Cornerstones of epidemiology: Observational Studies and Routine Data Flashcards
List the hierarchy of scientific evidence
Systematic reviews and meta-analysis Randomised controlled trials Cohort studies Case-control studies Descriptive studies (inc clinical audit) Case reports/series
Give an example of when you would use a cross-sectional study
Accuracy of diagnostic test
Give an example of when you would use a cohort study
Prognosis of disease
How could you measure the causes and risk factors of diseases
Various non-randomised designs
How could you determine the healthcare needs of the population
various, inc ecological (aggregate) studies
How could you determine treatment efficacy
Randomised trial
Why are observational studies important
They help us find the cause of diseases
Where do many observational studies obtain their data from
Routine data
What happens as you go up the hierarchy
The data becomes more robust and less prone to confounding and bias, but all are of value
Describe descriptive studies
Describe the distribution of factors or disease in relation to:
Person (e.g. age, sex, race, marital status, occupation, lifestyle)
Place (e.g. variation between and within countries/units – CQC does this)
Time (variation over time and season)
Why is standardised mortality ratio important
Used for comparing one area with a “standard population”, adjusting for age (often also sex)
Describe the standardised mortality ratio
Represents the ratio of the number of observed deaths (or cases of disease) (O) in a particular population to the number that would be expected (E), if that population had the same mortality or morbidity experience as a standard population, corrected for differences in age (and sex) structure.
Describe the types of data used in descriptive studies
Routine e.g. births, deaths
Survey e.g. Health Survey for England
Performance management: Quality and Outcomes Framework for GPs
Other study designs tend to collect their own data
What is meant by routine data
“Data that are routinely collected and recorded in an ongoing systematic way, often for administrative or statutory purpose and without any specific research question in mind at the time of collection”
The Law says that you have to collect them
Describe some of the different types of routine data
Healthcare use and outcome data e.g. deaths, hospital admissions, primary care consultations or prescriptions, immunisation uptake
Exposures and health determinant data, e.g., air pollution, crime statistics
Demographic data e.g. census
Geographical data e.g. health authority boundaries, location of GP practices
Health service provision, e.g. bed/staff counts
What are the advantages of routine data
Relatively cheap Already collected and available Standardised collection procedures Relatively comprehensive – population coverage, large numbers Wide range of recorded items Available for past years Experience in use and interpretation Can look at time trends as data is available from past years
Describe the disadvantages of routine data
May not answer the question (no information or not enough detail) Incomplete ascertainment (not every case captured) Variable quality (e.g. variable diagnosis fields) Validity may be variable (i.e. do they measure what you think they measure?) Disease labelling may vary over time or by area Coding changes may create artefactual increases or decreases in rates, e.g. ICD9 to ICD10 Need careful interpretation
Describe the coding changes
Free text- converted to code- these systems change over time and get updated- careful interpretation required.
Describe some health outcome data collected by routine studies
Mortality Cancer Notification of infectious diseases Terminations of pregnancy Congenital anomalies Hospital episode statistics GP data e.g. QOF (see later) Road Traffic Accidents
Describe cross-sectional studies
Useful for health care providers to allocate resources efficiently and plan effective prevention
Provide clues leading to hypotheses which can be tested in analytical studies
Describe status of individuals with respect to absence or presence of both exposure and disease assessed at the same point in time
…but cannot easily distinguish whether exposure preceded disease: chicken or egg?
Why don’t cross-sectional studies give you answers
Data only collected and analysed at one period of time, hence you can’t get answers but you can generate hypotheses.
