A patient recently asked me “what’s your numbers?”, “What’s your biopsy rates?” What she was referring to was my numbers for skin cancer biopsies. But what did she really mean and do the numbers make a difference?
SCARD is an online database and surgical audit tool used by millions of doctors and other healthcare practitioners around the world. It is used to facilitate auditing and record-keeping and is a fantastic resource for anybody involved in skin cancer.
It tracks all your specimens (that you submit) and allows easy simple audit, and provides you with the ‘numbers’. But what are those numbers and are they useful?
To see if the numbers are useful or not, we must first understand the numbers. What statistics are we referring to? And how are they useful?
Positive Predictive Value (PPV)
‘The probability that subjects with a positive screening have the disease’. In other words, if you test positive, what’s the likelihood that you actually do have the condition you are being tested for – the chance the positive results are correct.
Negative Predictive Value (NPV)
‘The probability that subjects with a negative screening truly don’t have the disease’ – what is the chance that if you test negative then the result is accurate and you really don’t have the condition. In other words, the negative results are correct.
Ideally, all tests should have a high PPV and high NPV but there’s more to it than this though.
This refers to the chance that if you have the disease, the test will pick it up.
The likelihood that if you are clear from the condition then it will give you a negative result.
Confused? Let’s use an example
Geoff is worried he has skin cancer. He visits me in the clinic. I do a biopsy because I think he has skin cancer.
PPV is the chance that I’m correct and he has skin cancer and sensitivity is the chance that if he does have skin cancer that I pick up his skin cancer with the biopsy.
What are the average results?
Your results are different – why?
This is the crux of the matter. The results by themselves are meaningless. They lack context. You could say that I am poor at picking up melanoma. You could be correct, or you could be very wrong.
If your numbers near 100%, you are either extremely good or extremely badUnknown
Let me explain. If all you biopsy are barn-door, obvious to anybody, skin cancers, then you will have a very high pick up rate. Look at these pictures.
If all you ever biopsy is the one on the right then you will have a very high PPV and sensitivity, but you will potentially be missing thousands of early melanomas. That is not good for patients, as by that point it could have grown and spread.
Conversely, if you take out every mole you see then you will have a low PPV and specificity, but you will find lots of skin cancers by virtue of removing so many moles.
Another variable is the training and the work the doctor is doing. Skin cancer doctors tend to have results similar to mine. There is a lot of selection bias and often the patients will want a biopsy regardless of what we suggest, mostly because they are so worried.
Skin cancer doctors and dermatologists will tend to have patients presenting much earlier as well when compared to GPs or family doctors, so it will be less obvious that they are skin cancers and we may have a lower PPV and specificity, even though they may actually be better at picking them up early.
What’s the best approach?
Impossible to say, and this is where the clinician must make a decision with the patient, explaining the risks and benefits and other options if there are any. Generally, you would want to be somewhere around the 65% mark but it depends what your level of experience and risk tolerance is. For melanoma, it is usually around 25%.
So do the numbers matter? Yes, but not in the way you think they might.