I'm writing an article on feline immunodeficiency virus and was trying to figure out the difference in sensitivity and specificity when discussing diagnostic testing. THIS.HELPED.SO.MUCH. thank you :)
Hi Physiotutors, I have great admiration for your work and your easy-to-understand on difficult concepts and pathophysiologies and myth debunking and all that; Thank you for all your contributions to the Physio world!! I have a question regarding the definition of Specificity in this video. The most commonly accepted definition of Specificity is "the proportion of patients without the disease who test negative" which is mathematically expressed as ( TN ) / ( TN + FP ). This was the same mathematical formula for Specificity in your "calculate Sensitivity and Specificity" video. However, in this video, the definition of Specificity you shared here is "the extent to which a positive test result really represents the condition/ disease of interest.." which would then be mathematically expressed as ( TP ) / ( TP + FP). However, this formula represents Positive Predictive Value. The definition for Specificity in this video also matches very closely to the definition of Positive Predictive Value in your "How to Calculate Positive (PPV) and Negative Predictive Values (NPV)" video; which I quote "PPV tells you exactly how likely it is that the patient has the disease after he tested positive". I hope you may clarify on this confusion when you can? I understand from your Instastories that you all are terribly busy. I hope to hear from you all soon!
Hi Javier, happy to hear and thanks for the compliment! Agree with your first definition and you are right that the definition we give is actually the PPV, so this is something we should probably fix in a remake! The problem with giving definitions in case of sens and specs confuses people a lot, which is why we tried to avoid this as much as possible in this video. In case of a 100% specific test, you will have no false positives so what we tried to focus on is the fact that in a 100% specific test you can be 100% sure that the person who tests positive will have the disease.
@@Physiotutors Hi Physiotutors! thanks for looking into this and clarifying my doubts! Yup, I totally agree that in the case of a 100% specific test, because there are no false positives, therefore 100% of the people who tests positive will have the disease. However, in a scenario where Specificity isn't 100%, the rate at which "a postive result means the person has the disease" (which is the worded definition for PPV) is not equal to the rate at which "the proportion of people who do not have the disease test negative" (which is the worded definition for Specificity) A simple illustration of the above point (based on the illustrations of coloured dots and circles in this video for Specificity): TP = 3 red dots with green circles FN = 2 red dots without green circles FP = 1 white dot with green circle TN = 2 white dot without green circle *compared to the video's illustration, only the numbers for FP and TN have been modified to represent 66% Specificity so, Specificity = TN / ( TN + FP ) = 2 / ( 2 + 1) = 2/3 = 66% PPV = TP / ( TP + FP) = 3 / ( 3 + 1 ) = 3/4 = 75% Yup, I was really confused about Specificity too before watfching this video and I see now from your reply what you mean that "in the case of a 100% specific test, ..100% of the people who tests positive will have the disease." is easier to understand compared to the true worded defintion. However, I think that a wrongly worded defintion will have implications in understanding the rate reflecting Specificity if it weren't 100% (as illustrated in the modifed example above and most test aren't 100% specific). But anyway, what eventually helped me understand was your explanation and illustration of Sensitivity and the pictorial illustration for Specificity was also clear and helpful! It was only the worded definition of Specificity that threw me off because it didn't seem to reconcile with your illustration but your reply here really helped, so I really appreciate you taking time to help me understand and reconcile the error :) (sorry for taking up your time; will be catching up on your vieos/ posts on myth busting and stroke now)
very strange informations.. why in other videos on youtube sensitivity is TP/TP+FN ?? This means, that sensitivity is good at detecting a disease! In your videos i saw that sensitivity is about rule out the disease.. so? Whats wrong? I think that many people dont understand "yours" sensitivity and specifity.. please answer
This is something I struggle to get my head round. Is my thinking correct in the example below?: To rule a condition/disease out use SNNOUT To rule out further an enhance validity use SPINN any help would be greatly appreciated
+Matt Shutt first part is correct. however, SPINN is used to confirm or rule in. In other words, a positive test outcome makes the condition u are testing way more likely.
Thanks for the quick response as always. So in your example at the end of the clip your not performing both tests in unison, one is an example of how to rule out due to it being highly sensitive and the other being highly specific so could be an example of a test to rule in?
