Biostatistics Flashcards
case control study
observational and retrospective comparison of people with a disease to people without
trying to find a risk factor
odds ratio
cohort study
observational and retro/prospective comparison of group with exposure/risk factor to a group without risk factor
see if exposure increases likelihood
relative risk
cross-sectional study
observational collection of data from a group to assess frequency at one time
cannot demonstrate causality
finds disease prevalence
twin concordance
measures heritability
adoption study
measures heritability vs environment
clinical trial phases
I: is drug safe on healthy individuals
II: does drug work in sick pts
III: does drug work better
IV: postmarketing surveillance
sensitivity
proportion of people with disease who test positive, probability test detects disease when present
SNOUT: sensitivity rules OUT
TP/(TP+FN) = a/(a+c) = 1-FN rate
specificity
proportion without disease who test negative, probability a test indicates non-disease when disease is absent
SPIN: specificity rules IN
TN/(TN+FP) = d/(d+b)
positive predictive value
proportion of positive results that are true positives, probability has disease when test is positive
=TP/(TP+FP) = a/(a+b)
varies directly with prevalence
negative predictive value
negative test results that are true negatives, probability person is disease free when test is negative
=TN/(FN+TN) = d/(d+c)
varies inversely with prevalence
incidence
new cases of a disease
=# new cases/population at risk
prevalence
looks at all current cases
=# existing cases/population at risk
~incidence rate*duration of disease
prevalence>incidence for chronic diseases
odds ratio
case control studies.
=odds of exposure in disease group/odds of exposure in healthy group
=(a/b)/(c/d)
relative risk
cohort
=risk of developing disease when exposed/risk in unexposed group
=[a/(a+b)]/[c/(c+d)]
if prevalence is low RR ~ OR
attributable risk
proportion of disease attributable to exposure
=a/(a+b)-c/(c+d)
=incidence exposed-incidence unexposed
absolute risk reduction
reduction in risk associated with a treatment as compared to a control
number needed to treat
# treated for 1 person to benefit =1/absolute risk reduction
number needed to harm
# patients exposed to risk factor before 1 patient is harmed =a/attributable risk
precision
reliability of values=reproducible findings
does not mean the values are correct, just close to being the same
random error reduces precision in a test
accuracy
trueness of a test measurement (validity)
systematic error reduces accuracy
confounding bias
when a third variable that affects the outcome and exposure is not taken into account
Ex: coffee drinkers are more likely to get cancer because smokers are more likely to drink coffee. Smoking increases cancer, not coffee
Pygmalion effect
when a researcher’s belief in efficacy changes the outcom
AKA observer-expectancy effect
Hawthorne effect
when behavior changes because person is being studied
positive skew
mean>median>mode
longer tail on right
