32. INTERNAL VALIDITY Flashcards
- According to the World Health Organisation (WHO), how would we define Interpersonal Education (IPE)?
THIS KIND OF EDUCATION OCCURD WHEN:
- two or more professionals learn about, from and with
each other
- this enables effective collaboration
- it improves the health outcomes
- What does Clinical and Public Health work involve?
- it involves a Multi-disciplinary team
THIS IS COMPOSED OF THE LIKES OF:
- doctors
- nurses
- psychologists
- environmental specialists
- legal representatives
- counsellors
- etc.
- Look at this study example.
How do you know if this is a good study and if it applies to your Patient Population?
- it adheres to the necessary ethical research guidelines
- it has a long and constant follow-up period
- hospital records and national cancer registries were
used in this study - the sample used was made up of Health Professionals
- this means that they have some knowledge on how to
fill out a Questionnaire correctly and in an unbiased
manner
- What is Internal Validity?
THIS RELATES TO:
- how well a study is conducted
- it looks at how the study is designed
- it looks at how the data is collected
IT LOOKS AT WHETHER THE STUDY RESULTS ARE VALID:
- in terms of whether the study has found the truth
about the specific source population
IT LOOKS AT IF:
- there is really an exposure-outcome association
- or if the results are explained by something else
- What is External Validity?
THIS RELATES TO:
- how applicable the findings are in the real world
- how applicable it is to your population or the
subgroup being analysed
IT LOOKS AT IF:
- the results can be generalised to the general
population
- or if there are other similar populations
IT LOOKS AT IF THE RESULTS CAN BE USED:
- for a broader group of patients
- How do we relate variables to Internal Validity?
INTERNAL VALIDITY IS THE MEASURED BY:
- the extent to which you are able to say that only the
independent variable has caused a change in the
dependent variable
THIS MEANS THAT:
- no other variable is responsible for this change
- In order to judge Validity, which three factors do we look at when it comes to their effect on the results?
- Chance
- Bias
- Confounding
- What is Chance?
- this is a type of Random (Sampling) Error
- it is due to the random nature of the sampling
THE ROLE THAT CHANCE PLAYS IN OUR RESULTS:
-is determined by the p-value
- and by the 95% Confidence Interval
- How do we determine if the Validity of our results has been affected by Chance?
- CHECK THE FINDINGS FOR STATISTICAL SIGNIFICANCE
- this is based on the p-value result
- it is also based on the Confidence Interval values
- IF THE EXPOSURE-OUTCOME ASSOCIATION IS
STATISTICALLY SIGNIFICANT- we can conclude that our finding is not by CHANCE
- there is in fact an exposure-outcome association in
the population
- What is Bias?
- this is a type of Systematic Error
- How do we minimise the influence of Chance in our findings?
- we take a large Sample
- this sample has to be representative of the Population
as a whole
A LARGE SAMPLE:
- will reduce the Standard Error
- it increases the study power
- it gives more precise estimates
- In this study, what can be said about the presence of Chance (Random Error)?
THE 95% CONFIDENCE INTERVAL:
- does not include the value 1
- this is in the case of All-cause Mortality
THE RESULTS IN THIS STUDY:
- are statistically significant
WE CAN CONCLUDE WITH HIGH CERTAINTY:
- that the observed association between the
Mediterranean- Diet and Prostate Cancer is NOT due
to Chance
THE VALIDITY OF THIS STUDY:
- is not compromised by Random Error
- What are the two types of Bias?
- Selection Bias
- Information Bias
- Define Selection Bias.
- these are errors
- they happen in the process of sampling
THEY ARE A RESULT OF:
- selecting a non-representative sample
- this leads to any derived estimate having a high
likelihood of being biased
- Define Information Bias.
- these are errors
- they happen in the process of data collection
THIS IS A RESULT OF:
- inaccurately assessing the exposure or outcome
variables
EXAMPLE:
- taking specific measurements in a study that are
inaccurate
- they are non-representative of the sample
- this can be done without the researcher realising
- What are some examples of Information Bias?
- RECALL BIAS
- this is when people struggle to remember information
- this can alter the results of the study - INTERVIEWER BIAS
- the researcher unintentionally brings in Bias by asking
questions in a specific way
- How is Bias determined in a study?
- this is not a straightforward process
THE RESEARCHER:
- needs to to be reflective and self-critical
- this is with regards to the appropriateness of the
sample they are choosing
- as well as with the measurements that they are taking
- What kind of samples are always expected to have Bias in them?
SAMPLES COLLECTED USING:
- Convenience Sampling
MEASUREMENTS RESULTING FROM:
- participant self-reports
- How do we minimise the influence of Bias in our findings?
- CHOOSING A SAMPLE:
- that is representative of the Source Population that
we want to investigate
- this will minimise Selection Bias - CHOOSING ASSESSMENT TOOLS
- that have high accuracy
- these tools need to be valid and reliable
- this minimised Information Bias - MAKING A THOROUGH INVESTIGATION
- of the accuracy of the data collected
- this should be done during the data collection process
- it should also be done during the Data Analysis - DATA CLEANING
- this is an analysis of the data during Data Analysis
- it removes any biased data
- it does this by looking at abnormalities and outliers
- makes the results more reliable
- In this study, what can be said about the presence of Bias?
- THE STUDY SAMPLE IS LARGE
- it is a Random Study of 47 867 individuals
- it was collected randomly
- this means that there is no Selection Bias - THE STUDY INVOLVES SELF-REPORTING
- this is done by the participants with regard to their
Mediterranean diet food intake
- Measurement Error (Bias) in the exposure assessment
is very likely - THE INFORMATION ABOUT THE PROSTATE CANCER
- was taken through National Health Registries
- and through Hospital Records
- Define Confounding.
- this is when there is a factor or combination of factors
- these explain all (or part) of the association between
an Exposure and Outcome
- How does Confounding effect the study?
- it can greatly compromise the validity of the study
- it can distort the estimate of the potential association
- it can mask a true association
- it can make a false association appear
- even if an association does not exist
- How do we determine if our results are affected by Confounding?
- UNDER NON-EXPERIMENTAL CONDITIONS:
- it should be expected that Confounding is present
- it will affect the results to some degree - IF THERE IS AN UNEXPECTED RESULT OBTAINED
- based on previous evidence
- we can assume that there is Confounding present
- IF THERE IS SOMETHING THAT DOES NOT MAKE
SENSE BIOLOGICALLY
- we can assume that there is confounding present
- What is an example of a Non-Experimental Condition?
- Observational studies
- How do we minimise the influence of Confounders in our studies?
- MAKE A LIST OF POTENTIAL CONFOUNDERS
- for a given exposure-outcome association - ADJUST THE POTENTIAL CONFOUNDERS
- this is done during Data Analysis
- it will give Confounder-adjusted estimates
- these are free from the confounding effects - ALWAYS BE CAUTIOUS OF RESIDUAL CONFOUNDING
- even after adjusting the data
- In this study, what potential confounders did the Researchers adjust their results for?
- age
- BMI
- physical activity
- smoking status
- ethnicity
- height
- diabetes
- family history of PCa
- vitamin supplement used
- In this study, what can be said about the presence of the Confounders??
- THE MOST IMPORTANT POTENTIAL CONFOUNDERS
- were identified and adjusted for - RESIDUAL CONFOUNDING
- could still be present
- this is due to the unknown confounders
- as well as any imprecise assessments that could have
happened with regards to the existing confounders - THE INTERNAL VALIDITY OF THE STUDY
- could be compromised
- by the possible presence of Residual Confounding
- it is not compromised to a large extent
