Study validity Flashcards
Confounding
Relationship between exposure and outcome
Distorted (magnified or reduced) d/t a 3rd variable associated w/ both the exposure and the outcome
Random error
Chance error
Decreasing bias and confounding
Randomized, double blinded, control group
Cross over study
Pt is own control
Both control and experimental group for period of time
Selection bias
Sample not representative of population
Ex: more late stage dz than early stage dz refered to trial
Berkson’s bias
Hospital/patient bias
Ex: studying a group of hospitalized pt, results skewed d/t risk factor hx contributing to hospital admission in first place
Self-selection bias
Those who volunteer are different from those who don’t, may have same sx
Healthy worker effect
Decreased rate of disease if working
Harmful effect of a toxin exposure among active employees underestimated because control group filled w/ people from general population
Recall bias
Parents of a diseased child can recall more details of child’s history than non-diseased child’s parents
Lead-time bias
Earlier detection of dz does not mean longer survival
Late-look bias
Sending a survey out to people with a fatal illness 5 yrs after dx will preferentially sample those with low-grade dz or few comorbidities
Procedure bias
Groups with in same study group not treated the same
Experimental group gets more education/counseling than control group
Interview bias
Treat control and experimental groups differently
Ask more question about perceived sx improvement to experimental group than control group
Loss to follow up
More people lost to follow up from the experimental group of a lipid tx study than control group because of side effects
Hawthorne effect
Change from knowing they are being observed
