Diagnostic Tests/Sensitivity/Specificity/ROC

Question 1

Screening Tests

Answer 1

• Medical tests or procedure performed on members (subjects)
• Defined, asymptomatic population or population subgroup
• Assesses likelihood of their members having a particular disease

Question 2

Screening tests DON’T…

Answer 2

• diagnose the illness

- Testing positive usually requires further evaluation with diagnostic tests/procedures

Question 3

Planning a Screening Program

Answer 3

• Disease/condition being screened should be a major medical problem
• Acceptable treatment should be available for individuals with disease
• Access to health care facilities/follow-up services should be available
• Disease should have a recognized course with early to late stages
• Tests/procedures should be accepted to general population
• Natural history of disease should be understood
• Policies and procedures should be determined to know who needs further testing/treatment
• Process should be simple enough to encourage large groups to participate
• Screening should be an ongoing activity

Question 4

“Perfect” Diagnostic Test

Answer 4

1. All individuals without disease give one uniform value
2. All individuals with disease give a different, uniform value
3. All test results coincide with being diseased or disease free

Question 5

Normal/”Reference” Range

Answer 5

• Not all values are the same in disease free individuals
• Disease free individuals will have a range of values (Reference Range)
• “Normal” misrepresents the range - distribution may not be normal and the individuals could be in range and not be healthy

Question 6

Reproducibility/Reliability of Test

Answer 6

• Ability to produce consistent results when repeated under same conditions
• Re-testing requires operator/observer to remain the same and blinded to results of 1st test

Question 7

Kappa

Answer 7

• K

- Index of agreement

Question 8

Precision

Answer 8

• Agreement of repeated measurements

- Range of results +/- SD

Question 9

Accuracy of Test

Answer 9

• Agreement of test values with true value
• Accuracy requires reliability or precision
• Results can be reliable without being accurate
• (T-) + (T+)/All

Question 10

Validity of Test Types

Answer 10

• Content validity - test represents what is trying to be measured
• Criterion Validity - Test results agree with other results measuring the same parameter
• Construct Validity - Real biological property being measured to explain why results vary among individuals

Question 11

Diagnostic Test Results

Answer 11

• Most often think or patients having or not having disease
• Nearly all disease processes are continuous
• Disease severity and number of symptoms tends to increase over time
• More difficult to diagnose early stages of disease while late stages are more distinct

Question 12

“The Gold Standard”

Answer 12

• Test whose results determine disease status
• Assumed to be 100% accurate/true
• Many diseases lack a test like this
• New tests are always compared to gold standard
• New drugs always compared to placebo control or current drug of choice (active control)
• Accuracy of new tests are always less or equal to gold standard
• New tests can outperform gold standards and become new gold standards

Question 13

Steps Determining Diagnostic Performance

Answer 13

1. Choose gold standard test
2. Perform gold standard test on a full spectrum of subjects
3. Test all subjects with new diagnostic test
4. Compare results of both tests in a 2x2 table
5. Calculate the proportions of accurate and inaccurate results of the new test

Question 14

Prevalence + Diagnostics

Answer 14

• If proportion with disease is high, easy to diagnose a disease positive
• If proportion with disease is low, easy to diagnose a disease negative
• Surrogates for positive/negative disease diagnosis determine if patients will respond to drug therapy and likely drug therapy failures

Question 15

Reporting Prevalence

Answer 15

• Percentage

- Number of cases per a total number of people

Question 16

Types of Prevalence

Answer 16

1. Point prevalence - proportion of people with disease at a specific point in time
2. Period prevalence - proportion of people with the disease during a time period
3. Lifetime prevalence - proportion of people with the disease at some point in their lifetime

Question 17

Prevalence

Answer 17

• Incidence rate X duration of disease
• Incidence - number of new cases in population over a period of time
• High incidence with long duration will have high prevalence

Question 18

Data Analysis

Answer 18

• Analysis of the comparison of the diagnostic test to gold standard requires independence of each test
• If both tests both misclassify the same patient, falsely high agreement with an overestimate of sensitivity and specificity
• If they independently misclassify the same patient, underestimate sensitivity/specificity

Question 19

Sensitivity

Answer 19

• Likelihood of a test reflecting that someone has a disease that we know they have
• (T+)/(T+) + (F-)

Question 20

Specificity

Answer 20

• Likelihood that a test will show that a patient doesn’t have a disease that they don’t have
• (T-)/(F+) + (T-)

Question 21

PPV

Answer 21

• Likelihood that a positive test result reflects a patient actually having a disease
• (T+)/(T+) + (F+)

Question 22

NPV

Answer 22

• Likelihood that a negative test result actually reflects a patient being disease free
• (T-)/(F-) + (T-)

Question 23

Predictive Values

Answer 23

• Specificity and sensitivity have no meaning to individual patients
• Patients’ true disease status is unknown at time of testing

Question 24

Prevalence + Predictive Values

Answer 24

• As prevalence goes down so does the PPV

- As prevalence goes down, NPV rises

Question 25

ROC Curve

Answer 25

- Receiver Operating Curve - Position of the "cut point" between normal and abnormal test results determine the test's sensitivity/specificity - Cut point usually determined by the reference interval - Sometimes altered to enhance either sensitivity or specificity

Question 26

ROC Curve Meaning

Answer 26

- Shows tradeoff between sensitivity and specificity - ROC curve following left-hand border and then top border shows greater test accuracy - The closer it approaches a 45 degree diagonal, lower the test accurcy - At the 45 degree angle, test is no better than a coin toss

Question 27

Pretest Probability

Answer 27

-Probability of disease before application of the results of a physical finding

Question 28

Post-test Probability

Answer 28

-Probability of a condition being present after a diagnostic test