The Use of Data

Question 1

define disease

Answer 1

symptoms, signs – diagnosis. Bio-medical perspective

Question 2

define illness

Answer 2

ideas, concerns, expectations – experience. Patients

perspective

Question 3

give an example of a disease with no illness

Answer 3

HTN

Question 4

factors affecting the uptake of care

Answer 4

concept of lay referral
sources of info
medical factors
non medical factors

Question 5

issues with a disease with no illness

Answer 5

believe to be health
physically fit
no need to take medications

Question 6

what are rates in epidemiology

Answer 6

Clinical medicine deals with the individual patient, epidemiology deals with
populations. It is essential to be quite clear which populations we are talking about
when we carry out a survey, conduct a study or formulate a hypothesis about disease
and risk. In order to do this we talk in terms of ratios :
Numerator = Events
Denominator Population at risk
eg, Deaths from IHD in men aged 55-64 in Grampian in 1990
All men aged 55-64 in Grampian in 1990
The numerator is the top line, the number of events (in this example
deaths). The denominator is the bottom line, the population at risk.
It is usual to convert such ratios into rates by expressing them in terms of a specified
time period (eg, per year) and a notional ‘at risk’ population of 10n (eg, %; per 1000;
per 100,000).
Note that the at risk part is crucial. What this means is that everyone in the
denominator must have the possibility of entering the numerator, and conversely those
people in the numerator must have come from the denominator population.

Question 7

what is relative risk

Answer 7

This is the measure of the strength of an association between a suspected risk factor and the
disease under study.
Relative risk (RR) = incidence of disease in exposed group
/incidence of disease in unexposed group

Question 8

source of epidemiological data

Answer 8

mortality data
hospital activity stats
reproductive health stats
cancer stats
accident stats
general practice morbidity
health and household surveys
social security stats
drug misuse databases
expenditure data from NHS

Question 9

what is health literacy?

Answer 9

people having the knowledge, skills, understanding and confidence to use health information to be active partners in their care, and to navigate health and social care systems

Question 10

describe studies

Answer 10

attempt to describe the amount and distribution of a disease in a given population. does not provide definitive conclusions about causation but may give glues to possible risk factors and candidate aetiologies
usually cheap, give and valuable initial overview

Question 11

what are descriptive studies useful in?

Answer 11

Identifying emerging public health problems through monitoring and surveillance of
disease patterns.
Signalling the presence of effects worthy of further investigation.
Assessing the effectiveness of measures of prevention and control (eg, screening
programmes).
Assessing needs for health services and service planning.
Generating hypotheses about disease aetiology.

Question 12

types of analytic studes

Answer 12

cross sectional
case control
cohort

Question 13

cross sectional studies

Answer 13

In cross-sectional studies, observations are made at a single point in
time.
Conclusions are drawn about the relationship between diseases (or
other health-related characteristics) and other variables of interest in a
defined population.
A strength of this method is its ability to provide results quickly;
however, it is usually impossible to infer causation.

Question 14

case control studies

Answer 14

In case control studies, two groups of people are compared: a group of individuals who have the disease of interest are identified (cases), and a group of individuals who do not have the disease (controls)
Data are then gathered on each individual to determine whether or not he or she has been exposed to the suspected aetiological factor(s). The average exposure in the two groups, cases and controls, is compared to identify significant differences, give clues to factors which elevate (or reduce) risk of the disease under investigation.
The results obtained from case control studies are expressed as ‘odds ratios’ or ‘relative risks’ (see above). Be aware that relative risks are also presented for cohort studies and randomised trials. Sometimes confidence intervals or ‘p values’ are presented as a guide as to whether the result could be a chance finding.

Question 15

cohort studies

Answer 15

In cohort studies, baseline data on exposure are collected from a group of people who do not have the disease under study. The group is then followed through time until a sufficient number have developed the disease to allow analysis.
The original group is separated into subgroups according to original exposure status and these subgroups are compared to determine the incidence of disease according to exposure. Cohort studies allow the calculation of cumulative incidence, allowing for differences in follow up time.

Question 16

how are results of cohory studies expressed

Answer 16

The results are usually expressed as relative risks (see above), with
confidence intervals or p values.

Question 17

trials

Answer 17

Trials are experiments used to test ideas about aetiology or to evaluate interventions. The “randomised controlled trial” is the definitive method of assessing any new treatment in medicine. Two groups at risk of developing a disease are assembled, a study (intervention) group and a control group. An alteration is made to the intervention
group (eg, a suspected causative factor is removed or neutralised), whilst no alteration is made to the control group. Data on subsequent outcomes (eg, disease incidence) are collected in the same way from both groups, and the relative risk is calculated. The aim is to determine whether modification of the factor (removing, reducing or increasing exposure) alters the incidence of the disease.
In a trial of a new treatment, the underlying design is the same: the intervention group receive the new therapy, the control group receive the current standard therapy (or a placebo) and the treatment outcomes (eg, reduction in symptoms) are compared in the two groups

Question 18

factors to consider when interpreting results

Answer 18

standardisation
standardised mortality ratio
quality of data
case definition
coding and classification
ascertainment

Question 19

standardisation

Answer 19

A set of techniques used to remove (or adjust for) the effects of differences in age or other confounding variables, when comparing two or more populations. An age-sex standardised rate represents what the unstandardised (crude) rate would have been in the study population if that population had the same proportion of males and females, and of people in different age groups, as the standard population. Rates can be standardised for any other relevant confounding factor (eg, social class). Comparisons of incidence or mortality rates in a population over time, or between two different populations, or between population subgroups, should always be based on standardised rates, never on crude rates.

