Observational Studies 2 (Sept 18)

Question 1

How should a case be selected?

Answer 1

• shoud select average cases that are representative of the average patient (not the sickest or healthiest)
• specific criteria
• prefer incident over prevalent cases (tend not to be long survivors with a disese- can recall exposure status better)
• healthy survivor bias; people who can live longer with disease than typical people can

Question 2

How should controls be selected?

Answer 2

• must come from same source population as cases (if cases come from hospital, select controls from hospital)
• do not have disease under investigation

Question 3

Matching (in case control study)

Answer 3

• want to make sure controls and cases are similar with respect to certain characteristics
• individual matching: get a case then find a control the same day that shares characteristics
• frequency (group) matching: group of controls is matched to a group of cases with respect to a particular characteristic

Question 4

Stratified analyses

Answer 4

• break up analysis into two groups and see if they are roughly the same
• since we’ve selected certain number of men and women, we can compare odds ratios between men and women (example)
• we make sure we have enough men in group and women in group to make sure this is possible

Question 5

Potential problems if we don’t match in case control study

Answer 5

• possible that we could end up with only cases of men and no male controls
• can’t compare odds ratios between men and women

Question 6

Disadvantages of matching in case control study

Answer 6

• inefficient (discard potential controls if they do not meet the criteria to match)
• hard to find controls if you match on many variables
• over matching: match on something we think is important to control for but ends up being an important determinant of disease (eg. SES- if you choose people in the same neighbourhood and match on SES, you can no longer look at association between SES and disease)
• don’t want to match on something you think in the beginning could influence the exposure and outcome relationship

Question 7

Advantages of case control studies

Answer 7

• useful for studying many risk factors (exposures)
• good for studying rare diseases (diseases that we don’t see often or take a long time to develop- we have a concentrated population in one hospital so we can study them there)
• pretty inexpensive (surveys, figure out exposure status, etc. and don’t have to follow through for many years)

Question 8

Disadvantages of case control studies

Answer 8

• can’t figure out incidence rates because if you match individually, by definition you have a set incidence rate (eg. in a population of 100 in the study, 50 have to be cases)
• no prospective follow up (can’t estimate if people are likely to develop disease because you start out with them already having developed it)
• can’t estimate exposure rates are in that population (defined sample that we’ve matched cases to controls) so we can’t figure out in general population what the frequency of exposure is
• temporality: ask them to recall exposure but it is possible for them to forget or things change

Question 9

Concordant pair

Answer 9

• both the case and control are exposed or unexposed
• doesn’t tell us about association between exposure and outcome
• non informative pair

Question 10

Discordant paid

Answer 10

• case is exposed, control is unexposed
• case is unexposed, control is exposed
• can learn something about exposure and disease

Question 11

Matched Pairs Odds Ratio

Answer 11

• include only discordant pairs in calculation
• set up 2x2 table with discordant and concordant pairs (cases exposed/unexposed as columns and controls exposed/not exposed as rows)
• odds ratio= (# of pairs with case exposed/control not exposed)/(# of pairs with case not exposed/control exposed)