Downloaded from http://emj.bmj.com/ on December 14, 2016 - Published by group.bmj.

com

54

RESEARCH SERIES

Observational research methods. Research design II:

cohort, cross sectional, and case-control studies
C J Mann
.............................................................................................................................

Emerg Med J 2003;20:5460

Cohort, cross sectional, and case-control studies are While an appropriate choice of study design is
collectively referred to as observational studies. Often vital, it is not sufficient. The hallmark of good
research is the rigor with which it is conducted. A
these studies are the only practicable method of checklist of the key points in any study irrespec-
studying various problems, for example, studies of tive of the basic design is given in box 1.
aetiology, instances where a randomised controlled trial Every published study should contain suffi-
cient information to allow the reader to analyse
might be unethical, or if the condition to be studied is the data with reference to these key points.
rare. Cohort studies are used to study incidence, causes, In this article each of the three important
and prognosis. Because they measure events in observational research methods will be discussed
with emphasis on their strengths and weak-
chronological order they can be used to distinguish nesses. In so doing it should become apparent
between cause and effect. Cross sectional studies are why a given study used a particular research
used to determine prevalence. They are relatively quick method and which method might best answer a
particular clinical problem.
and easy but do not permit distinction between cause
and effect. Case controlled studies compare groups COHORT STUDIES
retrospectively. They seek to identify possible predictors These are the best method for determining the
of outcome and are useful for studying rare diseases or incidence and natural history of a condition. The
studies may be prospective or retrospective and
outcomes. They are often used to generate hypotheses sometimes two cohorts are compared.
that can then be studied via prospective cohort or other
studies. Prospective cohort studies
A group of people is chosen who do not have the
.......................................................................... outcome of interest (for example, myocardial inf-
arction). The investigator then measures a variety
of variables that might be relevant to the develop-

C
ohort, cross sectional, and case-control
studies are often referred to as observa- ment of the condition. Over a period of time the
tional studies because the investigator sim- people in the sample are observed to see whether
ply observes. No interventions are carried out by they develop the outcome of interest (that is,
the investigator. With the recent emphasis on evi- myocardial infarction).
dence based medicine and the formation of the In single cohort studies those people who do
Cochrane Database of randomised controlled not develop the outcome of interest are used as
trials, such studies have been somewhat glibly internal controls.
maligned. However, they remain important be- Where two cohorts are used, one group has
cause many questions can be efficiently answered been exposed to or treated with the agent of
by these methods and sometimes they are the interest and the other has not, thereby acting as
only methods available. an external control.
The objective of most clinical studies is to
determine one of the followingprevalence, inci- Retrospective cohort studies
dence, cause, prognosis, or effect of treatment; it These use data already collected for other
is therefore useful to remember which type of purposes. The methodology is the same but the
study is most commonly associated with each study is performed posthoc. The cohort is
objective (table 1) followed up retrospectively. The study period
may be many years but the time to complete the
study is only as long as it takes to collate and ana-
lyse the data.
Table 1
.......................
Objective Common design Advantages and disadvantages
Dr C J Mann, Department The use of cohorts is often mandatory as a
of Accident and Emergency Prevalence Cross sectional
Medicine, Taunton and randomised controlled trial may be unethical; for
Incidence Cohort
Somerset Hospital, Cause (in order of Cohort, case-control, cross example, you cannot deliberately expose people to
Taunton, Somerset, UK reliability) sectional cigarette smoke or asbestos. Thus research on risk
Prognosis Cohort factors relies heavily on cohort studies.
Correspondence to: Treatment effect Controlled trial
Dr C J Mann;
As cohort studies measure potential causes
[email protected] before the outcome has occurred the study can
....................... demonstrate that these causes preceded the

www.emjonline.com
 Downloaded from http://emj.bmj.com/ on December 14, 2016 - Published by group.bmj.com

