Intended for healthcare professionals

Observations Ethics Man

How good a doctor do you need to be?

BMJ 2012; 345 doi: (Published 21 November 2012) Cite this as: BMJ 2012;345:e7858

Re: How good a doctor do you need to be?

I am responding to Dr Shyan Goh's response to my previous rapid response.

First of all my question was a more general one about when it might be considered negligent not to use clinical decision support tools rather than about the specifics of my own company's tool.

It would be more meaningful if Dr Goh had tested the tool itself with a range of cases than basing her judgement on just one sample case. The tool initially returns a good list of 10 possible diagnoses with the time sensitive "don't miss" diagnoses highlighted. If the clinician wants a longer list then he can ‘view all’ and will then be shown all the 34 as mentioned. A search in Google with the same clinical features returned 637,000 results! Readers may recall a study looking at Google as a diagnostic tool where it found an accuracy rate of 50% even after the clinicians had gone through several intermediate steps. An independent study of Isabel, under more realistic conditions, found an accuracy rate of 96%.

I am a little concerned at Dr Goh’s willingness to use ‘any internet search engine’ to help her with a differential diagnosis when these engines are essentially tools to search everything and are not designed to generate a differential diagnosis. Their results are skewed on a continual basis by new content that is added.

We certainly do not regard testimonials from doctors as validation of our tool although a very useful adjunct. Over the last 10 years there have been many papers published (over 30) covering the studies looking at Isabel. Dr Goh is most welcome to read most of them from this link:

Dr Goh’s assertion that a doctor who is well taught will not find the tool useful and will get distracted by ordering unnecessary investigations is not bourne out in practice. We have many very experienced clinicians who find the system invaluable and some cases where they tell us that it saved a patient’s life as they had not thought of a diagnosis that was prominently suggested. In many cases these are wise and forward thinking clinicians who realise the folly of relying on their memory alone and are only too happy to use tools that can help. So far, we have also not found clinicians ordering tests for every diagnosis suggested and, in fact, increasingly they are using the tool to help them do the thinking before ordering tests, thereby decreasing the amount of inappropriate tests ordered.

I would encourage Dr Goh and others to read the research on diagnosis errors to see why these tools should be used routinely. Even if most of the patients don’t have issues with getting a timely and correct diagnosis, an awful lot do. Peter Pronovost and colleagues estimated that 40,000-80,000 people die from misdiagnosis alone in the US and, speaking at the recent Diagnostic Error in Medicine conference, said he now thought that this was an underestimate.

Competing interests: Founder of Isabel Healthcare

26 November 2012
Jason Maude
Isabel Healthcare
Haslemere, Surrey, GU27 1AE