HaDPop Flashcards
Define a CENSUS
The simultaneous recording of demographic data by the government at a particular time, pertaining to all of the people who live in a particular territory.
What are the uses of a census?
Population projections and trends (age, ethnicity etc), allocation of resources for health, housing, employment, transport.
Define the Crude Birth Rate (CBR)
The number of live births per 1000 population (per year).
Define the General Fertility Rate (GFR)
The number of live births per 1000 females aged 15-44 (per year)
Define the Total Period Fertility Rate (TPFR), and how it is calculated.
The average number of children that would be born to a hypothetical woman in her lifetime. It is the sum of the current age specific fertility rates. (Short term measure, not long term).
What is the disadvantage of Crude Birth Rate?
Men and women who are not of fertile age do not give birth.
Give one advantage of General Fertility Rate and one reason why it might not be calculated.
It is more accurate for measuring birth rates as it only describes fertile female populations. However it is difficult to ascertain fertile females in some areas.
What does TPFR account for?
Confounding.
Define fecundity
Physical ability to reproduce
Define fertility
Realisation of the potential of fecundity as births
What affects fecundity?
Hysterectomy, sterilisation
What increases fertility?
Sexual activity, good economic climate
What decreases fertility?
Contraception, abortion
Conceptions =
Live births + miscarriages + abortion
Define Crude Death Rate (CDR)
Number of deaths per 1000 population per year
Define: Age Specific Death Rate
Number of deaths per 1,000 in age group (per year)
Define Standardised Mortality Ratio (SMR)
Compares the observed number of deaths with the number of expected if the age-sex distributions of populations were identical. Describes local population to general population.
Takes into account confounders
How do you calculate SMR?
observed no. of deaths/ expected no. of deaths x100
SMR of 100 is same as ref population
What does SMR account for?
Confounding
What do SMR values above and below 100 suggest?
SMR > 100 suggests excess mortality
SMR < 100 suggests less mortality
What is Birth Notification and when and who is it done by?
Notification by an attendant at birth (usually a midwife) for relevant services such as immunisation purposes etc. It must be done with 36 hours to local Child Health Registrar.
What is Birth Registration, and when and who is it done by?
For statisitical purposes to Local Registrar for Births, by parent, within 42 days.
What is Death Certification?
An obligation by an attending doctor to provide the likely cause of death, and notify a coroner if unsure or suspicious.
What is Death Registration?
To local Death Registrar, within 5 days by a qualified informant, usually a relative. Requires a Death Certificate.
What is a confounder?
Something that is associated with both the outcome and exposure of interest, but is not on the causal pathway between exposure and outcome.
They distort and give misleading results.
Define Incidence rate
The number of new cases of a disease, per 100 people per year (person years)
Define prevalence (proportion)
The amount of people who currently have a disease in a set population.
Denominator is persons, no time period
Define incidence rate ratio
Calculation?
Comparison of the incidence rates in two populations with differing levels of exposure. Relative risk.
IRR = RateB (exposed) / RateA (unexposed)
What is another use of IRR
Efficiacy of treatment.
What are some common confounders?
Age, sex, ethnicity
What is the difference between systematic and random variation?
Systematic: can be attributed to various factors e.g. age.
Random: when fluctuations can not be explained.
What is a confidence interval? (95%)
The range in which we can be 95% sure that the true value of the underlying tendency lies.
How do you calculate a confidence interval?
Lower bound = value / error factor
Upper bound = value x error factor
How would you carry out a hypothesis test?
- State the null hypothesis
- Work out the IR/IRR
- Work out the error factor
- Work out the 95% confidence interval
- Interpret your data
What is the null hypothesis?
There is no difference in risks between exposed and unexposed populations. The IRR is 1, the SMR is 100.
What values of p are statistically significant? (and therefore can reject the null hypothesis)
When p<0.05
When p>0.05, what conclusion can be drawn?
There is little or no evidence against the null hypothesis, however it is not proven, it can not be accepted.
What is the p value?
The probability that the data observed is simply due to chance and of obtaining a test statistic.
What are the limitations of hypothesis tests?
Rejecting a hypothesis is not always useful, the 0.05 value of p is arbitrary.
