# 05-5 Population health and epidemiology

# 5 Population health and epidemiology

Population health 
and epidemiology
H Campbell
DA McAllister
Global burden of disease and underlying risk factors 92
Life expectancy 92
Global causes of death and disability 92
Risk factors underlying disease 93
Social determinants of health 93
The hierarchy of systems – from molecules to ecologies 93
The life course 93
Preventive medicine 93
Principles of screening 94
Epidemiology 95
Understanding causes and effect 95
Health data/informatics 97


92 • POPULATION HEALTH AND EPIDEMIOLOGY
the population to older ages, and this is placing an increasing 
burden on health systems. For a few conditions (e.g. HIV/AIDS, 
diabetes mellitus and chronic kidney disease), age-standardised 
death rates continue to rise. Within this overall pattern, significant 
regional variations exist: for example, communicable, maternal, 
neonatal and nutritional causes still account for about two-thirds 
of premature mortality in sub-Saharan Africa.
GBD also provides estimates of disability from disease (Box 
5.2). This has raised awareness of the importance of conditions 
like depression, low back and neck pain, and asthma, which 
account for a relatively large disease burden but relatively few 
deaths. This, in turn, has resulted in greater health policy priority 
being given to these conditions. Since the policy focus in national 
health systems is increasingly on keeping people healthy rather 
than only on reducing premature deaths, it is important to have 
measures of these health outcomes.
It is also essential to recognise that, although these estimates 
represent the best overall picture of burden of disease, they 
are based on imperfect data. Nevertheless, the quality of data 
underlying the estimates and the modelling processes are 
The UK Faculty of Public Health defines public health as ‘the 
science and art of promoting and protecting health and well-being, 
preventing ill-health and prolonging life through the organised 
efforts of society’. This definition recognises that there is a 
collective responsibility for the health of the population that 
requires partnerships between government, health services 
and others to promote and protect health and prevent disease. 
Population health has been defined as ‘the health outcomes 
of a group of individuals, including the distribution of such 
outcomes within the group’. Medical doctors can play a role 
in all these efforts to improve health both through their clinical 
work and through their support of broader actions to improve 
public health.
Global burden of disease and 
underlying risk factors
The Global Burden of Disease (GBD) exercise was initiated by 
the World Bank in 1992, with first estimates appearing in 1993. 
Regular updated figures have been published since, together with 
projections of future disease burden. The aim was to produce 
reliable and internally consistent estimates of disease burden 
for all diseases and injuries, and to assess their physiological, 
behavioural and social risk factors, so that this information 
could be made available to health workers, researchers and 
policy-makers.
The GBD exercise adopted the metric ‘disability adjusted 
life year’ (DALY) to describe population health. This combines 
information about premature mortality in a population (measured 
as Years of Life Lost from an ‘expected’ life expectancy) and years 
of life lived with disability (Years of Life lived with Disability, which 
is weighted by a severity factor). The International Classification 
of Disease (ICD) rules, which assign one cause to each death, 
are followed. All estimates are presented by age and sex groups 
and by regions of the world. Many countries now also report 
their own national burden of disease data.
Life expectancy
Global life expectancy at birth increased from 61.7 years in 1980 
to 71.8 years in 2015, an increase of 0.29 years per calendar 
year. This change is due to a substantial fall in child mortality 
(mainly caused by common infections), partly offset by rises in 
mortality from adult conditions such as diabetes and chronic 
kidney disease. Some areas have not shown these increases 
in life expectancy in men, often due to war and interpersonal 
violence.
Global causes of death and disability
Box 5.1 shows a ranked list of the major causes of global 
premature deaths in 2015. Communicable, maternal, neonatal 
and nutritional causes accounted for about one-quarter of deaths 
worldwide, down from about one-third in 1990. In contrast, deaths 
from non-communicable diseases are increasing in importance 
and now account for about two-thirds of all deaths globally, 
including about 13 million from ischaemic heart disease and 
stroke, and about 8 million from cancer. The age-standardised 
death rates for most diseases globally are falling. However, despite 
this, the numbers of deaths from many diseases are rising due 
to global population growth and the change in age structure of 
5.2 Global disability: top 15 ranked causes, 20151,2
1. Lower back and neck pain (1)
2. Sense organ diseases (3)
3. Depressive disorders (4)
4. Iron deficiency anaemia (2)
5. Skin diseases (5)
6. Diabetes (9)
7. Migraine (6)
8. Other musculoskeletal conditions3 (7)
9. Anxiety disorders (8)
10. Oral disorders (11)
11. Asthma (10)
12. Schizophrenia (13)
13. Osteoarthritis (19)
14. Chronic obstructive pulmonary disease (14)
15. Falls (12)
1By Years of Life lived with Disability (YLD). 2Rank in 1990 is shown in brackets. 
3Not otherwise classified as specific conditions such as osteoarthritis.
5.1 Global premature mortality: top 15 ranked 
causes, 20151,2
1. Ischaemic heart disease (4)
2. Cerebrovascular disease (5)
3. Lower respiratory infections (1)
4. Neonatal preterm birth complications (2)
5. Diarrhoeal diseases (3)
6. Neonatal encephalopathy (6)
7. HIV/AIDS (29)
8. Road injuries (10)
9. Malaria (7)
10. Chronic obstructive pulmonary disease (12)
11. Congenital anomalies (9)
12. Tuberculosis (11)
13. Lung cancer3 (20)
14. Self-harm (16)
15. Diabetes (> 30)
1By Years of Life Lost (YLL). 2Rank in 1990 is shown in brackets. 3‘All cancers 
combined’ would rank in the top three causes.


