Biostats Flashcards
Incidence
new cases/total population
Prevalance
total with disease (new and old) / total population
Confidence interval range
Mean +/- (Z x SEM)
90% confidence Z = 1.645
95% confidence Z = 1.96
99% cofidence Z = 2.58
If CI for odds ratio or relative risk crosses 1 - no association between risk and the disease
If CI for study of two treatment groups crosses 0, no actual difference between two treatments
Standard deviations
1 SD - 68%
2 SD - 95%
3 SD - 99.7%
Case report/ case series
Report of an interesting/ unusual case (or a series of such cases) to describe the common features of the disease
Cross-sectional survey
Survey of a large number of people to determine disease prevalence and identify potential risk factors
Case control study
Retrospective - compare a group with the disease to group of healthy controls
Identify potential risk factors
Ideal for studying rare diseases
Allows calculation of odds ratio
Subject of late look bias (participants die) and recall bias
Cohort study
Follow a cohort that is exposed to a given risk factor to see how many of them developed a particular disease overtime
Usually prospective
Confirms risk factors (or risk reduction) and allows calculation of relative risk
Randomized clinical trial
Compares the effect of an intervention in an experimental group to a control group
Subjects are randomized into two groups
Usually double blinded to avoid bias
Meta-analysis
Pool the data for multiple studies to create a larger dataset
Can help resolve conflicts in the medical literature
Maybe subject to publication bias
Limited by the quality of the original studies
Type one error
Study shows relationship between two phenomena that does not exist
The study findings inappropriately reject null hypothesis (null is true)
Type II error
Study fails to show the relationship between two phenomena that does exist
The study fails to appropriately reject the null hypothesis (null is false)
Changing behavior that can result from a study group knowing that they are being observed
Hawthorne effect
Findings are influenced by the actions of the people collecting or interpreting the data
Investigator bias
Information is gathered too late to draw accurate conclusions about the disease or risk factor in the entire intended study population
Late look bias
Early detection of a disease increases the time that a person survives after diagnosis, even if the natural history of disease is unchanged
Lead time bias
Screening detects slowly progressive cases of the disease and misses rapidly progressing cases
Length bias
Study participants drop out of a study and cannot be contacted for follow-up
Loss to follow-up
Subjects awareness of being studied alters their reporting of subjective findings; maybe avoided by double blinding and placebo controls
Observation bias
Study groups are not treated the same
Procedure bias
Studies that show a difference between two groups are more likely to be published in studies that do not show difference
Publication bias
Difference and recall between the two study groups
Recall bias
Systemic difference in the way subjects are selected that makes the study population different from the entire eligible population
Selection bias
Patients with a certain medical history are more likely to participate in a study related to their condition, although they aren’t necessarily representative of the general population
Self selection bias
Confounding variables
Factors that affect both the control group in the experimental group
