epi/biostats Flashcards
Cross-sectional study
Design: collect data from group of people to assess frequency (and risk factors) of disease at one point in time
Measures disease prevalence; can not establish causality
Case-control study
Design: Retrospective. Compare group of people w/ disease to group without disease. Look at prior exposure/risk factor.
Measures odds ratio (OR)
Cohort study
Design: Prospective or retrospective. Compare group with exposure to group without exposure and then look at difference in disease.
Measures relative risk (RR)
Twin concordance study
Compare freq of both twins developing disease
Measures heritability and influence of environmental factors
Adoption study
Compares siblings raised by biological vs adoptive parents -> measures heritability and environmental influence
Ecological study
Analyzes populations instead of individuals for disease prevalance or risk factors etc
Sensitivity
True (+) rate -> probability the test detects disease when it is present
= (test+ w/disease)/(all with disease)
=TP/(TP+FN)
Rules OUT disease and indicates LOW FN
Used for screening
Specificity
True (-) rate -> probability negative result when disease not present
=(test- w/out disease)/(all without disease)
=(TN/TN+FP)
Rules IN disease and indicates LOW FP
Used for confirmation after screening
Positive predictive value (PPV)
PPV= TP/ (TP+FP)
Varies directly with prevalence or pretest probability
Negative predictive value (NPV)
NPV= TN/(TN+FN)
Varies indirectly with prevalence or pretest probability
Odds ratio
Used in case-control studies
Odds group w disease were exposed/ Odds group without disease were exposed
OR=(a/c)/(b/d)
a= +disease, +risk b= -disease, +risk c= +disease, -risk d= -disease, -risk
Relative risk
Used in cohort studies
Risk developing disease in exposed group/ risk of disease in unexposed groups
RR=[a/(a+b)]/[c/(c+d)]
= [(+disease and risk)/(all +risk)]/
[(+disease and -risk)/(all -risk)]
Attributable risk
AR= difference in risk btwn exposed and unexposed
=(a/a+b) - (c/c+d)
Relative risk reduction
proportion of risk reduction attributable to intervention
RRR=1-RR
Absolute risk reduction
ARR= difference in risk attributable to intervention = -AR= (c/c+d) - (a/a+b)
Number needed to treat (NNT)
NNT=1/ARR
Number needed to harm (NNH)
NNH=1/AR
Precision
random error decreases precision
Increased precision-> decreased std and increased statistical power (1-beta)
Accuracy
systemic error or bias decreases accuracy
Berkson bias
study pop. selected from hospital is less healthy than general pop.
recall bias
awareness of disorder alters recall by subjects. common in retrospective studies
procedure bias
subjects in different groups treated differently
Hawthorne effect
observer effect - modify behavior in response to awareness of being observed
confounding bias
factor is related to both exposure and outcome but not on causal pathway-> distorts or confuses effect of exposure on outcome
normal distribution
w/in 1 std= 68%
w/in 2 std=95%
w/in 3 std= 99.7%
power
1-beta;
beta=type2 error: stating no difference when there is a difference
Increase power and decrease beta:
- Increased sample size
- Increased expected effect size
- Increased precision of measurement
t-test
check difference btwn means of 2 groups
ex: mean BP btwn men and women
ANOVA
check differences btwn means of 3 or more groups
ex: mean BP btwn multiple ethnic groups
chi-square test
check differences between 2 or more categorical outcomes (not mean values)
Ex: comparing percentage of members of different ethnic groups who have essential HTN