Social determinants of health • 93

to higher risk of hypertension and type 2 diabetes in young 
adults, and of cardiovascular disease in middle age. It has been 
suggested that under-nutrition during middle to late gestation 
permanently ‘programs’ cardiovascular and metabolic responses. 
This ‘life course’ perspective highlights the cumulative effect on 
health of exposures to illness, adverse environmental conditions 
and behaviours that damage health.
Preventive medicine
The complexity of interactions between physical, social and 
economic determinants of health means successful prevention 
is often difficult. Moreover, the life-course perspective illustrates 
that it may be necessary to intervene early in life or even before 
birth, to prevent important disease later. Successful prevention 
is likely to require many interventions across the life course 
and at several levels in the hierarchy of systems. The examples 
below illustrate this.
improving over time and provide an increasingly robust basis for 
evidence-based health planning and priority setting.
Risk factors underlying disease
Box 5.3 shows a ranked list of the main risk factors that underlay 
GBD in 2015 and how this ranking has changed in recent years.
Social determinants of health
Health emerges from a highly complex interaction between a 
person’s genetic background and environmental factors (aspects 
of the physical, biological (microbes), built and social environments, 
and also distant influences such as the global ecosystem; Fig. 5.1).
The hierarchy of systems – from molecules 
to ecologies
Influences on health exist at many levels and extend beyond 
the individual to include the family, community, population and 
ecology. Box 5.4 shows an example of this for determinants 
of coronary heart disease and demonstrates the importance of 
considering not only the disease process in a patient but also its 
context. Health care is not the only determinant – and is usually 
not the major determinant – of health status in the population. 
The concept of ‘global health’ recognises the global dimension 
of health problems, whether these be emerging or pandemic 
infections or global economic influences on health.
The life course
The determinants of health operate over the whole lifespan. Values 
and behaviours acquired during childhood and adolescence have 
a profound influence on educational outcomes, job prospects and 
risk of disease. These can have a strong influence, for example, 
on whether a young person takes up damaging behaviour like 
smoking, risky sexual activity and drug misuse. Influences on 
health can operate even before birth. Low birth weight can lead 
Fig. 5.1 Hierarchy of systems that influence population health. 
Adapted from an original model by Whitehead M, Dahlgren G. What can be 
done about inequalities in health? Lancet 1991; 338:1059–1063.
Macro-economy,
politics, culture,
global forces
Other
neighbourhoods,
other regions
People
Age, sex and
hereditary
factors
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
L
i
f
e
s
t
y
l
e
 