+Matt Shutt no worries! You can still combine them and use both. However it is important to realize their potential and to be able to interpret outcomes, resp. Understand what it means if those tests are positive/negative
to make sure i am understanding if we have a 100% sensitive test this means that all diseased we be included in the positive tested group and all the tested negatives are really no diseased , but on the other hand if we have 60% specficity for this test that mean many undiseased will be included in the tested positive group?? hope i am right
I don't understand this... Surely if you have a test that is 60% specific, considering the SPPIN rule, if the test is positive then this will rule that condition in only 60% of the time. Therefore the other 40% of the time that the test is positive it is actually picking up a positive that is false - people that actually do not have the disease. So there is a 40% chance of a false positive (FP) outcome with the test? But then in the video, you talk about how you cannot have FP with a specific test.I could look at it the other way and see that as the specificity is only 60%, 40% of the time the test will not be positive. However, this 40% could indicate how the test is falsely missing the disease and is therefore falsely negative. Am I utterly lost? Please help? Thanks
Agree, the definition of sensitivity can be confusing. Imagine a very sensitive alarm that goes off with the tiniest bit of movement or noise - it will detect everything. So if this alarm doesn't go off (= negative test) then you can be sure that your house is safe.
I guess if you are accurately ruling out a disease, you could then on the reverse say that what you are then left with are those people that do have the disease? Does that work? I'm so confused...
Lemme explain using alarms. A sensitive alarm or test is one which picks up the slightest movement or slightest presence of disease. However anything can trip the alarm- a thief, a child or even a dog. A specific alarm would only trip up when a thief is present. Now for safety you want a sensitive alarm but you only want to shoot a thief not a loved one. So when the alarm rings you watch the CCTV feedback to see what tripped the alarm. So the CCTV is specific while alarm is Sensitive.
To explain the No false Positives for the Specific alarm here a CCTV. You don't want to shoot who you think is a thief but actually NOT a thief. So that is a False positive. Therefore for sensitivity you want reverse no False negatives. So if anything that seems like a thief even if its wind blowing pls Let the alarm ring!
Physiotutors so accdg to your example can we say, if a patient is (+)cross SLR he is (+) disease, and if he is (-) SLR he (-) disease? Am i right?? Pls enlighten me thanks
does it make sense to do tje cslr after a negative SLR? I don't think so. the -LR ( which is more important than validity btw) is so low that performing the cslr after a neg SLR is of no additional value. therefor: screening tests are meant to be highly sensitive.
Im still so confused by this! Why, if we had a test which was 100% sensitive would we not use it to rule people in and out? Why could we not just use a positive test as if its showing 100% sensitive, why would we need to only use for a negative test?
Coz with 💯 sensitivity you can still have lots of false positives. You don't want to confirm that someone has a disease while there is still a chance that they don't have it. It's easier to understand if you watch our video on the calculation of sens and specs
so I understand that when a test has a specificity of 100%, you can be 100% sure a patient has the disease when the test is positive. But how sure can you be when the test has a specificity of 98%? Can you be 98% sure? Or how sure can you be then? how to estimate that?
Hey Arne, specificity 100% --> 100% a patient has this disease (because the positive likelihood LR+ is unlimited) . A test with 100% specificity basically does not exist in real life. To make the calculation about how high the chance actually is that your patient has the disease (after a positive test) has to be made based on positive predictive value (PPV) or LR+, because that depends on sensitivity as well, as well as disease prevalene and findings from history. As you see, it's not that simple unfortunately. I suggest that you watch our two videos on PPV and likelihood ratios and then get back to us if you have a question. In this way we can answer more specifically. Maybe check out our post to fully (and finally) grasp this complex topic: www.physiotutors.com/how-statistics-will-make-you-a-better-physiotherapist/
Thats what I thought. But what is the clinical value of Spin and Snout then. For example: I often read something like: "If thesensitivity is extremely high, we can be sure that a negativetest will rule the disorder out." But if sensitivity alone, is only reliable to rule out a disease when the sensitivity is 100% (and in reality there are hardly any test with 100% sensitivity) doesnt the Snout Rule might be misleading?
