Calibrating cases

Question 1

what is case definition

Answer 1

criteria that determines whether someone has disease

Question 2

how do sensitivity/specificity + -ve/+ve predictive values differ

Answer 2

PPV/NPV depend on prevalence while sensitivity/specificity do not

sensitivity = how many actually have disease and tested +ve (few false -ve)
true +ve/ true +ve and false -ve
specificity = how many don’t have disease and tested -ve
true -ve/true -ve and false +ve

PPV = likelihood that +ve result actually has disease
true +ve/true +ve and false +ve
NPV = likelihood that -ve result actually doesn’t have disease
true -ve/true -ve + false -ve

Question 3

how does prevalence affect PPV/NPV

Answer 3

as prevalence increases, PPV increases (so less false +ve) + NPV decreases (so more false -ve)

Question 4

what is type 1 + 2 error

Answer 4

type 1 = false +ve, doesn’t have condition so is overtreated

type 2 = false -ve, has condition so is underrated

Question 5

what is case-control study design

Answer 5

pt picked based on outcome status - retrospective study

1 group has disease, other does not

useful if rare condition (low prevalence)
bad if rare exposure - do cohort study instead

Question 6

what is odds + RR + odds ratio

Answer 6

odds = with disease/total

odds ratio = odds in exposed vs unexposed
(can convert OR into RR but not vice versa)

RR = done in cohort studies as outcome unknown

cant calculate AR absolute risk in case-control as total population size unkown

Question 7

is odds-ratio ever true value

Answer 7

NO - always an overestimate