Introduction

The aim of this chapter is to provide you with a basic understanding of epidemiology, and to introduce you to some of the epidemiological concepts and methods used by researchers and practitioners working in public health. Epidemiologists are often described as ‘disease detectives’, and they play a key role in identifying and presenting the evidence that underpins policy and practice in both the clinical and public health settings. The ultimate goal of epidemiology is to contribute to the prevention of disease and disability and to delay mortality.

Epidemiology is fundamental to evidence-based medicine and public health policy and practice. Rather than examine health and illness on an individual level, as clinicians do, epidemiologists focus on communities and populations, where important information and insights can be gained regarding the health of populations, the distribution of disease and injury, and the determinants of these conditions, as well as the effectiveness of health interventions.

But why do we really need epidemiology and this population-level understanding of health? Our health, risk of disease, and chance of having an injury are all determined by a complex interaction between multiple factors related to our family history, where we live and how we live. These factors can be difficult to tease apart unless we have a systematic way of studying them and determining real causal associations. Only then can we determine the best ways to treat or prevent poor health outcomes. Epidemiology is a structured, logical framework for thinking, for unravelling complex problems, and piecing together the best evidence available. This is why epidemiology is the basis for evidence-based clinical practice, strategic planning, prioritising health issues and evaluating health services.

Here is one example of an association between an exposure of interest and a health outcome that we will use to help you understand how epidemiology works, and how it is useful in public health. Australia has the highest incidence of asthma in the world; it is commonly found that people who own cats are less likely to have asthma. Does this mean that owning a cat protects you from developing asthma? Or that people with asthma tend to not own cats as it exacerbates their asthma? Or is it possibly a bit of both? Should doctors tell women with a family history of asthma to get rid of their cat, if they are planning to get pregnant? Or should the public health message be to get a cat if you are thinking of getting pregnant, to reduce the chances of your child having asthma? Epidemiology attempts to answer such questions. We will come back to this example at different times in this chapter to illustrate different aspects of epidemiologic concepts and methods.

Defining epidemiology

Epidemiology can be defined as:

Although this may seem straightforward, it is worth taking a moment to think further about this definition. It succeeds in capturing the scope of epidemiology in a clear and concise manner. Look carefully at each of the bolded words in the definition of epidemiology given above.

Distribution refers to the pattern or frequencies of health events by person (who gets affected), place (where it happens) and also time (when it happens). We will talk more about this when we consider ‘person, place and time’ in more detail, together with the ways we capture and measure health outcomes. Determinant refers to both the causes of, and risk factors for, health events. These can include any aspect of the environment we live in (biological, physical, cultural, social, etc.) including living organisms (e.g. viruses and bacteria), physical entities (e.g. radiation, pollution and dangerous machinery), lifestyle (e.g. stress and diet), social factors (e.g. poverty), and genetic factors (e.g. inherited or changed genes that cause genetic diseases). We will revisit health determinants when we consider how we might measure these exposures when trying to understand the patterns of disease that we see in populations. The World Health Organization (WHO) defines health as ‘a state of complete physical, mental, and social wellbeing and not merely the absence of disease or infirmity’ (WHO 1948). Health could include a specific disease state, the absence of a disease state, a quality of life rating, life expectancy or the incidence of mental illness or physical injury. Population refers to a group of people with definable commonalities, for example, the people who reside in a certain city or country, a school community or a certain age or ethnic group. Control of health problems refers to reducing the burden of a health problem in a population. Epidemiology quantifies the burden of disease, the strength of association between exposures and health states, the magnitude of risk at the population level, and the potential benefit of interventions, and, through this, drives evidence-based health policy and practice.

There are a number of subspecialities now emerging in epidemiology: infectious or communicable disease epidemiology, social and environmental epidemiology (focusing on the social and environmental determinants of health) and genetic and molecular epidemiology (based on studies of disease inheritance and laboratory tests for genotype and phenotype) are the more common subdisciplines, but there are others, such as palaeoepidemiology (the study of disease in past populations). Subdisciplines have developed associated specialty methods, including advanced analytic methods appropriate to the health issues and exposures of interest.

In the last 20 years, the focus of epidemiology has begun to shift from a disease outcome focus to one where the emphasis is also on understanding the patterns and causes of resilience and good health, which is equally important in achieving the goals of better health and a lower disease burden. Whilst epidemiology is a stand-alone discipline in health, epidemiologists also draw upon other disciplines to achieve these outcomes, including statistics, the social and behavioural sciences, the biological sciences and clinical medicine.

Objectives of epidemiological studies

Epidemiological studies can fulfil three primary roles: description, analysis and intervention. Health status can be explored and described within and between populations, different geographic localities and at different times. This can be of great importance for understanding the burden of disease or the state of health of the population, and the health service needs (for example, determining the proportion of a population with asthma and whether there is sufficient emergency department access for those likely to suffer an acute asthma attack).

