Final Epidemiology Flashcards
Sensitivity
Proportion of persons w/ dz w/ + test
Specificity
Proportion of persons w/o dz w/ - test
Positive Predictive Value (PPV)
Proportion of pt w/ + test who have the dz
*PPV is closely related to the prevalence
Negative Predictive Value (NPV)
Proportion of pt’s w/ - test who do NOT have the dz
Sn-N-OUT Rule
Sensitivity is high, you can RULE OUT the dz if test comes back -
Sp-P-IN Rule
Specificity high then you can rule in the dz if test comes back +
LR+
Probability of + test for person w/ dz divided by probability of + test for person w/o the dz
= Sensitivity / (1 - Specificity)
LR-
Probability of - test for person w/ dz divided by probability of - test for person w/o dz
= (1 - Sensitivity) / Specificity
LRs to “rule in” dz
> 10: strong
5-10: moderate
2-5: weak
LRs to “rule out” dz
0.2 - 0.5: weak
0.1 - 0.2: moderate
< 0.1: strong
Parallel Testing
Order several tests @ once, useful for rapid assessment situations
- maximizes sensitivity and NPV
- e.g. - CP in the ED order CK and Troponin, etc.
Serial Testing
Order next test on basis of prior result, useful when assessment can be done over time or tests are expensive
- maximizes specificity and PPV
- e.g. - serial troponins for a CP pt in ED under obs
Benefits of Screening
Earlier detection leads to better outcome, test has good sensitivity/specificity, burden of dz in community warrants test, resources available to follow-up + tests
Primary Prevention
Identify risk factor for dz and reduce exposure, promote resistance
*e.g. - immunization
Secondary Prevention
Identify early dz before signs/sx’s start
*e.g. - screen for DM
Tertiary Prevention
Dz has been identified but prevent advanced development, side effects, and outcomes etc.
*e.g. - pt w/ DM educate to prevent loss of sight/limb
Cumulative Incidence (Risk)
Probability of an individual developing dz during a specific timeframe, using # of persons @ risk for the denominator
Incidence Density (person-time)
Probability of an individual developing a dz during specific period of time using total person time as the denominator
Case Reports
Describe the experience of a single pt (n = 1) w/ an interesting finding
*no comparison
Case Series
Describe the experience of a group of pt’s (n > 1) w/ similar dx’s
*no comparison group
Cross-Sectional Studies
Aka Prevalence Study, exposure and dz status assesed @ single point in time, individual is unit of observation and analysis, no cause-effect and no incidence known
- dz or no dz -or- exposed or unexposed
- common analysis is Chi square
Case-Control Study
Key comparison: Dz vs. No Dz
- no prevalence or incidence assessed; measure risk of exposure to risk factors in dz group vs the no dz group retrospectively
Cohort Study
Key comparison: Risk factor vs. No Risk Factor, groups followed over time to see who ends up developing the dz, assesses incidence and causality but not prevalence
Intention-to-treat Analysis
Analyze according to group initially assigned regardless of whether they received tx
Explanatory Analysis
Analyze according to tx actually received, regardless of randomization
Efficacy setting
Analyzing tx in ideal setting
Effectiveness setting
Analyze tx in ordinary circumstances; usual pt care
Relative Risk Reduction
Describes the magnitude of the effect; represents the % reduction in risk of studied outcome achieved by the use of the intervention
Absolute Risk Reduction
Describes the risk difference in outcome btwn pt’s who have undergone 1 therapy and those who have not
Number Needed to Treat
NNT = 1/AR
*if you’re looking @ the treatment
Number Needed to Harm
NNH = 1/AR
*if you’re looking @ exposure
Randomized Clinical Trial
Randomly allocated into “intervention” or “control” group to receive tx or not
- control can be placebo group or standard of care group to compare the tx group to
Measures used to describe Prognosis
- 5 yr survival
- median survival
- response to tx (improvement)
- time to recurrence
- case fatality
- dz-specific mortality
Prognosis
Set of probabilities that predict a worse outcome once a dz is present
Survival Curve
AUC is # of person-years lived by population studies (probability of events is dependent on the length of follow-up)
e.g. - time to recurrence, time to remission, time to illness
Censored
When pt is lost from a study @ any point in time and are no longer counted in the denominator from that point forward; they are removed from denominator before the probability of survival is calculated for a given period
Equipoise
Randomization is ethical when there is no compelling reason to believe that either of the randomly allocated tx’s is better than the other
Observational studies show association and:
- exposure precedes dev of dz
- strength of assoc
- dose-response relationship is apparent
- consistency of findings from study to study
- biological plausibility
- findings are c/w other studies and/or types of data
Odds Ratio
OR = 1: no association btwn risk and outcome
OR > 1: increased risk of exposure in cases compared to control
OR < 1: decreased risk of exposure in cases compared to control