Lecture 2 - Dx tests

Question 1

What is False Negative

How do you calculate it?

Answer 1

-Claiming to be healthy and disease-free – when it DOES have the disease!

%FN = (FN / D+)x100%

Question 2

What is False positive

How do you calculate it?

Answer 2

-Claiming to have a disease based on the result of a diagnostic test — when it does NOT have the disease.

%FP = (FP / D-)×100%

Question 3

What is sensitivity

Answer 3

-probability of a test to correctly ID those animals that are INFECTED.

Sn = TP / D+

Question 4

What is specificity

Answer 4

-probability of a test to correctly ID those animals that are NOT infected.

Sp = TN / D-

Question 5

What is reference standard

Answer 5

• a test or criterion that will unequivocally define a specific disease condition

Question 6

What is a Positivity predictive value?

What are the implications?

Answer 6

• proportion of individuals who test positive who actually have the disease

Implication

• values depend on disease prevalence in population.
• can only be inferred to target population if the sample is representative of the disease prevalence in the population –i.e. you don’t have 2:1 ratio of Dz’ed to non-Dz’ed

Question 7

What is the negative predictive value

Answer 7

• proportion of individuals who test negative who are actually free of the disease

Question 8

What is a test?

Answer 8

Any procedure that reduced uncertainty about the state of disease
- PE is also a type of “test”

Question 9

“If an animal tests positive, what is the probability that animal is diseased?” refer to PPV or Sn?

Answer 9

PPV

Question 10

“If an animal is diseased, what is the probability that animal will test positive?” refer to PPV or Sn?

Answer 10

Sn

Question 11

Is PPV the same as Sn?

Answer 11

NO, they are very different!

• PPV is greatly affected by the prior knowledge/ belief of dz prevalence
• PPV have a greater Clinical Application.

Question 12

What is it called when 2 bell curves overlap?

Answer 12

Partial Diagnostic Discrimination

Question 13

How are Sn and Sp related?

Answer 13

Inversely/ negatively correlated

increased Sn / Decreased Sp

Question 14

How is a cutoff value chosen?

Answer 14

Based on the purpose of the test:

1. Penalty of missing a case
   - High Sn used to minimize false negatives
2. Expense/invasiveness of confirmatory test
   - High Sp used to minimize false positives

**Youden’s index

Question 15

Why is the “top left” corner, considered the best cutoff?

Answer 15

“Best” cutoff in terms of making the fewest mistakes when:
- prevalence is about 50%
and
- when Sn and Sp are equally important.

Question 16

What does AUC (area under the curve) tell you?

What would be a perfect test

Answer 16

probability that a randomly selected diseased individual will be regarded with greater suspicion than a randomly selected non diseased individual
-Perfect test – AUC = 1

Question 17

What classifies a positive result, if using multiple tests in PARALLEL?

• What is the net Sn/Sp?
• Should you use a >Sp or >Sn test first?

Answer 17

• ANY positive result = POSITIVE!

Increase SENSITIVITY
Decreases Sp

• Use > Specific test first!

Question 18

What classifies a positive result, if using multiple tests in series?

• What is the net Sn/Sp?
• Should you use a >Sp or >Sn test first?

Answer 18

• ALL results are positive = POSITIVE

Decreased sensitivity
Increases SPECIFICITY

-Use > Sensitive test first!

Question 19

What does Validity mean?

What is a synonym?

Answer 19

• result is UNBIASED
• measuring what the test intends to measure

AKA -Accuracy-

Question 20

What does Precise mean?

What is a synonym?

Answer 20

• Small variability

AKA -Reliability-

Question 21

What can threaten validity?

Answer 21

1. Selection bias
2. Information bias
3. Specification bias
4. Confounding

Question 22

What factors affect precision?

Answer 22

1. Random variation (error)
2. Sample size
3. Study design (efficiency)