First Aid- Epidemiology and Biostatistics Flashcards
define cross-sectional study
assessing frequency of disease and frequency of risk-related factors- ex. Surveys
(what’s happening)
define case-control study
-compares group with disease and group without disease
-do odds of prior exposure/risk factors differ by disease state
(what happened)
define cohort study
- compares group with exposure to group without exposure
- is exposure related to LATER developed disease
- prospective or retrospective
define twin concordance study
- compares frequency of disease between identical and fraternal twins
- higher the concordance –> more genetic the factors, the lower –> more environmental factors
define adoption study
compares siblings raised by biological v adaptive parents to measure heritability and influence of environmental factors
define triple blind study
- single is patients don’t know what treatment they are getting
- double is patients and doctors not knowing
- triple is patients, doctors, and researchers/statisticians not knowing
purpose of phase I trials
(is it safe)
assesses safety, toxicity, pharmodynamics and pharmacokinetics (discover safe dose range)
purpose of phase II trials
(does it work)
assesses treatment efficacy, optimal dosing, adverse effects
purpose of phase III trials
(is it better than what there is)
compare treatment to current standard of care
purpose of phase IV trials
(can it stay)
- drug is on the market, although can be withdrawn
- detects rare, long-term adverse effects
define drug/therapy efficacy
how well the drug/therapy works under good experimental conditions
(internal validity)
define drug/therapy effectiveness
how well the drug/therapy works under real world conditions
generalizability, external validity
sensitivity, aka (1)
define SN-N-OUT
1- true positive rate
2- high sensitivity, if negative, r/o disease
specificity, aka (2)
define SP-P-IN
1- true negative rate
2- high specificity, if positive, rules in disease
pretest probability varies directly with…
positive predictive value, prevalence
pretest probability varies inversely with…
negative predictive value
LR+ formula and the desired/significant value
LR+ = sensitivity / (1 - specificity) = TP rate / FP rate
LR+ > 10 indicates useful diagnostic test
LR- formula and the desired/significant value
LR- = (1 - sensitivity) / specificity = FN rate / TN rate
LR- < 0.1 indicates useful diagnostic test
LRs (+ or -) multiplied by (1) estimates (2)
1- pretest odds
2- posttest odds
PostOdds = LR * PreOdds
prevalence = …
prevalence = (number of cases) / (total population)
[at a point in time]
incidence = …
incidence = (number of new cases) / (number of people at risk)
[during a given time period]
high prevalence indicates what values of PPV and NPV
(positive/negative predictive value)
- high PPV
- low NPV
in acute diseases prevalence is (=/>) incidence
in chronic disease prevalence is (=/>) incidence
Acute: P ~= I
Chronic P > I
define Odds Ratio (OR)
odds of certain exposure given an event (disease) / odds of exposure in absence of exposure
-usually used in case-control studies
define Relative Risk (RR) and the indications of all 3 results
risk of developing disease in exposed group / risk of developing disease in non-exposed group
-usually used in cohort studies
RR = 1, no association with exposure/disease
RR > 1, exposure associated w/ inc disease occurrence
RR < 1, exposure associated w/ dec disease occurrence
define Attributable Risk (AR)
difference in risk between exposed group and non-exposed group
define number needed to treat, what is ideal value
- number of people that need to be treated for 1 patient to benefit
- low number is best
define number needed to harm, what is ideal value
- number of people that need to be exposed to risk factor for 1 patient to be harmed
- high number is best
positive skew has peak on (R/L) side
compare values of mean, mode, median
- L sided peak (L modal) – R sided tail
- mean > median > mode
negative skew has peak on (R/L) side
compare values of mean, mode, median
- R sided peak (R modal) – L sided tail
- mode > median > mean
define null hypothesis (H0)
no association between disease and risk factor (hypothesis of no difference / relationship)
define alternative hypothesis (H1)
some association between disease and risk factor (hypothesis of some difference / relationship)
define selection bias
nonrandom sample or treatment allocation that is not representative of target population
- usually a sampling bias > treatment allocation
- during recruitment participants phase
define recall bias
awareness of disorder alters subject’s recall
- common in retrospective studies
- during actual performing of study phase
define measurement bias
info is gathered is systemically distorted manner: hawthorne effect, faulty equipment
-during actual performing of study phase
define Hawthorne effect
study participants change behavior (improved health behaviors) when they know they are being observed
define procedure bias
subjects in different groups are not treated same (treatment group has more specialized care than placebo group)
-during actual performing of study phase
define observer-expectancy bias
aka Pygamalion effect, researcher expects treatment group to do better and has more positive view or documents more positive outcomes
-during actual performing of study phase
define confounding bias
when a factor contributes to both an exposure and outcome, but not necessarily on the causal pathway
- ex. inc Pulmonary disease in coal miners > general population, but also higher smoking rates in in coal miners
- during interpretation of results phase
define lead-time bias
early detection of disease is confused with inc survival of people with disease
- fix by adjusting survival to severity of disease at time of diagnosis
- during interpretation of results phase
define length-time bias
screening test detects disease with long latency period, although those with shorter latency are symptomatic earlier
- ex. slow progressive cancer is more likely to be detected by screening test than rapid progressive cancer
- during interpretation of results phase
type I error
false positives (aka α error)
type II error
false negative (aka β error)
Power (of a study) = ….
P = 1 - β
β- false negatives
inc Power (dec β) by inc sample size, inc expected effect size, inc precision of measurement
what is the ideal/pre-judged α value, what is compared against, and what does it mean
α = 0.05
- indicates 95% confidence interval
- if p < α / 0.05 => 5% chance the data is random, statistically significant results
what is the ideal CI
(confidence interval)
- 95%
- Odds Ratio with confidence interval that doesn’t include => statistically significant results
define meta-analysis
statistical analysis of multiple studies to generate more precise estimate of the size of an effect, estimate heterogeneity between studies, and improve generalizability of study findings
define t-test (statistical test)
compares differences of means between two groups in a study (often men and women)
define ANOVA (statistical test)
compares differences in means between three or more groups (often ethnicity)
define Chi-square (statistical test)
compares differences between two or more groups by percentages/proportions of categorical outcomes
Pearson correlation coefficient = …… (include its characteristics)
r, between -1 and 1
- positive => positive correlation, negative => negative correlation
- r = 0 => no correlation
- r^2 = variance
