A Better Way of Explaining Risk

I recently came across the book Know Your Chances: Understanding Health Care Statistics. If you want to see through the misleading medical statistics provided by drug companies and featured in news reports, then this is the book for you. The authors have an excellent way of presenting risk which makes it easy to understand what effects (good or bad) any drug or treatment has. The point is to put every risk in perspective, along with related risks. For example, there is no point in knowing that a drug will decrease your risk of stroke by 50%. What you need first is a table:

The table shows the risk of death from various causes for a group of 1,000 50-year old women. The numbers in the table are the expected number of deaths from a particular cause. If an entry contains “-”, then less than 1 person in a group of 1,000 can be expected to die from that cause.

I like this table for a number of reasons:

• It uses whole numbers not probabilities. It’s easier to relate to 4 women in 1,000 than to a probability of 1/250 or 0.0004.
• It shows a selection of risks, so it’s easy to place an individual risk (such as AIDS or breast cancer) in perspective and prioritize it.
• It shows all causes. A new drug which reduces the risk of death from (say) stroke by 50% may sound significant, but this needs to be compared with the overall risk of death. Your chances of death as a 50-year old woman are now 36.5 in 1,000 not 37 in 1,000. If the side-effects of the preventative drug are bad, this may not be a sensible choice given your chances of dying due to other causes.
• It shows the risk for a group of comparable people (50-year old women). A table for the entire population, or for all women is not going to be as helpful as a more specific table. This is especially true of health statistics where probabilities vary with age and gender.

Our field of business risk management is not so very different from that of health. We can identify a number of risks our business faces. We have treatments which may possibly reduce some of those risks. Those treatments typically have side-effects (cost and inconvenience) which may cause us to hesitate before implementing them.

A table showing business risks, set out like the table above, is a good way to explain risks and to start determining what actions to take. Our figures may be approximate and lack the rigor of the Cochrane Collaboration meta-analyses of medical trials, but such a table still makes it far easier to understand what a risk means in practice.

When you see risk figures presented out of context you should be suspicious. Will the elimination or reduction of a risk make any practical difference? Should we be concerned about Risk X or Risk Y? Unless we are given the context, we just can’t say.

We should therefore always try and present our risk figures in their context. We should follow the excellent example above given by Woloshin, Schwartz and Welch, and not copy the more dubious practices of the pharmaceutical marketeers.