Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
By the end of this chapter the reader should:
Understand how disease and health is measured in populations or groups and be able to use measures of disease incidence and prevalence
Understand and be able to use measures of effect (e.g. relative risk, absolute risk and number needed to treat)
Know the main indices of population child health and their significance
Know the strengths and limitations of different epidemiological studies
Be able to take into consideration bias, confounding and chance when interpreting epidemiological data and understand the difference between statistical association and causality
Understand the concept of social determinants of health and wider determinants of health and how it affects the health of children
Understand what is meant by inequalities
Understand the concepts, definitions, objectives and uses of public health surveillance
Understand what a ‘health needs assessment’ is and why it is undertaken
Understand the principles of screening
The word epidemiology is derived from the Greek ‘epi’, ‘demos’ and ‘logos’ and literally translated means the study (logos) of what is among (epi) the people (demos). It is the study of the occurrence and distribution of health-related states/events in specified populations. Its uses are listed in Box 2.1 .
Epidemiology can be used to:
Describe the spectrum of disease
Describe the natural history of disease
Predict disease trends
Identify factors that increase or decrease the risk of acquiring disease
Elucidate mechanisms of disease transmission
Test the efficacy of intervention strategies
Evaluate intervention programmes
Identify the health needs of a community
The prevention of sudden infant death syndrome (SIDS) is one of the major success stories in epidemiology (see Box 1.1 ). Epidemiological studies, culminating in the New Zealand Cot Death Study, a three-year case-control study (1987–1990), identified three modifiable risk factors for SIDS, namely prone sleeping position, maternal smoking and lack of breastfeeding. A prevention programme (‘Back to Sleep’ campaign) was launched and resulted in a dramatic reduction in SIDS mortality. Australia and then the United Kingdom followed suit and the number of deaths in the UK fell from 1500 to 600 per year by the mid-1990s. The ‘Back to Sleep’ campaign has been implemented in many countries across the world with similar results.
Certain terms are used when measuring health and disease of groups or populations. It is usually unhelpful to state numbers of events alone. In order to interpret the number, they need to be related to a denominator. For example stating 10 children were diagnosed with pertussis does not give as much information as the statement that 10 out of 20 children who had not been immunized were affected or that 10 children had pertussis out of a total population of 100,000 children.
The numerator is the number of people known to have a specific disease or problem.
The denominator is the total number of people at risk in the population.
If 100 children attended a Christmas party and 10 of them develop vomiting, the numerator is 10 and the denominator is 100.
Three terms are usually used to relate the number of cases of a disease or outcome to the size of the source population in which they occurred:
A ratio compares values. A ratio says how much of one thing there is compared to another thing, for example the number of stillbirths per thousand live births.
Proportion is a type of ratio where those who are included in the numerator must also be included in the denominator, for example the number of fetal deaths out of the total number of births – here the numerator will be the number of fetal deaths and the denominator the number of fetal deaths plus the live births.
A rate is defined as a ratio in which there is a distinct relationship between the numerator and denominator and most importantly a measure of time is an intrinsic part of the denominator. For example, the number of colds per 1000 primary school children during a one-month period.
However, in medical literature the term ‘rate’ is used interchangeably to denote different demographic or epidemiological measures that could be true rates, proportions or ratios.
Infant mortality rate is the number of children who die aged less than one year old per 1000 births. Some key facts are shown in Box 2.2 .
In 2011, there were 3077 infant deaths, compared with 7899 in 1980.
The infant mortality rate per 1000 live births in 2011 was 4.2, the lowest ever recorded, compared with 12 in 1980 and 130 in 1910.
The infant mortality rate was lowest among babies of mothers aged 30–34 years (3.8 deaths per 1000 live births) and highest among babies of mothers aged ≥40 years (5.5 deaths per 1000 live births).
Infant mortality rate per 1000 live births for low birthweight babies (<2500 grams) was 37; for very low birthweight babies (<1500 grams) it was 173.
The infant mortality rate:
was highest for babies with fathers employed in semi-routine occupations (4.9 deaths per 1000 live births) and lowest for those employed in managerial and professional occupations (2.8 and 2.5 deaths per 1000 live births, respectively).
was higher for babies of mothers born outside the UK (4.4 deaths per 1000 live births) than those born inside the UK (4.1 deaths per 1000 live births).
was highest for babies of mothers born in the Caribbean (9.6 deaths per 1000 live births) and in Pakistan (7.6 deaths per 1000 live births).
