So, if you’ve read part 1 (and you should otherwise this may not make sense – you can find it here) then you will hopefully be wondering about probability as it applies to diagnosis. I hope that I have convinced you that when we label a patient with a diagnosis we are usually assigning that label on the basis that the likelihood of them having it is high….., but how high?
Well, to some extent that tipping point for labelling should depend upon the diagnosis we are looking at. At one extreme I often use the generic term for all non-specific paediatric illness (it’s a bit of a virus – and no antibiotics won’t work); at the other extreme I want to be pretty certain that my diagnosis of myocardial infarction is correct as the therapy for the condition (thrombolysis ot PCI) has a risk in itself. So setting the tipping point for labelling is very much disease specific, but as diagnosticians (let’s go back to using that word from now on) we need to have a good feel for the consequences of diagnosis.
The rather simplistic view of diagnostics could be summarised in the following flow chart.
The reality of course cannot be this simple though and there are two fundamental problems in this process, the first is that the diagnostic process is probabilistic and in the groups labelled ‘diagnosis made’ or diagnosis excluded we are really ,mixing up two groups of patients.
- The diagnosis made group includes those who really have the disease (True positives) and also those of who do not have the disease (false positives).
- Similarly in the diagnosis excluded group again have a mixed group of those who do not have the disease (true negatives) and those that we have missed (false negatives).
Now I know that you know this as it will be familiar from med school stats lectures. In fact we are really just talking about the traditional 2×2 diagnostic table that will be familiar to you all. Let’s just review that 2×2 table before we recognise that it’s bloomin’ useless to clinicians.
“Useless you say”, is that not heresy, the 2×2 table is the foundation of diagnostic studies, and to be honest I’ve produced plenty in my time as a researcher. The difficulty is that all 2×2 tables are created in the knowledge of the results of what the gold standard is….which is not what we have when using the new test. So from our perspective when we are using a clinical test we only get the results of A+B or C+D, we cannot differentiate without doing more tests.
This is a situation that we have desctibed as researchers having their head in the clouds, whereas we clinicians have our feet firmly on the ground. Time to put on your glasses and go 3D!
As the poor clinican we are left scratching our heads not really knowing whether our patient does or does not have disease.
In (micro)summary, when we look back at the simplistic flow diagram of diagnosis above we know that the first stage – the stage of making a diagnosis is flawed.
So we have yet another problem – and we’ve not even got onto therapy yet – we must accept that our diagnostic prowess is based on putting patients into categories that are inevitably mixed.
Next we need to think about the consequences that has for us in terms of therapy.
vb
Simon Carley
For all posts in this series click here
Risky Business Part 1 You are not a diagnostician.
Risky Business Part 2 What does a test result really mean?
Risky Business Part 3 Is it acceptable to miss some diagnoses?
Risky Business Part 4 How many patients do we miss (and does it matter)?
Risky Business Part 5 Does a correct diagnosis mean that the therapy will work?
Risky Business Part 6 How many steps to disaster?
Risky Business Part 7. Risk Proximity
Before you go please don’t forget to…
- Subscribe to the blog (look top right for the link)
- Subscribe to our PODCAST on iTunes
- Follow us on twitter @stemlyns
- Like us on Facebook
- Find out more about the St.Emlyn’s team
Pingback: Emergency Medicine - a risky business part 1 - St Emlyns
Pingback: Emergency medicine, a risky business part 7: Risk proximity - St.Emlyn's
Pingback: The BIG questions in EM. Part 2 - Technology. St.Emlyn's - St.Emlyn's
Pingback: When is a Door Not a Door? Bias, Heuristics & Metacognition