Biostats Flashcards
Cross sectional study?
Case control study?
Cohort study prospective
cohort retrospective?
Measures and examples for each.
Cross sectional study?
groups of people to assess frequency of a disease and risk factors at a point in time.
Disease prevalence
Risk factor association with disease
Case control study?
Group of people with disease to group of people without
Prior exposure of risk factor
Odds ratio
Cohort study prospective?
Group of people with a given exposure or risk factor to a group without
Does exposure incr. likelihood of disease
Relative risk
Who will develop disease?
cohort retrospective?
Same as above but who did develop disease?
What is size and purpose of phase I, II, III, and IV of clinical trial?
I-small number of healthy volunteers; is it safe?
II-small number of pts with disease; does it work?
III-large number of pts randomly assigned to treatment or placebo; is it as good or better?
IV-postmarketing surveillance; can it stay?
What is sensitivity? How is it calculated? Purpose?
proportion of all people with disease who test positive.
Probability that a test detects disease when disease is present
Out of all the people with the disease, how many does it catch?
TP/(TP + FN)
1- FN rate
SN-N-OUT
Sensitive test, when negative, rules out disease (no false negatives)
What is specificity? How is it calculated? Purpose?
Proportion of all people without the disease who test negative
Probability that a test indicates no disease when disease is absent
Out of all the people without the disease, how many does the test say don’t have the disease?
TN/ (TN + FP)
1-FP rate
SP-P-IN
Highly specific test, when positive, rules in (no false positives)
What is a PPV? Calculaton? How does prevalence/pretest prob affect it?
TP/(TP + FP)
Probabilty that person actually has disease given a positive test result
Varies directly with prevalence or pretest prob
What is a NPV? Calculaton? How does prevalence/pretest prob affect it?
TN/(TN+FN)
Probability that a person actuallly is disease free given a negative test result
Varies inversely with prevalence or pretest prob.
Describe a 2x2 contigency table for quantifying risk.
Disease on top axis, risk factor or intervention on side
+ -
+ ab
- cd
What is an odds ratio? When is it used? Calculation?
Case control
Odds that group with disease was exposed to risk factor (a/c) divided by odds that group without the disease was exposed (b/d)
(a/c)/(b/d) OR ad/bc
What is an relative risk? When is it used? Calculation?
Cohort studies
Risk of developing disease in the exposed group divided by risk of developing disease in the unexposed group.
a/(a+b)/c/(c+d)
21% chance/1% chance
What is AR? Calculation?
Attributable risk
Difference in risk b/w exposed and unexposed groups, or proportion f disease occureces that are attributable to exposure.
a/(a+b)-c/(c+d)
21% chance - 1% chance
What is the relative risk reduction? Calculation?
proportion of risk reduction attributable to intervention as compared to control
RRR=1-RR
What is ARR? Calculation?
c/(c+d) - a/(a+b)
Difference in risk, not proportion attributable to intervention as compared to control.
What is the NNT? calc? NNH? calc?
number of patients needing to be treated for one to benefit
1/ARR
Number of pts. treated before someone is harmed?
1/AR
How does precision affect SD? How does it affect power? What decreases accuracy in a test?
Incr. prec. decr. SD
Incr. prec. incr. statistical power
Systematic error or bias.
What is selection bias? Examples? Strategy to reduce?
Error in assigning subjects to study group resulting in unrepresentative sample. Most commonly a sampling bias
Berkson=selected from hospital, not general popul
healthy worker=study population healthier than general
non-response=participating subjects differ frm non particpating in meaningful ways
Randomization
Ensure the choice of the right comparison/reference group
What is recall bias? Examples? Strategy? Measurement bias? Same questions? Procedure? Observer-expectancy?
RECALL
Awareness f disorder alters recall by subjects; common in retrospective studies
Pts with disease recall exposure after learning of similar cases
Decr. time from exposure to F/U
MEASUREMENT
info in gathered in way that distorts it
Miscalibrated scale
Standard method of data collection
PROCEDURE
Subjects in groups not treated the same
Pts in treatment group spend more time in hospital units
Blinding, use of placebo
OBSERVER-EXPECTANCY
Researchers belief in the efficacy of treatment changes outcome
Pygmalion effect, self-fulfilling prophecy
Observer more like to document positive outcomes if he expects treatment to work
blinding, use of placebos
What is confounding bias? Examples? Strategy? Lead time bias?
CONFOUNDING
Factor related to both exposure and outcome, but not on causal pathway
Pulmonary disease is amore common in coal workers, but coal workers also smoke more
Multiple/repeated studies
Cross over studies
matching
LEAD TIME BIAS
Early detection is confused with incr. survival
Measure back end survival (adjust according to severity at time of diagnosis)
What is the mean? Median? Mode? Which is least effected by outliers? Most affected by outliers?
Mean=most affected by outliers
Median-middle value of list
mode=most common value, least affected by outliers
What is standard deviation? Standard error of the mean? Calculation? What causes SEM to decr ?
SD-how much variability exists from the mean in a set of values
SEM=an estimate of how much variability exists between the sample mean and the true population mean.
SD/sq.rt (sample size)
SEM decr. with incr. sample size
What is normal distributin? What percentage falls within 1 SD? 2SD? 3SD?
Mean=median=mode
1SD=68%
2SD=95%
3SD=99.7%
What is bimodal distribution? Positive skew? Negative skew?
Two different populations
Mean > median > mode
positive outliers cause a longer tail on right
mean< medi<mode
negative outliers cause longer tail on left
What is the null hypothesis? Alternative?
H0=Hypothesis of n difference or relationship
H1=Hypothesis f sme difference or relationship (not due to chance)
What is a type I error? What is alpha? What is p?
There is an effect or difference when none exists (null hypothesis rejected when it should not have been).
Alpha is probability of making a type I error
p is judged against a preset alpha. If p <.05, there is a less than 5% chance of making a type I error
False positve error
You sAw a difference that did not exist
What is a type II error? What is Beta? What is power? What can incr. power?
Stating there is no effect when there is one.
Beta is prob of making a type II error
Power=1-beta, which is the probability of not getting a type II error (finding an effect if there is one)
You were Blind to the truth
Incr power by:
Incr. sample size
Incr. expected effect size
incr. precision of measurement
