Module 1 Flashcards
Don't fail Midsems
Define occurrence
Transition from a non-dis-eased state to a dis-eased state.
Define population
A group of individuals sharing a common factor.
Ways to display numerical data in the GATE frame
Categorisation: Usually done to allocate people into EG or CG-eg: Blood pressure can be classified as high or low based on if it is higher or lower than a particular value.
Average: Used for EGO and CGO. All numerical data in a group summarised as a mean value.
Incidence
Used when occurrences can be easily observed. Counts the number of onsets over a specific period of time.
Point prevalence
Used when occurrences are not easily observable. Measures the number of people with dis-ease at
the specific point in time when the study is taken.
Prevalence pool
Number of people with the dis-ease at any point. Does not account for people who had the dis-ease, but doesn’t anymore due to death/cure.
Period prevalence.
Used to measure dis-ease onsets that can be easily observed but does not remain at regular intervals (ie: comes and goes).
Counts how many study subjects HAD the dis-ease (instead of the number of onsets).
Prevents misleading data caused a few subjects contributing to many occurrences of dis-ease (eg: asthma attacks-does not occur with any pattern).
Recruitment Errors
External validity error- Recruited study population does not accurately represent the eligible population. Results obtained not applicable outside of study population. Can be caused if not enough people respond to the study-<70% leads to significant recruitment error.
Selection bias: EG and CG taken from the same eligible population, but people in the groups have inherent differences besides exposure status. Occurs if exposure status is dependent on other factors that can contribute to outcome. CONFOUNDING ERROR.
Allocation Errors
Only occurs when participants are ALLOCATED BY MEASUREMENT. (RCTs are a good way to get around this kind of error.)
CONFOUNDING ERROR- Caused by differences beyond exposure status-cannot tell if these differences caused differences between EGO and CGO.
Allocation measurement errors- Inaccurate measurement of exposure leading to mis-allocation of participants. Can be due to subjectivity in measurement, or participants falsely reporting their exposure status.
Unconcealed allocation: Occurs in RCTs. When investigators interfere with the random allocation based on clinical biases-eg: certain patients will benefit more from new intervention.
Selection bias: The two groups are recruited from very different populations (office worker and construction worker example).
Maintenance Errors
Caused by movement of participants between EG and CG as exposure status is not maintained. Can be prevented in clinical trials by blinding participants so they continue to be motivated to remain on the intervention, and blinding investigators to prevent unfair additional interventions.
Unblinded participants are aware of their exposure status (especially relevant in clinical trials).
-Co-intervention: Participants with intervention more likely to make other changes to improve health. Participants on placebo more likely to request other forms of aid from third party or investigators.
- Compliance: Participants with the placebo less willing to comply. In cohort studies this also requires smokers to NOT quit (and nonsmokers to not start), which is hard to control. (see also contamination)
- Lost to followup: People can leave the study and reduce the sample size of group. Usually occurs more significantly in one group than another.
Blind and Objective Measurement
Measurement errors can occur when bias is introduced when measuring.
Objective means of measurements used as bias cannot affect the values. They are usually done by machines which cannot be influenced. Only source of error here is crap equipment.
Subjective measurement must be done with both participants and investigators blinded. Knowledge of exposure status can lead to patients interpreting their outcomes differently. If a patient knows that they were on the actual intervention, they would report greater benefit than there actually was.
Similarly investigators will measure through a lens that would better fit their hypotheses.
Regression to the Mean
If a test initially gave extreme results, more repetitions would make the results less and less extreme, since extreme values tend to be rare.
Random Sampling Error
It’s impossible for a sample to be perfectly representative, both by chance and the fact that the perfect representation would be the actual population. Can lead to variations between identical studies just because chance dictates that every sample will be slightly different.
Can be reduced by taking larger samples which are better representations, or more samples to account for as many aspects of the population as possible (aggregates to a ‘complete’ representation).
Types of Random Errors
Random sampling error
Random measurement error
Inherent randomness of biological phenomena
Random allocation error
Random Measurement Error
Environment can affect the ability for investigators or even measuring equipment to accurately measure biological data- eg: hand shaking, background noise preventing accurate detection of heartbeats etc.
Can be reduced by taking more measurements and taking a mean value.
