Flashcards in biostatistique Deck (145):
1
standard deviation normal distribution a 68%
SD a 1 from the mean
2
standard deviation normal distribution a 95%
SD a 2 from mean
3
standard deviation normal distribution a 99,7%
SD a 3 from the mean
4
example mean a 230 mg/dl pour le cholesterol and SD a 10,what's the value for cholesterol if observations are 95%
210-250
5
disease with increasing prevalence and stable incidence what happened(2)
any rx which prolongs life
good quality of care
6
Quid of PPV(2)
positive predictive value
pobability of having the disease if the result test is positive
7
quid of negative predictive value
probability of having the disease if the test is negative
8
quid of null hypothesis
no relationship between exposure and outcome
9
example of null hypothesis
CRP high and colon cancer ,no relationship
10
what to consider when evaluating the effectiveness of new trial drug
the natural history of the disease
11
example drug effectiveness and drug trial(2)
effectiveness of antiviral drug and flu
no exact conclusion can be drawn becausecommon cold is a self limited disease
12
in hazard ration interpretation what are the 2 key factors to consider(2)
control arm
rx arm
13
quid of hazard ratio <1
the event will occur in the control arm
14
quid of hazard ratio > 1
the event will occur in the rx arm
15
practice of hazard ratio
the chance for an event to occur during taking a drug for instance
16
interpretation of hazard ration close to one
no difference for an event to occur in the 2 groups rx arm and control arm
17
what information is critical to successfull randomization
baseline characteristics
18
what an ideal randomisation
process to minimize selection biais and achieves possibility of cofounding variables
19
a study is done using placebo and Rx,the assignment of the rx in teh two arm is done randomly using numbers generated by computer what randomization helps to do in ths case?
to eliminate confounder
20
what methods are used to eliminate cofounder during analytic stage of study
stratified analysis
modeling
21
what methods are used to eliminate cofounder during adesign stage of study(3)
matching
restriction
randomization
22
clue for randomization(2)
Rx group
Placebo group
23
when using randomization
during clinical trials
24
quid of restriction
when you limit your study to one group for example one sexe is considered
25
problem with restriction
you cannot generalize your finding in the population
26
when using matching
control studies
27
importance in using macthing(2)
when you want to study risk factor(
exposure and outcome
28
practice of macthing
you will take a group with a known risk factor and another one without it and compare
29
quid of false negative
the test is negative but you have the disease
30
consequence of raising the cut off value
false negative will increase
31
quid of precision
measure of random error in study
32
how to know a study is precise
when the confidence interval is tighter
33
what to do to increase precision of a study
increase the sample
34
objectif in using confidence interval
to measeure precision of a study
35
hazard ratio quid
ration of an event to occur in the Rx group compared to the non Rx group(arm group)
36
hazard ratio less than 1
event less likely to occur in the rx group
37
hazard ratio more than 1
event likely to occur in the rx group
38
in the USMLE when hazard ration is used
to determine the complication of a rx
39
quid a factorial design
when a study uses different interventions with two or more different variables
40
use of scatter plots
to demonstrate linear and non linaear patterns
41
any time you see a new screening test in the USMLE think of
lead time biais
42
lead time biais?