Depending on the study design, the descriptive information may then be used to look for relationships between possible causal factors and health outcomes, through statistical analysis. This can help determine whether a certain population or individual factors seem to be associated with certain health outcomes (for example, the inverse relationship between asthma and cat ownership described above). Finally, the outcomes of analytic studies can be used to develop and justify the implementation of further analytic studies to further explore these associations (e.g. clarifying what is cause and what is effect), or there may be sufficient information to drive interventions, such as health promotion programmes. Epidemiologic approaches are also integral to the evaluation of the effectiveness of such interventions.

A brief history of epidemiology (1700s onwards)

Looking at the history of epidemiological ideas allows you to understand how scientific discoveries have fundamentally changed the ways in which humans see themselves. These days, we take it for granted that many diseases can be prevented through, for example, good nutrition. However, this idea had not been proven prior to James Lind’s work in preventing scurvy among British sailors by including citrus fruit in their diets in 1747 (The James Lind Library, see ‘Useful websites’ at the end of the chapter). Perhaps the best-known historical examples of epidemiological studies are those of John Snow and his investigations of cholera epidemics in London during the mid 1800s.

John Snow was a British physician who is considered by many epidemiologists to be the father of modern epidemiology. His story was set in nineteenth-century London, where millions lived an impoverished life in overcrowded and unsanitary slums, susceptible to devastating outbreaks of infectious diseases such as cholera. During the early 1800s, it was thought that cholera was airborne; however, Snow did not believe in the ‘miasma’ (bad air) theory, and was certain that the infectious agent entered the body through the mouth. In 1849, he published an essay entitled On the Mode of Communication of Cholera and, in August 1854, when a cholera outbreak occurred in Soho, London, Snow used epidemiological investigative skills to identify a water pump in Broad Street as the source of the disease. He presented his findings to the local government officials, and they decided to remove the handle of the pump the next day. The cases of cholera immediately began to diminish. Although John Snow’s ‘germ’ theory was not accepted until after his death, he was a prominent physician and certainly one of the founders of epidemiology. Should you wish to read more about the life and times of John Snow, there are a number of online resources available (see ‘Useful websites’ at the end of the chapter).

Measuring the occurrence of exposures of interest and of health outcomes

Studying the occurrence of health conditions in human populations is the cornerstone of public health practice. Measuring the health of populations can help to answer some fairly simple yet incredibly important questions. For example, how much disease is present now? How fast are new cases occurring? How long do people remain ill? How does the rate of disease or death differ over time within a population or compared with another population? Who does the disease affect? Where and when are they getting sick? What strategies are effective at reducing the occurrence of a certain disease or condition? Do we have sufficient health services in place to cope? Health indicators are measurable characteristics of a person, population or environment that are indicative of one or more aspects of a population’s health (e.g. infant mortality rate).

Case definitions are fundamental in epidemiology; they must be clear, unambiguous, and consistently applied across populations and time in order to allow reliable reporting and comparison of health data. This is equally true of health outcome data, as well as exposures of interest. For example, asthma data can be collected in a variety of ways. Asthma may be self-reported from symptoms (Have you ever had a persistent wheezing cough?), self-reported from a doctor’s diagnosis of asthma (Have you ever been diagnosed by a doctor as having asthma?), from medication typically associated with asthma (Have you ever used a preventer or reliever puffer?), or directly from medical records.

You can see that these measures might have varying degrees of reliability, and you would not want to compare the frequency of asthma based on different measures between groups, as any differences found might simply reflect differences in the measurement techniques used, and not real differences in the proportion of the population who have had asthma. If the exposure of interest is ‘cats’, we must be equally careful to define whether this includes any cat ownership, or whether only ‘indoor cats’ will be counted, and/or whether the timing of cat exposure is important (Did you own a cat before you developed asthma?).

Epidemiologists can count disease events or, more usually, calculate rates and proportions so comparisons can be made between populations of health status, or over time. There are several measures of disease frequency that are employed. The simplest quantitative measure is a count, which merely refers to the number of people in a certain health state, who die or become ill from a specified cause. These data have limited value without information about the population size or the number of people at risk. For example, we might count 20 influenza-related deaths; however, unless we know how many flu cases there were (i.e., the number of people at risk of dying of flu), we cannot assess the mortality rate associated with this particular type of flu. Were there 20 deaths out of 100 000 cases, or 20 out of 200 cases? The implications are very different for managing this flu outbreak. For a count to be descriptive of a population, it needs to be reported as a proportion relative to the size of that population. There are several other measures also used in speciality areas in epidemiology – for example communicable disease epidemiology uses ‘attack’ rates and ‘infection’ rates (the proportion of those exposed who develop an infection).