Observational research methods 55

Box 1 100 A&E attenders with minor injuries for the outcome of
diabetes mellitus will probably produce only one patient with
Study purpose the outcome of interest. The efficiency of a prospective cohort
The aim of the study should be clearly stated. study increases as the incidence of any particular outcome
increases. Thus a study of patients with a diagnosis of deliber-
Sample ate self harm in the 12 months after initial presentation would
The sample should accurately reflect the population from be efficiently studied using a cohort design.
which it is drawn. Another problem with prospective cohort studies is the loss
The source of the sample should be stated. of some subjects to follow up. This can significantly affect the
The sampling method should be described and the sample outcome. Taking incidence analysis as an example (incidence
size should be justified. = cases/per period of time), it can be seen that the loss of a few
Entry criteria and exclusions should be stated and justified. cases will seriously affect the numerator and hence the calcu-
The number of patients lost to follow up should be stated and lated incidence. The rarer the condition the more significant
explanations given.
this effect.
Control group Retrospective studies are much cheaper as the data have
The control group should be easily identifiable. already been collected. One advantage of such a study design
The source of the controls should be explainedare they from is the lack of bias because the outcome of current interest was
the same population as the sample? not the original reason for the data to be collected. However,
Are the controls matched or randomisedto minimise bias because the cohort was originally constructed for another
and confounding. purpose it is unlikely that all the relevant information will
Quality of measurements and outcomes have been rigorously collected.
Retrospective cohorts also suffer the disadvantage that
Validityare the measurements used regarded as valid by people with the outcome of interest are more likely to remem-
other investigators?
ber certain antecedents, or exaggerate or minimise what they
Reproducibilitycan the results be repeated or is there a rea-
son to suspect they may be a one off? now consider to be risk factors (recall bias).
Blindedwere the investigators or subjects aware of their Where two cohorts are compared one will have been
subject/control allocation? exposed to the agent of interest and one will not. The major
Quality controlhas the methodology been rigorously disadvantage is the inability to control for all other factors that
adhered to? might differ between the two groups. These factors are known
as confounding variables.
Completeness
A confounding variable is independently associated with
Compliancedid all patients comply with the study? both the variable of interest and the outcome of interest. For
Drop outshow many failed to complete the study? example, lung cancer (outcome) is less common in people
Deaths with asthma (variable). However, it is unlikely that asthma in
Missing datahow much are unavailable and why?
itself confers any protection against lung cancer. It is more
Distorting influences probable that the incidence of lung cancer is lower in people
Extraneous treatmentsother interventions that may have with asthma because fewer asthmatics smoke cigarettes (con-
affected some but not all of the subjects. founding variable). There are a virtually infinite number of
Confounding factorsAre there other variables that might potential confounding variables that, however unlikely, could
influence the results? just explain the result. In the past this has been used to sug-
Appropriate analysisHave appropriate statistical tests been gest that there is a genetic influence that makes people want
used? to smoke and also predisposes them to cancer.
Validity The only way to eliminate all possibility of a confounding
variable is via a prospective randomised controlled study. In
All studies should be internally valid. That is, the conclusions
this type of study each type of exposure is assigned by chance
can be logically drawn from the results produced by an
appropriate methodology. For a study to be regarded as valid and so confounding variables should be present in equal
it must be shown that it has indeed demonstrated what it says numbers in both groups.
it has. A study that is not internally valid should not be Finally, problems can arise as a result of bias. Bias can occur
published because the findings cannot be accepted. in any research and reflects the potential that the sample
The question of external validity relates to the value of the studied is not representative of the population it was drawn
results of the study to other populationsthat is, the generalis- from and/or the population at large. A classic example is using
ability of the results. For example, a study showing that 80% employed people, as employment is itself associated with gen-
of the Swedish population has blond hair, might be used to erally better health than unemployed people. Similarly people
make a sensible prediction of the incidence of blond hair in who respond to questionnaires tend to be fitter and more
other Scandinavian countries, but would be invalid if applied motivated than those who do not. People attending A&E
to most other populations.
departments should not be presumed to be representative of
the population at large.
outcome, thereby avoiding the debate as to which is cause and How to run a cohort study
which is effect. If the data are readily available then a retrospective design is
A further advantage is that a single study can examine the quickest method. If high quality, reliable data are not
various outcome variables. For example, cohort studies of available a prospective study will be required.
smokers can simultaneously look at deaths from lung, cardio- The first step is the definition of the sample group. Each
vascular, and cerebrovascular disease. This contrasts with subject must have the potential to develop the outcome of
case-control studies as they assess only one outcome variable interest (that is, circumcised men should not be included in a
(that is, whatever outcome the cases have entered the study cohort designed to study paraphimosis). Furthermore, the
with). sample population must be representative of the general
Cohorts permit calculation of the effect of each variable on population if the study is primarily looking at the incidence
the probability of developing the outcome of interest (relative and natural history of the condition (descriptive).
risk). However, where a certain outcome is rare then a If however the aim is to analyse the relation between
prospective cohort study is inefficient. For example, studying predictor variables and outcomes (analytical) then the sample

www.emjonline.com
 Downloaded from http://emj.bmj.com/ on December 14, 2016 - Published by group.bmj.com

56 Mann

Figure 1 Study design for cohort

studies.

