Exam 6 - EpiStats Flashcards
1
Q
Incidence
A
- How many new contractions of disease
- How many are coming in
2
Q
Prevalence
A
- The % of population who have the disease
3
Q
Recurrence
A
- Those who get better and then get sick again
4
Q
Mortality
A
- Death
- Those leaving the prevalence pool
5
Q
P-value
A
- How likely these results would occur strictly by chance
- If calculated p-value is less than picked level…we reject the null hypothesis
- Example: p-value is 0.05… we would expect this result 5% of the time by chance alone
6
Q
EDAC
A
Emboli Detection and Classification
7
Q
How to set up grid
A
- Disease on top
- Test results on side
- Yes then No
- Sensitivity and Specificity on bottom (left to right)
- PPV and NPV on Right side (top to bottom)
8
Q
TP
A
- true positive
- have disease and test positive for it
9
Q
TN
A
- True negative
- Do not have disease and test negative for it
10
Q
FP
A
- False positive
- Don’t have disease but test positive for it
11
Q
FN
A
- False negative
- Do have disease but test negative for it
- WORST outcome
12
Q
Sensitivity
A
- Think positive
- % of those who are positive and test positive
- TP/(TP+FN)
- If high…. low chance of a FN
- Remain constant regardless of population
- High sensitivity = Good screening tests
13
Q
Specificity
A
- Think negative
- % of those who are negative and also test negative
- TN/(TN+FP)
- If high…. low chance of FP
- Remain constant regardless of population
14
Q
Type I error
A
- FP
15
Q
Type II error
A
- FN
16
Q
Ideal test conditions minimize what?
A
- Minimize the FN and FP
17
Q
PPV
A
- Positive predictive value
- TP/(TP+FP)
- % that there is disease in those who test positive for the disease
- measures the usefulness of a test
- can vary across populations
- HIGHLY dependent on prevalence
18
Q
NPV
A
- Negative predictive value
- TN/(TN/FN)
- % that no disease is present among those who have tested negative
- measures the usefulness of a test
- can vary across populations
- HIGHLY dependent on prevalence of disease
19
Q
CI
A
- Confidence Interval
- (1 - p value)
- Between 90-95% is standard in medicine ‘
- much higher in technology
- affected by smaller sample sizes
- small sample sizes have wider, varying intervals
20
Q
How to calculate a CI
A
- Subtract sensitivity from unity (1)
- Multiply result by sensitivity
- Divide result # of tests
- Square Root the result to get “standard error”
- Multiply standard error by “normal distribution” (given)
- Sensitivity +/- this result equals your CI