apparent survival in patients to whom a screening test is applied without changing the prognosis of the disease
43
way to interpret strengh of association and dose response relationship from a study
risk relatif RR
44
quid of null value of RR if you have null value outside
1
45
quid of relative risk
the probability for en event to occur in the exposed group compared to the probability in the non exposed group
46
quid of RR>1
the event occurs more frequently in the exposed group
47
quid of RR< 1
the event occurs more frequently in the exposed group
48
what's the best way to to compare the mean of 2 groups of subject
two sample test
49
during a case control of disease the oddd ration is 5.0,the scientist conclude that the relative risk is 5f higher for an habit what does that imply
it's a rare disease
50
best way to compare 3 or more means
two sample Z test
51
all the time RR=OR what to conclude
you are dealing with a rare disease
52
quid of correlation coefficient
helps to tell relationship between a a risk factor and an outcome
53
what coefficient correclaition doesn't not imply
causality
54
Math of CC
iiii
55
quid of NPV
proportion de personne negatif for a test reellement non malade
56
quid of PPV
proportion de personne positif pour un test et reellement malade
57
what happen to the PPV when it's applied to a population with high incidence of a disease
PPV will increase
58
quid of outlier(2)
an extreme and unusual value observed in a dataset
may be caused by error or mistake
59
what's extremely sensitive to outlier
the mean
60
what's extremely resistant to outlier
mode and median
61
quid of sensitivity
true positive/true positive+false negative
62
importance of sensitivity
a negative test rules out the disease
63
quid of specificity
True negative/true negative+false positive
64
importance of specificity
positive result rule in the disease
65
quid of screening test
test tres sensible
66
quid of confirmatory test
test est tres specifique
67
confounder
biais that can result when the exposure disease relationship is mixed with the effect of extraneous factor
68
a crude analysis wants to see relation ship between ca of liver and smoking
stratified analysis divides the smoker in two groups one wo drinks alcohol and another who drinks alcohol but no smoker
all the ca of liver were linked to alcohol
confounder
69
when 2 factors are combined and increases the risk for a disease
effect modification
70
example of effect modification
family history of breast cancer and USE of OCP
no family history of breast cancer and USE of OCP
Increases the risk of ca of breast++++
no cancer
71
confounding (2)
two habits most of the time linked together but never cause the disease
ex:most of the time smoker also drinks alcohol but only alchohol is associated with ca of liver
72
effect modification(2)
you have an essential factor ,when combined to another one ,the risk for a disease increases +++
ex:family history of cancer and USE of OCP
73
quid of confounder
extraneous factor linked with the the expsoure and the outcome of interest
74
how will be the NPV in patient who has high probability of having a disease
low
75
how will be the NPV in patient who has low probability of having a disease
High
76
when baseline characteristis are similar in thRx group and Placebo group what does that mean
randomisation is succesful
77
benefit of efrenzia over clopidogrel
decreased the risk of reccurent MI
78
concept of number needed to be treat
NNT=1/absolute risk reduction
79
quid of absolute risk reduction
ARR=incidence with one mediction-incidence of the event winth another med
80
quid of case control study
hepls to compare people with an exposure who develops the disease from the other exposed but don't develop the disease
81
mean used to measure case control studies
odd exposure ration
82
case control study ?(2)
you look the outcome first(the disease)
and then yo go studying the risk factor
83
retrospective cohort study(2)
you look the risk factor first
and then you go studying the risk factor retrospectively
84
goal of case study
asess frequency of risk factor
85
goal of retrospective cohort study
compared incidence of disease
86
goal of prospective cohort study
compared incidence of disease in the future
87
how to measure prevalence of a disease
cross sectionnal studies
88
how to compare outcome of interest
clinical trials
89
prevalence
total number of cases in a particular point of time
90
quid of P value
probability of abtaining a result by chance
91
quid of P value
probability of abtaining a result by chance alone
92
quid P=0,01
1% pobability that the resulted is obtained by chance
93
P value for a test to be considered statiscally significant
< 0,05 or Less than 5%
94
quid of susceptibility biais
type of selection biais where a rx regimen is selected for a patient based on the severity the condition without taking into account other possible confounder
95
best way to measure incidence of a disease
cohort studies
96
name the 2 cohort studies (2)
prospective
retrospective
97
what calculation of incidence disease allows us to do
to know the relative risk
98
calculation of attributable risk factor ARR