 
 
 
 
 
 
 
 
 
 
W
o
r
k
,
 
p
l
a
y
D
i
e
t
,
 
p
h
y
s
i
c
a
l
 
a
c
t
i
v
i
t
y
 
S
o
c
i
a
l 
c
a
p
it
a
l 
 
 
 
 
 
 
C
o
m
m
u
n
i
t
y
 
 
 
 
 
 
 
 
 
 
N
e
t
w
o
r
k
s
W
e
a
lt
h
 
c
r
e
a
ti
o
n
 
 
 
 
 
 
 
 
 
L
o
c
a
l 
e
c
o
n
o
m
y
 
 
 
 
 
 
 
 
 
 
 
 
M
a
r
k
e
t
s
B
ui
ld
in
g
s,
 
pl
a
c
e
s 
 
 
 
 
 
 
 
 
 
 
 
 
 
E
n
vi
r
o
n
m
e
n
t 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
A
ir
, 
w
at
e
r,
 l
a
n
d
Cl
i
m
at
e 
c
h
a
n
g
e 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Gl
o
b
al
 e
c
o
s
y
st
e
m
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Bi
o
di
v
er
si
ty
5.3 Global risk factors: top 15 ranked causes, 20151–3
1. High blood pressure (3)
2. Smoking/second-hand smoke exposure (5)
3. High fasting blood glucose (10)
4. High body mass index (13)
5. Childhood underweight (1)
6. Ambient particulate matter pollution (6)
7. High total cholesterol (12)
8. Household air pollution (4)
9. Alcohol use (11)
10. High sodium intake (14)
11. Low wholegrain intake (15)
12. Unsafe sex (20)
13. Low fruit intake (16)
14. Unsafe water (2)
15. Low glomerular filtration rate (21)
1By percentage of burden of disease they cause. 2Rank in 1990 is shown in 
brackets. 3All dietary risk factors and physical inactivity combined accounted for 
10% of global burden of disease. Low physical activity was ranked 21, iron 
deficiency 16 and suboptimal breastfeeding 22 in 2015.
5.4 ‘Hierarchy of systems’ applied to 
ischaemic heart disease
Level in the 
hierarchy
Example of effect
Molecular
ApoB mutation causing hypercholesterolaemia
Cellular
Macrophage foam cells accumulate in vessel wall
Tissue
Atheroma and thrombosis of coronary artery
Organ
Ischaemia and infarction of myocardium
System
Cardiac failure
Person
Limited exercise capacity, impact on employment
Family
Passive smoking, diet
Community
Shops and leisure opportunities
Population
Prevalence of obesity
Society
Policies on smoking, screening for risk factors
Ecology
Agriculture influencing fat content in diet


