Snw's Epidemiology & Biostatistics Review (Usmle Step 3)

Front 1

Epidemic spread

Back 1

common source: group of persons are all exposed to an infectious agent or toxin from same source

point source: type of common source where individuals become ill within one incubation period

propagative: occurs from transmission from person to person over more than 1 incubation period

Front 2

Adjusted estimation

Back 2

reported cases serve as lower bound

calculations based on a model determine upper bound

midpoint is best guess to how many cases were actually seen

Front 3

Types of prevention

Back 3

primary prevention: prevents onset of risk behaviors --> decreases incidence (eg health education that promote healthy living)

secondary prevention: screens for risk factors and early detection of asympt or mild dz --> decreases prevalence (eg community BP screening, MD support to quit smoking)

tertiary prevention: reduces long-term impairment or disabilities --> decreases recurrences and slows dz progression (graded aerobic activity during recovery from MI)

Front 4

Adjusted rates

Back 4

calculated after using statistical procedures to minimize demographic differences btwn populations being compared

Front 5

Prevalence rate

Back 5

proportion of individuals with existing dz at a point of time

prevalence = (ppl with existing dz / population of ppl at risk for dz) x multiplier

point prevalence --> prevalence at a given time

period prevalence --> prevalence over period of time

good for chronic dz

Front 6

Incidence rate

Back 6

proportion of individuals developing new dz during a period of time

can only be calculated over a period of time

incidence rate = (ppl with dz onset / population at risk for dz) x multiplier

good for acute conditions

Front 7

Prevalence pot

Back 7

determined by rates of incidence, recovery, and death

prevalence = incidence x duration

Front 8

Number needed to treat

Back 8

how many ppl do you need to treat to prevent 1 case

inverse of incidence rate

if incidence 16 per 1,000 (16/1000) then NNT = 1000/16 = 62.5

if comparing two modulaties it is the inverse of absolute risk reduction (ARR) -->

1 / (incidence A - incidence B) where ARR = incidence A - incidence B

Front 9

Number needed to harm

Back 9

1 / (incidence of exposed - incidence of non-exposed)

Front 10

Infant mortality

Back 10

# deaths within 1yr of age / total births

AA have highest rates of infant mortality bc of low birth weight & infections

Native Americans have highest rate of SIDS

Front 11

Infant mortality risk factors

Back 11

maternal age (younger worse)

poverty

single parent family

Front 12

Years of potential life lost

Back 12

sum of yrs that ppl would have lived had they experienced normal life expectancy

if life expectancy is 75 & person dies at 65 then YPLL is 10

Front 13

Positive predictive value

Back 13

true positive / (true positive + false positive)

if test is positive what is the probability that someone has the dz

this is dependent on specificity

Front 14

Negative predictive value

Back 14

true negative / (true negative + false negative)

if test is negative, what is probability that someone doesnt have dz

this is dependent on sensitivity

Front 15

Types of studies

Back 15

Front 16

Observational vs experimental studies

Back 16

observational: no intervention, just observe natural course

experimental: there is an intervention and study determines the effect of the intervention

Front 17

Different types of observational studies

Back 17

1) case report: single clinical subject or event with no control

2) case series: characterization of a group of clinical subjects

3) cross-sectional study: presence or absence of dz determined in each member of the study population or in a representative sample at a particular time --> interested in prevalence but cause and effect cannot be determined (study is at a given time) data analyzed with chi-squard test

4) case-control study: identifies a group of ppl with a dz and group w/o dz then compares them for risk factors --> mostly retrospective (study is of the past) data analyzed with odds ratio

5) cohort study: identify population who has been exposed to risk factor & is followed over time and compared with a group not exposed to the risk factor --> prospective (study is following through the future) data analyzed with relative risk

Front 18

Relative risk

Back 18

goal of observational study

incidence rate of exposed group / incidence rate of unexposed group

Front 19

Attributable risk

Back 19

how many more cases in one group?

incidence rate of exposed group - incidence rate of unexposed group

Front 20

Odds ratio

Back 20

odds of exposure for cases / odds of exposure for controls

used for case-control studies

Front 21

3 phases of clincial trial

Back 21

Phase I: safety in healthy volunteers

Phase II: protocols and drug levels in small groups of patient volunteers (couple of 100)

Phase III: efficacy and occurrence of side effects in larger group of patient volunteers

Front 22

Different types of randomized controlled clinical trials

Back 22

1) double blind: neither subjects nor researchers know whether the subjects are in the tx or control group

2) community trial: therapeutic regimen received by an entire community

3) crossover study: for ethical reasons, no group involved can remain untreated; all subjects receive intervention at different times

Front 23

Types of bias

Back 23

selection bias: sample is not representative of population (Berkson's bias)

measurement bias: information is gathered in a manner that distorts the information (Hawthorne effect -> subject knows they are being evaluated so that changes performance)

experimenter expectancy: experimenter's expectations inadvertently communicated to subjects (Pygmalion effect)

lead time bias: gives a false estimate of survival rates

recall bias: subjects fail to accurately recall events in the past

late-look bias: individuals with severe dz are less likely to be uncovered in a survey bc they die first

confounding bias: the factor being examined is related to other factors of less interest

Front 24

Variance and STD

Back 24

variance = (value - mean)^2

STD = sqrt (variance1 + ... + varianceN / N)

+/-1 STD cover 68% of values are within mean

+/- 2 STD cover 96% of values are within mean

Front 25

Standard error

Back 25

SE = STD / sqrt (N)

Front 26

Skewed curves

Back 26

negatively skewed --> mean < median < mode (tail skewed to left)

positively skewed --> mean > median > mode (tail skewed to right)

Front 27

Confidence interval and significance

Back 27

if 95% CI overlap then not significant

if there is no overlap, then significant

Front 28

Relative risk and confidence interval

Back 28

if 95% CI does not contain 1.0 then is significant

if RR is >1 and CI does not contain 1.0 then there is a significant increased risk

if RR is <1 and CI does not contain 1.0 then there is a significant decreased risk

Front 29

Type I (alpha) error

Back 29

rejecting null hypothesis when it is really true --> think something is significant when it's not

if p-value = 0.05 --> type I error is 5 in 100 (aka 1 in 20)

Front 30

Type II (beta) error

Back 30

fail to reject null hypothesis when it is false --> think something isnt significant when it is

type II error= 1-power

Front 31

Power of study

Back 31

capacity to detect a difference if there is one

Front 32

Chi-square test

Back 32

you know what to expect and testing whether you are seeing that or not

eg toss coin 10x, you expect 5 heads & 5 tails

Front 33

t-test

Back 33

independent group: test 2 independent groups

paired sample: each group is measured 2x, once before and once after intervention

one group: have only 1 group but want to compare to imaginary group

Front 34

Length time vs lead time bias

Back 34

both involve screening tests

length time - testing for a dz that can have rapid or indolent progression & screening just selects for those with an indolent course thereby leading to impression screening improves survival

lead time - advent of new screening test picks up test earlier but there is no change to actual duration to mortality (increasing sensitivity picks up dz earlier)

Front 35

Likelihood ratio

Back 35

Positive LR = sens / (1-spec)

Negative LR = (1-sens) / spec

prob of given test result occurring in a pt with a disorder compared to the prob of the same result occurring in a pt w/o the disorder

not dependent on disease prevalence as PPV and NPV are