Designing an Observational Study Cross Sectional and Casecontrol

Designing an Observational Study: Cross. Sectional and Casecontrol studies Roberto Cardarelli, DO, MPH, FAAFP Professor, Chief of Community Medicine

Cross-sectional studies • All measurements are taken on a single occasion • Looks at distributions within sample • Good for describing populations • Look at relationships (not causality) between predictor and outcome • Examples are national cross sectional surveys like NHANES, BRFSS

Example 8. 1

2 x 2 Table Yes Chlamydia No Chlamydia • • Yes OC 4 (a) 20% 16 (b) 80% No OC 8 (c) 10% 72 (d) 90% Prevalence of chlamydia infection in users= 4/20=20% Prevalence of chlamydia infection in nonusers= 8/80=10% Prevalence of chlamydia infection overall= 12/100=12% Relative prevalence 20%/10%=2. 0

8. 1 Statistics for expressing Disease Frequency

Strengths of crosssectional studies • • • No waiting for outcome to occur Fast and inexpensive No loss to follow up Can be first step to cohort study Give prevalence of a disease or risk factor Convenient for looking at networks of causal links

Weaknesses of Cross. Sectional Studies • Cannot establish causation • Impractical for studying rare disease – Ex. Infant mortality – Can be done if sample comes from population that has the specific outcome • Can only look at prevalence, not incidence

Prevalence vs. Incidence

Serial Surveys • Series of several CS studies at several points in time. – Ex. Census data • Good when looking at changes in population

Case-Control Studies • Generally retrospective • 2 groups – 1 with disease, the other without • Look backward to find differences in predictor variables • Provide descriptive information on characteristics of the cases • Provides estimate of the strength of the association btw predictor and presence of disease

Strengths of Case. Control • Can look at rare diseases • More efficient than other designs if disease is rare • Can look at a large number of predictor variables • Useful in generating hypothesis of causes of new outbreaks.

Weaknesses of Case. Control studies • Cannot directly estimate incidence or prevalence • Only one outcome can be studied • Greater susceptibility to bias – Sampling – Retrospective measure of predictors

How to control sampling bias • Sampling from patients with disease: – Not representative of all patients with disease – Leaving out undiagnosed, misdiagnosed, or dead

How to control sampling bias cont. • Selecting controls is bigger problem – Need to select from a group that is at risk of disease – Need to be similar to the cases • Four strategies: – – Hospital or clinic-based controls Matching Using a population-based sample Using two or more control groups

Hospital or clinic-based controls • If cases obtained from hospital, controls can be obtained from same place. • However, can look like prevalence of risk factor in control group is high if the risk factor also causes diseases for which patients seek care. • Balance between convenience and validity

Matching • Ensures cases and controls are comparable • Can match to factors related to disease but of no interest to investigator such as: – Age, or sex

Using a populationbased sample • Can use population based registries that track new cases of cancer, AIDS, birth defects. • They are specific to an area • Controls just need to be from the same area • Can use random-digit dialing – However, has problems

Using two or more control groups • The more control groups you have the less chances of bias • Ex. PHS study of Reye’s syndrome- used 4 control groups • Consistent findings=more validity

How to control differential bias • Bias that affects one group more than the other – Patients with cancer may recall greater exposure to factors that could cause cancer. • Use data recorded before outcome occurred – Dependent on availability of recorded information – Reliability of data • Use blinding – Both study subjects and investigators: difficult to do – Specific risk factor: easier- can include “dummy” questions – Observer: easy, should always be done

Choosing a design: Depends on the question

Exercise • Be preapred to discuss your project’s study designed and its strengths and weaknesses