Module 1 - Measuring Distribution & Determinnants Of Pop Health Flashcards
What is the main goal of epidemiology?
Main goal is to study the frequency of dis-ease in populations
Epidemiologists ALWAYS start with the population and then count the number of cases
Why measure the disease frequency in diff pops?
If dis-ease frequency (distribution) is diff in two pops, it can help us determine the causes (determinants)
How to calculate frequency ?
E = Numerator (no. Cases)/ Denominator (pop no.)
Epidemiologists definition of pop?
A group of people who share one or more common features
Definition of dis-ease?
Narrow - absense of death, dis-ease, disability
Broad - The ability to do what matters most to you
Numerator? Denominator?
Numerator - Cases of dis-ease
Denominator - Population
E = N/D
Why do epidemiologists often age standardise measures of dis-ease frequency before making comparisons between populations?
It’s critical epidemiologists age adjust/standardise populations dis-ease frequency before comparison - this allows for meaningful comparison - like for like.
E.g Sweden older pop than Panama = more deaths
Pakeha older pop than Maori = more pakeha die, however, more Maori die in each age range
Frequency at one point in time or over a period of time?
E = (N/D) / T
E = (cases of dis-ease / pop ) / Time
Define GATE?
G - Graphic
A - Approach
T - To
E - Epidemiology
What is PECOT?
It is an acronym for the components of the GATE framework P - Participants (Pop) E - Exposure Group C - comparison Group O - Outcome T - Time
EG & CG
EG - Exposure Group
CG - Comparison Group
Which exposure group can be chosen EG or CG?
Any, as long as you state it clearly
PECOT Outcomes?
A&B or (C & D) are numerators
Mostly use those with dis-ease as numerators A & B
Goal of epidemiology?
Calculate dis-ease frequency
But more specifically calculate:
Exposure group Occurrence (EGO)
Comparison group occurrence (CGO)
Comparison Groups tend to be ?
Your typical ‘normal’ group
If outcome is health use ?
C&D
Time arrows
Down - Over a period of time
Across - At one point in time
Cohort vs cross-sectional studies
Cohort - follow over time; can also measure prevalence - either beginning or at any point
Cross-sectional - relevant dis-ease events counted at the same time - can only measure prevalence
Death rate?
100% or 1per person per lifetime
What is Incidence?
Dis-ease occurrence over a period of time
E= N/D
What is prevalence ?
Where the number of people with dis-ease are counted at one point in time
Less perfect measurement - people can leak out of the prevalence pool - death or cured
When to use incidence or prevalence ?
Incidence - if easy to observe disease (death, road crash)
Prevalence - if hard to observe when disease occurs we measure IF It has occurred (psych disorder, obesity, diabetes)
Prevalence pool?
How much drizzle in the prevalence pool after the drizzle has fallen
Incidence and prevalence : categorical or numerical measurement ?
Incidence - Always involves counting (yes/no) disease EVENTS
Prevalence - involves counting categorical disease EVENTS or measuring numerical disease STATES
Prevalence - time measured backwards?
Measure a period of time E.g if someone has had a symptom in the last 3 months
Is time included in prevalence calculation?
No
Is time involved in incidence calculations ?
Yes
What determines incidence?
Numerator = Depends on number of events that happen during a specified time
What determines prevalence ?
Depends on how much drizzle (events - incidence) get into the pool AND how much leaks out (deaths or cured) before study
If you want to know how fast a car can go? You measure?
Incidence
If you want to know how worn out a car is ?
You measure prevalence
Can you measure incidence and prevalence in a cross-sectional study?
You can only measure prevalence
Can you measure prevalence and incidence in a cohort study?
Yes
2 numerators of prevalence?
- One at a specific point in time
2. One looking back in time (IF) it has occurred
How could a A HIGH incidence have a low PREVALENCE
If people in the prevalence pool died rapidly or cured quickly (common cold)
Populations have different age structures?
What do you do?
You need to adjust (or standardise) the age structures of each population so they can be meaningfully compared
How could A LOW incidence have a HIGH prevalence ?
If most people in the prevalence pool remained in the prevalence pool (obesity)
Disease frequency, risk, occurrence, distribution
Used interchangeably
Benefits of RCTs?
Participants have equal chance of being allocated to EG or CG, so any differences between the groups are likely to be due to effect of the drug they are given
How do you calculate risk difference?
EGO - CGO
If EGO = CGO the risk difference = 0
Risk ratio?
