9: Observational studies

Question 1

Cohort study
Cohort
How to chose cohort
How to measure expsoure
How to measure outcome
Mesures
Aim
Pros
Cons

Answer 1

Cohort
- groups of persons that are similar in one more more traits and move forward together as a unit

How to chose cohort
- select target population under risk (only persons without disease of interest)

1. select population, then meaasure exposure if all or
2. select exposed and not exposed persons directly (maybe xternal comparison group)

How to measure expsoure
- routine data, data from earlier investigations
- interview, questionaire
- examinate person or enviroment

How to measure outcome
- standardized assessment, same methology for both groups, obs. equality

Mesures
- incidence, risk ratio / relative risk
- survival analysis (Kaplan-Meier-Schätzer, days of exposure)

Aim
- determine natural history of disease, identify factors that explain its behavior, delay disease onset by prevention

Pros
Descriptive:
- incidence

Bradford-Hill:
- temporalioty
- dose response relationships

Study scope:
- multiple effects, rare exposures

Cons
Ressources:
- time/money
- large study population

timelyness:
- problem of latency period

Bias:
- changing conditions, drop oits, no exp. study

Question 2

Case control study
Structure
selection of groups

Answer 2

Selection of cases
- can use variety of sources, careful specification of criteria for eligibilty (incident vs prevalent, several vs one hospital)

Selction of controls
- control shouldnt have the disease
- random sample
- similar to cases (except exposure) by matching, but conceptual problems cause you cant analyse these factors
- non hospital controls or hospital controls (biased through confounders)

Measure exposure
- exsiting records
- interview, questionaire

Measures
no incidence, therefore no RR
only Odds ratio

Pros
resources:
- timesaving (latency)
- cost saving

scope of study:
- rare disease can be studied
- several expsoure in aetiology of disease

Cons
descriptive:
- incidence, RR

causation:
- selection bias
- observation bias
- recall bias,
- structural, obs., operation equality questionable

Question 3

Calc Odds ratio for Case control study

Answer 3

• Proxy of RR, is always more extreme
• OR = (A/B) / (C/D), odds to have the disease in exposed vs non exposed people
• also AD / BC

Question 4

Cross sectional study

Answer 4

Prevalence study
at one time point, measurement similar to RR, but we have prevalence data, therefore assocation of disease is maybe rather surviving than developing the disease

Pros
costs, time, latency does not impact study

Cons
no temporality at all, prevalence data