You will find that in clinical trials what we have called the cumulative incidence is often called the experimental event rate (EER) when it describes the cumulative incidence in the intervention or treatment group, and the
control event rate (CER) in the control or placebo group.
For example, a group in the USA investigated whether four weeks of aspirin treatment would reduce the risk of blood clots in patients being treated with antibiotics for infective endocarditis (an infection of the lining of the heart that usually affects the heart valves). In total 115 patients were enrolled in the study and assigned at random to receive aspirin (n= 60) or placebo (n = 55). During the study, 17 patients in the aspirin group (EER or CIIntervention= 17 ÷ 60= 28.3%) and 11 in the placebo group (CER or CIControl= 11 ÷ 55 = 20.0%) experienced blood clots (Table 2.3). The authors concluded that aspirin treatment did not reduce the risk of clots (Chan et al., 2003).
Table 2.3 Results of an RCT evaluating aspirin use for infective endocarditis. Number with Event rate (cumulative Total patients blood clots incidence of blood clots)
Aspirin 60 17 28.3%
Placebo 55 11 20.0%
(From Chan et al., 2003.)
have been higher than 40%. The maximum estimate of the cumulative incidence would assume that all three of the missing people developed the disease: CI= 7 ÷ 10 = 70% in seven years
Note that we could calculate an accurate cumulative incidence at two years – since we do have complete follow-up to that point:
CI= 1 ÷ 10 = 10% in two years
One type of study in which the study group is clearly defined and loss to follow- up is usually minimal is a clinical trial (see Chapter 4) and this means that the cumulative incidence is an appropriate and common measure of outcome in this type of study. However, the field of clinical epidemiology has developed its own terminology for what we call cumulative incidence (see Box 2.5).
Incidence rate
Although we do not know what happened to three people in the group, we do know that they had not developed the disease before they were lost to follow-up.
Measuring disease occurrence in practice: using routine data 45
We can use this information to help us calculate the incidence rate or what is sometimes called the incidence density (Equation (2.6)). This is the number of new cases of disease (four) divided by the total amount of person-time at risk of developing the disease. An individual is at risk of developing the disease until the actual moment when they do develop it (in practice, when they are diagnosed) or until they are lost to follow-up. In this example, individual number one would contribute seven years of person-time; individual number two would contribute five and a half years; individual number three would contribute two years, and so on.
What is the total amount of person-time at risk?
What is the incidence rate for this disease per 100 person-years? The total amount of person time is
7+ 5.5 + 2 + 3.5 + 4 + 2 + 7 + 1.5 + 5 + 7 = 44.5 person-years So the incidence rate is
4 cases÷ 44.5 person-years = 0.09 cases/person-year or 9 cases/100 person-years
Measuring disease occurrence in practice: using routine data
In practice most of our information about the occurrence of disease comes from routine statistics, collected at a regional, national or international level (we will discuss some of the sources of these data in Chapter 3), and in this for- mat they comprise the core of many published reports. The data are not based on specific information about individuals but relate the number of cases of disease (or deaths) in a population to the size of that population (often an esti- mate from a census). This can lead to problems when we try to relate the occur- rence of disease to potential causes. For example, if a region has a very high level of unemployment and also has a high incidence of suicide it might be tempting to jump to the conclusion that being unemployed drives people to commit sui- cide. However, we have no way of knowing from routine statistics whether it is the same people who are unemployed who are also those committing suicide. (This dilemma where we try to extrapolate from an association seen at the pop- ulation level to draw conclusions about the relation in individuals is often called the ecological fallacy and we will discuss it again in Chapter 3.)
A second drawback of routine data relates to the fact that in public health we often want to measure the incidence of disease – how quickly are people becom- ing ill? Unfortunately, it is often difficult to obtain reliable information about incidence because few illnesses are captured reliably in routine statistics. Some diseases, such as HIV infection and cancer, are ‘notifiable’ in many countries and,
therefore, all cases should be reported to a central body; however, these exam- ples are the exceptions rather than the rule and such data are not routinely avail- able for most diseases. Furthermore, even where reporting is mandated it does not always occur in practice. When HIV/AIDS first came to world attention and again during the 2003 SARS (severe acute respiratory syndrome) outbreak, some countries suppressed the real numbers of cases for both political and economic reasons.