Give some examples of cross-sectional studies
Health Survey for England: has core topics each year since 1991 + special topics
Census (last one was 2011)
General Lifestyle Survey (old GHS) since 1971: alcohol, smoking, household size etc
National surveys of NHS patients
NHS staff survey
Describe the census
Began in 1801. Every 10 years Population estimates Health question Other health indicators Unemployment Ethnicity Age Overcrowding
Describe vital registration in England and wales
Births, marriages and deaths
Only ask fathers in marriages- needs updating
Describe how mortality can be measured
Death certificates Local registrars of births and deaths ONS for coding and processing Produced as routinely published tables General DH1 By area DH5 By cause DH2 etc Public Health Mortality Files Data extracts
In 1984, why was there a massive increase in the number of deaths due to senile dementia
Due to coding change- cannot take data at face value
Describe cancer registrations
Voluntary notification to local cancer registry: now national system
Also from death certificates
Useful for both incidence and survival information
Increasingly being linked to hospital admissions data and national clinical audits
What were the 4 most commonly diagnosed cancers in the UK in 2001
Breast, Lung, Colorectal and prostate
List some infectious diseases that require notification
29 notifiable diseases: Acute encephalitis Plague Acute poliomyelitis Rabies Includes food poisoning, meningitis, tuberculosis and plague Reported by doctors
Describe the Quality and Outcomes Framework
Quality and Outcomes Framework (QOF) is a component of the new General Medical Services contract for GPs from April 2004
QOF rewards practices for the provision of quality care, and helps to fund further improvements in the delivery of clinical care
Collected in a national database system: Quality Management Analysis System
Practice-level data are published; being phased out in many areas
Determines how much money GPs have to reinvest in clinical services depending on how well they manage patients with chronic conditions.
What is currently happening with the Quality and Outcomes Framework
It is starting to be phased out, but it is still a useful data source.
Describe administrative hospitals admissions data
HES (Hospital Episode Statistics) covers all NHS hospitals + Treatment Centres in England
15m records annually + 60m OPD appts + 19m A&E atts
Coding uses ICD10 and OPCS
Data quality still improving
Many uses
What is meant by an episode
Finished Consultant Episode - the time spent under the continuous care of a specific consultant
What is meant by an admission
a patient’s stay in hospital, so comprises 1+ episodes and/or transfers between hospitals
What is meant by multiple episodes
The patient has been seen by different consultants
What is meant by observed morbidity and what does it depend on
Health service knows that you’re ill
How the health system is organised influences who observes you to be ill
Describe case-control studies
Case-control studies are commonly used in epidemiology
Relatively cheap and quick to conduct
Suitable for studying what might cause rare diseases
Describe the study designs of case-control studies
Appropriate control group- those free of the disease
Target group- those with the disease
Move back in time- ask questions about certain exposures, determine odds of exposure.
Outline the procedure for carrying out a case-control study
Procedure Example
1 Select cases with disease Cases: brain tumours
controls without disease Controls: from population without cancer
2 Obtain information on past Examine mobile phone exposures and other use to classify people factors into exposure categories
3 Compare proportions of Compare proportion of people exposed in frequent mobile phone cases and controls users in cases and controls
Describe the selection of controls
Selection of an appropriate comparison group is the most difficult and critical issue in the design of case-control studies
Controls are subjects free of the disease (or outcome of interest) during the same period of time in which the cases were identified.
They should be representative of the population of individuals who would have been identified and included as cases if they had also developed the disease
What are some of the sources of data for controls
General population Neighbourhood Friends/relatives Hospital or clinic-based These all vary in amount of recall bias, response rates, selection bias, cost
How do you calculate an odds ration
Odds of exposure in controls/ odds of exposure in cases
What are the advantages of case control studies
Good for rare diseases
Quick and cost-efficient
Can investigate many exposures simultaneously
What are the disadvantages of case-control studies
Problems with selection of controls (selection bias)
Subject to recall bias
uncertainty of exposure-disease time relationship
Poor for rare exposures
Cannot calculate incidence rates directly
Why are admission criteria sometimes different
They depend on the judgement of each doctor
What is meant by a cohort
A “cohort” is a group of people who have something in common:
All patients registered with the same GP
Everybody who has received the swine flu vaccine
People with an allergy to cats
A cohort represents the outcome-free population from which cases (people with the outcome) eventually arise
Describe the study design of a cohort study
Exposed and unexposed individuals in outcome-free individuals
Measure those who get the disease in each group
Calculate risk ratio
How do you calculate risk ratio
Risk factor of disease in exposure group/ risk factor of disease in unexposed group
Describe the advantages of cohort studies
Able to look at multiple outcomes
Able to follow through the natural history of disease
Good design to look at risks related to rare exposures
Incidence can be calculated
Can minimise bias in estimating exposure if prospective
Describe the disadvantages of cohort studies
Inefficient for studying rare diseases- you would need 1000s of participants for thorough analysis
Expensive and time consuming (if prospective)
Loss to follow-up may introduce bias
Healthy worker/volunteer effect may affect generalisability
Describe the healthy worker effect
Healthier people are more likely to volunteer to participate in cohort studies, hence bias is introduced and the findings are not as representative.
