Flashcards in Hadpop Deck (65):
Why is it essential to consider a population perspective of healthcare?
All treatments have an impact on others (even if just by cost)
Seeing trends in disease and treatment allows alteration to methods
Knowing needs of a community allows for better planning of services
What is a census?
A record of demographic information collected simultaneously by the government pertaining to all the people living in a particular territory.
Other than census how else might demographic information be gathered?
Smaller surveys - e.g. Household survey, health survey
National registration of births deaths and marriages.
What are the different ways of measuring birth rate?
Crude birth rate
General fertility rate
Total period fertility rate
What is crude birth rate? What use is it?
Number of live births per 1000 population
Used to monitor changes in population size
What is general fertility rate? What use is it?
The number of live births per 1000 women 15-44
Compares fertility of female populations
What is total period fertility rate? What use it?
Average number of children a women would have over her life, calculated by summing fertility rates of each age (year). Hypothetical and assumes rates will stay the same. Compares fertility between groups without interference by age group structure.
Differentiate fertility and fercundity
Fecundity is physical ability to reproduce (influenced by sterilisation, hysterectomy, infertility)
Fertility is the realisation of fecundity (influenced by behaviour, contraception, abortion, economic climate!)
What are the three ways of measuring death rate?
Crude death rate - number of deaths per 1000
Age specific death rate - CDR per age group
Standardised mortality ratio - observed deaths : expected deaths if age/sex distribution of compared populations were identical
What are examples of epidemiological numerator errors?
Changes in diagnostic methods
Changes in disease classification
Changes in protocol (e.g. Death certification)
What are examples of epidemiological denominator errors?
Indirect sources (e.g. Counting self reported sick days as illness)
What should be excluded before declaring an observed phenomenon real?
Numerator/denominator error (confounders, bias)
Number of new cases per unit time
= new cases / population*time
Number of people with a disease in a population
= number effected / population
What influences prevalence?
Incidence and length of disease (until cure or death, rapidly fatal diseases have lower prevalences than slow progressing diseases)
What is incidence rate ratio?
Rate in exposed / rate in unexposed
If >1 then exposure may be increasing risk of disease
What does an smr of 106% mean?
What does it not mean!?
An individual has 6% higher risk of death in that population
It does not mean there are 6% more deaths - the population may be younger etc. meaning deaths are actually lower - just that compared to an individual of the same demographics in the other group the risk is higher.
What is the effect of randomness?
The variation of the observed values from the true values. The observed value is our best estimate of the true value
How can observed values be used?
To test a hypothesis about the true value
To provide a probable range in which the true value lies
What is a p value?
The probability that a given result occurred by chance.
What can we infer from a p value of less than 0.05?
How can the two inferences be combined?
Something very unlikely has occured
The nul hypothesis was incorrect
We can say that it is reasonable to reject the nul
We can say that the data is consistent with the stated hypothesis
What can we say if p >0.05?
There is reasonable evidence to reject the stated hypothesis
There is substantive evidence for the nul hypothesis
What are drawbacks of p values?
Small data sets with proportionally the same results as large sets could reach different conclusions
It is an arbitrary value, there is little in actuality different between p=0.049 and p=0.051
Significance does not imply causality or importance
What are the advantages of confidence intervals?
It allows us to determine significance
It gives the range of values we are 95% certain the true value lies within
Reflect the data at level of measurement (i.e. Upper and lower limits given in units applicable to thing being measured)
How do you calculate a confidence interval?
Upper limit - observed x error factor
Lower limit - observed / error factor
Error factor = exp(2xSR(1/observed))
How does the error factor calculation vary between rates, ratios and smr?
Rates - 1/ the lone rate
Ratios - 1/observed a + 1/ observed b
Smr - 1/ observed deaths (ONLY - not expected deaths)
What is the general format of a cohort study?
Recruit disease free exposed and unexposed individuals and monitor for appearance of disease
What are advantages of cohort studies?
Can study rare exposures
Control over what data (demographics, level of exposure confounders etc. collected)
Ability to change data collected if new trends appear
Sequence from exposure to illness
What are disadvantages of cohort studies?
Bad for rare diseases
Risk of loosing participants (loss to follow up and survivor bias)
What are loss to follow up and survivor bias?
Most ill participants may not make it to follow up thus get discounted from results
People making it to the end of a trial may be more likely to have a trait that improves survival - this can act as a confounder - e.g returners for assessment being more health conscious and thus having a healthier lifestyle.
What are some variations on a cohort study?
Collect data immediately or delay start to compensate for delay from exposure to outcome (any changes that occur in the initial period are not likely to be related to exposure)
Collect data historically (easier and faster but less control)
Collect stratified data (as opposed to yes/no, demonstrates trends better)
Compare exposed group to an external population using smr. Less control but smaller error factor
If a cohort study compares against an external population what problems may occur? How do you compensate?
Ref. population will change with time - compensate with lexis diagram calculation
Healthy worker bias - if all of exposed population are in work they will be likely healthier than the general
What is a case control study?
Look at effected and unaffected populations and look backwards to see if they have been exposed
What are advantages of case control studies?