We can never completely rule out or rule in any disease - not even with radiographic imaging and other laboratory tests. This is just clinical uncertainty that we have to deal with. Sppin and Snnout are just a mnemonic to help people with the complex matter of statistics. Most often, with a very high sensitivity (>90%) you will have a low LR- if the specificity is not very low. The same is true for a very high specificity (>90%) --> usually very high LR+. There are some cases where we should apply those rules with caution. A good example is given by Pewsner et al. (2004): www.ncbi.nlm.nih.gov/pmc/articles/PMC487735/
Thank you very much. I now read the article on Wikipedia about sensitivity and spezificity. Spins and Snouts are discussed under Misconceptions. I think, sometimes it might not help to try to simplify complex topics with the tradeoff of risking false interpretation. Thank you very much for all the efforts. The videos and answering questions too.
thank you so much for explaining this in words rather than the 2x2 table!!!!
Me and Kiesh are sat in the cold watching this before our exam and it helped a lot! Cheers Kai
I'm writing an article on feline immunodeficiency virus and was trying to figure out the difference in sensitivity and specificity when discussing diagnostic testing. THIS.HELPED.SO.MUCH. thank you :)
Sir you are an angel sent from above for us Physios.Great lectures,very helpful..keep it up!!!
Thx a lot,happy you got it!
greatest lecture on specificity and sensitivity for a physio
Appreciate it! Thanks Tayab!
Thank you so much. I tried 4 videos and still did not understand the concepts until I saw this. Very clearly and easy to apply. *thumbs up*
Glad you found the video then :) maybe you have a couple friends who could benefit the same way!
Same. I finally get it! Loved the snout and spin mnemonics will help remembering these concepts. Thank you!!
Omg... thank you. I struggled so much with this in PT school
Glad it helped!
Hi Physiotutors, I have great admiration for your work and your easy-to-understand on difficult concepts and pathophysiologies and myth debunking and all that; Thank you for all your contributions to the Physio world!!
I have a question regarding the definition of Specificity in this video. The most commonly accepted definition of Specificity is "the proportion of patients without the disease who test negative" which is mathematically expressed as ( TN ) / ( TN + FP ). This was the same mathematical formula for Specificity in your "calculate Sensitivity and Specificity" video.
However, in this video, the definition of Specificity you shared here is "the extent to which a positive test result really represents the condition/ disease of interest.." which would then be mathematically expressed as ( TP ) / ( TP + FP). However, this formula represents Positive Predictive Value. The definition for Specificity in this video also matches very closely to the definition of Positive Predictive Value in your "How to Calculate Positive (PPV) and Negative Predictive Values (NPV)" video; which I quote "PPV tells you exactly how likely it is that the patient has the disease after he tested positive".
I hope you may clarify on this confusion when you can? I understand from your Instastories that you all are terribly busy. I hope to hear from you all soon!
Hi Javier, happy to hear and thanks for the compliment!
Agree with your first definition and you are right that the definition we give is actually the PPV, so this is something we should probably fix in a remake! The problem with giving definitions in case of sens and specs confuses people a lot, which is why we tried to avoid this as much as possible in this video.
In case of a 100% specific test, you will have no false positives so what we tried to focus on is the fact that in a 100% specific test you can be 100% sure that the person who tests positive will have the disease.
@@Physiotutors Hi Physiotutors! thanks for looking into this and clarifying my doubts! Yup, I totally agree that in the case of a 100% specific test, because there are no false positives, therefore 100% of the people who tests positive will have the disease.
However, in a scenario where Specificity isn't 100%, the rate at which "a postive result means the person has the disease" (which is the worded definition for PPV) is not equal to the rate at which "the proportion of people who do not have the disease test negative" (which is the worded definition for Specificity)
A simple illustration of the above point (based on the illustrations of coloured dots and circles in this video for Specificity):
TP = 3 red dots with green circles
FN = 2 red dots without green circles
FP = 1 white dot with green circle
TN = 2 white dot without green circle
*compared to the video's illustration, only the numbers for FP and TN have been modified to represent 66% Specificity
so,
Specificity = TN / ( TN + FP ) = 2 / ( 2 + 1) = 2/3 = 66%
PPV = TP / ( TP + FP) = 3 / ( 3 + 1 ) = 3/4 = 75%
Yup, I was really confused about Specificity too before watfching this video and I see now from your reply what you mean that "in the case of a 100% specific test, ..100% of the people who tests positive will have the disease." is easier to understand compared to the true worded defintion. However, I think that a wrongly worded defintion will have implications in understanding the rate reflecting Specificity if it weren't 100% (as illustrated in the modifed example above and most test aren't 100% specific).
