Medical Statistics - Sensitivity and Specificity

Anyone any good at the above topic of Sensitivity and Specificity? BIG ask I know but hey just throwing this out there....

I have read a huge amount of links on the internet so they do not help a great deal as I tend to need a human interface to guide me in some subjects and this is one.

Outline is that I am at present writing a medical abstract/poster and have data from a medical assessment clinic I run and wish to prove that you do not necessarily need a specialist consultant doctor to triage a specific group of patients and disease. Clinic A is the main triage clinic and clinic B is the consultants who finalise the disease.

Now the data is say

condition PRESENT / NOT PRESENT

Clinic A 175 / 27
Clinic B 160 / 46

need to if possible gain Sensitivity and Specificity on the 2x2 table above.

May just stick with %'s and fancy graphs! :major

 

LMAO Caliban, cheers for the reply, think I'm stretching my luck on MGs with this one....

Cheers and I have spent 1 week and being dragged to a Lecture on statistics this week (basic so I was told) yeah right! and searched the internet for info and have a excel Spreadsheet that may help if I actually know where I put the data into the correct area.

Think I have it right now, but I really need to chat to a statistician in the University next week.

Why I start things I never know.... seemed a good idea at the time to write an abstract on the clinic I run and its positive aspect to aiding Ophthalmology and ARMD diagnosis.

 

@rikky... its too heavy a subject for Monday and Tuesday also!

@Maxwell, many thanks for the post as that helps a great deal, yes you are correct its to check on what could be two clinics, as the clinic A is the one I run and clinic B will be a consultant panel what reviews all new patients into the service.

And in currently writing an abstract and poster, I wanted to check what I had sent to the panel as active disease (a specific one to my work) against what the panel think. I checked the data from last two years for disease I'm screening for in relation to the screen positives, and matched up every case against the review panels diagnosis

Yes clinic A and B do train together but are at different levels, A are more medical technical, and B are senior clinicians.

Good question on experience vs learning, part of my discussion and conclusion is that very question in that do you need to be a clinician or can you teach others to via medical imaging to diagnose eye disease (or a specific few that can be confidently diagnosed via imaging).

The imaging tests we use do have a pretty good confidence level of testing for specific eye conditions and then IF its the type we are screening for we can rule those in for further imaging tests that can narrow he type of disease and if its potentially active much better. NOW to some cases to is it active or not is where "experience" of a clinician comes in a little more.


The data I have in another table for what clinic A's initial findings are, are at 92% in concordance with clinic B, however I think initially I may have tried to segregate the various eye conditions found over the agreement levels,

**a bit of background to what I do is, is that while the clinic I run (Clinic A) is to screen for suspected active ARMD (nAMD) from Optometrists in the community, however in suspected nAMD comes various other eye conditions as Optometrists in the Community do not all have the scanning equipment as we do, so some false positives are going to happen (well its 48% positive vs 52% False Positives)

The FPs can be Macular Holes, Dry ARMD, Epiretinal Membranes, Cataracts, Diabetes, Cystiod Macular Oedema, Central Serous Retinopathy, etc so those are directed to the specialist clinics that deal with those.

Chatting to a Prof in work yesterday plus reading your links Maxwell, I need to look at the statistics in a different new way, which I'm currently doing.

BUT I am going to continue to get my head around Sensitivity and Specificity as I know at some time I will need to use them, Do have a program in work called SPSS that deals in stats, but need some training on its use.


Again that's folks for the input.

 

Just a quick update on this, I finally settled on a Kappa value statistic for the data as it was suited to it more as the data was a straight reliability test, based on agreement.

Had some help from one of the Phd's in gaining the statistic from SPSS

Plus seems as if the abstract and poster I'm writing may end up if accepted, with me going to Orlando next year for a Ophthalmic conference so I was informed at a research meeting we (I hosted) had the past 2 days.