RR-1/RR
99
quid of ARR
it's a percentage
100
in USMLE they will never ask what's the ARR they will say what's the percentage of colon cancer with patient with high fat diet could be attributable to their diet
ARR
101
quid of attrition biais
people are loss in follow up during study
102
appartenance of attrition biais
selection biais
103
quid of hawthorne effect
population studied change behavior because thye are aware of being studied
104
quid of generazibility of external validity
applicability of a result to an other population
105
if you study only middled age women what's the limitation of this study
you can generalize the study
106
clue for sectionnal study(2)
you assess the exposure and the outcome at the same time
snapshot study
107
using IFOBT to detect risk o colon cnacer what will happen if you increase the cutt off point in ROC curve
PPV will increase
108
what will happen when you raise the cutt point of screening test
you increase the number of criteria for the test to be positive
109
what will happen when you raise the cutt point of screening test
specicicity will increase
110
importance of interval de confiance (3)
when it's tighter
the sample size is bigger
test is more precise
111
median of dataset
the number that divides the left side from the right side
112
10 people with high CRP with cardio Vascular disease
10 people with high CRP but no vascular disease
you 20 people with CRP ,calculation 5 year cardiovascular risk
10/20=0,5
50% probability of having coronary disease with high CRP
113
you test a patient for diabete with a machine you obtain 3 results different one from another what can be said on this test
test is not reliable
114
null hypothesis of a study
no relationship between exposure end outcome
115
null value of relative risk also called confidence value
1
116
if you have confidence interval 1,2-1,5 for 95% of CI what's the value of P,null value outside the CI
<0,05
117
if you have confidence interval 1,2-1,5 for 99% of CI what's the value of P
<0,01
118
if you have CI 0,91-1,2
you say the null value is inside the confidence interval
119
if you have confidence interval 0,91-1,5 for 95% of CI what's the value of P Null value inside the CI
P> ou egal a 0,05
120
if you have confidence interval 0,91-1,5 for 99% of CI what's the value of P
P.> ou egal 0,01
121
example of hazard ratio
comparing the probability of having an adverse effect from two drugs
122
for a complication linkked to 2 drugs Haazard ratio =0,96 interpretation
there is no differance for the adverse reaction considering the 2 drugs because HR is closed to 0ne
123
quid of association of a strengh of a disease(3)
more a risk factro is present
more you risk to develop a disease
it's dose dependent
124
quid of null value in RR
1
125
importance of coefficient correlation
helps to determine the direction of an exposure and outcome in a linear fashion
126
example of coefficient correlation(3)
the risk with low HDL and carotid media thickness
r will be negative and P =positive
relation inverse between low HDL and carotid media thickness
127
if a test a negative what's the probability of having the disease calculation
1-negative predictive value
128
you palce a patient on statin 3 months later you assess the result ,and realise that the result is better after one year in teh prevention of CV event why?(2)
latency period
you need time to see good result when a risk modifier is used
129
quid of ROC curve(4)
x axis and y axis
extremity of both line are linked by an oblique line
space above the line is high sensitivity
space below the line is high specificity
130
quid of attrition biais
it's a selection biais
131
when attrition biais is demonstrated
when patient is lost in follow up during a study
132
observer biais
the investigator decision is adversely affectedd by knowledge of the exposure status
133
clue for chi square
to calculate proportions
134
representation of chisquare
you have a 282 table to compare expected value and oserved value
135
when you raise the cut off point consequence of that(2)
you increase the specificity
the FN
136
when you decrease the cut off point consequence of that(3)
sensitivity will increae
FP will increase
decrease positive predictive value
137
during crude analysis you found relationship between liver ca and smoke
further analysis divides the smoker in 2 groups alcoholic and non alcoholic
no association is found with liver cancer in either group
what this help us to explain
confounder
138
quid of founder(2)
extraneous factor associated with both a risk factor and outcome
most of the time alcoholic is smoker=confounder
139
when you want to determine risk factor liee a une maladie what type of study
case control studies
140
positive predictive value
TP/TP+FP
141
negative predictive value
TN/TN+FN
142
how will be the predictive positive value inside a zone with high prevalence of HIV if a screening test is used
HIGH
143
how will be the predictive negative value inside a zone with high prevalence of HIV if a screening test is used
Low
144
example of cross section studies
you measure simultaneously the the exposure and outcome
145