As a result, many common measures that you will come across will be mea- sures of mortality because death and cause of death are regularly and reliably recorded in many, but certainly not all, countries. Incidence and mortality rates have exactly the same form, but for incidence we count new cases of a disease whereas for mortality we count deaths. Mortality data are obviously uninforma- tive for many diseases that are not usually fatal – things like osteoarthritis, non- melanoma skin cancers, psoriasis and rubella (German measles) to name but a few. But even for those diseases from which a high proportion of cases die, mor- tality figures might not mirror the underlying incidence of disease, for example if a more effective treatment is introduced. Mortality data can also lag well behind changes in incidence, delaying identification of changes over time that may be important for planning or for providing clues as to the causes of the disease. We will take up these issues in more detail when we discuss the role and uses of surveillance in health planning and evaluation (see Chapter 13).
Crude incidence and mortality rates
As you saw above, when we conduct an epidemiological study we calculate the incidence rate as the number of cases of disease divided by the total person-time at risk of disease (where this is summed over all of the individuals in the study). This method is particularly useful when different people have been followed for different lengths of time but at the population level we may be dealing with mil- lions of people and it is clearly not feasible to calculate the person-time that each is at risk. Instead we usually work on the assumption that everyone is at risk for the whole of the year that we are interested in.
When we are working with routine data, therefore, an incidence rate is cal- culated by dividing the total number of new cases of a specific disease (or the number of deaths) in a specified period, usually one year, by the average number of people in the population during the same period (Equation (2.2)). This is then usually multiplied by 100,000 (105) and presented as a rate per 105people per year. The size of the population will, inevitably, change over a period of a year, so ideally we would use the number of people in the population in the middle of the year, the ‘mid-interval’ population, for our calculations. Depending on the data available, this may be calculated as the average of the size of the population at the start and at the end of the period of interest, or estimated from census data. Incidence rates may be calculated for a broad disease group (e.g. cancer) or a
Measuring disease occurrence in practice: using routine data 47
Table 2.4 Crude mortality rates (per 100,000 per year) for ischaemic heart disease (IHD) in males from selected countries, 1995–1998.
Crude IHD mortality rate
Country (per 105/year)
Germany 211 Australia 168 Canada 160 Singapore 118 Spain 116 Japan 50 Brazil 47
(Data source: Global Cardiovascular Infobase, www.cvdinfobase.ca, accessed on 23 September 2003.)
more specific disease (e.g. breast cancer). Similarly, mortality rates may include deaths from all causes (sometimes called all-cause mortality) or only those from a specific cause. These basic rates are called crude rates because they describe the overall incidence or death rate in a population without taking any other fea- tures of the population into account (in contrast to standardised rates – see below).
Table 2.4 shows crude mortality rates for ischaemic heart disease (IHD) in men in seven countries. We see that Germany, Australia and Canada have high mortal- ity rates, with intermediate rates in Singapore and Spain and low rates in Japan and Brazil. This correctly describes the total burdens that the different health systems have to cope with, but does it give an accurate picture of the level of mortality from IHD in each country? If we were to take this information at face value we would conclude that the death rate for IHD is three to four times higher in Western countries such as Germany, Australia and Canada than in countries like Japan and Brazil. Is that reasonable?
Age-specific incidence and mortality rates
A major disadvantage of crude rates is that they are just that – crude. They do not take into account the fact that different populations have different age struc- tures and that the risk of becoming ill or dying varies with age. Many diseases are more common among older people and the older a person is, the greater their risk of dying. Developed countries like Germany have a high proportion of older people whereas less developed countries like Brazil have a much greater proportion of young people, at a relatively lower risk of dying. Their contrasting population structures are shown in Figure 2.3. In the example above it turns out that we are trying to compare countries with very different age structures, so a
0–4 5–9 10–14 15–19 20–24 25–29 30–34 35–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75–79 80 Ag e (y ear s) Percentage of population 10 5 0 5 10 15
Figure 2.3 Age distribution of the population in Brazil (1995, solid bars) and Germany (1998, hatched bars). (Drawn from: Global Cardiovascular Infobase, www.cvdinfobase.ca, accessed 23 September 2003.)