Epi 1-3 Flashcards
threats to validity
Bias
Confounding
confounding aanpakken:
design:
analyse:
design: randomiseren, restrictie, matching
analyse: stratificatie, regression
relatie tussen odds en proportions
small proportions geeft gelijke odds, hoge getallen zal grote verschillen geven: prop 0.5 = OR 1.0
Prevalence =
proportie op 1 punt in tijd ziekte/totale populatie
s functie incidence & ziekte duration
problemen met cum incidence
- competing risk
- loss to follow upv (problamtscih als non random lost to follow up).
oplossing: incidence rate (IR / T).
MoA: RR, RD (attributable risk), OR
RR: R1/R0
RD = R1-R0
OR = (R1/1-R1)/[R0/1-R-]
unieke eigenschap OR:
Symmetrisch/reciprocal: Odds NOT develop disease = 1/odds develop disease.
Niet waar voor andere MoA
wat is meestal groter: RR of OR
OR
Attributatble fraction
+1portion of risk AMONG EXPOSED that is attirbutatble to exposure. = AR%
AR% = (RR-1)/RR = RD/R1
among xposed, ..% of risk is attibutable to hypertension.
Population attributable fraction
Tot AF: portiomn of risk in POPULATION that is attributable to exposure: PAR%
PAR% = [(p)(RR-1)][(p)(RR-1)+1].
p = prevalence of exposure = % exposed in population
PAF% uitgeschreven in zin
..% of risk in this population is attributable to E.
NNT
NNH
om 1 case te veroorzaken of voorkomen
NNT = 1/RD (als R1 < R0)
E verlaagt R
NNH = 1/RD (als R1 > R0)
(E verhoogt R)
In geval van competing risks welke MoA kan je wel gebruiken en welke niet?
RR, OR niet
Rate ratio wel!
Justification for RCT
equipoise: uncertainty about risk and benefits of treatment
meaning: sufficient expected benefit to gife treatment, sufficient doubt to withhold treatment from others
Waarom is RCT goede benadering echte ffect?
geeft best mogelijke weergave van countferfactual (incidence controle groep).
Oplossing dat er in kleine trials imbalance is:
random permutation (blocked randomization)
evt gecombineerd stratification: binnen stratum blocked randomization.
Effictiveness =
efficacy =
expected effect of a treatment in actual practice
expected effect of a treatment in an ideal situation.
Effect non compliance
observed effect of intervation towards 1.
Doel blinding
Masks treatment assignment, to ensure that conduct of study post-randomization is fair
hawthorne effect
subject acts differently when studied.
primaire analyse RCT is altijd:
ITT
secundair kan wel compliers only (dit is dan non randomized want patietnen zelf selecteren).
Pro cons factorial design
Can increase efficiency by studying multiple interventions in same populations, can examine whether effect of one treatment depends on another.
Cons: if there is interaction (effect modification) between treatments, decreased statistical power.
Cluster randomized trials
randmonizeren per groep
limit: individuele chars niet evenredig verdeld,