Prevalence is one of the most commonly used measures of disease occurrence in community ophthalmology, and unfortunately also one of the most misused. In this article the correct use and interpretation of prevalence is discussed.
Prevalence is a proportion
Prevalence is defined as the proportion of a population, or sub-population, that has a particular disease at a particular point in time. Prevalence of blindness is often reported as a percentage, such as 1.0 percent or 1.5 percent. These values refer to proportions: the number of blind per 100 persons in the population. For rarer diseases, such as childhood blindness, it is common practice to refer to the number of blind per 1000 population or per 10,000 population.
Occasionally, the term prevalence rate is used. A disease rate is a measure of disease occurrence per change in unit of time. Examples of disease rates include surgery rate (operations per year), birth rate (births per 1,000 women per year) and death rate (deaths per 1,000 persons per year). These measures are a count of the number of events per change in unit of time. The use of the term rate when referring to prevalence should be avoided as prevalence is strictly a proportion with no time element.
Measuring prevalence: the community survey
Prevalence of disease in the community can only be measured by going into the community and determining the disease status of all individuals in the community. If the population is too large for this to be undertaken, the disease status in a sample of individuals from the community can be measured (a sample survey).
While it appears straightforward and relatively easy, going out into a community and measuring disease status requires careful planning to ensure those individuals examined are representative of the community. If a sample of individuals is to be used rather than the entire community, then a careful and unbiased sampling technique needs to be used to make sure the sample is representative.
Special attention also needs to be given to ways of ensuring a high proportion of the sample is examined. If close to 100% of those selected for a survey are examined, then one can be confident about the reliability of the survey. If however, a low proportion of the sample is examined (<80%), then it is possible that bias may have been introduced, particularly if certain groups within the population are absent at the time of the survey.
When reporting prevalence, there are three different ways in which this can be done, and attention needs to be given to ensure the most appropriate is used for each situation.
1. Crude prevalence
This is the most widely used measure, and the one most people are familiar with. It is a summary measure:
p = n/N x 100
where (p) is the prevalence, as a percentage, (n) is the number of persons identified with the disease of interest and (N) is the total number of persons examined.
Crude prevalence is a useful statistic as it gives a summary of the burden of disease in the community.
For example, in the 1986 national survey of blindness in The Gambia,1 57 of the 8174 persons examined were blind (best visual acuity <3/60). The crude prevalence of blindness was therefore 0.70 percent.
p = n/N x 100 = 0.70%
2. Category-specific prevalence
It is often informative to present prevalence according to subgroups of the population. The most commonly used categories are age and sex. Shown in Table 1 is age-specific prevalence of blindness from the 1986 Gambian survey. Age-specific prevalence is calculated by dividing the total number of persons identified with the disease within each age category by the total number of persons examined within each age category.
Table 1. Prevalence of blindness
For example, among those aged 70 or older, 22 were blind out of 226 examined, giving an age-specific prevalence of 9.73 percent.
p = 22/226 x 100 = 9.73%
There are many other categories which could also be used to describe the distribution of disease prevalence. Other commonly used categories include geographic areas (such as health regions, local governments areas or suburbs of a city) urban-rural status, settlement sizes, and ethnic groups. Reporting disease prevalence with sub-populations may highlight differences in prevalence within the population.
It is useful to know this so that preventive and/or curative services can be targeted at those most affected.
Table 2. Prevalence of blindness by region
|Age||Western Region %||Central Region %||Eastern Region %|
Table 2 shows crude prevalence and age-specific prevalence of blindness from the 1986 Gambian survey, according to administrative regions. Crude prevalence ranged from 0.42 percent for the Eastern Region to 0.90 percent for the Central Region, a 2-fold variation in prevalence. Examination of age-specific prevalence highlights the variation in prevalence between regions. For example, among those aged 70 or older, age-specific prevalence ranged from 2.78 percent for the Eastern Region to 16.39 percent for the Central Region, a 6-fold variation in age-specific prevalence.
Relationship between crude prevalence and age-specific prevalence
It is important to be aware of the relationship between crude prevalence and age-specific prevalence. The crude prevalence is the sum of (S) each age-specific prevalence multiplied by the proportion of the population in that age group:
Crude prevalence = S (age-specific prevalence × proportion of population in age group)
Thus crude prevalence is dependent on two factors: the age-specific prevalences and the population’s age distribution.
Comparing crude prevalence between populations
A problem arises when comparing crude prevalence between populations, in that the populations may differ in their age structure. Even if two populations have identical age-specific prevalences, the crude rates will differ if there are differences in their age distributions. Thus, use of crude prevalence for comparing between populations may result in false conclusions being made.
Some of the observed variation between regions in crude prevalence in the Gambian survey was due to real differences in disease prevalence, and some was due to differences in age structure of the population in each region.
There are two ways of accounting for differences in age structure when comparisons are being made between populations. One is to base comparisons solely on the age-specific prevalences, as was done in Table 2. This is relatively simple, but may involve a large number of comparisons being made.
The other method is to present a summary measure of prevalence in which allowances have been made for differences in age structure. Such a measure is known as a standardised prevalence.
Table 3. Population age structures by region
|Age||All Regions %||Western Region %||Central Region %||Eastern Region %|
Table 3 shows the age distribution of the national population, and for the sample of persons surveyed from each region in the 1986 Gambian survey.
While the age distribution is similar for each region, there are minor variations. For example, age group 60-69 represents 4.2 percent of the Central Region’s population, but only 2.8 percent of the Western Region’s population. The higher proportion of older persons in one region may lead to an artificially elevated crude prevalence.
3. Age-standardised prevalence
The most commonly used standardised measure of disease occurrence in community ophthalmology is the age-standardised prevalence, a summary measure in which allowances have been made for differences in age structure between populations. The age-standardised prevalence is the weighted sum of the age-specific prevalences, with the weights taken from a standard population.
The use of crude prevalences for comparison between populations was confounded because there were differences in age structure of each region’s population. The calculation of age-standardised prevalence has removed these differences and enabled a valid comparison to be made between regions.
Table 4. Calculation of age-standardised prevalence for each region in The Gambia
|Age||National Age Structure %||Western Region
|A||B||A × B/100||C||A × C/100||D||A × D/100|
Table 4 shows an example of the calculations involved in determining the age-standardised prevalence of each of the three regions in the Gambian survey. The age distribution of the national population was used as the set of weights (column A) to which the age-specific prevalences of each region (columns B, C & D) were applied.
Adjusting for differences in the age distribution revealed that the highest prevalence of blindness occurred in the Western Region. This finding is clearly different from that in Table 2, where it was concluded that the Central Region had highest prevalence, based on the comparison of crude prevalences.
Which measure of prevalence is best?
This largely depends on how the prevalence data is to be used. Crude prevalence represents the actual burden of disease in the community and as such is useful for the allocation of resources and for programme planning.
The fact that crude prevalence is influenced by differences in underlying population age structures makes it difficult to make valid comparisons between communities, between regions or even between countries.
Age-specific prevalence is not influenced by differences in underlying population age structures and provides the most detailed information, but may require numerous comparisons to be made to fully interpret the data.
Age-standardised prevalence is also not influenced by differences in underlying age structures, provides a convenient summary measure of prevalence, and is particularly useful when comparing prevalence between populations.
In conclusion, there are several ways in which prevalence can be reported. Each has its own advantages and disadvantages. It is likely that a combination of these measures may best serve the reader.