But anyway, what eventually helped me understand was your explanation and illustration of Sensitivity and the pictorial illustration for Specificity was also clear and helpful! It was only the worded definition of Specificity that threw me off because it didn't seem to reconcile with your illustration but your reply here really helped, so I really appreciate you taking time to help me understand and reconcile the error :)
(sorry for taking up your time; will be catching up on your vieos/ posts on myth busting and stroke now)
Thank God for you
Great video! I have always been confused about these terms but now I have a quite clear picture.Thanks👍
Great to hear Fatema! That was our goal!
Thank you
very strange informations.. why in other videos on youtube sensitivity is TP/TP+FN ?? This means, that sensitivity is good at detecting a disease! In your videos i saw that sensitivity is about rule out the disease.. so? Whats wrong? I think that many people dont understand "yours" sensitivity and specifity.. please answer
11/10 10/10 for the explanation and +1 for the pig
This video is great!! Thank you!
You’re welcome 😇
Great job, Kai. Very well explained
Amazing video! Thanks
Thank you too!
This is something I struggle to get my head round. Is my thinking correct in the example below?:
To rule a condition/disease out use SNNOUT
To rule out further an enhance validity use SPINN
any help would be greatly appreciated
+Matt Shutt first part is correct. however, SPINN is used to confirm or rule in. In other words, a positive test outcome makes the condition u are testing way more likely.
Thanks for the quick response as always. So in your example at the end of the clip your not performing both tests in unison, one is an example of how to rule out due to it being highly sensitive and the other being highly specific so could be an example of a test to rule in?
+Matt Shutt no worries! You can still combine them and use both. However it is important to realize their potential and to be able to interpret outcomes, resp. Understand what it means if those tests are positive/negative
finally, i got it. thanks a lot
to make sure i am understanding
if we have a 100% sensitive test this means that all diseased we be included in the positive tested group and all the tested negatives are really no diseased , but on the other hand if we have 60% specficity for this test that mean many undiseased will be included in the tested positive group?? hope i am right
Yes that's correct!
I don't understand this... Surely if you have a test that is 60% specific, considering the SPPIN rule, if the test is positive then this will rule that condition in only 60% of the time. Therefore the other 40% of the time that the test is positive it is actually picking up a positive that is false - people that actually do not have the disease. So there is a 40% chance of a false positive (FP) outcome with the test? But then in the video, you talk about how you cannot have FP with a specific test.I could look at it the other way and see that as the specificity is only 60%, 40% of the time the test will not be positive. However, this 40% could indicate how the test is falsely missing the disease and is therefore falsely negative. Am I utterly lost? Please help? Thanks
If sensitivity is high then what is the need for specificity test?
Thank you very much for all of yours videos!
That was very helpful
great video! does anyone know the Name of the paper referenced?
great videos guys, love to learn with them
finally a good answer,thancs
glad you found what you were looking for!
Hello
If af sensitive test is good at ruling out a disease, why do we then say that it good at detecting a disease?
Isn't that a contradiction ?
Agree, the definition of sensitivity can be confusing. Imagine a very sensitive alarm that goes off with the tiniest bit of movement or noise - it will detect everything. So if this alarm doesn't go off (= negative test) then you can be sure that your house is safe.
I guess if you are accurately ruling out a disease, you could then on the reverse say that what you are then left with are those people that do have the disease? Does that work? I'm so confused...
Lemme explain using alarms.
A sensitive alarm or test is one which picks up the slightest movement or slightest presence of disease.
However anything can trip the alarm- a thief, a child or even a dog. A specific alarm would only trip up when a thief is present.
Now for safety you want a sensitive alarm but you only want to shoot a thief not a loved one. So when the alarm rings you watch the CCTV feedback to see what tripped the alarm. So the CCTV is specific while alarm is Sensitive.
To explain the No false Positives for the Specific alarm here a CCTV. You don't want to shoot who you think is a thief but actually NOT a thief. So that is a False positive.
