04 Medical diagnosis

Question 1

Diagnostic test

Answer 1

• used to rule in/out a disease
• motivated by symptoms
• used to specify the treatment

Question 2

Screening

Answer 2

• used without indication of symptoms
• test routinely for unrecognised disease (in healthy people)
• not intended to be diagnostic

Question 3

What makes a good diagnostic test?

Answer 3

• classify the correct health status of a patient
• high validity
• cheaper/faster/safer/less side effects

Question 4

What is the meaning of gold standard?

Answer 4

The gold standard is the closest we can get to the real, true health status. The aim is 100% but its often not.

Problems:
- new test -> no gold standard
- gold standard is not perfect itself

Question 5

Sensitivity vs. Specificity

Answer 5

Sensitivity:
- how many true positives are tested positive

Specificity:
- how many true negatives are tested negative

tests results may have psychological effects depending on the disease and treatment possibility -> higher sensitivity or specificity is more preferable

Question 6

What is a ROC curve?

Answer 6

• puts specificity and sensitivity into a relation (x = 1-spez -> false positive rate, y = sens -> true positive rate)
• area under the curve tells how good a test is
• at worst its a straight line -> 50% of results are correct
• 1-spec is the false positive part
• good cutoff values are at the shoulder of the curve

Question 7

Why is the ROC called an operator curve?

Answer 7

• radar in WW II were operated by signal detectors that had to distinguish between noise and enemy

Question 8

What is the predictive value?

Answer 8

• PPV (positive) -> probability of being ill when test is positive
• NPV (negative) -> probability of being healthy when test is negative
• depends on prevalence:
  
  • the higher the prevalence, the higher the PPV/lower the NPV

Question 9

Likelihood ratio

Answer 9

LR+: how much more likely is it that a person with disease has a pos. test than that a person without disease has a pos. test

LR-: how much more likely is it that a person with disease has a neg. test than that a person without disease has a neg. test

Question 10

What is the pre-test probability?

Answer 10

probability to have disease before the diagnostic test -> prevalence

Question 11

What is the post-test probability?

Answer 11

probability to have disease after the diagnostic test with a positive result

Question 12

Relation between odds and proportion

Answer 12

p = a/N -> odds = p/(1-p)

odds = p/(1-p) -> p = odds/(odds+1)

Question 13

From pre- to post-test probability

Answer 13

prevalence (p) -> pre-test odds = p/(1-p)

post-test odds = pre-test odds * LR

post-test probability (pp) = post-test odds/(1+ post-test odds)