Advertisement
If you have a new account but are having problems posting or verifying your account, please email us on hello@boards.ie for help. Thanks :)
Hello all! Please ensure that you are posting a new thread or question in the appropriate forum. The Feedback forum is overwhelmed with questions that are having to be moved elsewhere. If you need help to verify your account contact hello@boards.ie
Hi there,
There is an issue with role permissions that is being worked on at the moment.
If you are having trouble with access or permissions on regional forums please post here to get access: https://www.boards.ie/discussion/2058365403/you-do-not-have-permission-for-that#latest

Question about specificity and sensitivity

  • 24-03-2012 10:03pm
    #1
    Registered Users, Registered Users 2 Posts: 9,810 ✭✭✭


    Do specificity and sensitivity of a given test changes with the prevalence of a disease.

    What i want to know is:
    You know when they quote specificity and sensitivity for an investigation in medical texts. Shouldn't the prevalence of the disease that it applies to also be quoted (disease prevalence varies from place to place). Or is a given prevalence implicit in any quoted specificity and sensitivity figures/

    eg say a test has 99% specificity for picking up a disease.
    Say the disease's prevalence is 1/100.
    That means .99 (or approx 1) true +ve and .99 (or approx 1) false +ve.
    Does this reduce specificity from 99% to 50%

    Or does the 99% specificity figure take account of the 1/100 prevalence.
    If so, how would specificity be independent of varying prevalences (say the prevalence was 1/1000 or 1/10000 in the example above- you get different results)
    Thanks.


Comments

  • Registered Users, Registered Users 2 Posts: 2,320 ✭✭✭MrCreosote


    Sensitivity and specificity are independent of the prevalence.

    You're thinking of Positive predictive value and negative predictive values which definitely depend on prevalence.

    Your example is looking at the positive predictive value- what proportion of positive results are true positives. The prevalence or pretest probability is needed to get an accurate result for this.

    In the clinical work, PPV and NPV are more important, but a lot of the pretest probabilities are based on guesses or hunches.


  • Registered Users, Registered Users 2 Posts: 9,810 ✭✭✭take everything


    MrCreosote wrote: »
    Sensitivity and specificity are independent of the prevalence.

    You're thinking of Positive predictive value and negative predictive values which definitely depend on prevalence.

    Your example is looking at the positive predictive value- what proportion of positive results are true positives. The prevalence or pretest probability is needed to get an accurate result for this.

    In the clinical work, PPV and NPV are more important, but a lot of the pretest probabilities are based on guesses or hunches.

    So, specificity of a test can be quite crude when it comes to rare diseases (unless it's of an order of (opposite) magnitude similar to prevalence)? Sensitivity (false -ve divided by total -ve) is still accurate with rare diseases yeah?

    Given that problem with specificity. what is the best measure of a test's usefulness in picking up rare diseases.

    But the PPV. How do you calculate that. And should it be considered before choosing a test. Or should it only be applied after you get a test result.


  • Registered Users, Registered Users 2 Posts: 2,320 ✭✭✭MrCreosote


    So, specificity of a test can be quite crude when it comes to rare diseases (unless it's of an order of (opposite) magnitude similar to prevalence)? Sensitivity (false -ve divided by total -ve) is still accurate with rare diseases yeah?

    Given that problem with specificity. what is the best measure of a test's usefulness in picking up rare diseases.

    But the PPV. How do you calculate that. And should it be considered before choosing a test. Or should it only be applied after you get a test result.

    No- sensitivity and specificity depend on the test, not the disease. Even if the prevalence is one in a million, if the test is always positive for that one it is highly sensitive (100%).

    However unless the specificity is also very high, the rare true positives are going to be swamped by the false positives- in this scenario the PPV is very low. So you could argue that for a rare illness the specificity is more important, but it's still not directly related to the prevalence. Gold standard would be 100% sensitive, 100% specific of course, but in the real world...

    The wikipedia page is pretty good, especially the 2x2 graph:

    http://en.wikipedia.org/wiki/Positive_predictive_value


  • Registered Users, Registered Users 2 Posts: 869 ✭✭✭Icemancometh


    So, specificity of a test can be quite crude when it comes to rare diseases (unless it's of an order of (opposite) magnitude similar to prevalence)? Sensitivity (false -ve divided by total -ve) is still accurate with rare diseases yeah?

    Given that problem with specificity. what is the best measure of a test's usefulness in picking up rare diseases.

    But the PPV. How do you calculate that. And should it be considered before choosing a test. Or should it only be applied after you get a test result.

    Perhaps someone with a more thorough understanding of epidemiology than me may be more enlightening, but...

    The problem with rare diseases is that they are rare! As such, no one test is going to be fantastic for them (bar something like PCR, or culture, but then, that's only for microbial disease). In reality though, most diagnoses of rare disease take time, as the more likely causes of symptoms are investigated and outruled.


Advertisement