Evidence About Diagnosis Flashcards
What is pre-test probability?
The probability of a patient having a condition before assessing them (i.e. population prevalence)
What is a gold standard test (aka reference diagnostic test, reference standard)?
- The most accurate test available regardless of how invasive, expensive, or time consuming
- May be hindsight, where the correct diagnosis only becomes apparent with time (or post-mortem)
What is an index test?
- The test being investigated (diagnostic test of interest)
- More practical than the gold standard
What are the types of diagnostic study designs?
- Cohort study
- Cross-sectional studies
- Case-control
- Systematic review of diagnostic accuracy studies
What is a cohort study?
- All people with a particular clinical presentation, who present at a particular hospital, undergo both the gold standard and index test
- Stats done to compare how accurate index test was relative to gold standard
How can bias be minimised in a cohort study?
- True diagnosis not known until after enrolment
- Consecutive patients invited to participate in the study
- Gold standard test result is not known when conducting the index test & vice versa
What is a cross-sectional study?
A group of people with a particular clinical presentation are recruited, who then undergo both the gold standard and index test (i.e. a snapshot)
What is a case-control study?
- People who clearly have the disease & people who clearly don’t, undergo both the gold standard & index test
- Only has to discriminate between extreme cases; doesn’t reflect clinical reality
- Prone to bias
What does sensitivity of the index test refer to?
How many people who have the disease are correctly diagnosed with the index test (i.e. person with the disease with a positive test result, true positive)
What does specificity of he index test refer to?
How many people who don’t have the disease (who are well) are correctly identified with the index test (i.e. well person with a negative test result, true negative)
What does positive predictive value (PPV) refer to?
- The chance a positive index test result will be correct
- PPV will fall as disease prevalence falls
What does negative predictive value (NPV) refer to?
- The chance that a negative index test result will be correct
- NPV will rise as disease prevalence falls
What do likelihood ratios predict?
- The presence of a diagnosis
- Sensitivity & specificity cannot predict the risk of an abnormality (i.e. they refer to how abnormality predicts test results)
What does a positive likelihood ratio refer to?
- Used when a person has a positive index test
- How much more likely is a positive test result in a person with the condition than in a person without it?
= 1 is useless
= >2 helps rule in disease
= >10 extremely good for ruling in disease
What does a negative likelihood ratio refer to?
- How much more likely is a negative test result in a person without the condition than in a person with it?
- Used when a person has a negative index test
= 1 is useless
= <0.5 helps rule out disease
= <0.1 extremely good for ruling out disease
Which statement is correct?
Test A is highly sensitive (99%) for a condition (& has a high specificity of 80%); this means that …
a. Test A will identify the majority of people who have the condition (99%).
b. It is unlikely that someone with the condition will have a negative result on test A (1%).
c. Someone who does receive a negative result on test A is unlikely to have the condition (80%).
d. Test A is good at ruling out the condition
All statements are correct
Which statement is correct?
Sensitivity and specificity …
a) is never affected by the size of the sample
b) can be described using 95% confidence intervals
c) can have 95% CI that are unequal sizes
d) all of the above
e) b and c are correct
e) b and c are correct
How is the sampling theory affected by sensitivity & specificity?
- If we repeat a diagnostic accuracy study with a new sample the sample will be slightly different
- And the estimate of sensitivity and specificity will be slightly different - Sensitivity: Sample size = number with condition
- Specificity: Sample size = number without condition
When is 95% CI widest?
For proportion (point estimate) = 0.5 Narrows as proportion tends to 0 or 1
How are likelihood ratios calculated?
Positive likelihood ratio (PLR) = sensitivity/(1-specificity)
Negative likelihood ratio (NLR) = (1-sensitivity)/specificity
What are receiver operating characteristics used for?
Determining a cutoff point for tests on continuous scales (e.g. blood pressure)
The odds of the event are 1:3, what is the risk (probability) of the event occurring?
0.25 (25%)
An effective treatment has a number needed to treat (NNT) that is -?
Small
The risk (probability) of the event is 90%, what are the odds of that event?
0.9/0.1 = 9
What are the odds of flipping a coin and landing on tails?
1:1 (1)
True or false:
If the 95% confidence interval for the risk ratio of an intervention includes 1, this indicates no statistically significant effect
True
What are the odds of rolling a die (singular of dice) and landing on 6?
1:5 (0.2)