Statistical significance depends on sample size.
What is biasing?
The deviation of the results from the truth via certain process
What is selection bias?
Give 2 examples
Error due to systematic differences in the ways in which the two groups were collected.
Allocation bias, healthy worker effect
What is information bias?
Give 2 examples
Error due to systematic misclassification of subjects in the group
Recall bias, publication bias
What does a cohort study involve?
Recruiting disease free individuals and classifying them according to their exposure status.
They are followed up for extended periods, disease progress is monitored and incidence rates are calculated, then IRR.
What are the types of cohort study?
Prospective and retrospective.
What is a prospective cohort study?
Disease free individuals are recruited in the present and followed up in the future.
What is a retrospective cohort study?
DIsease free individuals are recruits, exposure status calculated from historical documentation, and followed up.
What are the two types of comparison?
Internal and external
Give 3 problems with internal comparison
- Sub cohorts may be of radically different sizes.
- Large studies may be needed.
- Sub cohorts may not be comparable in respect of confounding factors.
What is internal comparison?
Comparison between sub-cohorts within the original group, which are exposed and unexposed.
What is external comparison?
This occurs when the exposed population is compared against a reference population instead, using an SMR calculation.
Give 4 problems with external comparison
- There is often limited data pertaining to the reference population as you have not collected it.
- Often no incidence data.
- Usually have to make do with mortality data.
- Study and reference populations may not be compatable; selection bias, the healthy worker effect
What is the healthy worker effect?
When biasing of results in studies involving occupational cohrts are compared to a reference population.
Employed individuals are more likely to be healthy than an unemployed individual, as you have to be healthy to work.
When comparison is made, it should always be done against other workers in order to prevent bias.
What is sensitivity analysis?
A technique used to determine how different values of an independent variable will impact a particular dependent variable under a given set of assumptions
What are the underpinning concepts of HaDPop?
Truth
Bias
Confounding
Chance
Why is cancer registered?
Every new diagnosis
Tp monitor trends in incidence and prevalence, survival, screening and care quality
Define morbidity:
Any departure, subjective or objective, from a state of physiological or psychological well being.
Sickness, illness and morbid condition are similarly defined and synonymous
What areas is morbidity data held on in particular?
Cancer
Communicable diseases: 33 to notify
Congenital abnormalities
Abortion: notify cheif medical officer
What is population size determined by?
Births
Deaths
Migration in and out
What are sources of health information?
NHS: Hospital Episodes Statistics
GP registrations
Quality and Outcomes Framework
What are methodological issues with collecting health information?
Completeness/duplications
Accuracy
Variations in disease diagnosis over time
Numerator/denominator mismatch
Indirect nature of data
Confidentiality
Record Linkage
What can a low response rate to a study cause?
Non responder bias, innacurate representation of true value
What does prevalence equal?
Incidence x Mean Disease Duration
What is incidence rate a good measure of?
Give a specific type of incidence rate
Determines epidemics, effects of prevention programmes, good for short term illness, can compare risks between populations
Mortality rate
What is prevalence a good measure of?
Incurable, long standing diseases
How many people need to be cared for at a particular moment in time, services needed
What is the relationship between incidence and prevalence?
All prevalent cases were once incident
Higher incidence=higher prevalence
Prevalent cases always added to by incident cases
Prevalent cases depleted by death or recovery
New treatment which prolongs life = increased prevalence
More cured/death = decreased prevalence
Why is systematic variation useful?
It gives us clues about the aetiology (cause) of disease
Can compare levels of exposure in two groups to identify cause, prevent exposure and reduce incidence
What is variation?
What is the difference between systematic and random variation
Difference between observed and actual value
Systematic can be attributed to particular factors such as age
Random when fluctuations can not be explained
What is a hypothesis?
A statement that an underlying tendency of scientific interest takes a particular quantitative value
What is a hypothesis test?
Calculation of the probability of getting an observation as extreme as, or more extreme than the one observed, assuming the stated hypothesis is true
What are the limitations of the p value?
Rejecting a hypothesis is not always useful
0.05 value is arbitrary
Statistical significance depends on sample size, does not always correlate to clinical importance
Where does the observe value lie in a 95% confidence interval?