Key points Table 2

Cohort studies Questionnaire Interview
Cohort studies describe incidence or natural history. Cheap Expensive
They analyse predictors (risk factors) thereby enabling cal- Low response rate High response rate
culation of relative risk. Large sample size Smaller sample size
Cohort studies measure events in temporal sequence
thereby distinguishing causes from effects.
Retrospective cohorts where available are cheaper and
quicker.
Confounding variables are the major problem in analysing
cohort studies.
Subject selection and loss to follow up is a major potential
cause of bias.

should contain as many patients likely to develop the outcome

as possible, otherwise much time and expense will be spent
collecting information of little value.
Each variable studied must be accurately measured.
Variables that are relatively fixed, for example, height need
only be recorded once. Where change is more probable, for Figure 2 Study design for cross sectional studies
example, drug misuse or weight, repeated measurements will
be required.
To minimise the potential for missing a confounding because it influences considerably the likelihood of any
variable all probable relevant variables should be measured. If particular diagnosis and the predictive value of any investiga-
this is not done the study conclusions can be readily criticised. tion. For example, knowing that ascending cholangitis in chil-
All patients entered into the study should also be followed up dren is very rare enables the clinician to look for other causes
for the duration of the study. Losses can significantly affect the of abdominal pain in this patient population.
validity of the results. To minimise this as much information Cross sectional studies are also used to infer causation.
about the patient (name, address, telephone, GP, etc) needs to At one point in time the subjects are assessed to determine
be recorded as soon as the patient is entered into the study. whether they were exposed to the relevant agent and whether
Regular contact should be made; it is hardly surprising if the they have the outcome of interest. Some of the subjects will
subjects have moved or lost interest and become lost to follow not have been exposed nor have the outcome of interest. This
up if they are only contacted at 10 year intervals! clearly distinguishes this type of study from the other
Beware, follow up is usually easier in people who have been observational studies (cohort and case controlled) where ref-
exposed to the agent of interest and this may lead to bias. erence to either exposure and/or outcome is made.
The advantage of such studies is that subjects are neither
Examples deliberately exposed, treated, or not treated and hence there
There are many famous examples of Cohort studies including are seldom ethical difficulties. Only one group is used, data are
the Framingham heart study,2 the UK study of doctors who collected only once and multiple outcomes can be studied;
smoke3 and Professor Neville Butlers studies on British thus this type of study is relatively cheap.
children born in 1958.4 A recent example of a prospective Many cross sectional studies are done using questionnaires.
cohort study by Davey Smith et al was published in the BMJ5 Alternatively each of the subjects may be interviewed. Table 2
and a retrospective cohort design was used to assess the use of lists the advantages and disadvantages of each.
A&E departments by people with diabetes.6 Any study with a low response rate can be criticised because
it can miss significant differences in the responders and non-
CROSS SECTIONAL STUDIES responders. At its most extreme all the non-responders could
These are primarily used to determine prevalence. Prevalence be dead! Strenuous efforts must be made to maximise the
equals the number of cases in a population at a given point in numbers who do respond. The use of volunteers is also prob-
time. All the measurements on each person are made at one lematic because they too are unlikely to be representative of
point in time. Prevalence is vitally important to the clinician the general population. A good way to produce a valid sample

www.emjonline.com
 Downloaded from http://emj.bmj.com/ on December 14, 2016 - Published by group.bmj.com

Observational research methods 57

would be to randomly select people from the electoral role and

invite them to complete a questionnaire. In this way the Key points
response rate is known and non-responders can be identified. Cross sectional studies
However, the electoral role itself is not an entirely accurate
Cross sectional studies are the best way to determine
reflection of the general population. A census is another prevalence
example of a cross sectional study. Are relatively quick
Market research organisations often use cross sectional Can study multiple outcomes
studies (for example, opinion polls). This entails a system of Do not themselves differentiate between cause and effect or
quotas to ensure the sample is representative of the age, sex, the sequence of events
and social class structure of the population being studied.
However, to be commercially viable they are convenience
samplesonly people available can be questioned. This tech-
nique is insufficiently rigorous to be used for medical research.

