If evidence from health services research was crystal clear and comprehensive, policy making and implementation in healthcare would be straightforward. But health evidence comes in shades of grey, so policy makers, commissioners and providers have to make decisions with a large degree of uncertainty.

Academics, in the pursuit of knowledge and further grant income, often conclude their research confirms the need for more research and can go to great lengths not to be pinned down to firm conclusions. So it is little wonder policy makers, commissioners and managers are seduced by case studies of interventions demonstrating impressive before and after improvements, in spite of the inherent weaknesses of single case studies, which include regression to the mean and the Hawthorne effect.

During my time as the national clinical director for older people’s services in England (2000-08), I was struck by how research evidence was used in policy development and its implementation. To illustrate this, I will draw on observations from the development of stroke services and refer to the development of care coordination for older people, intermediate care and dementia services.

Organised stroke care is well accepted now, so many would be surprised that the first trials of stroke units in the early 1980s showed no benefit. By the mid-1990s further trials showed superior outcomes in stroke units but could not demonstrate whether the benefits related to better acute care, rehabilitation or both. So, in the National Service Framework for Older People (2001), the standard set for stroke services was there should be a stroke unit in every hospital treating people with stroke, with broad parameters for structure and function.

By 2003, evidence from trials showed benefit from early access to stroke units and international comparisons showed better outcomes in countries such as France, which treated stroke as a medical emergency. So we revised the stroke standard to encourage direct admission to specialist stroke units.

The National Audit Office then published a review of stroke services. It used weak evidence, mainly drawn from a single tertiary centre in Australia that suggested access to thrombolysis could reduce mortality in the acute phase. Evidence about the value of thrombolysis was mixed because of the risks and benefits of the treatment which, at that time, only had a licence for use in the UK within clinical trials. Nevertheless the NAO report suggested there was an excess of 5,000 deaths in the UK from failure to implement the treatment. This was seen as a national scandal.

The report was helpful to policy makers in the Department of Health as the treatment would require centralising stroke services into a smaller number of acute hospitals, supporting a wider programme of centralising acute hospital care. The report was the impetus for the development of the national stroke strategy, which continues to underpin quality improvement in stroke care across the whole pathway from prevention to long term support.

Does it matter that these changes were stimulated by the use of dodgy data and perhaps a covert agenda for centralising hospital services? Perhaps not, if judged by the improved outcomes for people with stroke.

People want to believe care coordination by community nurses reduces hospital admissions among frail older people, in spite of the fact that all controlled trials of this intervention had negative results in reducing hospital admissions - although there were other benefits, particularly patient satisfaction.

Failure to follow evidence in this area held back the development of interventions that are proven to reduce avoidable hospital care, including care coordination targeted at adults with specific long-term conditions, care home liaison services and investment in intermediate care services.

Blending the long-term conditions model with well run intermediate care services on a locality basis is, in my view, the best way to achieve reduction in non-elective care needs and improved outcomes for patients.

On a more positive note, evidence gathered by the Treasury showed the economic benefits of early recognition and diagnosis in dementia by reducing long term care costs. This was the breakthrough leading to dementia becoming a top clinical priority for the NHS, paving the way for the national dementia strategy.

Research evidence can, and does, inform policy in beneficial ways. Its impact is greatest when the message is simple and aligned with big policy themes. However, evidence is sometimes ignored or seized upon to fit wider interests, which can lead to disappointing results. Experience teaches health service leaders to be sceptical of the golden bullet and to use evidence from research and case studies wisely.

Ian Philp is honorary professor of healthcare for older people at Warwick University.