EGO/ CGO
IF RR = 0.67 then the risk has been reduced by 33%
Always out of 1
Strengths of Risk difference?
Gives more information than RR, it’s all about Risk difference
Strengths of Relative risk?
Easy to understand but deceptive ( what is the relative risk of what?)
Error in epidemiological studies?
Errors occur when wrong people are recruited into a study or right people are in the wrong GATE frame category
Random error?
If the error occurs by chance
Non- random error?
Errors occur due to poor study design, poor study process or poor study measurement
Study validity?
A study with only a small amount of random and non/random error
Where does non- random error occur?
R - recruitment A - allocation M - maintenance B - blind O - objective M - measurement
Recruitment?
Who was recruited
Are they a representative sample of a known pop
Allocation?
Two main ways
- By measurement or observation (cohort)
- By random Allocation (RCT)
- How well were the participants allocated
- how were the participants allocated
Confounding?
When the exposure is mixed with another factor (study has a bias)
How do you deal with confounding?
Divide (stratify) into sub studies so participants with the confounded are all in one sub study
What is RRR?
Relative risk ratio - when the relative risk for EGO is less than 1 compared to CGO, this is said to have been reduced. Therefore it can be claimed that the treatment reduces the risk.
What is RAMBOMAN?
Assesses non-Random error in cohort, cross sectional and RCT’S studies
Longitudinal study is also known as a ?
Cohort study
Where does allocation occur?
Between triangle and square
Do vitamins reduce risk of death?
- Evidence from cohort studies show that people who take vitamins have lower death rates than people who don’t take vitamins
- however evidence from RCT showed that people who take vitamins have higher death rates
This is due to confounding - people who take vitamins do other things to improve there health.
Maintenance?
- All about participants staying in the groups they were allocated to for the duration of the study
- All about participants staying in the study
- will the validity of the study be effected by how well they were MAINTAINED in EG & CG
B.O.M?
- If you measure by measurement BOM relates to CH & outcome
- Validity of study results in how well exposure & outcomes were measured
Objective - not influenced by personal interpretation
Subjective - how much do you drink? Pain? Movement? (People may not tell the truth)
Subjective?
People may not tell the difference not really reliable
Objective?
Not influenced by personal interpretation
Confounding is very common in what studies?
Ecological studies
Very difficult to do RCT on countries
Confidence interval?
There is about 95% chance that the true value in a population lies within the 95% confidence interval. (* assumes no non-random error or random error in the study)
4 main types of random error ?
Random measurement error
Random sampling error
Random allocation error
Random inherent
Random measurement error?
- Effects measurement of both exposure and outcomes
- Factor effecting the operator can change the measurements, the operators ability to measure accurately
- THE BEST way to reduce random measurement error is to do more measurements
Random sampling error?
- when recruiting a sample of the pop (children) despite the recruitment being done very well - the reps will never be a perfect rep of the whole pop.
- thus every study sample of the pop will have diff reps and therefore produce slightly diff results.
Random allocation error ?
- The exposure & comparison groups in a RCT may differ by chance alone, particularly if the trial is small - this type of random allocation error can also be reduced by undertaking a larger study so larger numbers randomised
The 95% Confidence interval?
- Is the measure of the amount of random error in our estimates of EGO, CGO, RR & RD in the whole pop when you have only done 1 study.
- CI describes the range of values (of EGO, CGO, RR, RD) that is likely to include the true value in the underlying population.
How Ca we reduce random sampling error?
We do a bigger sample
The bigger the sample, the less random error
How do we reduce random measurement error?
- We do more measurements
- Repeating measurements or studies with extreme results many times usually gives less extreme results “ regression to the mean”
If EGO & CGO CIs do not overlap what does that mean?
- The it is reasonable to assume that EGO & CGO are truly different from each other.
- When the RD = 0 (doesn’t overlap 0) we call that statistically significant
Line down the middle ?
No effect line
If EGO & CGO CIs do overlap what does this mean?
- There is probably no statistical significant difference between EGO & CGO.
- EGO = CGO: RR = 1
The wider the CIs?
The less confident I am about true value
The narrower the CIs?
The more confident I am in the true value
Can you sometimes have PEOT?
Yes - you do not always need to measure EG and CG
Relative risk ?
Allows us to compare the risks of an event between two groups
Risk difference?
Allows us to see the difference between the two groups - gives us more information