In 1911, 130 out of every 1000 children born in England and Wales would die before their first birthday. The decrease in infant deaths between 1911 and 2010 is because of considerable improvements in healthcare, including the control of infectious diseases and public health infrastructure over this time period. For the UK as a whole, infant mortality has been declining and is now only about a quarter of what it was in 1970.
Infant mortality rate is linked to several factors including access to healthcare services for mothers and infants, socio-economic status of the child's parents, the health of the mother, birth weight and the proportion of infants born preterm.
Terms used in child mortality statistics are:
Stillbirths and perinatal mortality rates – reported per 1000 total births (live and stillbirths)
Early neonatal, neonatal, postneonatal and infant mortality rates – reported per 1000 live births
Childhood (1–15 years) mortality rates – reported per 100,000 population of the same age
The precise definitions of these terms relating to newborn infants are included in Chapter 10, Perinatal medicine . Analysis of the data around deaths identifies that the majority of deaths in childhood occur before one year of age; 70% of infant deaths in England and Wales in 2011 were neonatal deaths – deaths at less than 28 days. The most common cause of death, in children as a whole group, is now related to perinatal problems and congenital abnormalities.
In England and Wales, stillbirths and neonatal deaths are registered using a special death certificate that enables reporting of relevant diseases or conditions in both the infant and the mother. The Office for National Statistics (ONS) has developed a hierarchical classification system (also referred to as the ONS cause groups), which allows the death to be assigned to a specific category, based on the likely timing of the damage leading to the death. This produces broad causation groups to enable direct comparison of neonatal and postneonatal deaths. The following are mutually exclusive categories:
Congenital anomalies
Antepartum infections
Immaturity-related conditions
Asphyxia, anoxia or trauma
External conditions
Infections
Other specific conditions
The linkage of birth and infant death records has been conducted since 1975 to obtain information on the social and biological factors relating to the baby and parents; it is collected at birth registration. Death registration gives only a limited amount of information about the parents of the deceased infant; for example, occupation of parent. However, a considerable amount of information is given at birth registration. This includes: age of each parent, number of previous children born (the mother's parity), country of birth of parents, place of birth and whether the baby was a singleton or multiple birth. Linking the infant death record to the birth record improves understanding of the key characteristics of the baby’s parents as further information is provided by the birth record. In 2010, 98% of infant deaths in England and Wales were successfully linked to their corresponding birth registration record. The potential use of such data is outlined in Box 2.3 .
Childhood mortality rates are a way of evaluating policies concerning the health and welfare of children, as they give an indication of the quality of support available for both children and families. One example is a review which systematically examined mortality and morbidity in children and young people between 1 and 18 years of age in the UK and linked death certificate data with hospital admissions.
It found there had been an overall reduction in all causes of child mortality since 1980 in all age groups. Injury was the most frequent cause of childhood death (31–48% of deaths in children aged 1–18 years old). The highest rates of death due to injury were found amongst boys aged 15–18 years of age. Mortality rates secondary to injury were three times higher in boys compared to girls. Two-thirds of injury mortality among 10–18 year olds was unintentional, with transport accidents accounting for 77% of this.
Higher mortality rates in both infancy and throughout childhood were seen in children whose mothers were less than 20 years old. This association persisted even after taking birth weight into account. Young maternal age was an ongoing risk factor for child death throughout early childhood.
Approximately two-thirds of childhood deaths take place in hospital (of causes other than injury). Two-thirds of children who died were identified as having a chronic condition, with mental and behavioural conditions being the most common.
From a public health perspective, the report highlighted a number of groups in whom preventative policies could be targeted in order to reduce childhood mortality rates, in particular children born to young mothers, those at risk of injuries or those with an underlying chronic condition. Policies designed to address these risk factors may result in lower childhood mortality.
Birth weight is an important indicator of overall health and is influenced by a number of factors including smoking and drinking during pregnancy, low parental socio-economic status, education levels, low income and inadequate living conditions. In the UK nearly 8% of births are preterm.
