Odds Ratio= Tells the risk of a disease for those exposed as compared to those not
exposed
Data collected via case-control study
Can estimate the relative risk in this type of study
(Exposed + disease * not exposed disease) / (not exposed + disease *
exposed disease)
Example the risk of lung cancer is X times higher for smokers than
nonsmokers
Number Needed to Harm= # of people that must be exposed to a risk factor for
one person to be harmed that wouldnt have otherwise been harmed
1 / attributable risk (inverse of attributable risk)
Number needed to treat (NNT)= # of people who need to be treated for one
person to benefit from a treatment
1 / absolute risk reduction
Absolute risk= # w/ disease / pop
Absolute risk reduction= absolute risk of placebo group absolute risk of
treatment group
Biased test/study= constructed so one outcome is more likely that the other to
occur
Selection Bias=if subjects or investigator allowed to choose group assignments
Sampling Bias= subjects to not represent the population being studied- results
may not be generalizable
Recall Bias= knowledge of the presence of a disorder alters the way the subject
remembers his/her history (ex mothers of children w/ cleft palate ay over estimate
how much meds they took during preg..can lead to false finding that certain meds
related to cleft palate)
Lead-time Bias=early detection of disease is confused with inc survival or inc
incidence rate
Surveillance bias=people that are aware they are being followed for the
development of a disease are more likely to seek testing and thus be identified
early
Late-look Bias=people who already died from the disease are not included in the
sample. symptoms of disease seem less severe
Single blind=subject doesnt know if getting treatment or not
Double Blind=neither subject nor evaluator knows the assignments
z Score (standard normal value)= difference btwn individual score and the pop
mean in units of standard deviation
Standard Error= estimate of the quality of the sample
standard deviation / sqrt of # of scores in the sample
Confidence Interval= specifies the limits btwn which a given % of the pop should
fall
For example with a CI = 95% if conduct the clinical trail 100 times 95% of
results will fall within the calculated range
Narrower intervals give greater preision
Limits calc by mean of sample +/- z score * standard error
@ 95% CI z score=2
@99% CI z score =2.5
@99.7% CI z score=3
Normal distribution= bell curve/ Gaussian curve theoretical distribution of
scores in which the mean, median and mode are equal. ~68% of pop falls w/in 2 SD
of the mean; ~95% of scores w/in 2 SD of mean; ~99.7% of scores w/in 3 SD of
mean
Highest pt in distribution of scores is the modal peak
Precision=degree to which the mean is resistant to random variation
Accuracy=likelihood of bias
Skewed distribution=model peak shifts to one side
Positive skew(right skew)= tail towards right, peak towards left=scores cluster
at lower end
2)
3)
4)
5)
Study designs from least rigorous and most bias to most rigorous and
least bias
Case series and Case reports-made of collection of reports on treatment of
individual patients or report on a single patient.
No control group
Little statistical validity
Case Control Studies-patients that already have a specific condition are
compared with people that do not have the condition
Look back to identify risk factors and exposures
Rely on records and/or patient recall
Statistical relationship doesnt mean 1 factor caused the other
Cohort Studies- use group of patients already taking a particular treatment or
have an exposurefollow them over timecompare their outcomes with similar
group not affected by the treatment or exposure
Odds Ratio=those with the outcome are X times more likely to have been exposed
than where those in the control group
(exposed + disease / not exposed + disease) / (exposed disease/ not
exposed disease)