94 • POPULATION HEALTH AND EPIDEMIOLOGY
of obesity, therefore, we not only need to help those who are 
already obese but also develop strategies that impact on the 
whole population and reverse the obesogenic environment.
Poverty and affluence
The adverse health and social consequences of poverty are well 
documented: high birth rates, high death rates and short life 
expectancy. Typically, with industrialisation, the pattern changes: 
low birth rates, low death rates and longer life expectancy. 
Instead of infections, chronic conditions such as heart disease 
dominate in an older population. Adverse health consequences 
of excessive affluence are also becoming apparent. Despite 
experiencing sustained economic growth for the last 50 years, 
people in many industrialised countries are not growing any 
happier and the litany of socioeconomic problems – crime, 
congestion, inequality – persists.
Many countries are now experiencing a ‘double burden’. They 
have large populations still living in poverty who are suffering from 
problems such as diarrhoea and malnutrition, alongside affluent 
populations (often in cities) who suffer from chronic illness such 
as diabetes and heart disease.
Atmospheric pollution
Emissions from industry, power plants and motor vehicles of 
sulphur oxides, nitrogen oxides, respirable particles and metals are 
severely polluting cities and towns in Asia, Africa, Latin America 
and Eastern Europe. Burning of fossil and biomass fuels, with 
production of short-lived carbon pollutants (SLCPs – methane, 
ozone, black carbon and hydrofluorocarbons), contributes 
to increased death rates from respiratory and cardiovascular 
disease in vulnerable adults, such as those with established 
respiratory disease and the elderly, while children experience 
an increase in bronchitic symptoms. Developing countries also 
suffer high rates of respiratory disease as a result of indoor 
pollution caused mainly by heating and cooking using solid 
biomass fuels.
Climate change and global warming
Climate change is arguably the world’s most important 
environmental health issue. A combination of habitat destruction 
and increased production of carbon dioxide and SLCPs, caused 
primarily by human activity, seems to be the main cause. 
The temperature of the globe is rising, and if current trends 
continue, warming by 4°C is predicted by 2050. The climate 
is being affected, putting millions of people at risk of rising sea 
levels, flooding, droughts and failed crops These have already 
claimed millions of lives during the past 20 years and have 
adversely affected the lives of many more. The economic costs 
of property damage and the impact on agriculture, food supplies 
and prosperity have also been substantial. Global warming 
will also include changes in the geographical range of some 
vector-borne infectious diseases. Currently, politicians cannot 
agree an effective framework of actions to tackle the problem, 
but reducing emissions of CO2 and SLCPs is essential.
Principles of screening
Screening is the application of a test to a large number of 
asymptomatic people with the aim of reducing morbidity or 
mortality from a disease. The World Health Organisation (WHO) 
Alcohol
Alcohol use is an increasingly important risk factor underlying GBD 
(see Box 5.3). Reasons for increasing rates of alcohol-related 
harm vary by place and time but include the falling price of 
alcohol (in real terms), increased availability and cultural change 
fostering higher levels of consumption. Public, professional and 
governmental concern has now led to a minimum price being 
charged for a unit of alcohol, tightening of licensing regulations 
and curtailment of some promotional activity in many countries. 
However, even more aggressive public health measures will 
be needed to reverse the levels of harm in the population. The 
approach for individual patients suffering adverse effects of alcohol 
is described elsewhere (e.g. pp. 1184 and 880).
Smoking
Smoking is one of the top three risk factors underlying GBD (see 
Box 5.3). It is responsible for a substantial majority of cases of 
chronic obstructive pulmonary disease (COPD) and lung cancer 
(pp. 573 and 598), and most smokers die either from these or 
from ischaemic heart disease. Smoking also causes cancers 
of the upper respiratory and gastrointestinal tracts, pancreas, 
bladder and kidney, and increases risks of peripheral vascular 
disease, stroke and peptic ulceration. Maternal smoking is an 
important cause of fetal growth retardation. Moreover, there is 
evidence that passive (‘second-hand’) smoking has adverse 
effects on cardiovascular and respiratory health.
The decline in smoking in many high-income countries has 
been achieved not only by warning people of the health risks 
but also by increasing taxation of tobacco, banning advertising, 
legislating against smoking in public places and giving support for 
smoking cessation to maintain this decline. However, smoking 
rates remain high in many poorer areas and are increasing 
among young women. In many developing countries, tobacco 
companies have found new markets and rates are rising.
A complex hierarchy of systems interacts to cause smokers 
to initiate and maintain their habit. At the molecular and cellular 
levels, nicotine acts on the nervous system to create dependence 
and maintain the smoking habit. There are also strong influences 
at the personal and social level, such as young female smokers 
being motivated to ‘stay thin’ or ‘look cool’ and peer pressure. 
Other important influences include cigarette advertising, with the 
advertising budget of the tobacco industry being much greater 
than that of health services. Strategies to help individuals stop 
smoking (such as nicotine replacement therapy, anti-smoking 
advice and behavioural support) are cost-effective and form an 
important part of the overall strategy.
Obesity
Obesity is an increasingly important risk factor underlying GBD (see 
Box 5.3). The weight distribution of almost the whole population 
is shifting upwards: the slim are becoming less slim while the fat 
are getting fatter (p. 698). In the UK, this translates into a 1 kg 
increase in weight per adult per year (on average over the adult 
population). The current obesity epidemic cannot be explained 
simply by individual behaviour and poor choice but also requires 
an understanding of the obesogenic environment that encourages 
people to eat more and exercise less. This includes the availability 
of cheap and heavily marketed energy-rich foods, the increase 
in labour-saving devices (e.g. lifts and remote controls) and the 
rise in passive transport (cars as opposed to walking, cycling, or 
walking to public transport hubs). To combat the health impact 