In England and Wales, around 700,000 babies were born in 2013 of which approximately 50,000 were low birth weight (<2.5 kg). Of all those babies who were born with a low weight, 62% were preterm.
This is collected internationally and allows comparison between countries. In the UK in 2013, the under-five mortality rate was 4.9 per 1000 live births, the highest in western Europe; in Iceland it was 2.4, Sweden 2.7, Spain and Germany 3.6, France and Italy 3.7. It was worse in the UK than many eastern European countries, including Serbia, Estonia and Croatia. This means that 2000 more children die in the UK each year than if they lived in Sweden.
Data show that more children die in adolescence than in any period other than infancy. The World Health Organization (WHO) classifies deaths into communicable and non-communicable disease (NCD). Deaths due to communicable disease are very low in the UK. However, for NCD deaths, for almost all ages the UK does worse than its comparators. Up to 74% of childhood deaths in the UK occur in children with co-morbidities, i.e. a long-term condition, of which the most common is a neurological or sensory condition.
Well-being is increasingly the focus of policy making and evaluation. It is now largely accepted that what children become in their adult life is largely a product of their experiences in the early stages of their lives. Particularly important are issues of health and safety, material and emotional security, education and socialization.
According to the 2011 Census, there were just over 10.5 million children aged 0 to 15 in England and Wales – about one in five of the population compared with one in three in 1911. Whilst the proportion of children in the population has declined, the proportion of the elderly has risen. Over the same period there has been a change in attitudes to children which has arguably improved their well-being. For example, all children in the UK are expected to be in compulsory education until they are at least 16 years old; and in England in some form of education or training until 18 years.
Many measures of children's well-being are used in surveys. For effective policy making and evaluation, it would be preferable for a systematic and uniform way of measuring children's well-being to be used so that there could be meaningful comparison both between sources and also over time.
As part of the Measuring National Well-being programme, the Office for National Statistics has worked with other government departments, academics and third sector organizations to examine measures of children's well-being. The aim is to understand the data that already exist to measure children's well-being and evaluate its limitations. A framework has been developed based on responses to the national debate, research findings and expert opinion. From this, the 10 domains proposed to measure national well-being for the UK are ‘Individual well-being’, ‘Our relationships’, ‘Health’, ‘What we do’, ‘Where we live’, ‘Personal finance’, ‘Education and skills’, ‘The economy’, ‘Governance’ and ‘The natural environment’. For most of these domains, some of the measures proposed for adults are also appropriate for children, either as listed above (for example the measures in ‘The economy’ domain) or by analysis specifically for children (for example, individuals living in poverty in the ‘Personal finance’ domain).
Some specific aspects of these domains for children aged 0 to 15 include circumstances in which they live, what they feel about their relationships, what they do and also decisions that adults make on their behalf. These domains (examples in Box 2.4 ) measure what children think and feel about their lives. They focus on:
How many children there are in England and Wales
Children's health
Poverty and its relationship with parental economic activity
Education and skills
Children's relationships and their well-being
Use of technology and social media
Where children live
89% of children said that they were relatively happy with their lives overall, 4% reported being relatively unhappy.
A much higher percentage reported being completely happy with their friends and family than with their school, their school work or, particularly, their appearance.
Boys more often reported being happy with their life overall, their friends and their appearance than girls, while girls more often reported being happy with their school work.
Data from children is collected through the Youth Module of the UK Longitudinal Study, a self-completed questionnaire answered by those aged 11 to 15. The questions, employing a seven-point scale, from completely happy to not at all happy, analyse measures of health and well-being in children, including how they feel about:
Life as a whole
Schoolwork
Appearance
Family
Friends
School
Another measure of health and well-being in children in England is the Good Childhood Index published by The Children's Society. It is also produced using surveys. The main measure of overall subjective well-being consists of five statements to which children are asked to respond on a five-point scale from ‘strongly disagree’ to ‘strongly agree’:
My life is going well
My life is just right
I wish I had a different kind of life
I have a good life
I have what I want in life
Children's responses to each question are coded on a scale from zero (‘strongly disagree’) to four (‘strongly agree’), to create an overall scale.