Therefore for sensitivity you want reverse no False negatives. So if anything that seems like a thief even if its wind blowing pls Let the alarm ring!
Thanks for another solid video!
Glad you enjoyed it!
Physiotutors so accdg to your example can we say, if a patient is (+)cross SLR he is (+) disease, and if he is (-) SLR he (-) disease? Am i right?? Pls enlighten me thanks
Yes, with a positive crossed SLR the patient is likely to have the disease and with a negative SLR he is unlikely to have the disease!
does it make sense to do tje cslr after a negative SLR? I don't think so. the -LR ( which is more important than validity btw) is so low that performing the cslr after a neg SLR is of no additional value. therefor: screening tests are meant to be highly sensitive.
hello,how can we resolve the issue of false negatives in the specific test that are sent home
Can only be "resolved" in a test that is highly sensitive as well
Great video
this helped so much! thank you :)
thx for this!
Im still so confused by this! Why, if we had a test which was 100% sensitive would we not use it to rule people in and out? Why could we not just use a positive test as if its showing 100% sensitive, why would we need to only use for a negative test?
Coz with 💯 sensitivity you can still have lots of false positives. You don't want to confirm that someone has a disease while there is still a chance that they don't have it. It's easier to understand if you watch our video on the calculation of sens and specs
@@Physiotutors Thanks very much for getting back to me. I will give the other video a watch too
Thank youuuuuuuuuuuu
You’re welcome 😊
Thank u brother
+sameer mohd no worries l!
Nice voice
So there is definitely a trade off between sensitivity and specificity
Definitely. Check our video on likelihood ratios that combine the two
Really I appreciate all the efforts you exert ❤️, perfect content
Thank you 😍
so I understand that when a test has a specificity of 100%, you can be 100% sure a patient has the disease when the test is positive. But how sure can you be when the test has a specificity of 98%? Can you be 98% sure? Or how sure can you be then? how to estimate that?
Hey Arne, specificity 100% --> 100% a patient has this disease (because the positive likelihood LR+ is unlimited) . A test with 100% specificity basically does not exist in real life.
To make the calculation about how high the chance actually is that your patient has the disease (after a positive test) has to be made based on positive predictive value (PPV) or LR+, because that depends on sensitivity as well, as well as disease prevalene and findings from history.
As you see, it's not that simple unfortunately.
I suggest that you watch our two videos on PPV and likelihood ratios and then get back to us if you have a question.
In this way we can answer more specifically.
Maybe check out our post to fully (and finally) grasp this complex topic:
www.physiotutors.com/how-statistics-will-make-you-a-better-physiotherapist/
Thats what I thought. But what is the clinical value of Spin and Snout then. For example: I often read something like: "If thesensitivity is extremely high, we can be sure that a negativetest will rule the disorder out."
But if sensitivity alone, is only reliable to rule out a disease when the sensitivity is 100% (and in reality there are hardly any test with 100% sensitivity) doesnt the Snout Rule might be misleading?
We can never completely rule out or rule in any disease - not even with radiographic imaging and other laboratory tests.
This is just clinical uncertainty that we have to deal with.
Sppin and Snnout are just a mnemonic to help people with the complex matter of statistics. Most often, with a very high sensitivity (>90%) you will have a low LR- if the specificity is not very low.
The same is true for a very high specificity (>90%) --> usually very high LR+.
There are some cases where we should apply those rules with caution. A good example is given by Pewsner et al. (2004): www.ncbi.nlm.nih.gov/pmc/articles/PMC487735/
Thank you very much.
I now read the article on Wikipedia about sensitivity and spezificity. Spins and Snouts are discussed under Misconceptions.
I think, sometimes it might not help to try to simplify complex topics with the tradeoff of risking false interpretation.
Thank you very much for all the efforts. The videos and answering questions too.
Great video. Thanks for keeping everything research based, even in the comments. Cheers!
can you come to my uni to teach by any chances
So the higher the specificity the more reliable the test is?
Sensitivity and specificity deal with the VALIDITY of a test. The higher the values (ideally both) increases the validity of a test.
a couple is 2, not 4. Not trying to be "that guy" but sort of important when talking about numbers.
He looks like Michael Fassbender.
I thought specificity was the true negative rate
The video ended prematurely
I love his body just saying
👍👍
is he single
Thank you so much sir