Good for rare diseases
What are disadvantages of case control studies?
Bad for rare exposures
Unable to use irr - must use odds ratio
Prone to bias
What is selection bias?
Selection of non representative cases or controls for study. For instance choosing controls in a lung cancer study from a resp ward - most of these patients will be smokers even though they don't have cancer.
What is information bias?
How can it be classified?
Non differentiated misclassification
- random inaccuracies with assigning cases/controls to exposed or non exposed groups. It should be the same proportion in both directions thus causes shrinkage towards the nul.
- non random misclassification. Can be caused by assessor bias or recall bias
Give an example of why recall bias may occur. What sort of bias is it?
Systemic information bias.
Could occur as an ill patient is more likely to record something they attribute to their illness eg. Lung ca pt. more likely than a healthy patient to remember smoking as a teenager.
What are the bradford hills criteria?
Strength of association
Fits current theories
What are henle koch postulates and why are they no longer considered relevant?
Exposure is - necessary for outcome, sufficient alone for outcome and specific for outcome
We now recognise disease as a multifactoral process with causes being more 'risk factors'
What are rcts?
A planned experiment involving patients to determine the best method of treating future patients with a given medical condition.
What must rcts be?
Large enough to avoid chance
What should be decided on prior to the start of a RCT?
Disease of interest
Treatments to compare
Outcomes to be measured
Eligible and excluded patients
What are the different grades of blinding in a RCT?
Single blind (usually pt unaware)
Double blind (usually pt and another unaware)
Triple blind (all of pt, clinician and assessor unaware)
What can make blinding difficult?
Surgery - unethical to perform fake surgery, clinician would always be aware
Physical interventions (psychotherapy, physiotherapy etc.) - clinician and pt would be aware.
Lifestyle interventions - pt has to be aware
Life threatening condition necessitating intervention - blinding must be lifted
What is the placebo effect?
The improvement in a patients condition when they believe they are receiving an intervention that they think will do them good even if it having no physical effect.
What is a placebo?
An inert treatment given to one randomised group to measure the degree of placebo effect in a trial so it can be subtracted from the results of the true treatment given to another randomised group.
What are the ethics surrounding placebo treatment?
Should only be used if there is non alternative medicine
Patients should be aware they may receive a placebo
Why should RCT outcome measures be pre defined?
What should the outcomes consist of?
To avoid data trawling - you will eventually find something that is significant if you look through the data enough ways
Pathophysiological result (e.g. Change in blood test values)
Clinical result (e.g. Improvement in rate of disability)
Patient focussed result (e.g. Increased quality of life)
When should monitoring occur during a trial?
During the trial - stop if one treatment is obviously disadvantaged or harmed.
Why do patients back out of trials?
Different to what they expected
Feeling that they are being disadvantaged by treatment
How do you avoid inappropriate loss of participants in a rct?
Properly consent and explain trial
Don't coerce pts
Don't recruit on false pretences
What is the difference between appropriate and inappropriate loss to follow up in a rct?
Appropriate - withdrawal due to clinical condition - usable data as if high in one group then suggestive of inferior treatment
Inappropriate - withdrawal due to other reasons (inconvenience, frustration) - not useful
What is the difference between an explanatory and pragmatic rct analysis? What else can this be called?
Explanatory - also as treated analysis - excludes losses, e.g. Drug a causes 'x' effect when taken correctly
Pragmatic - intention to treat analysis - includes losses, e.g. Drug a causes 'x' effect however has poor compliance due to 'y' therefore has less effect.
What are individual ethics? How do they contrast with rcts? How is this justified?
Autonomy (in RCT pt doesn't get choice of treatment)
Justice (in RCT pts receive different treatments)
RCTs are justified by collective ethics - that all pts. should have effective treatments that can only be ascertained by rcts. If a patient is receiving a treatment they have a moral obligation to help further clinical science?
How should rcts be kept ethical?
Clinical equipoise (genuine uncertainty over beat treatment)
Recruited from population that will benefit (i.e. not testing drug on people who will never be able to receive it)
Scientifically robust (should produce a usable outcome)
Fully consent patients
No coercion of patients
What is the difference between meta analysis and systematic review?
Meta analysis is quantitative
What must a meta analysis do?
Collate a large number of studies with the same hypothesis
Account for variation
Provide a single quantifiable result
How is weight of studies decided in meta analysis?
Size of study
Error factor of study
In some circumstances based on quality of study
How are the different studies represented in a meta analysis?
On a forrest plot
How can heterogeneity between studies be dealt with in meta analysis?
Fixed effect model - assume true value for all studies is the same
Random effect model - assume that the true value for all studies is similar
What is the consequence of using the random effect model in a Forrest plot? When should it be used?
Small studies gain weigh and large studies loose weight. Better used when heterogeneity is higher, otherwise use fixed effect.
Other than modelling how can study heterogeneity be accounted for?
Sub group analysis - dividing the studies up
How can the variable quality of studies be accounted for?
What are issues with doing this?
Standard for inclusion
Score based on criteria to change weighting
Who decides? Introduces bias