Epidemiology • 95

has identified a set of (‘Wilson and Jungner’) criteria to guide 
health systems in deciding when it is appropriate to implement 
screening programmes. The essential criteria are:
• Is the disease an important public health problem?
• Is there a suitable screening test available?
• Is there a recognisable latent or early stage?
• Is there effective treatment for the disease at this stage 
that improves prognosis?
A suitable screening test is one that is cheap, acceptable, 
easy to perform and safe, and gives a valid result in terms of 
sensitivity and specificity (p. 4). Screening programmes should 
always be evaluated in trials so that robust evidence is provided 
in favour of their adoption. These evaluations are prone to 
several biases – self-selection bias, lead-time bias and length 
bias – and these need to be accounted for in the analysis. 
Examples of large-scale programmes in the UK include breast, 
colorectal and cervical cancer national screening programmes 
and a number of screening tests carried out in pregnancy and 
in the newborn, such as the:
• diabetic eye screening programme
• fetal anomaly screening programme
• infectious diseases in pregnancy screening programme
• newborn and infant physical examination screening 
programme
• newborn blood spot screening programme
• newborn hearing screening programme
• sickle-cell and thalassaemia screening programme.
These are illustrated in Figure 5.2.
Problems with screening include:
• over-diagnosis (of a disease that would not have come to 
attention on its own or would not have led to death)
• false reassurance
• diversion of resources from investments that could control 
the disease more cost-effectively.
An example of these problems is the use of prostate-specific 
antigen (PSA) testing as a screening test for the diagnosis of 
prostate cancer (p. 438).
Epidemiology
Epidemiologists study disease in free-living humans, seeking 
to describe patterns of health and disease and to understand 
how different exposures cause or prevent disease (Box 5.5).
Chronic diseases and risk factors (e.g. smoking, obesity etc.) 
are often described in terms of their prevalence. A prevalence is 
simply a proportion: e.g. the prevalence of diabetes in people 
aged 80 and older in developed countries is around 10%.
Events such as deaths, hospitalisations and first occurrences 
of a disease are described using incidence rates: e.g. if there 
are 100 new cases of a disease in a single year in a population 
of 1000, the incidence rate is 105 per 1000 person-years, not 
100, because of the effect of ‘person-time’. Person-time is the 
sum of the total ‘exposed’ time for the population and in this 
example is 950 person-years. The reason person-time is less 
than 1000 is that 100 people experienced the event. These 100 
people are assumed to have had an event, on average, halfway 
through the time period, removing 100 × 0.5 person-years from 
the exposure time (as it is not possible to have a first occurrence 
of a disease twice). Hence, the incidence per 1000 person-years 
is 105, not 100.
A similar measure is the cumulative incidence or risk, which 
is the number of new cases as a proportion of the total people 
at risk at the beginning of the exposure time. If, in the example 
above, the same 1000 people were observed for a year (i.e. 
with no one joining or leaving the group), then the 1-year risk is 
10% (100/1000). The time period should always be specified.
These rates and proportions are used to describe how diseases 
(and risk factors) vary according to time, person and place. 
Temporal variation may occur seasonally (e.g. malaria occurs 
in the wet season but not the dry) or as longer-term ‘secular’ 
trends (e.g. malaria may re-emerge due to drug resistance). 
Person comparisons include age, sex, socioeconomic status, 
employment, and lifestyle characteristics. Place comparisons 
include the local environment (e.g. urban versus rural) and 
international comparisons.
Understanding causes and effect
Epidemiological research complements that based on animal, cell 
and tissue models, the findings of which do not always translate 
to humans. For example, only a minority of drug discoveries 
from laboratory research are effective when tested in people.
However, differentiating causes from mere non-causal 
associations is a considerable challenge for epidemiology. 
This is because while laboratory researchers can directly 
manipulate conditions to isolate and understand causes, 
such approaches are impossible in free-living populations. 
Epidemiologists have developed a different approach, based 
around a number of study designs (Box 5.6). Of these, the 
clinical trial is closest to the laboratory experiment. An early 
example of a clinical trial is shown in Figure 5.3, along with ‘effect 
measures’, which are used to quantify the difference in rates 
and risks.
In clinical trials, patients are usually allocated randomly to 
treatments so that, on average, groups are similar, apart from 
the intervention of interest. Nevertheless, for any particular 
trial, especially a small trial, the laws of probability mean that 
differences can and do occur by chance. Poorly designed or 
executed trials can also limit comparability between groups. 
Allocation may not be truly random (e.g. because of inadequate 
concealment of the randomisation sequence), and there may 
5.5 Calculation of risk using descriptive epidemiology
Prevalence
• The ratio of the number of people with a longer-term disease or 
condition, at a specified time, to the number of people in the 
population
Incidence
• The number of events (new cases or episodes) occurring in the 
population at risk during a defined period of time
Attributable risk
• The difference between the risk (or incidence) of disease in exposed 
and non-exposed populations
Attributable fraction
• The ratio of the attributable risk to the incidence
Relative risk
• The ratio of the risk (or incidence) in the exposed population to the 
risk (or incidence) in the non-exposed population


