can be mapped as a pyramid, stratified from top to bottom as follows:
nearly completely implemented
partially implemented
proven effective
possibly efficacious
basic science knowledge
and the huge bottom area underneath is health, disease and treatments unknown,

Most of day-to-day medicine is in the bottom area: complex, with a great array of choices, and unexpressed patient preferences which should (but do not) shape decisions made. Yet when a rate of action or prescription is low in this domain, it is viewed as rational - although we know empirically that certain treatments should be done, and populations treated - at the same time, we tend to think where volume is low, this is rational behaviour. The speaker opined, ‘God help us if that’s true, as we’re truly doomed’.

What is preference-based care?
The speaker proposed that delegates imagine a queue for back surgery: the push is to reduce the queue. That appears to make perfect sense, the patient is uncomfortable, clinicians feel they should have surgery, patients believe clinicians (2005 data on inpatient back surgery enrolees per 1,000 Medicare enrolees). Which is the right rate of surgery? They reviewed randomised controlled trials (RCTs) for back surgery at Dartmouth across many academic medical centres, and found that the average patient benefit is very slight if detectable: outcomes were much more largely determined by whether patients received treatment chose using formal decision-making aids presenting risks and benefit impartially and allowing considered decision-making.

A variable 'right' rate
The speaker and his colleagues assert that the ‘right’ rate will vary across different places, but a low rate is not necessarily the right rate. He cited work on arthoplasty in Canada – by any measure, those studied were mostly eligible for major joint replacement, but once they were given fair impartialadvice about joint replacement, only 25% decided that they wanted it at that time.

In understanding the idea of variation, physician preferences and beliefs imparted to the local population are significant.

So healthcare provision is often not aligned with patient and family preference. Hospice use in cancer decdents by hospital referral regions varies hugely. Most patients don’t want to die in acute settings, and would prefer hospice care when they are very ill and the chances of life-extending care are low. The speaker and colleagues show that the US map of what happens to this patient group does not represent preferences, and can prove that the cost of care is inversely associated with hospice use. A lot of money is therefore being spent to get patient’s preferences wrong, not to treat them well and that represents waste. Most money in healthcare is spent on acute care in the last 6 months of life.

Preference-sensitive care
Preference-sensitive care acknowledges that:
healthcare involves trade-offs
decisions should be based on the patient’s own preferences
the effect of supply is variable

The effect of supply is significant. By place, these differences are specific and dramatic. The region where the speaker and colleagues work has low hospice use and less end-of life healthcare intensity.

The speaker concluded that:
1. there is unwarranted variation in supply of healthcare resources
2. healthcare capacity is not located where needs are greater
3. difficulties in matching healthcare resource and need are common
4. mortality metrics can help us to compare end results of care with funding and capacity across health economies, and ask whether there is equity of distribution of capacity in terms of health status achieved.