CAT Cohort & CC Flashcards
DId the study address a clearly focussed issure? (1)
DId the authors use an apt method to answer the question?
- retrospective cohort:
-
- PICO
- Retrospective cohort = uses pre-existing secondary research to examine relationship between exposure and outcome
- Prospective cohort = starts with exposure and measures outcomes, follows individuals over time
Good match between study design and clinical question?
Gold standard is RCT for assessing cause & effect.
Not practical/ethical.
Cohorts allow analysis of outcomes by starting with exposures.
Good for rare exposures.
Cannot prevent changing behaviour of patient as they are studies, and have to wait a long time for results (lag between exposure and outcomes).
What is clinical equipose?
Genuine uncertainty about the value of each approach/intervention in a study.
Would not be ethical to allocate a participant into a study arm that is willingly more harmful.
Was cohort recruited in an acceptable way?
- Use of relevant databases - e.g. national health
- cohort should be representative of the defined population
- to remove selecton bias
Was the exposure accurately measured to minimise bias?
- exposure measured directly from the database - avoids recall bias
- subjects classified into exposure groups using the same procedure, avoiding procedural bias
Risk of Measurement/observational errors: - avoiding non-differential error: information collected about both groups is inaccurate (bias towards null –> type II error)
- and differential error: info collected about one group is inaccurate –> bias results in one direction (spurious or obscuring true associations, type I or II error)
Was the outcome accurately measured to avoid bias?
- Subjective or objective measurements
- Are the measurements validated?
- Does the outcome reliably detect all the cases?
- Were measurement methods similar across the groups?
- Blinding? - may not be possible; does this matter?
Confounding factors
Have all been identified?
Have they been taken account of in design/analysis?
Confounding - when the association between two factors is linked by a third
Limiting confounding in a cohort study
1. Restriction - limiting participants if they have possible confounders
2. Matching & Stratification - make comparison groups with and without the confounder (e.g. age/sex - subgroup analyses ) and then adjust for the confounder –>
3. Multiple variable regression - a type of sensitivity analysis to establish coefficients for the confounder groups
E.g.
The study may identify confounders but not take them into account in analysis.
Was the follow-up of subjects complete enough?
- Latency periods - was the time period long enough for the disease to present itself
- was there a trend in the patients that were lost to follow-up? - e.g. was there anything special about the people who left?
CC
Key points for:
* Case definition
* Control definition
Defining cases
* Appropriate to the clinical question
* Quantity - power calculation?
* Reliable system
* Incident > prevalent cases
* Affects selection bias - validity
Defining controls
* representative of both the general population; and of the population likely to be exposed?
* Aka matched, population based or random?
* sufficient number - power calcuation?
CC
What are the challenges to measuring exposure in case-controls?
Methods of measuring exposure can be through recall (e.g quesionnaire, interview) or through biomarkers (may be less reliable).
- - Recall bias: over/underexaggerating of the exposure in one group over the other
- Measurement bias - subjective/objective measurements of exposure?
- Variable exposure: patient’s envrionment may ave changed
- Unavailable data: due to inabilty to remember; medical records unavailbale/incomplete/inaccurate
- Blinding: was this incorporated where feasible?
Phrases for external validity
- Reasons to advocate change
- Uncertainty warrants further research
- Use of a different study type
- Ethical implications for a different study type
*
