epidemiology Flashcards
case fatality rate - what is it?
proportion of people w/a condition than end up dying from the condition
how is case fatality rate different from mortality rate?
mortality has a denominator of entire population (even those w/o the condition) so is generally smaller than CFR
attack rate - what is it?
proportion of pl who develop an illness out of total population at risk
median survival time - what is it?
length of time it takes for 1/2 study population to die
standardized mortality ratios - what are they?
used to determine if the observed number of deaths in a group exceeds what would expect in a similar group (age, gender) excluding the variable of interest (smoking, cholesterol status)
SnNout - what does this mean?
in sensitive test, negative result helps rule OUT
SpPin - what does this mean?
in specific test, positive helps rule IN
PPV - equation?
TP / (TP + FP)
PPV - what does it mean?
probability that a positive test correctly identifies person w/dz (usually correlates best with highest SPECIFICITY)
does PPV depend on prevalence?
YES, increases with increasing prevalence
NPV - what does it mean?
probability that a negative test correctly identifies someone w/o the dz
NPV - equation?
TN/(TN +FN)
absolute risk reduction (ARR) - how calculate?
subtract risk of relapses in treatment group from placebo group
number needed to treat - how calculate?
1/ARR (absolute risk reduction)
odds ratio - how calculate?
probability divided by (1 minus probability) P/(1-P)
odds ratio - how calculate w/ 2x2 plot?
(a/c)/(b/d)
RR (relative risk) - what is it?
a ratio obtained by dividing 2 values (can divide 2 RRs to get cumulative RR)
if the RR of an outcome in group A as compared to B is x, what is the RR in group B as compared to group A?
1/x
likelihood ratio - what is it?
probability of a given test result occurring in a pt w/a disorder compared to the probability of same result in pt w/o the d/o
positive LR - equation?
sensitivity / (1-specificity)
negative LR
(1-sensitivity) / specificity
are sensitivity and specificity independent or dependent on disease prevalence?
independent
what can you use to calculate post-test odds?
LR
post-test odds - equation?
= pre-test odds * LR
which test can be used to compare the results of multiple diagnostic tests?
LR
factorial study design - what is it?
study that uses >= 2 interventions and all combinations of these interventions (ie A, B, A and B, or nothing)
crossover study - what is it?
study where patients are exposed to different treatments/exposures sequentially
nested study - what is it?
retrospective observational study where subsets of controls are matched to cases and analyzed for variables of interest
is case-control study prospective or retrospective?
retrospective
primordial prevention
prevention of risk factors themselves
primary prevention
prevention of complications from a risk factor (ie someone with hld, start statin to prevnt MI)
secondary prevention
attempts to halt dz rogression beore irreversible changes take place
tertiary prevention
disease progressed beyond early stages, trying to limit impairments and disabilities (ie CABG)
quaternary prevention
ie preventing unnecessary repeat cardiac cath by using shared med record
sensitivity analysis
repeating primary analysis calculations after modifying variable ranges to see if modifications significantly affect results initialy obtained
propensity scoring
weighs different variables (ie severity of different comorbidities) in tx and control groups to ensure variables are balanced between both groups
lead time bias
happens when a test diagnoses a dz earlier than another test so “apparent survival period” appears longer
length time bias
subjects with rapidly progressive dz are less likely to be detected by screening compared to those w/slowly progressive dz
informed consent applies to which ethical principle of medicine?
autonomy
fairness in deciding who gets scarce resources applies to which ethica principle of medicine?
justice
not taking actions with respect to a patient that may be harmful or doubtful benefit - which medical ethical principle?
nonmaleficence
needed to harm (NNH)
inverse of the absolute risk increase (difference in incidence rate between exposed and non-exposed groups)
needed to treat (NNT)
inverse of the absolute risk reduction (difference in incidence rate between exposed and non-exposed groups)
net clinical benefit
possible benefit minus possible harm
intention to treat analysis - what is the main intent?
maintain randomization (ie avoid effects of crossover and dropout)
observer bias
observer may be influenced by prior knowledge or details of the study that affect the results
what do you do to avoid observer bias?
blinding
type II error
failure to reject the null hypothesis when it is false
why does type II error happen?
need good power of study (depends on sample size)
ppv calculation
TP / (TP + FP)
false positive rate
FP / (FP + TN)
NPV
TN / (TN + FN)
sensitivity calculation
TP / (TP + FN)
specificity
TN / (TN + FP)
PPV
TP / (TP + FP)
NPV
TN / (TN + FN)
positive LR
sensitivity / (1 - specificity)
negative LR
(1-sensitivity) / specificity
specificity - definition in words
probability of a nondiseased person testing negative
sensitivity - definition in words
probability of a diseased person testing positive
criteria for lung ca screening?
- age 55-80
- 30+ pack years
- quit less than 15 years ago or current smoker
attributable risk percentage - calculation?
(risk in exposed - risk in unexposed) / risk in exposed
OR
(RR-1)/RR
population attributable risk percent - how different than attributable risk percentage?
PARP = measure of excess risk in the total population, not just those exposed
population attributable risk percent - equation?
(risk in total population - risk in unexposed) / risk in total population
risk in total population - equation?
(risk in exposed)(proportion exposed) + (risk in nonsmokers)(proportion of nonsmokers)
type II error
study fails to reject a null hypothesis that is false (false negative)
equation for power
power = 1-B
type I error
study rejects null hypothesis that is true (false positive)
standardized mortality ratio - calculation?
observed # of deaths in pop of interest / expected # of deaths in total pop
negatively skewed distribution?
mean < median < mode (mea, med, mo)
positively skewed distribution?
mode < median < mean (mo, med, mea)
in skewed distributions, what is most valid measure of central location?
median
what is internal validity?
are we measuring what we think we are measuring
what is the major threat to internal validiy?
confounding
LR+
probability of pt w/dz testing positive divided by the probability of a pt without the disease testing positive
LR-
probability of a pt with the dz testing negative divided by the proability of a pt without the disease testing negative
NNT - equation?
NNT = 1/ARR ((ARR = control rate - treatment rate))
absolute risk reduction (ARR) - equation?
ARR = control rate = treatment rate
