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
16. Ischaemic heart disease (4)
17. Cerebrovascular disease (5)
18. Lower respiratory infections (1)
19. Neonatal preterm birth complications (2)
20. Diarrhoeal diseases (3)
21. Neonatal encephalopathy (6)
22. HIV/AIDS (29)
23. Road injuries (10)
24. Malaria (7)
25. Chronic obstructive pulmonary disease (12)
26. Congenital anomalies (9)
27. Tuberculosis (11)
28. Lung cancer3 (20)
29. Self-harm (16)
30. 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