Observational Studies (Sept 13) Flashcards
Observational study design
- natural experiments
- investigator does not manipulate treatment or exposure to disease, but rather observes natural variation
- observe what exposures they choose and whether these exposures lead to disease
Why use observational study design?
- some things are not ethical to tell people to do (can’t randomize someone to an experimental drug that is known to be harmful)
- can observe people who choose to smoke, etc. and study their disease outcomes
- usually applicable to the world
Problems with observational studies
- prone to bias
- something about study design or population makes it difficult to be certain that what we saw was the truth
- eg: if you are doing a study on drinkers and lung cancer, you must know whether or not the participants are smokers (confounding)
Case Report
- 1 patient
- unusual symptoms/syndrome
- don’t know cause
- first, describe and speculate about what might be causing the disease to occur
- can do lab tests, imaging, histology and describe what is happening
- describing how a person presents, what you did to treat, and what the outcomes are
Why are case reports important?
- highlight unusual/novel findings
- bring new diseases to the attention of the medical community
- new side effects of drugs can be discovered
- identifying new genetic diseases that are very rare
- birth defects
- treatment failures
Limitations of case reports
- case may not be generalizable (if it’s something that only strikes health sci students but then it strikes an eng student, we don’t necessarily know if this will apply)
- not systematic (if you have a unique case others might present differently and require different treatment)
- causes or associations may have different explanations (eg. someone may be stressed from an underlying heart condition and not because they don’t cry and process their feelings)
Case series
- group or series of case reports
- involves patients with a specific presentation who were given similar treatment
What is contained in a case series?
- detailed information about the patients (exposure pattern)
- demographic information
- diagnosis, treatment, response to treatment, and follow up after treatment
- defined protocol and inclusion/exclusion criteria (what type of people will be included in the study)
- try to pick up on typical cases (set definition and choose cases that fit that criteria)
SARS case series example
- people who presented with SARS and another comorbid disease did worse than those without a comorbid disease
- this case study was important as when people first started coming in, it was unknown who would have a better outcome
Why are case series important?
- same reasons as case reports but with more people
- can help identify or characterize rare conditions or treatment courses
Why are case series limited?
- if the series is retrospective it will depend on the availability and accuracy of the data records (no control over how the data was collected)
- subject to selection bias because the clinician or researcher self selects the cases
- findings reported may not be generalizable
- it is often impossible to know what would have happened to the cases if they had not been treated
Counterfactual
- if we could go back in time with people that we treated and not treat them, what do we think would have happened to them?
- take a group of people similar to them who got a disease but we didn’t treat them (control group)
Case control studies
- control group gives context
- two groups of people and we compare what happens to them
- control group allows you to estimate what would have happened to the treated group had they not been treated
Method for case control study
- define cases and recruit them into study
- every time a case joins the study, find someone who does not have the disease and recruit them into the study
- ask both participants the same questions about their exposure history
- compare exposure histories
- classify people as cases or controls and as exposed or not exposed
How to select a case for a case control study
- if the interest is looking at the average patient, make sure you are not selecting the sickest patient so they are representative
- specify clear inclusion criteria of cases (DSM criteria, test score, etc.)
- incident cases preferable to prevalent (prevalent have often had disease for a long period of time and have survived- survivorship bias- tend to be not typical cases of that disease. Also can be hard to separate exposure from disease as if they’ve lived with it for a long time, they mgiht not remember if the disease came first or other unhealthy exposures were first)
Measures of Association
-want to know if our exposure is associated with the disease
Concept of Odds
-probability that an event will occur/probability that an event will not occur
-eg: attending counselling regularly leads to 25% probability of being stressed
odds of being stressed: 0.25/(1-0.25)= 0.33
Odds Ratio
OR= odds of a case being exposed/odds of a control being exposed (can do this calculation by cross multiplying the 2x2 table)
OR=1 : no association between exposure and outcome
OR>1 : exposure is more “harmful”
OR<1 : exposure is “protective”
Selecting controls
- controls must come from same source population as cases
- do not have disease under investigation but would be included in the study if they got the disease
Individual matching
- for every male case, select a male control
- for every one male case, select 4 female controls
Group matching
-select based on proportion of a characteristic in the cases
What are advantages of matching?
-remove known risk factors that may not be distributed evenly across groups (counfounding things)
What are disadvantages of matching?
- can be inefficient if you have a man but can’t find a man to be a control (increased time and cost)
- hard to find controls if you match on many variables
- can overmatch; inadvertently matching on variables that are not of interest
- can not study the effect of matched variables
