Biostats_6_Sensitivity, Specificity, Predictive value, Screening tests

Question 1

While analyzing the distributions for an appropriate cut-off value for a screening test, what would cause the distribution curves to be tighter? What would be the impact on the sensitivity and specificity?

Answer 1

Increasing the sample size will make the distributions tighter and alter both the sensitivity and specificity. The blue curves have a smaller sample size when comparing the red curves, and appear to improve the accuracy by tightening the distribution. This can be done with a larger sample size. The main effect is seen with the tails crossing the threshold, represented by “X” (to the right indicates false positives and the left negatives). A decrease in their size occurs when analyzing the area under the curve while comparing the blue curve to the red curve. Therefore, the red curves are associated with higher sensitivity and specificity because the magnitudes of false positive and false negative values are reduced.

Question 2

What effects occur when the threshold is decreased (towards point A) ?

Answer 2

The amount of false negative values decrease, leading to an increase in sensitivity.

vvvvvvvvvvvv

The amount of false positive values increase, leading to a decrease in specificity.

vvvvvvvvvvvv

The positive predictive value will decrease, but the negative predictive value will increase.

Question 3

What effects occur when the threshold is increased (towards point B) ?

Answer 3

The amount of false negative values increase, leading to a decrease in sensitivity.

vvvvvvvvvvvv

The amount of false positive values decrease, leading to an increase in specificity.

vvvvvvvvvvvv

The positive predictive value will increase, but the negative predictive value will decrease.

Question 4

Point “A” represents 100% ________ ?

Answer 4

100% sensitivity cutoff value

Question 5

How is sensitivity measured?

Answer 5

TP / ( TP + FN )

Question 6

Another name for sensitivity is the _______ rate

Answer 6

Sensitivity = True-positive rate = 1 – FN rate

This is because sensitivity is the proportion of all people with disease who test positive, or the ability of a test to correctly identify those with the disease.

Question 7

If the sensitivity is high, then the _______ are low

Answer 7

If the sensitivity is high, then the false negatives are low

Question 8

Point “B” represents 100% ________ ?

Answer 8

100% specificity cutoff value

Question 9

How is specificity measured?

Answer 9

TN / ( TN + FP )

Question 10

Another name for specificity is the _______ rate

Answer 10

Specificity = True-negative rate = 1 – FP rate

This is because specificity meansures the proportion of all people without disease who test negative, or the ability of a test to correctly identify those without the disease.

Question 11

If the specificity is high, then the _______ are low

Answer 11

If the specificity is high, then the false positives are low

Question 12

In a population of 100,000 with a disease prevalence of 1%, what are the false positives if the test specificity is 95 % ?

Answer 12

Specificity represents the probability of testing negative in patients without the disease.

vvvvvvvvvvvvvv

In this population of 100,000 people, there are 1,000 people with the disease (100,000 x .01) based on the prevalence. Therefore 99,000 people are free of the disease. If the test has a specificity of 95% then the test would be negative in 95% of these people, which is 94,050 (99,000 x .95). The amount of false positives are found in the remaining 4,950 people (99,000 x.05 or 99.000 - 94,050).

vvvvvvvvvvvvvv

A test with high specificity is typically used as a confirmatory test

vvvvvvvvvvvvvv

A positive test result on a highly specific test would rule in the disease
(SpIN = Specificity, Positive, rule In).

Question 13

In a population of 100,000 with a disease prevalence of 1%, what are the false negatives if the test sensitivity is 90 % ?

Answer 13

Sensitivity represents the probability of testing positive in patients with the disease.

vvvvvvvvvvvvvv

In this population of 100,000 people, there are 1,000 people with the disease (100,000 x .01) based on the prevalence. Therefore 900 people with the disease will test test positive if the test has a sensitivity of 90% (1,000 x .90). The amount of false negatives are found in the remaining 100 people (1,000 x.10 or 1.000 - 900).

vvvvvvvvvvvvvv

A test with high sensitivity is typically used as a screening test

vvvvvvvvvvvvvv

A negative test result on a highly sensitive test would rule the disease out
(SnOUT = Sensitivity, negative, rule out).

Question 18

What is the relevence of the positive likelihood ratio ?

Answer 18

The likelihood ratio is an indicator of test performance.

vvvvvvvvvvvvvv

The positive likelihood ratio is calculated by dividing sensitivity by (1-specificity).

vvvvvvvvvvvvvv

For example, a positive likelihood ratio of 9 indicates that a positive test result is seen 9 times more frequently in patients with the disease than in patients without the disease.

Question 19

Is there any relationship between the disease prevalence and the likelihood ratio?

Answer 19

Unlike predictive values, the likelihood ratio is independent of disease prevalence.

Question 20

What is on the Y-axis in a receiver operating characteristic (ROC) curve?

Answer 20

true-positive rate (sensitivity)

Question 21

What is on the X-axis in a receiver operating characteristic (ROC) curve?

Answer 21

false-positive rate (1 – specificity)

Question 22

ROC curve demonstrates … ?