In the middle
How should a cohort study analysis be done?
Null hypothesis
Calculate IRR
Calculate error factor
Confidence interval
Conclusion
How to write a conclusion for a test analysis
Sufficient evidence to reject null hypothesis? Yes/No
Explain with reference to confidence interval
Statistically significant? Yes/no, give p value
Explain how many times likely to develop disease
Define person years
The sum of the total time of everybody followed up in a study
No of people exposed x time exposed for
What are advantages of cohort studies?
Can study exposures and personal characteristics that are not routinely collected
Good for rare exposures
Opportunity to look for different potential outcomes at once from varying exposures
Obtain detailed info on outcomes and exposures
Collect additional data on potential confounders, can adjust
Establish that exposures precede outcomes
Specific calculation of absolute risk
What are disadvantages of cohort studies?
Large, resource intensive, expensive
Take a long time (historical less), ethics
Rigorous definitions of disease can require expensive and invasive investigatio
Risk of high number of losses to follow up - survivor bias
Not good for rare diseases - too few cases
Diffculty with unknown confounders
What is a case control study?
Recruit disease free individuals (controls)
Recruit diseased individuals (cases)
Then determine their exposure status
Compare level of exposure in both
Analyse by calculating an odds ratio
Generally retrospective
How is an odds ratio calculated?
AD/BC
Odds being exposed of cases/odds being exposed controls
Table with cases and controls on top
Exposed and unexposed down side
How many controls should be used?
4 to 6 times the number of cases
As precision of OR affected by total number of controls as well as cases
Decreases the error factor
What isa nested case control?
Case control within a cohort
Good as can calculate IR
Population for sampling controls already defined
Need detailed info only on a few individuals
What are the issues with a case control study?
Selection bias: cases should be representative of all cases, not just in specialist centres
Information bias: recall bias, exposure status incorrectly determined. Different questionnaires, random inaccurate measurement
How can confounding be minimised in a case control study?
Match cases and controls using important confoundeds (age, sex, dates of birth)
Adjusted with analysis using logistic regression
Describe advantages of case control studies
Good for rare diseases
Can look for different potential exposures at once
Quicker so cheaper than cohort studies
Describe disadvantages of case control studies
Can’t obbtain absolute risks (unless nested)
Heavily affected by selection and recall bias
Not good for rare exposures
In terms of Henle Koch postulates, what 3 things must a cause be?
Agent must be present in every case of the disease
A cause always precedes the disease (NECESSARY)
Cause is absent in other diseases (SPECIFIC)
A cause alone can lead to the disease, can be recovered from experiemental inoculated animals (SUFFICIENT)
Give examples of causes and diseases which are necessary, sufficient, both or neither
Necessary, not sufficient: TB and mycobacterium TB
Sufficient, not necessary: Lung cancer and radon gas
Necessary and sufficient: measles and the measles virus
Neither necessary or sufficient: TB and poor living conditions
What does disease result from?
An interplay of host, environment and agent
Define a cause
Why is it important to identify causes?
An exposure or factor that increases the probability of disease
They do not have to be necessary or sufficient to be importnat causes
Aim is to use the knowledge gained to remove, avoid, or protect against harmful factors
What can apparent associations be caused by?
Chance, p value measure this
Bias, selection or information
Confounding, known or unknown
Due to common cause
Reverse causailty
Could be true causal association
What are the bradford hill criteria?
Strenght of association
Specificity of association
Consistency of association
Temporal Sequence
Dose response
Reversibility
Biological plausibility
Analogy
Cohernece of theory
How is an RCT carried out?
Identify eligible patients
Invite
Allocate treatment randomly
Follow up identically
Minmise losses to follow up, maximise compliance
Analyse date, obtain results
Why is a non random trial weak?
Selection bias, by patient, clinician or investigator
Confounding
What is the advantage of randomisation in a trial?
Minimal allocation bias
Minimal confounding, similar size groups and characteristics by chance
Why might differences between groups occur in an RCT?
Knowledge of which patient is receiving which treatment
Patient may change behaviour/expectations
Clinician may make different secondary care choices
What does blinding do in an RCT?