Question 20

Standardised Mortality Ratio (SMR)

Answer 20

This is a special kind of standardisation which you may encounter in your reading. It is simply a standardised death rate converted into a ratio for easy comparison. The figure for a standard reference population (eg, Scotland) is taken to be 100 and the standardised death rates for the comparison (study) populations (eg, Grampian) are expressed as a proportion of 100. A figure below one hundred means fewer than expected deaths, and above 100 means more. For example, an SMR of 120 means that 20% more deaths occurred than expected in the study population, allowing for differences in the age and sex structure of the standard and study populations and an SMR of 83 means 17% fewer deaths occurred.

Question 21

quality of data

Answer 21

In working with data about health and disease, we must be careful to ensure that the data are trustworthy. There are some questions you can ask yourself which can help you decide whether to believe the results of analyses based on the data.

Question 22

case definition

Answer 22

The purpose of case definition is to decide whether an individual has the condition of interest or not. It is important in because not all doctors or investigators mean the same thing when they use medical terms. Differences in incidence of disease over time or in different populations may be artefact, due to differences in case definition, rather than differences in true incidence.

Question 23

coding and classification

Answer 23

This is related to the issue of case definition. When data are being collected routinely (eg, death certificates), it is normal to convert disease information to a set of codes, to assist in data storage and analysis. Rules are drawn up to dictate how clinical information is converted to a code. If these rules change, it sometimes appears that a disease has become more common, or less common, when in fact it has just been coded under a new heading.

Question 24

ascertainment

Answer 24

Is the data complete - are any subjects missing? If researchers in one country look harder for cases of a given disease than researchers in any other, it might not be surprising that they come up with higher incidence rates.

Question 25

what is bias

Answer 25

Bias is any trend in the collection, analysis, interpretation, publication or review of data that can lead to
conclusions that are systematically different from the truth. There are very many types of bias which can
creep into epidemiological studies. Four important types are described below

Question 26

types of bias

Answer 26

selection
information
follow up
systematic error

Question 27

selection bias

Answer 27

Occurs when the study sample is not truly representative of the whole study population about which conclusions are to be drawn. For example, in a randomised controlled trial of a new drug, subjects should be allocated to the intervention (study) group and control group using a random method. If certain types of people (eg, older, more ill) were deliberately allocated to one of these groups then the results of the trial would reflect these differences, not just the effect of the drug.

Question 28

information bias

Answer 28

arises from systematic errors in measuring exposure or disease. For example, in a case control study, a researcher who was aware of whether the patient being interviewed was a ‘case’ or a ‘control’ might encourage cases more than controls to think hard about past exposures to the factors of interest. Any differences in exposure would then reflect the enthusiasm of the researcher as well as any true difference in exposure between the two groups.

Question 29

follow up bias

Answer 29

arises when one group of subjects is followed up more assiduously than another to measure disease incidence or other relevant outcome. For example, in cohort studies, subjects sometimes move address or fail to reply to questionnaires sent out by the researchers. If greater attempts are made to trace these missing subjects from the group with greater initial exposure to a factor of interest than from the group with less exposure, the resulting relative risk would be based on a (relative) underestimate of the incidence in the less exposed group compared with the more exposed group.

Question 30

systematic error

Answer 30

A form of measurement bias where there is a tendency for measurements to always fall on one side of the true value. It may be because the instrument (eg, a blood pressure machine) is calibrated wrongly, or because of the way a person uses an instrument. This problem may occur with interviews, questionnaires etc, as well as with medical instruments.

Question 31

what are confounding factors

Answer 31

A confounding factor is one which is associated independently with both the disease and with the exposure under investigation and so distorts the relationship between the exposure and disease. In some cases the confounding factor may be the true causal factor, and not the exposure that is under consideration.
Age, sex and social class are common confounders. There are several ways to deal with confounding, depending on the particular study design: In trials, the process of randomisation (in effect the play of chance leads to similar proportions of subjects with particular confounding in the intervention and control groups). Restriction of eligibility criteria to only certain kinds of study subjects . Subjects in different groups can be matched for likely confounding factors. Results can be stratified according to confounding factors. Results can be adjusted (using multivariate analysis techniques) to take account of suspected confounding factors.

Question 32

criteria for causality

Answer 32

Criteria for Causality
Strength of association
As measured by relative risk or odds ratio.
Consistency
Repeated observation of an association in different populations under different circumstances.
Specificity
A single exposure leading to a single disease.
Temporality
The exposure comes before the disease.
Biological gradient
Dose-response relationship. As the exposure increases so does the risk of disease.
Biological plausibility
The association agrees with what is known about the biology of the disease.
Coherence
The association does not conflict with what is known about the biology of the disease.
Analogy
Another exposure-disease relationship exists which can act as a model for the one under investigation.
For example, it is known that certain drugs can cross the placenta and cause birth defects
- it might be possible for viruses to do the same.
Experiment
A suitably controlled experiment to prove the association as causal - very uncommon in human
populations.