Answer 22

how well a diagnostic test can distinguish between 2 groups

(eg, disease vs healthy).

Question 23

What does curve “A” represent?

Answer 23

The area under curve “A” is near 1 and the test is very accurate.

Question 24

What does curve “B” represent?

Answer 24

The area under curve “B” is near 1/2 and the test lacks predictive value.

Question 25

What is the purpose of this two receiver operating characteristic (ROC) curve?

Answer 25

A receiver operating characteristic (ROC) curve iillustrates the tradeoff between sensitivity and specificity which is made when choosing a cutoff value for positive and negative test results. The area under ROC represents accuracy of the test (the number of true positives plus true negatives divided by the number of all observations). An accurate test would have area under the ROC close to 1.0 (rectangular shape) whereas a test with no predictive value would be represented by a straight line.

vvvvvvvvvvvvvv

For example, a cuttoff at point “X” would have a high sensitivity and low specificity, a cutoff a point “Y” would have low sensitivity and high specificity. “X” is closer to 100% sensitivity than to point “Y” and “Y” is closer to 100% specificity than to point “X”. Based on these observations, it can be concluded that “X” would require a lower serum marker for a positive test result.

Question 26

For a receiver operating characteristic (ROC) curve, the better performing tests will have … ?

Answer 26

a higher area under the curve (AUC), with the curve closer to the upper left corner.

Question 27

What is the calculation for the positive predicitve value?

Answer 27

PPV = TP / (TP + FP)

vvvvvvvvvvvvvv

Probability that a person who has a positive test result actually has the disease.

Question 28

Does the prevalence impact predicitve values?

Answer 28

Yes. the PPV varies directly with prevalence while the NPV varies indirectly with prevalence.

vvvvvvvvvvvvvv

A high prevalence corresponds to a positive test more likely being a true positive (PPV is high).

vvvvvvvvvvvvvv

A low prevalence corresponds to a negative test more likely to be a true negative (NPV is high).

Question 29

How does a high pretest probability impact the positive predicitve value?

Answer 29

PPV varies directly with pretest probability (baseline risk, such as prevalence of disease).

vvvvvvvvvvvvvv

A high pretest probability corresponds to a high PPV.

vvvvvvvvvvvvvv

For example, you test two patients with HIV. The first patient has many risk factors, the seond patient doesn’t. The first patient will have a higher PPV with a positive test result because the pretest probability is higher. The Second patient without risk factors will not have such a high PPV because the pretest probability is not as high as the the first patient.

Question 30

What is the calculation for the negative predicitve value?

Answer 30

NPV = TN / (TN + FN)

vvvvvvvvvvvvvv

Probability that a person with a negative test result actually does not have the disease.

Question 31

Test efficiency is ….

Answer 31

(TP + TN)/(TP + FN + FP + TN)

Question 32

Prevalence based on TN and FN is …

Answer 32

Prevalence = [(TP + FN) / (TP + FN + FP + TN)]

Question 33

What is the term used for a test that has is interpreted increased survival, but is really detected earlier and the disease course has not changed … ?

Answer 33

Lead- time bias.

vvvvvvvvvvvvvv

This can be rectified by measuring the “back-end” survival (adjust survival according to the severity of disease at the time of diagnosis).

Question 34

What is the testing bias where the test only picks up the less aggressive forms of the disease, with long latency periods, while those with shorter latency period become symptomatic earlier and are picked up, when these are compared the survival time appears to be longer for the less aggressive form of the disease?

Answer 34

Length-time bias

vvvvvvvvvvvvvv

This is a phenomenon whereby a screening test preferentially detects less aggressive forms (with a longer-latency period) of a disease and therefore increases the apparent survival time. If a new screening test detects more non-aggressive diseases and fewer aggressive ones than the previous method of diagnosis, this may appear to increase survival.

vvvvvvvvvvvvvv

This can be rectified by a randomized controlled trial assigning subjects to the screening program or to no screening.

Question 36

When patients are enrolled on basis of ease of contact, this is called?

Answer 36

Convenience sampling and is a form of sampling bias.

vvvvvvvvvvvvvv

Sampling bias is a nonrandom sampling or treatment allocation of subjects such that the study population is not representative of target population.

vvvvvvvvvvvvvv

Fix with randomization

Question 37

What is the bias where the cases and/or controls are selected from hospitals (bedside bias) and are less healthy and have different exposures?

Answer 37

Berkson bias and is a form of sampling bias.

vvvvvvvvvvvvvv

Selecting control subjects for a case-control study from hospitalized patients can potentially bias the results because the exposure frequency in hospitalized patients does not necessarily reflect that of the general population. This type of selection bias is called Berkson fallacy. Patients in a university hospital may have more severe illness and higher mortality rates than individuals with the same condition in a community hospital.

vvvvvvvvvvvvvv

Fix with randomization.

Question 38

What distortion occurs with sampling bias?

Answer 38

Selection bias results from selection of study subjects that are not representative of the study population.