96 • POPULATION HEALTH AND EPIDEMIOLOGY
Fig. 5.2 UK NHS Pregnancy and Newborn Screening Programmes: optimum times for testing. (GA1 = glutaric aciduria type 1; HCU = 
homocystinuria; IVA = isovaleric acidaemia; MCADD = medium-chain acyl-CoA dehydrogenase deficiency; MSUD = maple syrup urine disease; 
PKU = phenylketonuria; T13, 18, 21 = trisomy 13, 18 and 21) Based on Version 8.1, March 2016, Gateway ref: 2014696, Public Health England.
Blood for sickle cell
and thalassaemia
Commence
folic acid
Pre-conception
Pregnancy
Newborn
Blood for haemoglobin,
group, Rhesus and
antibodies as early as
possible, or as soon
as a woman arrives for
care, including labour
Blood for syphilis, hepatitis B
and HIV as early as possible,
or at any stage of the
pregnancy, including labour
Reoffer screening for
infectious diseases if
initially declined
Hepatitis B vaccination
± immunoglobulin
within 24 hours
Repeat
haemoglobin
and antibodies





































+1
+2
+3
+4
+5
+6
Birth
Key
Week
Blood for T21,
T18 and T13
(combined test)
Blood for T21
(quadruple test)
Newborn
physical
examination
by 72 hours
Newborn blood spot screens
(ideally on day 5) for
 sickle cell disease (SCD),
cystic fibrosis (CF),
congenital
hypothyroidism (CHT)
and inherited metabolic
diseases (PKU, MCADD,
MSUD, IVA, GA1 and HCU).
NB: babies who missed the
screen can be tested up to
1 year (except CF offered up
to 8 weeks)
Early pregnancy scan to
support T21, T18 and
T13 screening
Women with type 1 or type 2
diabetes are offered diabetic
eye (DE) screening annually.
In pregnancy women with
type 1 or type 2 diabetes are
offered a DE screen when
they first present for care
Give screening
information as
soon as possible
Give and discuss
newborn screening
information
Follow-up DE screen for
women with type 1 or 2
diabetes found to have
diabetic retinopathy
Further DE screen for
women with type 1 or 2
diabetes
Detailed ultrasound
scan for structural
abnormalities, including
T18 and T13
Newborn
hearing screen
Infant physical
examination
at 6–8 weeks
T21, T18, T13 and fetal
anomaly ultrasound
Sickle cell and
thalassaemia
Newborn and infant
physical examination
Newborn blood spot
Infectious diseases
in pregnancy
Diabetic eye
Newborn hearing


