Epidemiology and ethics Flashcards
Sensitivity
TP/(TP+FN)
True positve rate, all negatives are TN
To screen disease, if negative, SNOUT
Specificity
TN/(TN+FP)
True negative rate, all positive are TP
To confirm disease, if positive, SPIN
Case-control study
Observation, retrospective
“What happened?”
Uses ODD RATIO (OR)
Cohort study
Observation, either prospective or retrospective
Who will develop disease, or who developed disease?
Uses RELATIVE RISKS (RR)
Cross-section study
Observation
“What’s happening”
Determines disease prevalence (DP)
***association but not necessarily causation
Clinical trials
1: few healthy volunteers: safety, toxicity, PK
2: few diseased pts: efficacy, optimal dosing, adverse effect
3: large number of pts: new tx vs current standard of care
4: postmarketing surveillance: detects rare, long term effect
PPV and NPV
PPV: high pretest prob => high PPV
NPV: high pretest prob => low NPV
Incidence vs. prevalence
Incidence rate: # new cases in a time/ pop at risk during the same time period
Prevalence: # exisiting cases/ pop at risk
Prevalence ~ incidence rate x average disease duration
Prevalence > incidence for chornic dz
Odds ratio (OR)
Used in case controlled (obs/retro)
dz expos/dz unexpo)/(healthy expo/healthy unexpo
Relative risk (RR)
Used in cohort (obs/retro or pros)
(dz/all exposed)/(dz/all unexposed)
If prevalence low, RR~OR
Attributable risk
Diff in risk btwn exposed and unexposed
If lung cancer risk 21% in exposed, 1% in unexposed,
then 20% is ATTRIBUTABLE RISK to smoking
*1/AR= number needed to be exposed (for 1 pt to be harmed)
Absolute risk reduction (ARR)
Absolute reduction in risk due to tx compared to control
8% of placebo vs. 2% vaccinated develop flu = 6%
*1/ARR= number needed to treat (for 1 pt to benefit)
Precision vs. accuracy
Precision: consistent and reproducible/reliable, absence f random variation in the test
High precision=> low SD
Accuracy: trueness of test measure, validity, absences of systematic erros or bias in a test
Biases…
Selection: non random assignment (loss to f/u)
Recall: knowledge of presence of disorder alters recall by subjects; common in RETROSPECTIVES.
Sampling: subjects not generalizable to population
Late-look bias: info gathered at an inappropriate time (using a survey to study fatal dz, but only those alive will be able to answer the survey)
Procedure bias: subjects not treated the same; more attention given to tx group, so better adherence)
Confounding bias: occurs when factor is related to both exposure and outcome, distorting the effect of exposure
Lead-time bias: early detection confused with increased survival, seen with improved screening, but natural hx of dx is not changed, but early detection makes it seem AS THOUGH survival has increased
Observer-expectancy effect: researcher’s belief in the efficacy of a tx changes the outcome of the tx
Hawthorne effect: the group being studying changes its behavior due to the knowledge of being studied. Dr. Hawthorne is watching you.
Normal distribution
Gaussian, aka bell-shaped
Mean=median=mode
Mode is least affected by outliers
Positive skew: means>median>mode, longer tail on right
Negative skew: mean<mode, longer tail on left