Knowledge of treatment allocation not available
Can be single (patient)
Or double (patient and clinician)
Ideally investigator too
Why is it important ot blind clinicians in an RCT?
Doctors more likely to place healthier patietns in new drug group
Show more postive results
For statistically significant results
What is it difficult to blind for in an RCT?
Surgery
Psychotherapy vs anti depressant
Alternative medicine vs western medicine
Lifestyle interventions
Prevention programmes
What is a placebo?
An inert substance made to appear identiceal in ever way to the active formulation to which it is being compared
What is the aim of a placebo?
To cancel out any placebo effect that may exist in the active treatment
What is the placebo effect?
The psychological effect/benefit that derives from being looked after more carefully or differently, or from being on a new special treatment
Can result in improvement of clinical condition due to making you feel better psychologically
What are the ethical issues of using a placebo?
Should only use when no standard treatment is available
Form of deception
Essential that all participants are informed that they may receive a placebo
What is a primary outcome?
What is a secondary outcome?
Primary: preferably only one, used in sample size calculations
Secondary: other outcomes of interest, often side effects, can be promoted wrongly when statistically significant
What are outcome types?
Pathophysiological - objective
Clinically defined
Patient focused - subjective
What do RCTs measure?
Efficacy and harm of treatments
How should trials be analysed?
Intention to reat: non compliers included in results to give an idea of how the drug would do in real life clinical practice. Preserves randomisation
As treated only compares physiological effects, may miss side effects, loses randomisation
How can losses be minimised and compliance maximised in a clinical trial?
Follow up appointments at practical times
No coercion or inducements
Honesty to patient
Simplify instructions
Let patient ask Qs
Cool off period
Make simple and accessible to patients
Measure compliance with blood/urine tests
Describe RCT ethics
Clinical equipoise: reasoble uncertainty on which drug is better for the patient, so not subjecting patients to a known less effective treatment or known harm (loss of autonomy compensated by greater monitoring, care and placebo)
Scientificially robust: address relevant valid issue of benefit, pursuit of knowledge for the good of the population. Appropriate study design and protocol, large enough to find clinical importance, stop if harmful
Ethical recruitment: no inclusion of participants in region unlikely to benefit (e.g AIDS), or those witha high risk of harm , no innapropriate exclusion of those who are not in ideal homogenous group, those with co morbidites, elderly, children
Valid consent and voluntariness: no coercion, knowledgable informant,cooling off period.
What are the 4 principles with any medical role?
Principle of
Benefiance (do good)
Non-Malefiance (do no harm)
Autonomy (let the patient decide)
Justice (Be fair)
What must be the primary consideration of a research study according to a research ethics committee?
Dignity, rights, safety and well being of participants
Focus of design, conduct, recruitment, care, confidentiality, consent, benefits after study
What is stratified random sampling?
Random but selecting certain age groups
What is a cross over trial?
Both arms of the study get both interventions
What is study contamination?
Patients in the other control group are subject to the new intervention
What is a sample size calculation?
FOr a good study, tels us what the calculation is based on
e.g. a 10%different in outcome,and what size study they need as a result
What is the difference between bias and confounding?
Bias is any deviation of results from the true value, or processes leading to this, due to something systematically wrong with the study.Feature of study being conducted, different samples or treated differently, selection and information bias. Does not correctly reflect population it purports to measure.
Confoudning gives rise to a spurious assoication because of another causal factor, which is due to the structure of the population being studied.Feature of population, influences measure of association, can lead to misinterpretation. Something associated with both the outcome and exposure of interest, but is not on the causal pathway between exposure and outcome
What are the limitations of using evidence from individual studies?
Can’t read all individual studies
Statistical significance may be due to chance
Different types of study have different weaknesses
On trial can’t definitively answer a hypothesis
What are the weaknesses of expert reviews?
They make implicit assumptions
Opaque methodology
Biased
Subjective
Not reproducible
What qualities should studies used in systematic review be?
Transparent
Reproducible
Explicit
What is a systematic review?
