Predictive value: an introduction
Posted on 26th May 2015 by Dabean Faraj
The terms ‘positive predictive value’ and ‘negative predictive value’ seem like difficult and confusing concepts; not least because of how similar they sound. Fundamentally, they are terms used when analysing the data in screening programmes.
In reality, when people undergo screening programmes, the tests arenâ€™t 100% fool proof. There is room for error. ThisÂ significantly affects the screening programme.Â How? Imagine a whole group of people who are screened for a disease â€“ some will test â€˜positiveâ€™ (suggesting they have the disease). And some will test â€˜negativeâ€™ (suggesting they donâ€™t have the disease).
But because the screening tests arenâ€™t 100% â€˜correctâ€™, you cannot assume that someone who has the disease will definitely test positive. Nor can you assume that someone who definitely doesnâ€™t have the disease will testÂ negative. This is where the predictive valueÂ comes in handy.
But what does this mean?
The positive predictive value (PPV) tells you how likely it is for someone who tests positive (screen positive) to actually haveÂ the disease (true positive). It answers the question, â€œI tested positive. Does this mean I definitely have the disease?â€
Equally the negative predictive value (NPV) tells you how likely it is for someone who tests negativeÂ (screen negative) to not sÂ haveÂ the disease (true negative). I.e. it answers the question â€œI tested negative. Does this mean I definitely donâ€™t have the disease?â€
Calculation and interpretation
The following table shows you how to calculate the predictive value. ItÂ can help further understand the concept. After calculating the predictive value, interpreting it is simple. The higher the value for the PPV/NPV, the more accurate the test is. And therefore, the better the screening programme!
As I said before,Â I think that these two concepts can be easily confused. However, the table above illustrate. My advice would be to draw out that table quickly in a situation where you think you may need to discuss or work out the predictive value; I have certainly done this and it has been superÂ helpful!
I found this link really helpful and recommend checking this out as well!