Lecture 4. Screening for Disease

Question 1

Because diagnosis a population-level problem, what do we need to know?

Answer 1

How individual values are distributed (among populations and within a population)

Question 2

What is a case definition?

Answer 2

A set of criteria used to decide whether an individual has the disease of interest

Question 3

What are case definitions a combination of?

Answer 3

Clinical signs (e.g. diarrhoea, rash, rapid breathing, broken limb)
Diagnostic tests (one or many) e.g. bacterial culture, ELISA, X-ray, MRI scan, PCR

Question 4

What is the aim of defining cases?

Answer 4

To create one unique set of criteria that defines a disease with 100% accuracy in an individual on every occasion
Also make sure every country defines diseases in the same way

Question 5

What are the limitations to case definition?

Answer 5

Misclassification - diagnostic tests are often not 100% accurate
Often not one unique set of criteria (required when multiple infections occur together – e.g diarrhoea – even if Salmonella typhimurium is identified we cannot be sure these are no other causes involved)
Sometimes no known cause e.g TSEs (transmissible spongiform encephalopathies, prions)
If case definition relies on post-mortem findings, can only make the diagnosis after death
Definitions, symptoms etc. can all change

Question 6

What limit the use/development of perfect case definitions?

Answer 6

Scientific knowledge
Time
Money

Question 7

What is epidemiology used to make decisions about?

Answer 7

Groups not individuals
A case definition might be good enough despite errors depending on the purpose

Question 8

What do we do when we know how a test is wrong?

Answer 8

Account for the error when estimating incidence and prevalence

Question 9

What is a dichotomous result?

Answer 9

Can only be positive or negative (e.g pregnancy or covid)

Question 10

What is a true positive result (a)?

Answer 10

Have disease and have positive test result

Question 11

What is a false negative result (c)?

Answer 11

Have disease but have a negative test result

Question 12

What is a false positive result (b)?

Answer 12

No disease but have a positive test result

Question 13

What is a true negative result (d)?

Answer 13

No disease and have a negative test result

Question 14

Ideally, which groups do we want all samples to fall into?

Answer 14

a and d

Question 15

Since we don’t really know who really has the disease, how do we evaluate the new diagnostic tests?

Answer 15

We evaluate new diagnostic tests by comparing them to a gold standard test (=best available test for disease, often invasive)

Question 16

What is the sensitivity (Se) of a test?

Answer 16

The probability of a positive test result given the presence of the disease - How good is the test at identifying diseased
a/(a+c)

Question 17

What is the specificity (Sp) of a test?

Answer 17

The probability of a negative test result given the absence of the disease - How good is the test at identifying non-diseased
Properties of a specificity test shouldn’t change
d/(b+d)

Question 18

What is the test accuracy of a test?

Answer 18

The probability of correct diagnoses provided by a test – varies with disease prevalence as well as test sensitivity and specificity
(a+d)/n where n = total

Question 19

What is true prevalence?

Answer 19

The proportion of the population that is truly affected with the disease
(a+c)/n

Question 20

What is apparent prevalence?

Answer 20

The proportion of the population that is diagnosed as having the disease based on the results of a diagnostic test
(a+b)/n

Question 21

How to calculate Apparent Prev

Answer 21

(Se * True Prev) + (1 - Se)*(1 - True Prev)

Question 22

How to calculate True Prev

Answer 22

(App Prev + Sp - 1)/(Se + Sp - 1)

Question 23

To estimate prevalence what is required?

Answer 23

Se and Sp

Question 24

In what setting are sensitivity, specificity and accuracy used?

Answer 24

In a public health setting

Question 25

In what setting are positive and negative predictive values used?

Answer 25

In a clinical setting

Question 26

What is the predictive value positive (PVP) of a test?

Answer 26

The probability that a subject has the disease given that the subject has a positive test result
a/(a+b)

Question 27

What is the predictive value negative (PVN) of a test?

Answer 27

The probability that a subject does not have the disease given that the subject has a negative test result
d/(c+d)

Question 28

What is the calculation for PVP utilising Se ,Sp and Prev?

Answer 28

(Se * Prev) / (Se x Prev + (1-Sp) x(1-Prev))

Question 29

What is the calculation for PVN utilising Se ,Sp and Prev?

Answer 29

(Sp * (1-Prev)) / ((Sp * (1-Prev) + (1-Se) * Prev))

Question 30

What is the relationship between prevalence and predictive values?

Answer 30

Sensitivity and specificity are constant for any given test but PVP and PVN vary with prevalence

Question 31

As prevalence increases, which predictive value becomes higher?

Answer 31

PVP

Question 32

If specificity is high, what prevalence value can we be more confident at diagnosing at?

Answer 32

At a lower prevalence value

Question 33

For testing a population for a rare infection (screening), what should be used to get a good PVP?

Answer 33

A very specific test

Question 34

The relationship between PVP and prevalence means a testing program is most efficient when directed at which group?

Answer 34

A high-risk target group

Question 35

What does the ‘ideal’ testing program depend on?

Answer 35

How the results of the tests will be used
The epidemiological context (prevalence)
The costs associated with false positive and false negative results

Question 36

When do you choose a test with high sensitivity?

Answer 36

The consequences of a false negative are worse than a false positive

Question 37

What are examples of highly sensitive tests?

Answer 37

HIV testing of human blood for donations (don’t mind throwing it away by mistake, don’t want to give it to somebody incorrectly)
An eradication program where it is important to ensure that all positive animals are culled even at the expense of culling some healthy false positive animals

Question 38

When do you choose a test with high specificity?

Answer 38

The consequences of a false positive are worse

Question 39

What are examples of highly specific tests?

Answer 39

When an cow tests positive for bTB, the herd is depopulated
Pregnancy testing (you do not want to tell people they are pregnant when they are not) - maybe not pregnancy

Question 40

What does simultaneous testing mean?

Answer 40

Individuals are considered to be positive if they are positive to any of the tests
Increases net sensitivity but tends to decrease net specificity of the combined tests

Question 41

What is an example of simultaneous testing?

Answer 41

Testing cattle for brucellosis (which has been eliminated from UK, so really want to find the positives)

Question 42

What is sequential testing?

Answer 42

Only run second test if individuals test positive on the first test. Individuals are considered to be positive if they have positive results to all tests
Increases net specificity of the combined tests at the expense of net sensitivity

Question 43

What is an example of sequential testing?

Answer 43

Used for HIV and cancers – only the positives are tested with the next test to reduce the false positive rates

Question 44

When is sequential testing typically used?

Answer 44

When one test is less accurate but cheaper or less invasive