Module 1 (Lectures 1-11) Flashcards
Clinical medicine: focus, education, rights
Individual, biomedicine model, cure>prevention, individual rights of patient
Popl hlth: focus, education, rights
Popl (max benefit for max no. of people) epidemiology, human rights
Epidemiology starts with _____
Describing a popl
Frequency of a disease =
No. of cases of disease ÷ no. of people in popl
Why measure freq. of disease in diff. popl?
To help identify causes / determinants
Population
A group of people who share one or more common features
Why do we need age standardisation?
For meaningful comparison
Epidemiology (definition)
The study of the FREQUENCY (AND CAUSES) of disease in a POPULATION(s) at ONE POINT or over a PERIOD OF TIME
Epidemiology = (formula)
Numerator (disease) ÷ denominator (population) ÷ time
Describe the gate frame
Triangle - no of participants/populations. Circle - divided into exposure group and comparison group. Square - outcome. Arrows - time.
Participants
Triangle starts broad, gets narrower eg study setting -> eligible popl-> participants
EG and CG are _____ (what part of formula?)
Denominators
Disease outcomes are _____ (what part of formula?)
Numerators
Exposure Group Occurence (EGO) =
a ÷ EG
Comparison Group Occurence (CGO) =
b ÷ CG
Ecological cohort study
‘Popl’ of countries
Time is measured ____ or ____
Over a period of time or at one point in time
Incidence (definition and EGO / CGO formula)
When no. of disease events that occur are counted OVER A PERIOD IF TIME. E/CGO = a/b ÷ E/CG ÷ T
Prevalence (definition and EGO / CGO formula)
When no. of people with disease are counted AT ONE POINT IN TIME. E/CGO = a/b ÷ E/CG
Occurence / Event
When move from state of no disease to state of disease
If it is easy to measure when disease occurs, usually measure ____
Incidence
If it is hard to observe when disease occurs, then we measure if it has occurred. We measure ____
Prevalence
Numerical Data
Described in numbers eg heart rate
Categorical data
Described in categories eg death (yes/no)
Incidence involves what type of data? (numerical / categorical)
Only involves counting categorical disease events
Prevalence involves what type of data? (numerical / categorical)
Either counting categorical disease events or measuring numerical disease states
Cohort study (time and incidence / prevalence)
exposures measured, then disease events counted after a period of time, generally measure incidence, but can measure prevalence (at any point)
Cross-sectional study (time and incidence / prevalence)
exposures and outcome measured at same point in time, only prevalence
What determines incidence?
Depends only on number of events during a specified time period
What determines prevalence?
Can count ‘incidence drizzle’ but miss out on ‘death drips’ and ‘cure cloud’
Incidence (2 strengths, 1 weakness)
Strengths - only determined by disease risk in popl (‘clean’ measure); includes events (N), population (D) and time (T). Weakness - can be difficult to measure (need to measure events over time)
Prevalence (2 weaknesses, 1 strength)
Weaknesses - only includes events (N) and population (D), not time (T); determined by incidence, cure rate and death rate (‘dirty’ measure). Strength - relatively easy to measure (‘stop time’ and count)
Epidemic
occurrence of disease in excess of normal
Pandemic
epidemic occurring in many countries
two types of numerator when measuring prevalence
measure at one point in time (eg obesity) or measure by looking back (eg asthma)
Randomly controlled trials
Like cohort studies, except participants are randomly allocated to EG or CG
Double blind
neither participants nor investigators know which intervention was given to which participant
two reasons why it is often not possible to do RCT
unethical or impractical
Risk Difference, and if EGO = CGO
EGO - CGO, RD is 0
Risk Ratio (Relative Risk), and if EGO = CGO
EGO ÷ CGO, RR is 1
units of RD
has units eg deaths / 100 people / 5 years
units of RR
no units
Random Error
if error occurs by chance
Non-Random Error
errors due to poor study design, processes or measurement
RAMBOMAN
Recruitment, Allocation, Maintenance, Blind or Objective Measurement, ANalyses
Questions to think about for Recruitment
are participants representative of whole population? what was the response rate?
Questions to think about for Allocation
was allocation to EG and CG successful / accurate? were EG and CG similar or was adjustment needed?
Questions to think about for Maintenance
did participants remain in initial exposure group? (most danger in long term studies)
Questions to think about for Blind or Objective Measurement
was there blind or objective measurement? will validity be affected by how well EG and CG were measured?
Confounding - what is it and how to deal with it
when exposure is mixed with another factor that is also associated with the outcome; adjustment (do sub studies, with all confounders in one sub study), randomisation
How to reduce random error
Repeating
Random Measurement error
participants are always moving, so identical measurements of exposures / outcomes in same / similar people can change
Random Sampling Error
cannot study everybody, only a sample, so every study will have a different EGO / CGO
95% Confidence Level (good enough definition)
about a 95% chance that the true value in a popl lies within the 95% confidence level (assuming no non random error)
95% Confidence Level (actual definition)
in 100 identical studies using samples from the same popl, 95/100 of the 95%CI will include the true value of the popl
RD is statistically significant when ____
CI of EGO and CGO don’t overlap, if RD doesn’t overlap zero
how to reduce random sampling error?
use a bigger sample
how to reduce random measurement?
do more measurements
meta-analysis
combination of 95% CI in 4 or more identical studies
How many parts to epidemiological studies?
usually 5, sometimes only 4 ( no comparison group)
Cross-Sectional Studies (3 pros, 1 con)
Pros - useful for investigating prevalence of risk factor / disease, no maintenance error, relatively easy / cheap / quick. Con - not useful for studying benefits of intervention (confounding, reverse causality)
Randomised Control Trials (3 Pros, 1 Con)
Pros - very good if can keep maintenance error low, reduces confounding, if participants blind to intervention maintenance usually similar in EG and CG. Con - in long term studies, maintenance error can be high