Thursday, March 27, 2008

Cost accounting does not call for randomized statistical method.

Dear Colleagues

I have recently been sent a copy of a report "Constructing the evidence base on the social determinants of health: A guide". It is more than 300 pages long, and I cannot pretend to have absorbed all of its content by a long shot.

But I was interested to find the following:
One of the most influential British texts in the history of evidence based medicine appeared in 1972. This was Archie Cochrane’s essay Effectiveness and Efficiency: Random Reflections on Health Services. Cochrane, himself an eminent physician, argued that health services have a tendency towards inefficiency because of organizational, institutional, demographic and technical factors and a variety of other things including human failure. His principal concern was that there was no agreed way to determine what worked or did not work, and therefore it was not possible to tell whether interventions did more harm than good, or had neutral effects. He also complained that no one could tell how much anything cost, so there was no way of telling what was good value for money and what was not. He advocated the use of the clinical trial and argued that economic appraisal of medical interventions must be undertaken. The randomized controlled trial (RCT), as Cochrane realized, was the most precise way to determine the effectiveness of an intervention. With subjects properly randomized and with investigators blind to which is the experimental group and which is the control group, it provides the best way to determine whether something works and allows bias of various kinds to be controlled to a large extent.
This is very interesting ... both in regard to the lack of understanding about cost ... and the lack of understanding about effectiveness.

More than three decades later the use of randomized trials is commonplace ... and, I would argue, has its place in understanding the science of medicine.

But the same techniques are not, in my view, appropriate in determining costs. Rather it is accounting, and specifically cost accounting that should be used, and this does not make use of randomized statistical methods. Such methods have their place in audit ... but not in accounting.

Furthermore, I would argue that where there is large scale deployment of a public health intervention ... such as IRS or bednets in a malaria control campaign ... the costs should be easy to ascertain, and the outcomes quite easily observable and measurable. If the results are hardly obvious ... and the area is a malaria endemic area, the presumption has to be that the intervention is not appropriate.

Where there are serious health crises ... and the disease is as common and well known as malaria ... it would seem that the result of costly interventions should be easily and quickly observable. If not, then something is wrong.


Peter Burgess

No comments: