Category: Epidemiology

Categories
Epidemiology

Chapter 3. Comparing disease rates

More chapters in Epidemiology for the uninitiated

“Is this disease increasing in incidence? Does it occur with undue frequency in my local community? Does its incidence correlate with some suspected cause? Has the outcome changed since control measures were instituted?” To answer such questions means setting two sets of rates side by side and making some sense of the comparison. This chapter examines some of the problems that may arise.

Terminology and classifications of disease

Diagnostic labels and groupings are many and various, and in continual flux: in the interests of communication some standardisation is necessary, even though no single system can meet all requirements.

The ICD system

The International Classification of Diseases, Injuries, and Causes of Death, published by the World Health Organization, assigns a three character alphanumeric code to every major condition. Often a fourth character is added for more exact specification: for example, ICD C92 is myeloid leukaemia”, which may additionally be specified as C92.0 (“acute”) or C92.1 (“chronic”). Broader groupings are readily formed – for example, ICD C81-C96 consists of all malignant neoplasms of lymphatic and haematopoietic tissue. This system is used for coding death certificates. It determines the presentation of results in the registrar general’s reports and in the diagnostic registers of most hospitals.

The system has to be revised periodically to keep pace with medical usage. The ninth revision came into general use in 1979, and has now been superseded by the 10th revision for many applications. When the classification alters from one revision to the next, disease rates may not be directly comparable before and after the change. For example, the eighth revision included separate categories for gastric ulcer and for peptic ulcer of unspecified sites, whereas in the seventh revision this distinction was not made. In this situation some aggregation of categories is needed before valid comparisons can be made.

Measures of association

Several measures are commonly used to summarise comparisons of disease rates between populations, each with its special applications. The definitions given here assume that rates in an “exposed” population are being compared with those in “unexposed” people. The exposure might be to “risk factors” suspected of causing the disease (for example, being bottle fed or owning a cat) or of protecting against it (for example, immunisation). Parallel definitions can be used to compare disease rates between people with different levels of exposure to a risk factor (for example, people with high or low aluminium concentrations in their drinking water).

Attributable risk is the disease rate in exposed persons minus that in unexposed persons. It is the measure of association that is most relevant when making decisions for individuals. For example, in deciding whether or not to indulge in a dangerous sport such as rock climbing, it is the attributable risk of injury which must be weighed against the pleasures of participation.

Relative risk is the ratio of the disease rate in exposed persons to that in people who are unexposed. It is related to attributable risk by the formula: Attributable risk= rate of disease in unexposed persons x ( relative risk- 1)

Relative risk is less relevant to making decisions in risk management than is attributable risk. For example, given a choice between a doubling in their risk of death from bronchial carcinoma and a doubling in their risk of death from oral cancer, most informed people would opt for the latter. The relative risk is the same (two), but the corresponding attributable risk is lower because oral cancer is a rarer disease.

Nevertheless, relative risk is the measure of association most often used by epidemiologists. One reason for this is that it can be estimated by a wider range of study designs. In particular, relative risk can be estimated from case-control studies (see Chapter 8) whereas attributable risk cannot. Another reason is the empirical observation that where two risk factors for a disease act in concert, their relative risks often come close to multiplying. Table 3.1 shows risks of lung cancer in smokers and non-smokers according to whether or not they had worked with asbestos. Risk in smokers was about 10-fold more than in non-smokers, irrespective of exposure to asbestos. Attributable risk does not show this convenient invariance as often as relative risk.

Table 3.1 Relative risks of lung cancer according to smoking habits and exposure to asbestos
Exposure to asbestos Cigarette smoking
No Yes
No 1.0 10.9
Yes 5.2 53.2

Closely related to relative risk is the odds ratio, defined as the odds of disease in exposed persons divided by the odds of disease in unexposed persons. People who bet on horses will be aware that a rate or chance of one in 100 corresponds to odds of 99 to one against; and in general a rate of one in x is equivalent to odds of one to x – 1. In most circumstances, the odds ratio is a close approximation to relative risk.

Population attributable risk = attributable risk x prevalence of exposure to risk factor in population Population attributable risk measures the potential impact of control measures in a population, and is relevant to decisions in public health. Attributable proportion is the proportion of disease that would be eliminated in a population if its disease rate were reduced to that of unexposed persons. It is used to compare the potential impact of different public health strategies.

Confounding

In an ideal laboratory experiment the investigator alters only one variable at a time, so that any effect he observes can only be due to that variable. Most epidemiological studies are observational, not experimental, and compare people who differ in all kinds of ways, known and unknown. If such differences determine risk of disease independently of the exposure under investigation, they are said to confoundits association with the disease.

For example, several studies have indicated high rates of lung cancer in cooks. Though this could be a consequence of their work (perhaps caused by carcinogens in fumes from frying), it may be simply because professional cooks smoke more than the average. In other words, smoking might confound the association with cooking.

Confounding determines the extent to which observed associations are causal. It may give rise to spurious associations when in fact there is no causal relation, or at the other extreme, it may obscure the effects of a true cause.

Two common confounding factors are age and sex. Crude mortality from all causes in males over a five year period was higher in Bournemouth than in Southampton. However, this difference disappeared when death rates were compared for specific age groups (Table 3.2). It occurred not because Bournemouth is a less healthy place than Southampton but because, being a town to which people retire, it has a more elderly population.

Table 3.2 Deaths in males in Bournemouth and Southampton during a five year period
Age group (years) Bournemouth Southampton
No of deaths Population Annual death rate per 100 000 No of deaths Population Annual death rate per 100 000
<1 116 919 2524 223 1897 2351
1-44 204 34616 118 332 64090 104
45-64 1252 19379 1292 1728 24440 1414
65+ 4076 11760 6932 3639 9120 7980
All ages 5648 66674 1694 5922 99547 1190

The above example shows the dangers of drawing aetiological conclusions from comparisons of crude rates. The problem can be overcome by comparing age and sex specific rates as in Table 3.2, but the presentation of such data is rather cumbersome, and it is often helpful to derive a single statistic that summarises the comparison while allowing for differences in the age and sex structure of the populations under study. Standardisedor adjusted ratesprovide for this need. Two techniques are available: Direct standardisation

Direct standardisation entails comparison of weighted averages of age and sex specific disease rates, the weights being equal to the proportion’ of people in each age and sex group in a convenient reference population. Table 3.3shows the method of calculation, based on mortality from coronary heart disease in men in the USA aged 35-64 during 1968. Table 3.4 gives standardised rates for men and women in the ensuing years, calculated in the same way, and shows a remarkable fall.

Table 3.3 Example of direct standardisation, based on mortality from coronary heart disease (CHD) in men in the USA aged 35 – 64, 1968
Age (years) CHD deaths/100 000 (1) % of reference population in age group (2) (1)    X (2)
35 – 44 93 34.4 3 199.2
45 – 54 355 360 12 780.0
55 – 64 961 29.5 28 349.5
Total 100 443 28.7 /100=443

 

Table 3.4 Coronary heart disease in American men and women aged 35-64: changes in age standardised mortality (deaths/100 000/year) during 1968-1974
1968 1969 1970 1971 1972 1973 1974
Men 443 430 420 413 408 399 377
Women 134 126 126 124 120 118 111

The direct method is for large studies, and in most surveys the indirect method yields more stable risk estimates. Suppose that a general practitioner wants to test his impression of a local excess of chronic bronchitis. Using a standard questionnaire, he examines a sample of middle aged men from his list, and finds that 45 have persistent cough and phlegm. Is this excessive? The calculation is shown in.

Table 3.5 Example of indirect standardisation
Age (years) No in study (1) Symptom prevalence in preference group (2) Expected cases = (1) x (2)
35 44 150 8% 12
45 54 100 9% 9
55 64 90 10% 9
Total 30

First the numbers of subjects in each age class are listed (column 1). The doctor must then choose a suitable reference population in which the class specific rates are known (column 2). (In mortality studies this would usually be the nation or some subset of it, such as a particular region or social class; in multicentre studies it could be the pooled data from all centres.) Cross multiplying columns 1 and 2 for each class gives the expected number of cases in a group of that age and size, based on the reference population’s rates. Summation over all classes yields the total expected frequency, given the size and age structure of that particular study sample. Where 30 cases were expected he has observed 45, giving an age adjusted relative risk or standardised prevalence ratio of 45/30 = 150%. (Conventionally, standardised ratios are often expressed as percentages).

A comparable statistic, the standardised mortality ratio (SMR) is widely used by the registrar general in summarising time trends and regional and occupational differences. Thus in 1981 the standardised mortality ratio for death by suicide in male doctors was 172%, indicating a large excess relative to the general population at the time. To analyse time trends, as with the cost of living index, an arbitrary base year is taken.

Other methods of adjusting for confounders

The techniques of standardisation are usually used to adjust for age and sex, although they can be applied to control for other confounders. Other methods, which are used more generally to adjust for confounding, include mathematical modelling techniques such as logistic regression. These assume that a person’s risk of disease is a specified mathematical function of his exposure to different risk factors and confounders. For example, it might be assumed that his odds of developing lung cancer are a product of a constant and three parameters – one determined by his age, one by whether he smokes, and the third by whether he has worked with asbestos. A computer program is then used to calculate the values of the parameters that best fit the observed data. These parameters estimate the odds ratios for each risk factor – age, smoking, and exposure to asbestos, and are mutually adjusted. Such modelling techniques are powerful and readily available to users of personal computers. They should be used with caution, however, as the mathematical assumptions in the model may not always reflect the realities of biology.

Chapters

Categories
Epidemiology

Chapter 2. Quantifying disease in populations

More chapters in Epidemiology for the uninitiated

What is a case?

Measuring disease frequency in populations requires the stipulation of diagnostic criteria. In clinical practice the definition of “a case” generally assumes that, for any disease, people are divided into two discrete classes – the affected and the non-affected. This assumption works well enough in the hospital ward, and at one time it was also thought to be appropriate for populations. Cholera, for instance, was identified only by an attack of profuse watery diarrhoea, which was often fatal; but we now know that infection may be subclinical or cause only mild diarrhoea. Similarly for non-infectious diseases today we recognise the diagnostic importance of premalignant dysplasias, in situ carcinoma, mild hypertension, and presymptomatic airways obstruction. Increasingly it appears that disease in populations exists as a continuum of severity rather than as an all or none phenomenon. The rare exceptions are mainly genetic disorders with high penetrance, like achondroplasia; for most acquired diseases the real question in population studies is not “Has the person got it?” but “How much of it has he or she got?”

One approach, therefore, is to use measures that take into account the quantitative nature of disease. For example, the distribution of blood pressures in a population can be summarised by its mean and standard deviation. For practical reasons, however, it is often helpful to dichotomise the diagnostic continuum into “cases” and “non-cases”. In defining the cut off point for such a division, four options may be considered:

Statistical – “Normal” may be defined as being within two standard deviations of the age specific mean, as in conventional laboratory practice. This is acceptable as a simple guide to the limits of what is common, but it must not be given any other importance because it fixes the frequency of “abnormal” values of every variable at around 5% in every population. More importantly, what is usual is not necessarily good.

Clinical – Clinical importance may be defined by the level of a variable above which symptoms and complications become more frequent. Thus, in a study of hip osteoarthritis cases were defined as subjects with a joint space of less than 2 mm on xray, as this level of narrowing was associated with a clear increase in symptoms.

Prognostic– Some clinical findings such as high systolic blood pressure or poor glucose tolerance may be symptomless and yet carry an adverse prognosis. Sometimes, as with glucose tolerance, there is a threshold value below which level and prognosis are unrelated. “Prognosticate abnormal” is then definable by this level.

Operational– For some disorders, none of the above approaches is satisfactory. In men of 50, a systolic pressure of 150 mm Hg is common (that is, “statistically normal”), and it is clinically normal in the sense of being without symptoms. It carries an adverse prognosis, with a risk of fatal heart attack about twice that of a low blood pressure, but there is no threshold below which differences in blood pressure have no influence on risk. Nevertheless, practical people require a case definition, even if somewhat arbitrary, as a basis for decisions. An operational definition might be based on a threshold for treatment. This will take into account symptoms and prognosis but will not be determined consistently by either. A person may be symptom free yet benefit from treatment or alternatively may have an increased risk that cannot be remedied.

Each of these four approaches to case definition is suitable for a different purpose, so an investigator may need to define the purposes before cases can be defined.

Whatever approach is adopted, the case definition should as far as possible be precise and unambiguous. A standard textbook of cardiology proposes these electrocardiographic criteria for left bundle branch block: “The duration of QRS commonly measures 0.12 to 0.16 seconds… V5 or V6 exhibits a large widened R wave…” (our italics). As a basis for epidemiological comparisons this is potentially disastrous, because each investigator could interpret the italicised words differently. By contrast, the epidemiological “Minnesota Code” defines it like this: “QRS duration  0.l2 seconds in any one or more limb leads and R peak duration  0.06 seconds in any one or more of leads, I, II, aVL, V5, or V6; each criterion to be met in a majority of technically adequate beats.” If different studies are to be compared, case definitions must be rigorously standardised and free from ambiguity. Conventional clinical descriptions do not meet this requirement.

It is also essential to define and standardise the methods of measuring the chosen criteria. An important feature in diagnosing rheumatoid arthritis, for example, is early morning stiffness of the fingers; but two interviewers may emerge with different prevalence estimates if one takes an ordinary clinical history whereas the other uses a standard questionnaire. Cases in a survey are defined not by theoretical criteria, but in terms of response to specific investigative techniques. These, too, need to be defined, standardised, and reported adequately. As a result, epidemiological case definitions are narrower and more rigid than clinical ones. This loss of flexibility has to be accepted as the price of standardisation.

Measures of disease frequency
For epidemiological purposes the occurrence of cases of disease must be related to the “population at risk” giving rise to the cases. Several measures of disease frequency are in common use.

Incidence

The incidence of a disease is the rate at which new cases occur in a population during a specified period. For example, the incidence of thyrotoxicosis during 1982 was 10/100 000/year in Barrow-in-Furness compared with 49/100 000/year in Chester.

When the population at risk is roughly constant, incidence is measured as:

Number of new cases: Population at risk×time during which cases were ascertained

Sometimes measurement of incidence is complicated by changes in the population at risk during the period when cases are ascertained, for example, through births, deaths, or migrations. This difficulty is overcome by relating the numbers of new cases to the person years at risk, calculated by adding together the periods during which each individual member of the population is at risk during the measurement period.

It should be noted that once a person is classified as a case, he or she is no longer liable to become a new case, and therefore should not contribute further person years at risk. Sometimes the same pathological event happens more than once to the same individual. In the course of a study, a patient may have several episodes of myocardial infarction. In these circumstances the definition of incidence is usually restricted to the first event, although sometimes (for example in the study of infectious diseases) it is more appropriate to count all episodes. When ambiguity is possible reports should state whether incidence refers only to first diagnosis or to all episodes, as this may influence interpretation. For example, gonorrhoea notification rates in England and Wales increased dramatically during the 1960s, but no one knows to what extent this was due to more people getting infected or to the same people getting infected more often.

Prevalence

The prevalence of a disease is the proportion of a population that are cases at a point in time. The prevalence of persistent wheeze in a large sample of British primary school children surveyed during 1986 was approximately 3 per cent, the symptom being defined by response to a standard questionnaire completed by the children’s parents. Prevalence is an appropriate measure only in such relatively stable conditions, and it is unsuitable for acute disorders.

Even in a chronic disease, the manifestations are often intermittent. In consequence, a “point” prevalence, based on a single examination, at one point in time, tends to underestimate the condition’s total frequency. If repeated or continuous assessments of the same individuals are possible, a better measure is the period prevalence defined as the proportion of a population that are cases at any time within a stated period. Thus, the 12 month period prevalence of low back pain in a sample of British women aged 30-39 was found to be 33.6%.

Mortality

Mortality is the incidence of death from a disease

Interrelation of incidence, prevalence, and mortality
Each new (incident) case enters a prevalence pool and remains there until either recovery or death: {recovery
– prevalence             {  death If recovery and death rates are low, then chronicity is high and even a low incidence will produce a high prevalence:

Prevalence = incidence x average duration

In studies of aetiology, incidence is the most appropriate measure of disease frequency. Mortality is a satisfactory proxy for incidence if survival is not related to the risk factors under investigation. However, patterns of mortality can be misleading if survival is variable. A recent decline in mortality from testicular cancer is attributable to improved cure rates from better treatment, and does not reflect a fall in incidence.

Prevalence is often used as an alternative to incidence in the study of rarer chronic diseases such as multiple sclerosis, where it would be difficult to accumulate large numbers of incident cases. Again, however, care is needed in interpretation. Differences in prevalence between different parts of the world may result from differences in survival and recovery as well as in incidence.

Crude and specific rates

A crude incidence, prevalence, or mortality (death rate) is one that relates to results for a population taken as a whole, without subdivision or refinement. The crude mortality from lung cancer in men in England and Wales during 1985-89 was 1034/million/year compared with 575/million/year during 1950-54. However, this bald fact masks a more complex pattern of trends in which mortality from lung cancer was declining in younger men while going up in the elderly.

mortality from lung cancer in men in England and Wales, 1950-89, by five year age group

Mortality from lung cancer in men in England and Wales, 1950-89, by five year age groups

It is often helpful to break down results for the whole population to give rates specific for age and sex, but it is frustrating if results are given for 35-44 years in one report, 30-49 in another, and 31 to 40 in another. When feasible, decade classes should be 5-14, 15-24, and so on, and quinquennia should be 5-9, 10-14, and so on. Overlapping classes (5-10, 10-15) should be avoided.

Extensions and alternatives to incidence and prevalence

The terms incidence and prevalence have been defined in relation to the onset and presence of disease, but they can be extended to encompass other events and states. Thus, one can measure the incidence of redundancy in an employed population (the rate at which people are made redundant over time) or the prevalence of smoking in it (the proportion of the population who currently smoke).

Some health outcomes do not lend themselves to description by an incidence or prevalence, because of difficulties in defining the population at risk. For these outcomes, special rates are defined with a quasi population at risk as denominator.

Some special rates
Birth rate:  Number of live births Mid-year population   Fertility rate: Number of live births Number of women aged 15-44 years

Infant mortality rate: Number of infant (< 1 year) deaths Number of live births

Stillbirth rate: Number of intrauterine deaths after 28 weeks Total births

Perinatal mortality rate:  Number of stillbirths + deaths in 1st week of life Total births 

NB These rates are usually related to one year

Sometimes the population at risk can be satisfactorily defined, but it cannot be enumerated. For example, a cancer registry might collect information about the occupations of registered cancer cases, but not have data on the number of people in each occupation within its catchment area. Thus, the incidence of different cancers by occupation could not be calculated. An alternative in these circumstances would be to derive the proportion of different types of cancer in each occupational group. However, care is needed in the interpretation of proportions. A high proportion of prostatic cancers in farmers could reflect a high incidence of the disease, but it could also occur if farmers had an unusually low incidence of other types of cancer. Incidence and prevalence are preferable to proportions if they can be adequately measured.

Chapters

Categories
Epidemiology

Chapter 1. What is epidemiology?

More chapters in Epidemiology for the uninitiated

Epidemiology is the study of how often diseases occur in different groups of people and why. Epidemiological information is used to plan and evaluate strategies to prevent illness and as a guide to the management of patients in whom disease has already developed.

Like the clinical findings and pathology, the epidemiology of a disease is an integral part of its basic description. The subject has its special techniques of data collection and interpretation, and its necessary jargon for technical terms. This short book aims to provide an ABC of the epidemiological approach, its terminology, and its methods. Our only assumption will be that readers already believe that epidemiological questions are worth answering. This introduction will indicate some of the distinctive characteristics of the epidemiological approach.

All findings must relate to a defined population
A key feature of epidemiology is the measurement of disease outcomes in relation to a population at risk. The population at risk is the group of people, healthy or sick, who would be counted as cases if they had the disease being studied. For example, if a general practitioner were measuring how often patients consult him about deafness, the population at risk would comprise those people on his list (and perhaps also of his partners) who might see him about a hearing problem if they had one. Patients who, though still on the list, had moved to another area would not consult that doctor. They would therefore not belong to the population at risk.

The importance of considering the population at risk is illustrated by two examples. In a study of accidents to patients in hospital it was noted that the largest number occurred among the elderly, and from this the authors concluded that “patients aged 60 and over are more prone to accidents.” Another study, based on a survey of hang gliding accidents, recommended that flying should be banned between 11 am and 3 pm, because this was the time when 73% of the accidents occurred. Each of these studies based conclusions on the same logical error, namely, the floating numerator: the number of cases was not related to the appropriate “at risk” population. Had this been done, the conclusions might have been different. Differing numbers of accidents to patients and to hang gliders must reflect, at least in part, differing numbers at risk. Epidemiological conclusions (on risk) cannot be drawn from purely clinical data (on the number of sick people seen).

Implicit in any epidemiological investigation is the notion of a target populationabout which conclusions are to be drawn. Occasionally measurements can be made on the full target population. In a study to evaluate the effectiveness of dust control measures in British coal mines, information was available on all incident (new) cases of coal workers’ pneumoconiosis throughout the country.

More often observations can only be made on a study sample, which is selected in some way from the target population. For example, a gastroenterologist wishing to draw general inferences about long term prognosis in patients with Crohn’s disease might extrapolate from the experience of cases encountered in his own clinical practice. The confidence that can be placed in conclusions drawn from samples depends in part on sample size. Small samples can be unrepresentative just by chance, and the scope for chance errors can be quantified statistically. More problematic are the errors that arise from the method by which the sample is chosen. A gastroenterologist who has a special interest in Crohn’s disease may be referred patients whose cases are unusual or difficult, the clinical course and complications of which are atypical of the disease more generally. Such systematic errors cannot usually be measured, and assessment therefore becomes a matter for subjective judgement.

Systematic sampling errors can be avoided by use of a random selection process in which each member of the target population has a known (non-zero) probability of being included in the study sample. However, this requires an enumeration or censusof all members of the target population, which may not be feasible.

Often the selection of a study sample is partially random. Within the target population an accessible subset, the study population, is defined. The study sample is then chosen at random from the study population. Thus the people examined are at two removes from the group with which the study is ultimately concerned:

Target population – study population – study sample

This approach is appropriate where a suitable study population can be identified but is larger than the investigation requires. For example, in a survey of back pain and its possible causes, the target population was all potential back pain sufferers. The study population was defined as all people aged 20-59 from eight communities, and a sample of subjects was then randomly selected for investigation from within this study population. With this design, inference from the study sample to the study population is free from systematic sampling error, but further extrapolation to the target population remains a matter of judgement.

The definition of a study population begins with some characteristic which all its members have in common. This may be geographical(“all UK residents in 1985” or “all residents in a specified health district”); occupational(“all employees of a factory,” “children attending a certain primary school”, “all welders in England and Wales”); based on special care(“patients on a GP’s list”, “residents in an old people’s home”); or diagnostic (“all people in Southampton who first had an epileptic fit during 1990-91”). Within this broad definition appropriate restrictions may be specified – for example in age range or sex.

Oriented to groups rather than individuals
Clinical observations determine decisions about individuals. Epidemiological observations may also guide decisions about individuals, but they relate primarily to groups of people. This fundamental difference in the purpose of measurements implies different demands on the quality of data. An inquiry into the validity of death certificates as an indicator of the frequency of oesophageal cancer produced the results shown in Table 1.1

Inaccuracy was alarming at the level of individual patients. Nevertheless, the false positive results balanced the false negatives so the clinicians’ total (53 + 21 = 74 cases) was about the same as the pathologists’ total (53 + 22 = 75 cases). Hence, in this instance, mortality statistics in the population seemed to be about right, despite the unreliability of individual death certificates. Conversely, it may not be too serious clinically if Dr. X systematically records blood pressure 10 mm Hg higher than his colleagues, because his management policy is (one hopes) adjusted accordingly. But choosing Dr. X as an observer in a population study of the frequency of hypertension would be unfortunate.

 

Table 1.1 Cause of death diagnosed clinically compared with at necropsy
Diagnosis of oesophageal cancer No.
Diagnosed by clinician and confirmed by pathologist 53
Diagnosed by clinician and not confirmed by pathologist 21
First diagnosed post mortem 22

Conclusions are based on comparisons
Clues to aetiology come from comparing disease rates in groups with differing levels of exposure – for example, the incidence of congenital defects before and after a rubella epidemic or the rate of mesothelioma in people with or without exposure to asbestos. Clues will be missed, or false clues created, if comparisons are biased by unequal ascertainment of cases or exposure levels. Of course, if everyone is equally exposed there will not be any clues – epidemiology thrives on heterogeneity. If everyone smoked 20 cigarettes a day the link with lung cancer would have been undetectable. Lung cancer might then have been considered a “genetic disease”, because its distribution depended on susceptibility to the effects of smoking.

Identifying high risk and priority groups also rests on unbiased comparison of rates. The Decennial Occupational Supplement of the Registrar General of England and Wales(1970-2) suggested major differences between occupations in the proportion of men surviving to age 65:

Table 1.2 Men surviving to 65, by occupation
Farmers (self employed) 82%
Professionals 77%
Skilled manual workers 69%
Labourers 63%
Armed forces 42%

These differences look important and challenging. However, one must consider how far the comparison may have been distorted either by inaccurate ascertainment of the deaths or the populations at risk or by selective influences on recruitment or retirement (especially important in the case of the armed forces).

Another task of epidemiology is monitoring or surveillance of time trends to show which diseases are increasing or decreasing in incidence and which are changing in their distribution. This information is needed to identify emerging problems and also to assess the effectiveness of measures to control old problems. Unfortunately, standards of diagnosis and data recording may change, and conclusions from time trends call for particular wariness.

The data from which epidemiology seeks to draw conclusions are nearly always collected by more than one person, often from different countries. Rigorous standardisation and quality control of investigative methods are essential in epidemiology; and if an apparent difference in disease rates has emerged, the first question is always “Might the comparison be biased?”

Chapters