Question 39

What is a referral bias?

Answer 39

Referral bias results when patients are sampled from specialized medical centers and therefore they do not represent the general population.

Question 40

What bias results from participants lost to follow up have a different prognosis than those who complete the study?

Answer 40

Attrition bias and is a form of sampling bias.

vvvvvvvvvvvvvv

Fix with randomization

Question 41

In what type of studies does lose to follow up tend to occur?

Answer 41

Loss to follow-up is a form of selection bias. This occurs in cohort studies. If people from one group (exposed or unexposed) who are lost to follow-up are more likely to develop the outcome in question than those lost to follow-up from the other group, then selection bias results. A high rate of follow-up loss creates a high potential for selection bias in prospective studies.

Question 42

When given a choice of whether or not to participate in a study, what bias tends to occur?

Answer 42

Non-response bias may occur when study design allows subjects to decide whether or not to participate in the study. Health surveys conducted by a random selection of phone numbers are a prime example. The phone numbers selected are called and people are interviewed using a standardized questionnaire. There are always people who would refuse to participate in the survey. If the refusal is somehow related to their health status (e.g., they are sicker than the general population), then non-response selection bias results.

Question 43

What is susceptibility bias?

Answer 43

Susceptibility bias occurs when the treatment regimen selected for a patient depends on the severity of the patient’s condition. An example can be made with acute coronary syndrome, where the healthier patients may be preferentially selected for coronary intervention, while sicker patients may instead be selected for medical therapy. This may create bias whereby outcomes from coronary intervention appear superior to medical therapy simply because the subjects who underwent coronary intervention were healthier.

Question 44

What bias occurs when the incidence of a disease is estimated based on the prevalence?

Answer 44

Prevalence bias (Neyman bias) may occur when incidence of a disease is estimated based on prevalence, and data become skewed by selective survival. For example, diabetics are more likely to die from myocardial infarction than are non-diabetics. If living patients who have sustained myocardial infarction are asked about their diabetes status, it is likely that diabetics will be under-represented because non-diabetics ‘selectively survived’ their cardiovascular events.

Question 45

Information that is gathered in a systemically distorted manner in a research study is a form of ________ bias.

Answer 45

Information that is gathered in a systemically distorted manner is a form of measurement bias.

Question 46

Recall bias tends to problematic in what type of study?

Answer 46

Recall bias is a typical example of measurement bias which should always be considered as a potential problem in case-control studies.

vvvvvvvvvvvvvv

Recall bias can result in overestimation of the effect of exposure. An example is with a study that looks at women who have children with a neural tube defect. They are more likely to report use of the drug than women whose children are healthy. This over-reporting is likely due to psychological trauma induced by the birth of the baby with a congenital abnormality and search for the potential explanation of the problem.

Question 47

How do you alleviate recall bias, where the patient’s awareness of a disorder alters their recall in some manner or patients with disease recall exposure after learning of similar cases?

Answer 47

This bias is common in retrospective studies and can be mitigated by decreasing the time from exposure to follow-up, or use of medical records as sources.

Question 48

When participants change their behavior upon awareness of being observed is called … ?

Answer 48

Hawthorne effect

vvvvvvvvvvvvvv

This is a form of measurement bias.

Question 49

What is the implication of a measurement bias?

Answer 49

Measurement (information) bias results from inaccurate estimation of exposure and/or outcome.

vvvvvvvvvvvvvv

Measurement bias implies that exposure and/or outcome data are systematically misclassified (e.g., exposed cases are labeled as unexposed).

vvvvvvvvvvvvvv

Misclassification can be ** differential ** (e.g., outcome in the exposed subjects is misclassified) or ** non-differential ** (e.g., outcome in all groups is misclassified).

Question 50

Any form of measurement bias can be mitigated with … ?

Answer 50

Measurement bias can be mitigated with use of objective, standardized, and previously tested methods of data collection that are planned ahead of time. Also with use placebo group.

Question 51

When subjects in different groups are not treated the same, this is a ________ bias

Answer 51

Procedural bias

vvvvvvvvvvvvvv

This can be mitigated by blinding (masking) and use of placebo reduce influence of participants and researchers on procedures and interpretation of outcomes as neither are aware of group assignments.

Question 52

When the researcher’s belief in the efficacy of a treatment changes the outcome of that treatment (Pygmalion effect), what bias has occurred?

Answer 52

Observer-expectancy bias

vvvvvvvvvvvvvv

Observer bias (ascertainment bias, detection bias or assessment bias) is a form of measurement bias that occurs when the investigator’s decision is adversely affected by knowledge of the exposure status. For example, while evaluating kidneys, some pathologists’ may have their decisions skewed by the fact that hypertensive nephropathy is a common cause of end-stage renal disease in black patients.

vvvvvvvvvvvvvv

This can be mitigated by blinding (masking) and use of placebo reduce influence of participants and researchers on procedures and interpretation of outcomes as neither are aware of group assignments.