How to run a cross sectional study

Formulate the research question(s) and choose the sample
population. Then decide what variables of the study popula-
tion are relevant to the research question. A method for con-
tacting sample subjects must be devised and then imple-
mented. In this way the data are collected and can then be
analysed

Advantages and disadvantages Figure 3 Study design for case-control studies.

The most important advantage of cross sectional studies is
that in general they are quick and cheap. As there is no follow CASE-CONTROL STUDIES
up, less resources are required to run the study. In contrast with cohort and cross sectional studies, case-
Cross sectional studies are the best way to determine preva- control studies are usually retrospective. People with the out-
lence and are useful at identifying associations that can then come of interest are matched with a control group who do not.
be more rigorously studied using a cohort study or ran- Retrospectively the researcher determines which individuals
domised controlled study. were exposed to the agent or treatment or the prevalence of a
The most important problem with this type of study is dif- variable in each of the study groups. Where the outcome is
ferentiating cause and effect from simple association. For rare, case-control studies may be the only feasible approach.
example, a study finding an association between low CD4 As some of the subjects have been deliberately chosen
counts and HIV infection does not demonstrate whether HIV because they have the disease in question case-control studies
infection lowers CD4 levels or low CD4 levels predispose to are much more cost efficient than cohort and cross sectional
HIV infection. Moreover, male homosexuality is associated studiesthat is, a higher percentage of cases per study.
with both but causes neither. (Another example of a Case-control studies determine the relative importance of a
confounding variable). predictor variable in relation to the presence or absence of the
Often there are a number of plausible explanations. For disease. Case-control studies are retrospective and cannot
example, if a study shows a negative relation between height therefore be used to calculate the relative risk; this a prospec-
and age it could be concluded that people lose height as they tive cohort study. Case-control studies can however be used to
get older, younger generations are getting taller, or that tall calculate odds ratios, which in turn, usually approximate to
people have a reduced life expectancy when compared with the relative risk.
short people. Cross sectional studies do not provide an expla-
nation for their findings. How to run a case-control study
Rare conditions cannot efficiently be studied using cross Decide on the research question to be answered. Formulate an
sectional studies because even in large samples there may be hypothesis and then decide what will be measured and how.
no one with the disease. In this situation it is better to study a Specify the characteristics of the study group and decide how
cross sectional sample of patients who already have the to construct a valid control group. Then compare the
disease (a case series). In this way it was found in 1983 that of exposure of the two groups to each variable.
1000 patients with AIDS, 727 were homosexual or bisexual
men and 236 were intrvenous drug abusers.6 The conclusion Advantages and disadvantages
that individuals in these two groups had a higher relative risk When conditions are uncommon, case-control studies gener-
was inescapable. The natural history of HIV infection was then ate a lot of information from relatively few subjects. When
studied using cohort studies and efficacy of treatments via there is a long latent period between an exposure and the dis-
case controlled studies and randomised clinical trials. ease, case-control studies are the only feasible option.
Consider the practicalities of a cohort study or cross sectional
Examples study in the assessment of new variant CJD and possible aeti-
An example of a cross sectional study was the prevalence ologies. With less than 300 confirmed cases a cross sectional
study of skull fractures in children admitted to hospital in study would need about 200 000 subjects to include one
Edinburgh from 1983 to 1989.7 Note that although the study symptomatic patient. Given a postulated latency of 10 to 30
period was seven years it was not a longitudinal or cohort years a cohort study would require both a vast sample size and
study because information about each subject was recorded at take a generation to complete.
a single point in time. In case-control studies comparatively few subjects are
A questionnaire based cross sectional study explored the required so more resources are available for studying each. In
relation between A&E attendance and alcohol consumption in consequence a huge number of variables can be considered.
elderly persons.9 This type of study is therefore useful for generating
A recent example can be found in the BMJ, in which the hypotheses that can then be tested using other types of study.
prevalence of serious eye disease in a London population was This flexibility of the variables studied comes at the expense
evaluated.10 of the restricted outcomes studied. The only outcome is the

www.emjonline.com
 Downloaded from http://emj.bmj.com/ on December 14, 2016 - Published by group.bmj.com

