final Flashcards
(112 cards)
1
Q
measures the disease burden in a population
A
prevalence
2
Q
number of new cases in a population over a given period of time
A
incidence
3
Q
most precise measure of incidence
A
Incidence density
4
Q
rate of appearance
A
incidence
5
Q
total of # diseases/#total population
A
Prevalence
6
Q
new cases/ total population of # at risk
A
incidence
7
Q
new cases in a specified time period/ # units of person-time
A
incidence density
8
Q
what happens with people HIV are living longer but new cases is decreasing
A
prevalence is increasing but incidence is decreasing
9
Q
descriptive of a rare disease–> no stats
A
Case study
10
Q
establish the exposure and outcome at the same time period leading to an unclear temporal relationship
A
Cross-sectional
11
Q
exposure and unexposed at baseline not outcome
A
prospective cohort study
12
Q
outcome determined as well as exposure but it looks back in time from exposure to outcome
A
retrospective cohort study
13
Q
Outcome determined, exposure is assessed
A
case control study
14
Q
matching
A
case control study
15
Q
good for rare exposures
A
cohort
16
Q
good for rare outcomes
A
case control
17
Q
multiple outcome can be assessed
A
cohort
18
Q
multiple exposures can be assessed
A
case-control
19
Q
bad for rare outcomes
A
cohort
20
Q
bad for rare exposures
A
case-control
21
Q
RR or AR cannot be used, Only Odds ratio
A
case control
22
Q
allocation to a treatment group based on chance
A
RCT
23
Q
statistical analysis comparing groups as randomized regardles of the actual treatment given
A
Intention to treat analysis
24
Q
avoids ‘undoing’ randomization
A
intention to treat analysis
25
reduces bias by both investigators and volunteers
placebo
26
RR: 1>
significant causative
27
RR: 1
significant protective
28
RR: 0
nothing to say
29
RR of GI bleeding aspiring vs placebo 1.2-->
Women on aspirin are 20% more likely to have a GI bleeding over toys compared to women not on aspirin
30
RR of GI bleeding aspirin vs placebo 0.83
women on aspiring are 17% (1-0.83) less likely to have a stroke over 10yrs compared to women not on aspirin
31
risk difference, excess risk and absolute risk reduction
attributable risk
32
estimate what risk would be if I prevent exposure
prevention
33
1/absolute risk reduction
number needed to treat
34
to prevent one stroke, MDs need to treat # women with low dose aspiring
number needed to treat
35
incidence in smokers - incidence in non-smokers
attributable risk
36
incidence of stroke in aspirin- incidence of stroke in placebo
absolute risk reduction
37
ad/bc
odds ratio
38
freedom from bias
internal validity
39
generalizable
external validity
40
[3] criteria for confounding
1. unbalanced in exposure
2. risk for outcome
3. not a mediator
41
co co cr ad
compare confounding crude vs adjusted
42
EM Co Sub
effect modification compare across subgroups
43
does C trumps EM?
No! EM trumps C
44
volunteers know the study hypothesis and thus does that have the exposure and the outcome will apply but not the people who have the exposure but not the outcome
selection bias
45
toward the null
random miss-classification
46
towards or away from the null
non-random miss-classification
47
to avoid this bias we need to use well- defined and precise measurement
random missclassification
48
bias where both exposed and unexposed/ disease or non-diseased groups that have the same issue
random miss-classification
49
to avoid bias use blinding of the investigators and volunteers
non-random miss-classification
50
bias where only one group either the unexposed or exposed group show bias
non-random
51
how to avoid bias in case control
chose cases and controls independently of exposure
52
how to avoid bias in a retrospective cohort study
chose exposed and un-exposed independently of outcome
53
9 bradford hill
1. strong
2. consistency
3. specific
4. precede
5. gradient
6. plausible
7. coherent
8. experimental
9. analogy
54
data:
| dead/alive or HIV positive/ negative
nominal
55
breast cancer staging
ordinal
56
number of joints with arthritis
discrete
57
weight, age, cholesterol
continous
58
data that uses bar chart
ordinal
59
data is used for histrogram
continous
60
most common value in dataset
mode
61
always has units
standard deviation
62
68 % of observations are within
1 SD
63
95% of observation are within
2 SD
64
99% of observation are within
3 SD
65
34% are above and below the mean
1SD
66
number of standards deviations that the value is above or below the mean
Z score
67
value-mean/ SD
z score
68
how far that value is from the mean
Z score
69
horseshoe game
the value is fixed (the pole) and the Confidence interval is the horseshoe . there is a chance that the CI will have the value
70
sample mean (point estimate) +/- 2 x (SE)
confidence interval
71
SD/ square root of samples size
standard error (SE)
72
is the standard deviation of many samples means
SE
73
deviation of a single sample
standard deviation
74
things that make the confidence interval wider [2]
less precise and wider net
75
increasing the level of confidence from 95 to 99
confidence interval wide
76
smaller sample size
confidence interval wide
77
larger standard error.... more variability among observations
confidence interval wide
78
probability that the TX is effective when it is not
null hypothesis
79
type I error
alpha
80
P-value< alpha
reject null
81
P-value > alpha
do not reject null
82
where does the P-value come from?
from the the tail end of the area under the curve
83
interpretation of the p value
Given the null hypothesis is true, the probability of obtaining a result at least as extreme as the one observed
84
p value is only interpretable under the assumption that the
null hypothesis is true
85
interpretation of the p value:
RR of GI bleeding aspirin vs placebo= 1.2 (p< 0.001)
Given there is no difference in risk of GI bleeding between women taking or not taking low dose aspiring, the probability of obtaining a result at least as extreme as 20 % higher risk in women on aspirin is less than 1 / 1,000
86
P value does not tell us if [3]
1. HO or HA is true
2. clinical significance
3. any bias
87
does not cross zero
CI
88
does not cross one
RR, OR
89
appropriate when exposure is binary and outcome is continuous
two-sample t test
90
used to compare two means
two-sample t test
91
appropriate when both exposure and outcome are nominal
chi-square
92
used to compare two or more proportions (percentages)
chi-square
93
used to assess significance of rr and OR
cho-square
94
we can never say that we committed an error only that out conclusion might have an error
yes
95
if you conclude that tx is effective when it is not
type I error
96
conclude if TX. is not effective when it is
type II error
97
Beta power
type II error
98
statistical power increased [2]
1. sample size
| 2. difference between groups
99
Pearson when there is a curve
no correlation
100
coefficient of determination
R squared
101
best fit line
data above and below contain same amount of data
102
percentage of diseased subjects with a positive test
sinsitivity
103
percentage of disease free subjects with a negative test
specificity
104
percentage of subjects with a positive test who have the disease
pvp
105
percentage of subject with a negative test who are disease free
pvn
106
TP / TP + FN
sensitivity
107
TN / TN + FP
specificity
108
TP / TP + FP
PVP
109
TN / TN + FN
PVN
110
as prevalence increases
PVP increases and PVN decreases
111
remain unaffected by prevalence
sensitivity and specificity
112
Avid Clown's Pursuing Epidemiology Commonly Behind The Silly Samples
```
A- analogy
C- consistency
P- precedent
E- experimental
C- coherence
B- biochemical gradient
T- temporal
S- strength
S- specificity
```
