Introduction to epidemiology by https://padlet.com/helen_stagg/tneclobqxazv Course review en-us 2019-10-14 13:19:17 UTC 2023-05-29 22:25:43 UTC hello@padlet.com Purpose of epidemiology studies to accurately measure: helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397404746 2019-10-14 13:21:31 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397404746 STUDY DESIGNS helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397407601 2019-10-14 13:27:06 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397407601 INTERPRETING DATA helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397407681 2019-10-14 13:27:13 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397407681 1) Case-control helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397413155  Observational Study. Used to determine if exposure is associated to outcome. Retrospective.

Advantages are- Cheap to do, Not time consuming, Useful for rare diseases or diseases with long latency periods, Can look at multiple exposures.

Disadvantages- Cannot establish causality, Cannot look at multiple outcomes,  Cannot infer incidence 
Can be subject to selection bias.

Odds ratio used to interpret data.

]]> 2019-10-14 13:37:13 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397413155 2) Cohort helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397413256 what is it: longitudinal study looking at sample over time. Can be retrospective or prospective
 
what is it used for: Determine cause, rare exposures.
advantages: multiple outcomes can be looked at, estimate incidence, temporal relationship can be determined
disadvantages: Bad for rare outcomes,
key sources of error:  information bias- medical records as a source of information (more reliable than recall)
other key data:
Retrospective cohort and case control are differentiated by recall bias.]]> 2019-10-14 13:37:24 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397413256 4) Cross-sectional helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397413363 1.what is it
  •  survey of exposure and outcome status at a single point in time 
  • study of individuals instead of populations

2.what is it used for?
  • Observing the prevalence of population exposures and outcomes

3.Advantages
  • Easy & cheap
  • Routine data often used
  • Good to generate  hypothesis
  • can have a large population

4. disadvantages
  • Prone to bias- we can't always tell if the exposure is tied to the outcome
  • No test of causal relationships 
  • Prone to selection bias 
  • No temporal information

5.key sources of error
  • Selection bias 
  • Recall bias (about exposures)
  • Response bias 
  • Confounding factors that could contribute to the outcome
]]> 2019-10-14 13:37:37 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397413363 3) Ecological helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397413525 1. what is it
it is a population based, environmental, and observational study.
2. what is it used for?
To show incidence, prevalence
3. Advantages: rare diseases, cheap, fast, compare population in the places at the same time, values the different characters of population; Disadvantages: ignores confounding, no causality
4. key sources of errors
individuality ]]> 2019-10-14 13:37:54 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397413525 5) Intervention (RCT) helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397413586 What is it?
Experimental study design, EX: Randomized Control Trial
What is it used for?
to study the effects of a particular intervention
Advantages
reduced bias, can be double blinded, not as long as a cohort, stonger methodology, strict study design (very applicable to tested group)
Disadvantages
time, money, ethically difficult, limited generalizability, strict study design (can under rep minorities)
Sources of error
limited bias, but hard to generalize
Key info
gold standard]]> 2019-10-14 13:38:00 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397413586 6) Information bias helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397413970 1. Distortions of key variables. Bias in measurements or interpretation of data. 

2. Important because lack of awareness can lead to interpretations results that are not actually reflective of the underlying 
3. Occurs for a range of reasons
Incorrect/inappropriate usage of measurement tools 
Recall bias - Information remembered incorrectly
Observer bias - Misinterpreting due to expectations 

Avoiding information bias - measures such as double blinding, ensuring consistent use of analysis tools are important ways of avoiding information bias. Recall bias cannot be entirely avoided.

]]> 2019-10-14 13:38:42 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397413970 9) Selection bias helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397414451 Selection bias can occur during study design and follow-up.

Selection bias is where the study sample population is not representative of the population of interest.

Good selection reduces chance of error in findings and increases the internal validity of the study.

It can be avoided by using good sampling methods and having sufficient understanding of the population of interest. A pre-study can also be performed.]]> 2019-10-14 13:39:36 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397414451 Misclassification helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397414545 2019-10-14 13:39:46 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397414545 7) Differential helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397414774 What is it?
A form of information bias which occurs due to observant bias, respondent bias, recall bias, uncompleted medical records, recording errors, misinterpretation of records 

Why is it important? 
This causes misclassification of individuals into the wrong category and is unevenly distributed between categories which can skew data, effect the odds ratio and relative risks.  

How to avoid it: developing standardised protocol, blinding & double blinding to avoid interview bias, training of interviewers/researchers

]]> 2019-10-14 13:40:16 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397414774 8) Non-differential helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397414864 Information bias affecting all our groups- cases and controls equally. Usually affects OR or RR towards null.
Example:
In a type 2 diabetes study in obese patients with two groups- cases (people with diabetes) and controls(people without diabetes), there can be a non-differential information bias if the glucometer used to record blood glucose levels does not work properly.]]> 2019-10-14 13:40:25 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397414864 10) Chance helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397417277 Chance is a random error that may give misleading results, it can not be fully eliminated.

do not indicate the true population result.
can have an impact on internal validity

-need the large sample to represent the whole population, to do that  you need the large sample with similar characteristic 
-need adhere to a strictly analysis procedure 
-assess the chance by using p value and CI
]]> 2019-10-14 13:44:07 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397417277 11) Confounding helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397417518 Confounding is the distortion of the effect of exposure on outcome, due to a third party factor/factors. These factors have to meet 3 criteria: 
1. Associated with exposure
2. Risk or protective factor for outcome
3. Not on causal pathway between exposure and outcome 

Why is it important?
Can make results look like there is an association that is actually not there (between exposure and outcome) or can make it look like there is none, even though there is an association. 

How does it occur? How can we avoid it? 
1. Associated with exposure
2. Risk or protective factor for outcome
3. Not on causal pathway between exposure and outcome 
Avoided: You cannot avoid but adjust for it through restricting sample, randomisation and matching. During the analysis stage it can be controlled through stratification, multiple variable analysis and logistic regression. 

Other key information: 
Background knowledge of topic may be beneficial to identify potential confounders and associations. Often you are unable to identify all possible confounders, but you need to acknowledge these limitations in your study. 
]]> 2019-10-14 13:44:30 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397417518 12) External validity helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397418454 aka Generalisability

-Can you apply the results obtained from a small sample group to a larger, external population?
-Importance of CONTEXT: How might the study population differ (especially in the outcomes' determinants) from other populations to whom we might want to apply results?
-Can you apply data to policy-making?

]]> 2019-10-14 13:45:49 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397418454 13) Causation helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/397419705
  • Is an association actually causal?
  • Association and causation not the same thing- pirates do not cause global warming
  • Bradford Hill criteria
  • Koch's postulates
  • Chance, bias, confounding
    • ]]>     2019-10-14 13:47:46 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/397419705     14) EPIDEMIOLOGICAL MEASURES helen_stagg https://padlet.com/helen_stagg/tneclobqxazv/wish/398897924 DESCRIPTIVE
    Frequency
    Proportion/percentage
    Prevalence, incidence

    ASSOCIATION
    Odds ratio
    Risk ratio
    Rate ratio
    (Relative)
    Absolute
    (Population) attributable risks]]>     2019-10-17 08:53:00 UTC https://padlet.com/helen_stagg/tneclobqxazv/wish/398897924