58 Mann

presence or absence of the disease or whatever criteria was

chosen to select the cases. Key points
The major problems with case-control studies are the Case-control studies
familiar ones of confounding variables (see above) and bias.
Bias may take two major forms. Case-control studies are simple to organise
Retrospectively compare two groups
Sampling bias Aim to identify predictors of an outcome
The patients with the disease may be a biased sample (for Permit assessment of the influence of predictors on outcome
via calculation of an odds ratio
example, patients referred to a teaching hospital) or the con-
Useful for hypothesis generation
trols may be biased (for example, volunteers, different ages, Can only look at one outcome
sex or socioeconomic group). Bias is an major problem
Observation and recall bias
As the study assesses predictor variables retrospectively there
is great potential for a biased assessment of their presence and Blinding cases to their case or control status is usually
significance by the patient or the investigator, or both. impracticable as they already know that they have a disease or
illness. Similarly observers can hardly be blinded to the pres-
Overcoming sampling bias ence of physical signs, for example, cyanosis or dyspnoea.
Ideally the cases studied should be a random sample of all the As a result of the problems of matching, bias and
patients with the disease. This is not only very difficult but in confounding, case-control studies, are often flawed. They are
many instances is impossible because many cases may not however useful for generating hypotheses. These hypotheses
have been diagnosed or have been misdiagnosed. For example, can then be tested more rigorously by other methods
many cases of non-insulin dependent diabetes will not have randomised controlled trials or cohort studies.
sought medical attention and therefore be undiagnosed. Con-
versely many psychiatric diseases may be differently labelled Examples
in different countries and even by different doctors in the Case-control studies are very common. They are particularly
same country. As a result they will be misdiagnosed for the useful for studying infrequent events, for example, cot death,
purposes of the study. However, in reality you are often left survival from out of hospital cardiac arrest, and toxicological
studying a sample of those patients who it is possible to emergencies.
recruit. Selecting the controls is often a more difficult problem. A recent example was the study of atrial fibrillation in mid-
To enable the controls to represent the same population as dle aged men during exercise.11
the cases, one of four techniques may be used.
(1) A convenience samplesampled in the same way as the USING DATABASES FOR RESEARCH (SECONDARY
cases, for example, attending the same outpatient depart- DATA)
ment. While this is certainly convenient it may reduce the Pre-existing databases provide an excellent and convenient
external validity of the study. source of data. There are a host of such databases and the
(2) Matchingthe controls may be a matched or un- increasing archiving of information on computers means that
matched random sample from the unaffected population. this is an enlarging area for obtaining data. Table 3 lists some
Again the problems of controlling for unknown influences is
common examples of potentially useful databases.
present but if the controls are too closely matched they may
Such databases enable vast numbers of people to be entered
not be representative of the general population. Over match-
into a study prospectively or retrospectively. They can be used
ing may cause the true difference to be underestimated.
to construct a cohort, to produce a sample for a cross sectional
The advantage of matching is that it allows a smaller sam-
ple size for any given effect to be statistically significant. study, or to identify people with certain conditions or
(3) Using two or more control groups. If the study demon- outcomes and produce a sample for a case controlled study. A
strates a significant difference between the patients with the recent study used census data from 11 countries to look at the
outcome of interest and those without, even when the latter relation between social class and mortality in middle aged
have been sampled in a number of different ways (for exam- men.12
ple, outpatients, in patients, GP patients) then the conclusion
is more robust. Advantages and disadvantages
(4) Using a population based sample for both cases and These type of data are ordinarily collected by people other than
controls. It is possible to take a random sample of all the the researcher and independently of any specific hypothesis.
patients with a particular disease from specific registers. The The opportunity for observer bias is thus diminished. The use
control group can then be constructed by selecting age and sex of previously collected data is efficient and comparatively
matched people randomly selected from the same population inexpensive and moreover the data are collected in a very
as the area covered by the disease register. standardised way, permitting comparisons over time and
between different countries. However, because the data are
Overcoming observation and recall bias collected for other purposes it may not be ideally suited to the
Overcoming retrospective recall bias can be achieved by using testing of the current hypothesis, additionally it may be
data recorded, for other purposes, before the outcome had incomplete. This may result in sampling bias. For example, the
occurred and therefore before the study had started. The suc- electoral roll depends upon registration by each individual.
cess of this strategy is limited by the availability and reliability Many homeless, mentally ill, and chronically sick people will
of the data collected. Another technique is blinding where not be registered. Similarly the notification of certain commu-
neither the subject nor the observer know if they are a case or nicable diseases is a statutory responsibility for doctors in the
control subject. Nor are they aware of the study hypothesis. In UK: while it is probable that most cases of cholera are reported
practice this is often difficult or impossible and only partial it is highly unlikely that most cases of food poisoning are.
blinding is practicable. It is usually possible to blind the sub-
jects and observers to the study hypothesis by asking spurious Causes and associations
questions. Observers can also be easily blinded to the case or Because observational studies are not experiments (as are
control status of the patient where the relevant observation is randomised controlled trials) it is difficult to control many
not of the patient themselves but a laboratory test or external variables. In consequence when faced with a clear
radiograph. and significant association between some form of illness or

www.emjonline.com
 Downloaded from http://emj.bmj.com/ on December 14, 2016 - Published by group.bmj.com

