|
What/Why is a measure of disease occurence for?
|
Quantify the likelihood of disease development in a single group.
1)Risk
2)Odds
3)Incidence rates
|
|
|
4 purpose of cohort analysis
|
1)Measure of the occurence of D
2)Measure the effect of E on the occurrence of D
3)Control for confounding by other differences between E+ and E-
4)Test/Identify effect modification
|
|
|
Risk
|
The probability that a person who initially does not have the disease will develop it by the end of fixed or defined period
|
|
|
3 major components of risk
|
1)Probability (between 0 and 1)
2)simple: calculate proportion who develop D
3)Can use life-table method
|
|
|
Synonyms of risk
|
Cumulative incidence
Cumulative incidence rate
Cumulative incidence proportion
|
|
|
Odds
|
Odds = p/(1-p)
p=risk
|
|
|
Incidence Rate
|
likelihood of occurerence for binary outcomes that quantifies the instantaneous rate of disease development.
|
|
|
3 What NOT to do for variable follow-up
|
1)Ignore and estimate risk as proportion
2)Take shortest period of follow-up
3)Never allow - always do fixed period studies
|
|
|
What to with variable follow-up
|
Estimate incidence rate and calculate risk if strict assumptions hold true
|
|
|
Synonyms for Incidence
|
1)Incidence rate
2)Incidence density
3)Hazard function
4)Force morbidity/mortality
|
|
|
4 characteristics of Incidence rate
|
1)Varies between 0 and inf
2)[events/time]
3)Magnitude of rate depends upon time unit
4)often not constant over time
|
|
|
Can use person-time methods with 3 assumptions
|
1)Incidence is constant over time
2)Incidence is homogenous among persons in the population
3)Loss to follow-up or censoring is not associated with D
|
|
|
Incidence eqn if assumptions hold
|
I=(# of events)/(total person-time)
|
|
|
2 equations for estimating risk based on incidence
|
I must be constant
1)Risk(t) = 1-exp(-I*t)
2)If I*t < 0.1: Risk(t)=I*t
|
|
|
For non-constant I
|
Use survival analysis (life tables)
|
|