A ratio describes the magnitude of one group relative to another. For example, if, of the 20 people who died of influenza, 16 were men and 4 were women, then the sex ratio would be 4 : 1 male to female. Another frequently used measure is a rate, which is a measure of the frequency of occurrence of an event. A rate differs from a proportion in that it involves units of time in its calculation (e.g. the number of influenza cases in a given year). The attack and infection ‘rates’ mentioned above do not always explicitly include a time factor, but they are usually time limited as they often refer to a specific outbreak event.

Two of the most widely used measures of risk calculated from the frequency of a health outcome (e.g. asthma) or exposure (e.g. cat ownership) are prevalence and incidence.

Prevalence refers to the number of people in a defined population who have a specific disease or condition or exposure at a certain point in time (e.g. at the time the data were collected in a national health survey). Prevalence data can provide an indication of the extent of a health problem (number of cases and severity of disease) and thus assist in the planning of health services. Measuring prevalence basically involves counting cases. However, data on prevalence are much more meaningful if converted into a proportion, by dividing the count by the total number of people in the population from which the cases arose (e.g. the number of people who completed the health survey). Prevalence can be reported as a percentage (i.e. multiply by 100), or per 10 000 or 100 000 in the population if it is a rare event (Example 4.1).

Prevalence can refer directly to a specific point in time (point prevalence) but can also relate to a period of time (period prevalence). For example, we might count the number of motor vehicle accidents occurring on 31 December, or we might count the number of accidents that occurred between 1 January and 31 December.

Prevalence is calculated using the following equation:

Multiplying by 10ⁿ allows you to adjust the reporting units so that prevalence is expressed in the same units for comparison across populations or time. It also allows you to report very low prevalence as cases per 100 000 (13 case of meningococcal disease per 100 000) so that prevalence is not reported as fractions of cases (0.13 cases per 1000 population), which does not make much sense!

Incidence refers to the number of new cases of disease, injury or death in a population during a specified time period. For chronic conditions that are not that common and can last a lifetime (e.g. tuberculosis), there may be a large difference between incidence and prevalence measures, as most cases in a population at any point in time are old cases and there are few new, or incident, cases. However, for acute diseases (e.g., influenza), there will be little difference between incidence and prevalence estimates over a flu season, as all cases are incident cases.

Unlike prevalence, incidence is a true rate as it always specifies a unit of time in its calculation. There are two major measures of incidence, incidence rate and cumulative incidence. Incidence rate is a more precise measure that describes the rate at which new cases occur in a population over a specified period of time. Cumulative incidence is a simpler measure of the occurrence of disease or death, and tells us the proportion of a population, from the total at risk, that develops a disease during a specified time period. As with prevalence, cumulative incidence can be expressed as a proportion, a percentage, or as the number of cases per population.

It is worth taking a little time to think about why we need both incidence measures. Many epidemiological studies follow people over time to see who develops certain health outcomes in a population deemed to be ‘at risk’. Not everyone will be followed up for the full study period (some drop out of the study, die from other causes, etc.), and we need to take this into account so that we do not underestimate disease incidence relative to time at risk. For example, say in a group of 100 people who have had a coronary artery bypass graft, 24 have had a myocardial infarction (MI) after 2 years. This gives a 2-year incidence risk of 24%. This could also be expressed as a risk of 12 MIs per 100 person-years. However, that assumes that everyone was followed up for the full 2 years; but instead you find that on average, the follow-up time was really only 18 months. This should then be reported as 12 MIs per 150 person-years (which is the same as saying 16 MIs per 100 person-years). So you can see we may underestimate the disease incidence, or disease risk if we did not take into account the variation in the individual follow-up periods (Example 4.2).

Incidence rate and cumulative incidence are calculated using the following equations:

Note: ‘Person-time’ represents the sum of each participant’s individual time at risk (i.e. duration of follow-up) and can be expressed in any time unit depending on the context; for example person years, person months or person days.

If you would like to read further on the concept of person-time see Essential Epidemiology (Webb et al. 2011 pp 41–45).

Mortality rates and life expectancy are important health indicators globally. In Australia a major public health effort is being put in to ‘closing the gap’ to reduce the discrepancy in the average age at death seen in the Aboriginal population compared with the rest of the Australia population. In 2011, the reduced life expectancy is still estimated to be 11.5 years for Aboriginal males, and 9.7 years for females. In settings where other forms of disease surveillance or disease registers are nonexistent, such as resource-poor countries or historically, the statistics available on death are often more reliable than any other disease or health outcome available.