An overview of primary studies that use explicit and reporducible methods
Usually RCT - maybe cohort or case control
Large amount of studies initially identified then narrowed down to most reliable and credible
What is a meta analysis?
Wtihin a systematic review, used to provide quantitative analysis of the primary studies’ results (that addressed the same hypothesis in the same way)
Proveds an overall value with associated confidence intervals (estimate of the true value)
Can show results via a forest plot
What is the purpose of a meta analysis?
Facilitate synthesis of many studies
Systematically collate study results
Reduce problems in interpretationdue to variations in sampling
Quantify effect sizes and their uncertainty as a pooled estimate
Sits within a systematic review
How is a forest plot interpreted?
ORs and CIs combined to give a pooled estimate
Studies weighted according to size and OR certainty (ef size)
Squares give ORs, lines give CIs for each study
Size of square proportional to weight given
Solid line = null hypothesis
Diamond is the pooled estimate with its width the 95% CI
Ideally, what should studies in a systematic review be?
Homogenous: similar in terms of study design, participant profile, treatments/exposures, oucomes measured, statisitcal analysis used
Though in practice no two studies are identical
Usually heterogeneous
What are the two models used to show meta analysis?
Fixed effects model
Random effects model
Point estimate (OR) is usually similar
Often both methods are used
What is the fixed effects model?
Assumes the studies used are homogenous and any variation between data comes from within study variation
Estimating the same effect size
ONE TRUE EFFECT
Weight by study size
What is the random effects model?
Assumes the studies are heterogenous
Estimating similar, not same effect size, range of true effects which average to a mean value
Variation in data from within study and between study variation
Gives more weight to smaller studies, used when large variance between studies, more equal weights
95% CI usually wider
How is heterogeniety tested for?
Hypothesis test for heterogeneity
Weak test, low statistical power
Often at 10% significance level
How can heterogeniety be explained?
Sub group analysis
Stratification
by study characteristics
by participant profile
How is quality assessed in a systematic review?
Basic quality standard, cochrane only uses RCTs
Score quality, used to weight studies
Sub group analysis
Meta regression analysis
More than one assessor to handle disagreements
Can try and blind assesors, stops favoruitism, but difficult
List study types in increasing order of how prone to bias they are
RCTs
non RCTs
Cohort studies
Case control
What is publication bias?
Studies are only likely to be published if results are
statistically significant
or have a large sample size
If systematic reviews only include published studies, results will be biased
Towards demonstration of effect
What are systematic reviews reliant on?
Quality of primary studies - garbage in, garbage out
Publication bias, should use both published and unpublished studies
How can publication bias be determined?
Funnel pot
Check metaanalysis protocol for searching
How is a funnel plot interpreted?
WEll balanced systematic review will show a funnel shape in its studies when plotted
Biased systematic review will vary in shape
Plot of size of effect (OR) on x axis, vs size of study (1/ef) on y axis
Smaller studies vary from the centre more
What are advantages of a systematic review?
Explicit methods reduce bias and exclusion of poor quality studies
Mata analysis provides overall figure for studies, and statistical significance
Large amounts of information can be assimilated quickly by healthcare professionals
Reduction in time between research discovery and implementation of clinical use
Used in evidence based practice guidelines
What is risk difference vs risk ratio?
Risk difference: continuous outcome: e.g. systolic blood pressure
. Not mortality which is one of two outcomes, reduction or increase, - numbers. Null hypothesis is 0.
Risk ratio: compared to null hypothesis of 1.
What is a representative sample?
When measuring something about a large population, but would be expensive and impractical
Take a sample with representative characteristics
What is a random sample?
Not intended to be representative of the population
Everyone has an equal chance of selection
Random selection vs random allocation
Seleciton: recruitment in a clinical trial, affects external validity, e.g. generalisability
Allocation: allocation between treatment groups in a trial, affects internal validity e.g. fair comparison
Crude vs standardised
Crude: describes what was measure, unadjusted
Standardised: what would be measured if the groups were identical in respect to variables being adjusted for
As treated vs intention to treat
As treated: what the trial participants did, asses physiological effects of treatments studied
Intention to treat: analysis according to original allocation, preserves randomisation, reflects what would happen in practice