Overview. Prevalence is the number of cases of a condition per population at risk at one time or in a relatively short period of time. In a cross-sectional study, prevalence rates of disease among those with varying levels of exposure are measured and sometimes compared between groups. Cross-sectional studies can be used descriptively, to describe differences in prevalence between groups, or analytically, to test hypotheses. Cross-sectional study designs are generally less useful in studying disease causation but are very important in public health planning and evaluation.
Limitations in Etiologic Research. The main problem with cross-sectional study designs in etiologic research is the difficulty of sorting out temporal relationships. For example, a cross-sectional study of the relationship between calcium consumption and osteoporosis might find that those with high calcium consumption were more likely to have osteoporosis than those without, but this finding might be because people who know that they have osteoporosis increase their calcium consumption when told of this diagnosis. Interpretation of cross-sectional studies in terms of etiology is clear only for potential risk factors that will not change as a result of the disease, such as ABO blood groups or HLA antigens.
A second limitation of cross-sectional studies in etiologic research occurs because cross-sectional studies include both new and old cases. Accordingly, the case group will have more than its fair share of individuals with disease of long duration, because those who die or recover quickly will not be included. If the etiologic exposure affects disease duration or the likelihood of dying, relationships between exposure and disease will be distorted. In the most extreme instance, which is both hypothetical and unlikely, those with the disease who are exposed all die immediately upon developing the disease but those developing the disease who are not exposed live for a long time. No cases included in a cross-sectional study that assesses disease and exposure will be exposed (they are all dead) and, the exposure will appear to protect against the disease.
Uses in Public Health Planning and Evaluation. Cross-sectional studies are very important in public health planning and evaluation. They are widely used in these settings for a variety of purposes. For example, if a public health administrator wants to obtain an idea of how many and what sort of facilities are needed to treat people with a certain disease at a given point in time, knowledge of the prevalence of the disease in the community is important. Often prevalence rates are needed for specific segments of the population or according to the severity of the disease, since different methods of treatment and types of facilities may be needed for people with various stages of the disease. Cross-sectional studies are also used to help set research priorities based on consideration of the burden of the disease. For example, a study of the prevalence of chronic gynecologic conditions among US women of reproductive age found that the most common conditions are menstrual disorders, adnexal conditions, and uterine fibroids. This information suggests that not only are more effective treatments for these disorders needed, but also that more research on their causation would be highly desirable (Kjerulff et al. 1996).
Sampling Issues in Cross-Sectional Study Design. Determining the prevalence of a disease (or other condition) in a community often involves sampling people in the community and measuring the occurrence of disease by questionnaire, physical examination, or other method. When taking a sample of the population, it is important to use scientifically sound sampling methods. One may be tempted, for instance, to save money by asking for volunteers to be in a study (e.g., convenience sampling). Volunteers, however, almost always have different characteristics from the population as a whole. Volunteers for health surveys tend to be overrepresented with the "worried well" and with people who believe they will benefit from participation. One should also avoid the temptation to save money by letting study personnel choose the people to be included in a survey. These personnel may select the most accessible people or those who they think are most likely to benefit from participation.
When deciding which of several scientifically sound sampling methods to employ in a given situation, it is important to keep in mind that the basic purpose of sampling, rather than making measurements on an entire population, is to save time and money. Sampling may also result in greater accuracy of measurements, since more effort can be spent on ensuring that the measurements are of high quality if a manageable number of people is included.
Perhaps the most familiar method of sampling is simple random, sampling, in which each member of the population has an equal chance of being included in the sample. Although simple random sampling is simple to carry out and easy to understand, it may not be the most efficient method of sampling in many instances.
Another type of sampling is systematic sampling, in which the units are sampled at equal intervals; for instance, every 10th person on a list may be selected for the sample, or every fourth house on the block may be included. Systematic sampling is often simpler to administer under field conditions than simple random sampling, and it also does not require a list of the entire population in advance.
In stratified sampling, the population is divided into strata, or groups of units having certain characteristics in common (such as males and females, or geographic areas within a city), and a simple random sample is then taken from each stratum. Stratified sampling, which is frequently used in practice, offers several advantages over simple random sampling. With stratified sampling, one can be certain that members of each stratum are represented in the overall sample. If the strata are more homogenous than the population as a whole, then more precise overall estimates for the entire sample can be made. Finally, the strata can be constructed so that those that are least expensive to include, or in which there is the most variability, can be sampled more heavily, simultaneously reducing cost and increasing precision.
In cluster sampling, clusters (e.g., schools) rather than individuals (e.g., children) are sampled and then measurements are made on all children within the school. Cluster sampling can offer two major advantages. First, it is not necessary to enumerate the entire population, only the individuals in the clusters that are selected. Second, cluster sampling can allow the inclusion of larger numbers of individuals for a given cost, because once one has access to the unit (e.g., the school), one can efficiently make measurements on all of individuals in the unit.
In multistage sampling, combinations of these methods are used within the same survey. Stratified sampling might be used to ensure that schools representing different socioeconomic areas of a large city are represented in the sample, and cluster sampling of classrooms within the selected schools might then be employed for efficiency.
The different methods of sampling require different statistical procedures in order to obtain estimates of measures that are applicable to the population from which the sample was taken and to provide the correct estimates error associated with these measures. Estimates made on the basis of cluster sampling must take into account that the individuals sampled within a cluster tend to be more alike than individuals from different clusters. Estimates made from stratified sampling involve combining in an appropriate way the estimates made from the individual strata. The reader is referred to sampling textbooks (e.g., Levy and Lemeshow 1991) and specialized software programs (e.g., SUDAAN 1991) for descriptions of methods of estimation when various sampling schemes have been used and for more detailed descriptions of the sampling methods described above.
Was this article helpful?