Data on children's subjective well-being was also gathered in the 2013 NatCen study; the relationship between well-being and emotional and behavioural problems are considered in Chapter 24, Emotions and behaviour .
Which of the following BEST describes the prevalence of a disease?
That part of the population that are at risk of the disease
The incidence in the population divided by the time in years
The incidence minus the mortality rate
The incidence in the population multiplied by the duration of the disease
The number of new cases that occur during a specified period of time in a population
D. The incidence in the population multiplied by the duration of the disease.
The prevalence and incidence are often confusing. How these two terms are related is described below.
Two key elements are measured in many epidemiological studies:
Exposure : the risk factors that are being investigated, that may or may not be the cause
Outcome : disease, event or health-related state of interest.
The measures of disease frequency used most often in epidemiology are:
Incidence : the number of new cases that occur during a specified period of time in a defined population
Prevalence : the proportion of the population at risk that have the condition, where:
Population at risk : that part of the population which is susceptible to a disease.
For example, to calculate the prevalence of all childhood cancers in the age group 0–4 years in the UK:
The numerator would be the number of children aged 0–4 years living in the UK at that time who were diagnosed with cancer
The denominator would be the number of children aged 0–4 years living in the UK at that time (population at risk). The mid-year population is usually used.
Data on incidence and prevalence becomes more useful if converted into rates. The rates are usually expressed as per 1000, per 10,000, per 100,000, etc.
Example: As childhood cancers are rare (between 2008–2010 there was an average of 1603 new cases of childhood cancer each year in the UK), 883 (55%) in boys and 720 (45%) in girls, the rates are expressed per million – the crude incidence rate shows that there are 160 new cancer cases for every million boys in the UK, and 137 for every million girls.
Several factors can influence prevalence:
The number of new cases (incidence) : if the number of new cases per year is high (high incidence) this will result in the prevalence being higher, e.g. up to 400 children are diagnosed with acute lymphoblastic leukaemia (ALL) every year compared to fewer than 15 children per year with chronic myeloid leukaemia (CML).
The severity of the illness : if many children who develop the disease die, the prevalence rate is low, e.g. the survival rate for children with ALL is around 90% compared to 60% in children diagnosed with CML.
The duration of the illness : if a disease lasts for a short time, its prevalence rate is lower than if it lasts for a long time, e.g. the duration of chickenpox compared with type I (insulin dependent) diabetes.
There are three common indices of risk: absolute risk, relative risk and attributable risk. The number needed to treat (NNT) is the inverse of attributable risk. The terms are defined in Box 2.5 and an example is shown in Question 2.2.
A study on the prevalence of asthma in primary school children was undertaken in two towns (A and B). In town A, a questionnaire was sent to all parents of children in primary schools. In Town B, the same questionnaire was sent to a random sample of parents of children in primary schools. Parents were asked whether doctors had ever diagnosed wheezy bronchitis, asthma or bronchitis in their child. If they answered ‘Yes’ to either wheezy bronchitis or asthma, they were classified as having asthma. If they answered ‘Yes’ only to bronchitis, they were classified as having bronchitis.
Based on the data in Table 2.1 , which of the following statements are true (T) or false (F)?
Town A | Town B | ||||
---|---|---|---|---|---|
Total sample | 1500 | 8000 | |||
Questionnaire returned | 1125 (75%) | 6960 (87%) | |||
Prevalence | 95% CI | Prevalence | 95% CI | p-value | |
Asthma | 7% | 5%–9% | 9% | 8.5%–10% | NS |
Bronchitis | 30% | 26%–34% | 15% | 14%–17% | <0.01 |
The prevalence of asthma and bronchitis is significantly higher in Town A.
The prevalence of asthma and bronchitis is significantly higher in Town B.
There is no significant difference in the prevalence of asthma between Towns A and B.
There is no significant difference in the prevalence of bronchitis between Towns A and B.
The prevalence of bronchitis in Town A is twice the prevalence in Town B.
A. False; B. False; C. True; D. False; E. True.
A. and B. It is tempting but incorrect to add together the prevalences of asthma and bronchitis here. We cannot make assumptions about a composite end point. Therefore, A and B are incorrect (false).
C. This is correct and the information is given in the table. (NS is ‘Not significant’).
D and E. We do have the data that allows us to conclude that there is a significant difference in the prevalence of bronchitis. It is significantly higher in Town A. It is twice that in Town B.
