Diagnostics 1&3 Flashcards

1
Q

Definition of a diagnostic

A

A characteristic of an illness or phenomenon (e.g symptoms, signs and biomarkers)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Why do we do tests?

A

To include or exclude disease, support management, prognosis, monitoring or measure of general health

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

5 approaches to defining normality

A

Statistical/Gaussian normality, percentile, risk factor, diagnostic, therapeutic

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What is Gaussian distribution?

A

Normal distribution - defining normality by proximity to the mean in the population (bell curve)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

Percentiles

A

Defining normality by proximity to the median - useful if mean is effected by data skew

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Defining normal by diagnosis

A

Comparing values found in healthy to those found in disease -often overlap though and often levels often raised in sick people without the disease

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

100% sensitivity cut off

A

Positive result in all cases of disease (can get false positives)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

100% specificity cut off

A

Negative in all those without disease (can get false negatives)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

Compromise cut off

A

Minimising false positives and negatives

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Defining normal by therapeutic benefit

A

Normal if no benefit from treatment - must also consider harm of treatment so look at net benefit - useful if immediate treatment not necessary

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What is D-Dimer?

A

Small protein fragment found i blood after a clot. Degradation product of fibrin not found in normal blood

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

Examples of gold standard tests

A

Histology, imaging and genetics

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

Problems with histology

A

Interpretation and observer bias

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Problems with imaging

A

Interpretation, observer bias and incidental findings

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Problems with genetic tests

A

Normal variation, variable penetration and phenocopy (Other genes can cause same phenotype)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

True positive rate (sensitivity) formula

A

TP/(TP+FN)

17
Q

True negative rate (specificity) formula

A

TN/(TN+FP)

18
Q

What is positive predictive value?

A

How likely they are to have the disease if they get a positive result

19
Q

Positive predictive value formula

A

TP/(TP+FN)

20
Q

What is negative predictive value?

A

Chance of not having the disease if they get a negative result

21
Q

Negative predictive value formula

A

TN/(TN+FN)

22
Q

Best measure of test peformance

A

Likelihood ratios

23
Q

What is positive likelihood ratio? Normal cut off?

A

Likelihood of a positive result occurring in disease vs no disease. Normal cut off is 10

24
Q

Positive likelihood ratio formula

A

sensitivity/(1-specificity)

25
Q

What is negative likelihood ratio? Normal cut off?

A

The likelihood of a negative result occurring in disease vs no disease. Less than 0.1 is a good test

26
Q

What is an odds ratio in a binary test? Cut offs? Formula?

A

Odds that a positive result is due to disease vs no disease. 1=useless, more than 1 is useful
(TP/FN)/(FP/TN)

27
Q

Receiver operating curves - axis? grandient?

A

y=sensitivity x=1-specificity - gradient is therefore the positive likelihood ratio at that cut off

28
Q

Roc area =?

A

Discriminatory power of test

29
Q

How do you determine best cut off from roc curve?

A

Apex of graph

30
Q

Generic likelihood ratio formula

A

probability of finding in disease/probability of finding in healthy

31
Q

Bayes theorem

A

The pre-test probability must be considered alongside the test results to calculate the post test probability

32
Q

Clinical gestalt

A

Seeing the patient as a whole and using your experience and intuition

33
Q

Does a test have more or less effect on post test probability if pre-test probability is high/low or uncertain?

A

Uncertain

34
Q

What is the pre-test probability in rare diseases?

A

Always very low - eg phaeo still being very unlikely even after positive plasma metanephrites

35
Q

Do likelihood ratios apply to apply populations and clinical settings?

A

No they must be applied to appropriate clinical settings and populations