Thread by @MDaware

@PA_Sertraline @j_thePA say you have a seismograph (earthquake detector). I don't know how they actually work but I assume it's fundamentally some sort of mechanism with a spring that detects the ground shaking --> draws a squiggly line

@PA_Sertraline @j_thePA for our purposes, it's an "earthquake test"

squiggle line = earthquake detected

no squiggle line = no earthquake

@PA_Sertraline put it in the desert in the Nevada where I dunno I guess there are earthquakes and if the ground shakes -> spring bounces, earthquake detected

it SENSES ground shaking and says YES EARTHQUAKE because it is SENSITIVE to earthquakes

@PA_Sertraline based on how tight you set the spring or whatever, of all the earthquakes, it will detect a certain portion

make the spring super sensitive, and it will detect all the earthquakes

@PA_Sertraline thats the equation for sensitivity

(earthquakes detected) / (all earthquakes) = sensitivity

depends on how sensitive the seismograph is to earthquakes

@PA_Sertraline and (in general/kinda simplified terms) once you set the spring, that sensitivity is now an inherent property of the test, and wherever you put it, it will detect that same % of earthquakes

@PA_Sertraline which works great in the NV desert where any ground shaking is an earthquake

now, take the same seismograph and dont change anything and put it on a Michigan Ave in downtown Chicago (back when we had traffic and such)

@PA_Sertraline it will still detect any earthquakes we have

but there's a problem

lots of cars & trucks will drive past

and any shaking it detects isn't specifically an earthquake

@PA_Sertraline specificity is kinda the opposite of sensitivity:

of all the cars that drive by, how many does it call an earthquake?

@PA_Sertraline (patients without disease with a negative result) / (patients without disease) = specificity

@PA_Sertraline like sensitivity, specificity is (basically) an inherent property of a test (how tight I set the spring)

in NV desert that doesnt matter as much because no cars

on michigan ave that matters a lot because lots of cars (normal patients) will set off the seismograph (false pos)

@PA_Sertraline here's a 2x2 table

major issue is sensitivity only includes patients with disease (pink; earthquakes)

spec only includes patients without disease (purple; cars)

@PA_Sertraline and in the real world, we usually have more cars than earthquakes so seemingly good specificity (95%!) can mean a lot of false positives

@PA_Sertraline in this example we have 1000 patients, 10% have the disease

90& sensitivity and 90% specificity

and if you calculate it out, only 1 in 3 positives is a true positive!

(aka positive predictive value = 33%)

@PA_Sertraline keep the test the same but lower prevalence (testing a lower risk population) and it's even lower!

@PA_Sertraline which we understand and deal with all the time, like pulse ox alarms (usually its just the probe isnt in the right spot) or car alarms, which we are just annoyed by and dont even think about theft

@PA_Sertraline I think thats it for now, does that help?