Biostats Flashcards
Accumulation effect
Risk associated w duration of exposure: mom cumulative exp necessary for exp to have a significant effect on risk
Lead time bias
Dx of a disease at an earlier stage = prolongs survival BUT inc survival time d/t early detection
Observer bias
Misclassification of data d/t individual diff or preconceived expectations
Reduced by:
- Double-blind study
- Multiple observers encode n verify
Attrition bias
Form of selection bias
unequal loss of participants from a randomized controlled trial (RCT).
Allocation vs selection bias
Allocation = pts nonrandomy assigned to tx and control grps
Selection = pts not representative of the intended population. eg only cancer pts enrolling in a experimental trial
Referral / admission rate bias
Eg hospital specializing in treating asbestos may report inc relation bw asbestosis and lung ca
Length-time bias
rapidly progressing diseases are less likely to be detected
Length time vs Lead time bias
Length time bias: rapidly progressing diseases less likely to be detected
Lead-time bias: screening leads to inc detection
Detection bias
Pts w RF (eg. Smoking) undergo more testing/ screening
Preventing bias
- Double blind studies
- Randomized
- Placebo
- Cross-over studies
Case-control
Measures exposure status
Control - Normal grp
Study designs
Past:
Present:
Future:
Past:
- Retrospective cohort : incidence, RF
- Case-control: RF frequency, OR
Present:
1. Cross-sectional: Prevalence
Future:
- Prospective cohort: RF, incidence
- Clinical trial: Tx + control, outcome of interest
How is confounding controlled in case-control?
Matching
Cross-sectional study
Prevalence study
Exposure and outcome
Errors
α error:
β error:
1-β:
α error: FP
p<0.05
β error: FN
β = 0.2
1-β: Power = 80%
Prospective cohort
Exposed group followed over time for dev of disease
OR calculated in?
CohORt (uw)
Case-control (FA)
Prevalence calculated in
Cross-sectional
RR calculated in
Cohort
Precision vs accuracy
Precise = how close together it is
Accuracy = how close to the target value
RCT
Compares Tx
Drug vs placebo
Cumulative incidence
New cases/ total people at risk @ a specific time
ARP
ARPexposed = 100 × [(risk in exposed – risk in unexposed)/risk in exposed]
ARP = (RR-1)RR X 100
How is the accuracy of screening tests determined?
Area under curve (AUC) of receiver operating characteristic (ROC) ~ 1 w/ high SN + SP