Mortality rates measure the risk of dying in a certain time period. Mortality patterns can be described using crude rates, age-specific rates, sex-specific rates and cause-specific rates (deaths attributed to a certain disease). Crude mortality rates (CMR) are derived by the equation:

Crude mortality rates are affected by a number of population characteristics, particularly age structure. For example, in 1990, Sweden’s annual death rate was 11 per 1000. This rate was higher than that of Guatemala (8 per 1000), even though life expectancy in Sweden (78 years) was greater than in Guatemala (63 years). The difference in crude mortality rates between the two countries was mainly due to differences in age structure: 18% of Sweden’s population was aged over 65 years, compared with only 3% of Guatemala’s population. With our ageing population, Australia would be similar to Sweden. The differences in the age structure of a population, together with the fact that risk of mortality varies with age, can cause misleading conclusions about comparative health status. A range of approaches are used to take this into account when comparing populations, including:

Most introductory epidemiology textbooks discuss the measures described above; if you are interested in reading further see Epidemiology (Gordis 2004).

Epidemiological study design

Different types of epidemiological study design are used to answer different research questions, and each has associated advantages and disadvantages in terms of the costs involved in carrying out the study, the quality of data the study generates, and the strength of the conclusions that can be drawn. The research questions themselves are usually based on existing hypotheses built on a biological understanding of the natural history of the disease of interest, and/or observations made by clinicians or public health practitioners from individual cases (case series), or observational studies that seem to suggest a possible association. The research questions must be biologically plausible.

Studies can be classified as either observational or experimental, and the distinction is an important one. Observational studies allow nature to take its course and the investigator simply observes events in different populations/groups. The investigator then seeks information about the patterns of diseases and potential risk factors, or exposures of interest. In contrast, in experimental studies (also known as intervention studies) the investigator actively manipulates an exposure to judge its effect on a health outcome. This is a much more powerful study design for isolating the effects of an exposure and making causal inferences about the relationship between the exposure under study and the health outcome of interest.

Studies can also be classified as descriptive or analytical and, in some cases, a study can be both. Descriptive studies are used to describe and measure health indicators or the burden of disease within a population, whereas analytical studies are carried out to evaluate the association between one or more exposures and the development of a particular disease or health state, or a number of health outcomes.

Study designs differ with respect to the number of observations made, whether data are collected prospectively or retrospectively, the data collection procedures used, whether individuals or groups are studied, and the availability of subjects or existing data. Epidemiologists use a range of study designs, as illustrated in Figure 4.1.

Fig. 4.1 Main types of epidemiological study design.

Observational epidemiology

For ethical reasons, many factors thought to influence disease, or protect against it, cannot be imposed upon a study population. Instead, researchers make use of naturally occurring situations, and observe and measure exposures and patterns of health in naturally occurring groups. The investigator does not intervene in any way. Among observational research designs, descriptive studies look at patterns of disease and measure the incidence or prevalence of disease and/or risk factors in a population. Analytical studies make comparisons between groups of people with and without disease (Did they have the same prevalence of risk factors?), or people with and without exposures (Did they have the same prevalence of disease?), to identify potential associations between risk factors and health outcomes.

Ecological design is the simplest form of design, which is often very quick and cheap to run, and is based on the examination of existing data. In some cases, you can design a study that uses different data sets to extract information on exposures and disease outcomes for particular populations. For example, there was some concern that mobile phone use may increase the risk of brain cancer. The first step in investigating this was to carry out studies that compared mobile phone usage data at the population level to routinely collected cancer data for the same populations. If there was a strong association, then it could reasonably be expected that there would be a positive association between mobile phone uptake at the population level and population cancer rates. However, even if such an association was observed, you would not know whether the excess cases of cancer actually occurred in mobile phone users or not.

This is called an ecological design because all data on exposure and outcome are collected at the population level. There are no data on individuals and, therefore, care must be taken not to overinterpret any associations found. These studies are usually hypothesis generating, rather than providing definitive evidence of a relationship between an exposure and health outcome. Another example would be a study where the proportion of children with asthma in 10 regions is compared with the proportion of cat ownership in each region, where nothing is known about the exposure status or other characteristics of each individual child. Asthma rates are higher in developed countries such as Australia, and if the keeping of pet cats indoors is also more frequent in Australia, then we might conclude from an ecological study that exposure to cats is associated with more asthma, yet a closer look might reveal the pet cats are more likely to be found in the asthma-free homes. You could not see this from ecological study data alone.

Activity

• Consider the ecological study of cat ownership described above. The reliability of the exposure and outcome variables must always be considered in ecological studies, as the data have generally been collected for other purposes. For example, cat ownership might be determined from cat registration databases. How might asthma data be accessed? If you were comparing two populations, say Sydney, Australia, and New Delhi, India, are the exposure and outcome data likely to differ between the populations being compared? Write a short paragraph to address these questions.

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

1. The impact of globalisation on health
2. History and development of public health
3. Biological and environmental determinants
4. Evidence-based practice

Stay updated, free articles. Join our Telegram channel

Join