Bias and Cofounding

Bias

systemic error in design that results in mistaken estimate of exposure’s effect on disease risk

Selection Bias

results from differences in characteristics between those in study and out

controlled by defining criteria of selection

Interviewer or Observer Bias

probe differently about past exposures if subjects known

controlled by proper training and double blinding

Recall/Report bias

subjects forget/lie about past exposure if have disease

controlled by surrogates giving more info

Confounding

outside variable is associated with disease outcome

clouds effect of risk factor on disease

controlled by randomization/restriction/matching and stratified/multivariable

Confounding Control Design -Randomization

equal distribution of confounders between study and control groups

Confounding Control Design - Restriction

subjects restricted by levels of known confounder

Confounding Control Design - Matching

potential factors are kept equal between the study groups

Confounding Control Analysis - Stratified

analysis for various levels of potential confounders

Confounding Control Analysis - Multivariable

analysis for various levels of potential confounders

Random or Chance Error

association between exposure and outcome may be result of chance

minimize chance by increasing sample size and significance

Casual interference - strength of association

effect measure (OR/RR) away from unity

farther from 1 the better it is

Casual interference - Consistency

observed repeatedly by different investigators, populations, and study designs

meta-analysis good to asses

Casual interference - Dose response

probability of the outcome increases as the exposure level increases

Casual interference - Temporality

exposure must lead to disease

only criteria needed to judge

Casual interference - Biological Plausibility

possible biological reason for how exposure can lead to disease

Casual interference - Analogy

similar cause-effect relationships that a previous and new exposure can cause similar outcomes

Casual interference - Experiment

interventions have predictable effects on disease occurrence

Casual interference - Coherence

casual relationships do not conflict with other facts regarding disease