2 - Association in Medicine Flashcards
Understand the concepts of rate and risk.
RATE: A measure of event frequency; the speed with which events happen, relative to the size of the population experience observed
RISK: The probability that an event will occur in a defined period of time
- -For uncommon events, a rate of X per 100 person-years implies a one-year risk of X percent
- -Neonatal/infant mortality is a RISK, not a rate
Understand the different uses and implications of relative risk and risk differences.
RELATIVE RISK: risk exposed/risk unexposed
- -The farther RR is from 1.0, the stronger the evidence for a causal relationship
- -Tend to be similar across populations, but does NOT indicate causality unless very high (-> increases suspicion)
- -Doesn’t capture what you stand to gain or lose
- -No units!
RISK DIFFERENCE: risk assumed (or avoided) with risk factor [risk exposed - risk unexposed]
- -Measures clinical effect (“you increase your risk by 2.5%” rather than “you’ve doubled your risk”)
- -Smaller if background risk is small
- -Larger if background risk is large
- -Always has units (e.g. per 1,000 person years)
Understand how confounding can distort assessment of associations; describe the strategies that can be used in research to deal with confounding.
Ex: alcohol intake is associated with smoking. If you are looking at smoking as a relative risk for cancer, but alcohol use is very common in smokers (>3x as common as in non-smokers, perhaps) and alcohol increases your risk of cancer by a lot, then alcohol is a confounding factor that could explain all of the risk for cancer.
–Lifestyle habits run in packs, so you need to deal with confounding
Dealing with confounding:
- -MATCH on confounder: match population (people who drink 3x a day + people who smoke and drink 3x a day, or age matching, etc.)
- -ADJUST for confounder: using statistical methods or weighted stratification
- -RESTRICT analysis: study effects of coffee-lung cancer association only among life-long non-smokers so smoking doesn’t confound risk
Define adjusted analysis.
An analysis in which statistical account is made for potential confounding factors, so that an estimate of the independent effect of a risk factor can be made
Define cohort.
A group of subjects followed over time
Define competing risks.
Events that prevent the observation of a possible endpoint
Define confounding.
An indirect association of an exposure with a disease
Due to the association of the exposure with a confounding factor which is also related to the disease
Define effect.
The magnitude of a difference or relationship
Define event. List four examples.
A clinical outcome of importance
–Ex: onset of a disease, onset of a particular symptom, disease recurrence, or death
Define incidence.
The risk or rate or occurrence of new cases of a disease
Inherent time element!
–Ex: incidence of HTN is low, but prevalence is quite high
Define matched analysis.
Choosing exposed and unexposed subjects to have the same or similar values of some trait or exposure
Typically done to control confounding
Define prevalence.
The proportion of individuals who have a particular disease or trait in a given population
EXISTING cases at a point in time
Cross sectional - affected by disease incidence and duration
Define rate.
A measure of event frequency
The speed with which events happen, relative to the size of the population experience observed
Define relative risk (aka rate ratio, risk ratio)
A ratio of risks
Also used to refer to any relative measure of association
Numerator = occurrence of events (risk, rate, odds, hazard) in an exposed group
Denominator = occurrence of events in an unexposed group
Define restricted analysis.
Confining the analysis to one group of subjects to minimize confounding
Define risk.
The probability that an event will occur in a defined period of time
Define risk difference.
The difference between the risks of an exposed group and an unexposed group during a defined period of time
The risk of disease attributable to the exposure
Define risk factor.
An exposure associated with the occurrence of a disease or outcome
Define stratum.
A grouping of subjects, typically formed for the purposes of adjustment
–Ex: age groups
What is confounding by indication?
Disease severity associated with treatment
–Ex: use treatment Y for very severe disease and treatment Z for mild disease. Treatment Y will be associated with worse outcomes
T/F: Risk factors strongly related to a disease are virtually always causal.
FALSE - can just be associated
T/F: Risk difference is the most scientific way to understand an exposure/disease relationship.
FALSE - only important in communicating with patients
T/F: An association with a high relative risk will always be clinically important.
FALSE - must be meaningful (can take you from 0.00000001 to 0.0000001)
T/F: Risk differences tend to be more or less constant across populations with different disease rates.
FALSE. Increase with increasing disease rate, decrease with decreasing disease rate
T/F: A risk factor with a high risk difference will almost always be causal.
FALSE - can have a strong association only or confounding factors
–It is inappropriate to look at risk differences as causal relationships
T/F - An association confounded in real life cannot ever be investigated properly.
FALSE - can adjust via matching, adjusting, or restricting analysis
T/F: The number of babies in NH born with neural tube defects per 1000 live births in 2009 is a measure of prevalence not incidence.
TRUE. Birth conditions reported as PREVALENCE at birth.