Observational research methods 59

Table 3
Demographic and lifestyle data Census, General Household Survey, social trends, annual reports
Morbidity data GP morbidity statistics, communicable disease surveillance, hospital
inpatient inquiry, hospital activity analysis, cancer registration
Mortality data OPCS mortality statistics
Health services data Immunisation data, screening levels, district and annual reports,
confidential inquiry reports.
Specific databases OPCS survey of disability, OPCS longitudinal study, abortion data,
congenital abnormalities, workmens compensation data
Other Social Security statistics, commissions of inquiry

Table 4
The strength of the association The greater the magnitude of the association the greater the likelihood
that it is causal
Consistency If the association is observed at different times, places and by different
researchers it is more credible
Specificity The more specific the disease and the groups of people affected the
greater the likelihood of causality
Temporal relation Does the suspected cause precede the disease
Biological gradient Is there a doseresponse type relation
Biological plausibility Do the findings fit with plausible biological and disease mechanisms
Coherence of the evidence The cause and effect interpretation should fit with other known facts
regarding the natural history and biology of the disease
Prevention Does avoidance or removal of the cause decrease the incidence of the
disease
Reasoning by analogy Does the evidence mirror or match another cause and effect relation

cause of death and some environmental influence a judge- Subjects are selected without regard to the outcome of
ment has to be made as to whether this is a causal link or sim- interest
ply an association. Table 4 outlines the points to be considered Less expensive
when making this judgement.13
They are the best way to determine prevalence
None of these judgements can provide indisputable
evidence of cause and effect, but taken together they do Quick
permit the investigator to answer the fundamental questions The principal summary statistic of cross sectional studies is
is there any other way to explain the available evidence? and the odds ratio
is there any other more likely than cause and effect? Weaker evidence of causality than cohort studies
Inaccurate when studying rare conditions
SUMMARY
Qualitative studies can produce high quality information but Case-control studies look back at what has happened to each
all such studies can be influenced by known and unknown subject
confounding variables. Appropriate use of observational stud- Subjects are selected specifically on the basis of the outcome
ies permits investigation of prevalence, incidence, associa- of interest
tions, causes, and outcomes. Where there is little evidence on Cheap
a subject they are cost effective ways of producing and inves- Efficient (small sample sizes)
tigating hypotheses before larger and more expensive study
designs are embarked upon. In addition they are often the Produce odds ratios that approximate to relative risks for
only realistic choice of research methodology, particularly each variable studied
where a randomised controlled trial would be impractical or Prone to sampling bias and retrospective analysis bias
unethical. Only one outcome is studied
Cohort studies look forwards in time by following up each
subject GLOSSARY OF TERMS
Subjects are selected before the outcome of interest is Bias
observed The inclusion of subjects or methods such that the results
obtained are not truly representative of the population from
They establish the sequence of events
which it is drawn
Numerous outcomes can be studied
They are the best way to establish the incidence of a disease Blinding
The process by which the researcher and or the subject is
They are a good way to determine causes of diseases ignorant of which intervention or exposure has occurred.
The principal summary statistic of cohort studies is the
relative risk ratio Cochrane database
An international collaborative project collating peer reviewed
If prospective, they are expensive and often take a long time
prospective randomised clinical trials.
for sufficient outcome events to occur to produce meaning-
ful results Cohort
Cross sectional studies look at each subject at one point in Is a component of a population identified so that one or more
time only characteristic can be studied as it ages through time.

www.emjonline.com
 Downloaded from http://emj.bmj.com/ on December 14, 2016 - Published by group.bmj.com

60 Mann

Confounding variable Relative risk

A variable that is associated with both the exposure and out- This is the ratio of the probability of developing the condition
come of interest that is not the variable being studied. if exposed to a certain variable compared with the probability
if not exposed.
Control group
A group of people without the condition of interest, or unex-
posed to or not treated with the agent of interest. Response rate
The proportion of subjects who respond to either a treatment
False positive or a questionnaire.
A test result that suggests that the subject has a specific
disease or condition when in fact the subject does not. Risk factor
A variable associated with a specific disease or outcome.
Incidence
Is a rate and therefore is always related either explicitly or by
implication to a time period. With regard to disease it can be Validityinternal
defined as the number of new cases that develop during a The rigour with which a study has been designed and
specified time interval. executedthat is, can the conclusion be relied upon?

