Measuring Preventable Harm
Pangioti et al (1) are on a hiding to nothing when they hope to measure the extent of preventable therapeutic harm. To measure something you need to recognise it. Unfortunately the nature of adverse events in therapy is that they can be invisible, even in plain sight. I’d like to tell you how frequent invisible adverse events are, but their invisibility renders that impossible.
This is a central problem in evaluating the net benefits of treatment. It is easy to measure the intended therapeutic outcome of a treatment because it is (usually) unimodal and predictable. You apply a treatment in order to achieve a specific outcome. If the outcome is achieved, the treatment is successful (having used a control to account for spontaneous remission). If it doesn’t, the treatment is ineffective. Not all effective treatments are totally successful, of course, and you may have to measure the outcome on a scale, but generally there is one scale and the measured criterion is obvious.
This is not so of adverse events, because:
1. They may not relate to the pathology or even the system… or sometimes the individual treated (remember thalidomide?).
2. They may result from the interaction of other factors not related to the problem or its treatment, such as other pathology, diet, substance use, environmental hazards, change in weight, pregnancy etc.
3. They may be multiple.
4. They may not relate to the known mechanism of action of the intervention.
5. They may be disconnected in time from the treatment, appearing months or years later.
6. The patient may not recognise that they are caused by treatment.
7. The patient may not report them.
8. The therapist might not spot them.
9. The therapist might spot them but not recognise them as treatment-related.
10. The adverse effect in question may never have been reported, and therefore dismissed as non-treatment related.
11. The adverse effect(s) may be indistinguishable from non-treatment-related pathology.
It would be nice to imagine that only relatively minor preventable harms go unrecognised. But I am always brought back to the following story:
An elderly man had assumed that his recent inability to complete the Times Crossword related to his advancing years. He was unable to take his blood pressure pills for a few days. He was able to complete the crossword! A switch of pills allowed him to continue to solve it.
The upshot of this is that all measures of preventable harm (or even of unpreventable harm) resulting from therapy underestimate the prevalence. This is not in due to inadequate science or the shortage of studies. It is inherent in the nature of the problem.
Finding the exact incidence of preventable harm is therefore impossible. This is not a reason not to be vigilant. But, as always, when contemplating or discussing treatment, it is important to allow for the unknown unknown.
By the way, Papanicolas and Figuero (2) (perhaps inadvertently) use the word “error” when describing preventable harm. I think this is unfortunate. Most preventable harms do not result from “errors”. They are just unfortunate happenstance that occurs even with painstaking clinicians who adopt standard guidelines.
Hon Senior Visiting Research Fellow
University of Leeds
(1) Panagioti M, Khan K, Keers RN et al. BMJ 2019; 366:l4185
(2) Papanicolas I and Figueroa JF. BMJ 2019; 366:l4611
Competing interests: No competing interests