Health data/informatics • 97

or more often practical, considerations. Epidemiologists therefore 
seek to minimise bias and confounding by good study analysis 
and design. They subsequently make causal inferences by 
balancing the probability that an observed association has been 
caused by chance, bias and/or confounding against the alternative 
probability that the relationship is causal. This weighing-up 
requires an understanding of the frequency and importance of 
different sources of bias and confounding, as well as the scientific 
rationale of the putative causal relationship. It was this approach, 
collectively and over a number of years, that settled the fact that 
smoking causes lung cancer and, subsequently, heart disease.
Health data/informatics
As patients pass through health and social care systems, 
data are recorded concerning their family background, lifestyle 
and disease states, which is of potential interest to healthcare organisations seeking to deliver services, policy-makers 
concerned with improving health, scientific researchers trying to 
understand health, and also pharmaceutical and other commercial 
organisations aiming to identify markets.
There is a long tradition of maintaining health information 
systems. In most countries, registration of births and deaths is 
required by law, and in the majority, the cause of death is also 
recorded (Fig. 5.4). There are many challenges in ensuring such 
data are useful, especially for comparisons across time and place:
• A system of standard terminologies is needed, such as the 
WHO International Classification of Diseases (ICD-10), 
which provides a list of diagnostic codes attempting to 
cover every diagnostic entity.
• These terms must be understood to refer to the same, or 
at least similar, diseases in different places.
• Access to diagnostic skill and facilities is required.
• Standard protocols for assigning clinical diagnoses to 
ICD-10 codes are needed
• Robust quality control processes are needed to maintain 
some level of data completeness and accuracy.
Many countries employ similar systems for hospitalisations, to 
allow recovery of health-care utilisation costs or to manage and 
plan services. Similar data are rarely collected for communitybased health care, nor are detailed data on health-care processes 
generally included in national data systems. Consequently, there 
has been considerable interest in using data from information 
technology systems used to deliver care, such as electronic 
be systematic differences (biases) in the way people allocated 
to different groups are treated or studied.
Such biases also occur in observational epidemiological study 
designs, such as cohort, case–control and cross-sectional studies 
(Box 5.6). These designs are also much more subject to the 
problem of confounding than are randomised trials.
Confounding is where the relationship between an exposure 
and outcome of interest is confused by the presence of some 
other causal factor. For example, coffee consumption may be 
associated with lung cancer because smoking is more common 
among coffee-drinkers. Here, smoking is said to confound the 
association between coffee and lung cancer.
Despite these limitations, for most causes of diseases, 
randomised controlled trials are not feasible because of ethical, 
Fig. 5.3 An example of a clinical trial: streptomycin versus bed rest 
in tuberculosis. Both prevalences and risks are, in fact, proportions, and 
are therefore frequently expressed as odds. The reasons for doing so are 
beyond the scope of this text. 
Enrolled 107 patients
with tuberculosis
Effect measures
Risk ratio (relative risk, RR)
Odds ratio (OR)
Absolute risk reduction (ARR)
Relative risk reduction (RRR)
Number needed to treat to prevent
one death (NNT= 1/ARR)
Random allocation
Streptomycin
55 patients
Bed rest
52 patients
Follow-up and count deaths
Events 

