Skip navigation

Risk

Bandolier 26 raised the question "should we have and use a "risk ladder" to help us, the public, and politicians understand the level of risk associated with a particular action or event". Recently, with BSE and other issues we have been told that risks are "very small" - but without attempting quantification.

Bandolier is therefore indebted to Dr Peter Iredale, Chairman of Oxfordshire Health and a member of Wolfson College for drawing our attention to a useful book based on the Wolfson lectures on risk held in 1984 [1]. The book, "Risk: man-made hazards to man", with contributions from Hermann Bondi and Richard Southwood, is good reading. The chapter by William Inman (for many years connected with safety of medicines in the UK) on risks in medical interventions is compelling reading.

Rules

One useful quotation that Inman gives is from Lord Rothschild in his 1978 Dimbleby lecture. Rothschild said that when anyone makes a statement about an accident or the number of people involved in it, ask two simple questions:

  1. Is the risk stated in a straightforward language that I can understand, such as 1 in 1,000? If not, why not?
  2. Is the risk stated per year, per month, per day, or per some period of time?
If not, I shall ignore the information.

Scale of risk

Inman also proposes and defines a risk ladder which can take into account the very wide confidence limits that should be put on any guesstimate of risk. This makes a logarithmic scale useful. The downside is that since risk has to start with certainty - 1 in 1 - this anchors the scale so that higher numbers on the risk scale represent lower orders of risk.

The scales would, for instance, put the risk of death for the UK population in any one year from any cause at 1 in 90 in risk level 2, with road traffic accident death (1 in about 14,000) in level 5, and being struck by lightning in level 8.
Risk level 1 in: Range (per year) Risk of death from disease Violent/accidental death/other
1 1-9
2 10-99 Any cause of death
3 100-999 Cancer, stroke, coronary
4 1,000-9,999 Peptic ulcer
5 10,000-99,999 Arthritis, asthma, diabetes, cirrhosis Road accident, burns, falls, suicide
6 100,000-999,999 Pregnancy, STD Homicide, railways, aircraft
7 1,000,000-9,999,999 Tetanus, measles, whooping cough Falling objects
8 10,000,000-99,999,999 Acute rheumatic fever. Acquiring "new" form of CJD Lightning, animal/plant venom, Winning the lottery
Data taken from mortality statistics in England & Wales in 1981
Of course the risk levels in the table are only crude and refer to the whole population of England & Wales. Risks can be modified by behaviour (there is a great quote in the book about keeping the working place safe so employees can go hang-gliding at the weekend), or by circumstance. The risk of dying in any one year is modified by age or by having or not having a particular disease.

In patients with specified diseases, the risks can increase, and Inman goes on to show how the use of the table of risk levels can be used to look at risks of treatments. There are some well reasoned pages about reactions to adverse drug events which are well worth a read.

BSE

Where does the BSE risk come on the scale? There were 4 or 5 cases of the "new" disease in humans in 1994 and 1995 - so the crude risk is about 1 in over 10,000,000 - just inside the level 8 band and about the same as being struck by lightning or winning the lottery. That crude risk is independent of causation, of course, as well as being the earliest indication of possible risk level that could change with time.

Remember 3/n

Bandolier 23 carried a short article on calculating the risk of something happening that we haven't yet seen. If none of 100 patients receiving an intervention has a problem that concerns us, then we can be 95% confident that the chance of this occurring is at most 3 in 100 (3/n).

Reference:

  1. MG Cooper (ed). Risk. Man-made Hazards to Man. Clarendon Press, 1985. ISBN 0-19-854154-6 or ISBN 0-19-854154-4 (paperback).



previous or next story in this issue