Absolute risk : Incidence of disease in any defined population. It is the number of events ÷ total population at risk.
Relative risk : Ratio of the incidence rate in the exposed group to the incidence rate in the non-exposed group. It is the risk in the exposed group ÷ risk in the unexposed group
Attributable risk : Difference in the incidence rates in the exposed and non-exposed group. It is the risk in the exposed group minus risk in the non-exposed group
Numbers needed to treat (NNT): Inverse of attributable risk.
Epidemiologists use the triad of ‘when, who and where’ to study patterns of health and disease within and between populations. Two main types of epidemiological study are used – observational and experimental ( Fig. 2.2 ).
Observational studies are further subdivided into ‘descriptive’ and ‘analytical’, which can be retrospective, use existing data, or prospective with ongoing data collection.
A descriptive observational study is typified by basic population data such as census results and surveys, but can also include case reports and case series. They cannot be used to test any hypothesis, as there is no comparison between different groups.
An analytical observational study attempts to identify subpopulations that differ (for example, presence or absence of a risk factor), and the presence or absence of disease. Examples include cohort and case-control studies.
Experimental studies (also called intervention studies) allocate an exposure of interest (e.g. a drug) to a subgroup, before following them up and assessing outcome. This is typified by randomized controlled trials.
The type of study chosen depends on what question(s) are being addressed, data already available, practicality of design, ethics, and also cost. The main types of study are discussed in more detail in Chapter 37, Clinical research .
An association between an exposure or risk factor and an outcome or disease does not imply that the former causes the latter ( Box 2.6 ). Three possible factors are important when considering whether a causal association really exists:
Is this association due to a chance occurrence?
Is it due to a flaw in the methodology ( bias )?
Is it due to some other factor linked to both exposure and outcome ( confounding )?
Several observational studies have documented a correlation between early antibiotic usage and childhood asthma. However, correlation does not necessarily imply causality and so it is important to consider all possible explanations.
Is it chance? | The effect observed may simply be due to random error. This can be a particular problem in observational studies. |
Is it causation? | One postulated mechanism is via the hygiene hypothesis. Early use of antibiotics alters the gut flora, thereby altering the immune response to known pathogens resulting in an increase in atopic disorders. |
Is it bias? | Most studies demonstrating a correlation between antibiotic usage and asthma were retrospective in design, requiring parents to remember what antibiotics their children had been prescribed early in life. Recall bias is therefore likely – parents of children with asthma are much more likely to remember these early events and prescriptions compared with parents of children who were not subsequently diagnosed with asthma. |
Is it confounding? | One prospective study found that while antibiotic use in the first nine months of life was associated with an increased prevalence of asthma by the age of five years, the increase in prevalence was also associated with an increased number of illness visits to the doctor regardless of antibiotic use. In fact, when adjusted for the number of illness visits, the relationship between antibiotic usage and diagnosis of asthma disappeared. This led the authors to conclude that the apparent association between early antibiotics and asthma was actually due to a confounding factor of illness visits. |
Chance is mainly determined by sample size – the larger the sample, the smaller the risk that the finding is due to chance alone. This is usually expressed as a ‘p-value’, usually with ‘confidence intervals’. Strictly speaking, the p-value is the likelihood of incorrectly rejecting the null hypothesis. Confidence Intervals (CIs) provide a range within which the true answer will lie at a population level (for more detail, see Chapter 38, Statistics ).
Bias is a systematic error in the methodology of the study that affects the results. There are several types of bias, such as selection bias, etc., which affect epidemiological studies:
Selection bias occurs when the two groups being compared differ systematically. That is, there are differences in the characteristics between those who are selected for a study and those who are not selected, and where those characteristics are related to either the exposure or outcome under investigation.
Example of selection bias: In Dr Andrew Wakefield's discredited paper on the MMR (measles, mumps, rubella) vaccine, he took 12 children who had behavioural disorders and attempted to link them to vaccines. There was no comparison group employed of children without behavioural disorders to see if their exposure to the MMR vaccine was greater or less.
Become a Clinical Tree membership for Full access and enjoy Unlimited articles
If you are a member. Log in here