Latency Validityexternal
A period of time between exposure to an agent and the devel- The usefulness of the findings of a study with respect to other
opment of symptoms, signs, or other evidence of changes populations.
associated with that exposure.

Matching Variable
The process by which each case is matched with one or more A value or quality that can vary between subjects and/or over
controls, which have been deliberately chosen to be as similar time
as the test subjects in all regards other than the variable being
studied.

Observational study REFERENCES

A study in which no intervention is made (in contrast with an
1 Fowkes F, Fulton P. Critical appraisal of published research: introductory
experimental study). Such studies provide estimates and guidelines. BMJ 1991;302:113640.
examine associations of events in their natural settings with- 2 Lerner DJ, Kannel WB. Patterns of coronary heart disease morbidity and
out recourse to experimental intervention. mortality in the sexes: a 26 year follow-up of the Framingham population.
Am Heart J 1986;111:38390.
3 Doll R, Peto H. Mortality in relation to smoking. 40 years observation on
Odds ratio female British doctors. BMJ 1989;208:96773.
The ratio of the probability of an event occurring to the prob- 4 Alberman ED, Butler NR, Sheridan MD. Visual acuity of a national
ability of non-occurrence. In a clinical setting this would be sample (1958 cohort) at 7 years. Dev Med Child Neurol 1971;13:914.
5 Davey Smith G, Hart C, Blane D, et al. Adverse socioeconomic
equivalent to the odds of a condition occurring in the exposed conditions in childhood and cause specific mortality: prospective
group divided by the odds of it occurring in the non-exposed observational study. BMJ 1998;316: 16315.
group. 6 Goyder EC, Goodacre SW, Botha JL, et al. How do individuals with
diabetes use the accident and emergency department? J Accid Emerg
Med 1997;14:3714.
Prevalence 7 Jaffe HW, Bregman DJ, Selik RM. Acquired immune deficiency in the
Is not defined by a time interval and is therefore not a rate. It US: the first 1000 cases. J Inf Dis 1983;148:33945.
may be defined as the number of cases of a disease that exist 8 Johnstone AJ, Zuberi SH, Scobie WH. Skull fractures in children: a
population study. J Accid Emerg Med1996;13:3869.
in a defined population at a specified point in time. 9 van der Pol V, Rodgers H, Aitken P, et al. Does alcohol contribute to
accident and emergency department attendance in elderly people? J
Randomised controlled trial Accid Emerg Med1996;13:25860.
Subjects are assigned by statistically randomised methods to 10 Reidy A, Minassian DC, Vafadis G, et al. BMJ 1998;316:16437.
11 Karjaleinen, Kujala U, Kaprio J, et al. BMJ 1998;316:17845.
two or more groups. In doing so it is assumed that all variables
12 Kunst A, Groenhof F, Mackenbach J. BMJ 1998;316:163642.
other than the proposed intervention are evenly distributed 13 Hill AB, Hill ID. Bradford Hills principles of medical statistics. 12th edn.
between the groups. In this way bias is minimised. London: Edward Arnold, 1991.

www.emjonline.com
 Downloaded from http://emj.bmj.com/ on December 14, 2016 - Published by group.bmj.com

Observational research methods. Research

design II: cohort, cross sectional, and
case-control studies
C J Mann

Emerg Med J 2003 20: 54-60

doi: 10.1136/emj.20.1.54

Updated information and services can be found at:

http://emj.bmj.com/content/20/1/54

These include:

References This article cites 11 articles, 8 of which you can access for free at:
http://emj.bmj.com/content/20/1/54#BIBL

Email alerting Receive free email alerts when new articles cite this article. Sign up in the
service box at the top right corner of the online article.

Notes

To request permissions go to:

http://group.bmj.com/group/rights-licensing/permissions

To order reprints go to:

http://journals.bmj.com/cgi/reprintform

To subscribe to BMJ go to:

http://group.bmj.com/subscribe/