Risk 
7.3%
Odds 
0.068
Events 

Risk 
28.8%
Odds 
0.224
0.25
0.30
21.6%
74.8%
4.6
5.6 Epidemiological study designs
Design
Description
Example
Clinical trial
Enrols a sample from a population and compares outcomes 
after randomly allocating patients to an intervention
Medical Research Council (MRC) Streptomycin Trial – 
demonstrated effectiveness of streptomycin in tuberculosis
Cohort
Enrols a sample from a population and compares outcomes 
according to exposures
Framingham Study – identified risk factors for cardiovascular 
disease
Case–control
Enrols cases with an outcome of interest and controls 
without that outcome and compares exposures between 
the groups
Doll R, Hill AB. Smoking and carcinoma of the lung. British Medical 
Journal 1950 – demonstrated that smoking caused lung cancer
Cross-sectional
Enrols a cross-section (sample) of people from the 
population of interest; obtains data on exposures and 
outcomes
World Health Organisation Demographic and Health Survey – 
captures risk factor data in a uniform way across many countries


98 • POPULATION HEALTH AND EPIDEMIOLOGY
Further information
Books and journal articles
GBD 2015 Disease and Injury Incidence and Prevalence Collaborators. 
Global, regional, and national incidence, prevalence, and years lived 
with disability for 310 diseases and injuries, 1990–2015: a 
systematic analysis for the Global Burden of Disease Study 2015. 
Lancet 2016; 388:1545–1602.
GBD 2015 Mortality and Causes of Death Collaborators. Global, 
regional, and national life expectancy, and cause-specific mortality 
for 249 causes of death, 1980–2015: a systematic analysis for 
the Global Burden of Disease Study 2015. Lancet 2016; 
388:1459–1544.
GBD 2015 Risk Factors Collaborators. Global, regional, and national 
comparative risk assessment of 79 behavioural, environmental and 
occupational, and metabolic risks or clusters of risks, 1990–2015: a 
systematic analysis for the Global Burden of Disease Study 2015. 
Lancet 2016; 388:1659–1724.
Kindig D, Stoddart G. What is population health? Am J Public Health 
2003; 93:380–383.
Websites
fph.org.uk UK Faculty of Public Health: What is public
health?
gov.uk UK Government: population screening
programmes.
patient records, drug-dispensing databases, radiological software 
and clinical laboratory information systems.
Data from such systems are, of course, much less structured 
than those obtained from vital registrations. Moreover, the 
completeness of such data depends greatly on local patterns 
of health-care utilisation, as well as how clinicians and others 
use information technology systems within different settings. As 
such, deriving useful, unbiased information from such data is a 
considerable challenge.
Much of the discipline of health informatics is concerned 
with addressing this challenge. One approach has been to 
develop comprehensive standard classification systems such as 
SNOMED-CT, ‘a standardised, multilingual vocabulary of terms 
relating to the care of the individual’, which has been designed 
for electronic health-care records.
An alternative has been to use statistical methods such as 
natural language processing to derive information automatically 
from free text (such as culling diagnoses from radiological reports), 
or to employ ‘machine learning’, in which software algorithms 
are applied to data in order to derive useful insights. Such 
approaches are suited to large, messy data where the costs 
of systematisation would be prohibitive. It is likely that such 
innovations will, over the coming years, provide useful information 
to complement that obtained from more traditional health 
information systems.
Fig. 5.4 Completed death certificate. International Classification of Diseases 10 (ICD-10) codes are appended in red. WHO ICD-10, vol. 2; 1990. 
Available at https://commons.m.wikimedia.org/wiki/File:International_form_of_medical_certificate_of_cause_of_death.png.
Cause of death
Approximate
interval between
onset and death
I
Disease or condition directly
leading to death*
INTERNATIONAL FORM OF MEDICAL CERTIFICATE OF CAUSE OF DEATH
due to (or as a consequence of)
due to (or as a consequence of)
due to (or as a consequence of)
(a)
(b)
(c)
(d)
I21.9
E78.0
J47
*This does not mean the mode of dying, e.g. heart failure, respiratory failure.
It means the disease, injury, or complication that caused death.
Antecedent causes
Morbid conditions, if any,
giving rise to the above cause,
stating the underlying
condition last
II
Other significant conditions
contributing to the death, but
not related to the disease or
condition causing it