Debate About the Safety of Oral Contraceptives

Excelsior College PBH 321

Page 1

ECOLOGICAL ST UDIES AND CROSS-SECTION AL STUDIE S Apart from experimental studies, all other study designs in epidemiology are observational designs. We now begin a summary of two such designs. Ecologic Studies Ecologic studies examine disease rates in relation to a factor (exposure) described on the population level. Ecologic studies often identify groups by place (e.g., county, state, and country) or time period, and often both. This means the units of analysis are groups of people rather than individuals, and comparisons are usually of place or time. The population-level factor summarizes information for all individuals within a specific population. For example, an ecologic study might examine the relation between county-level air pollution and mortality rates. We may also compare disease rates for a geographically defined population over time, such as HIV seroprevalence over time and needle exchange programs (yes/no). Exposure and outcome are measured for the same time period. For example, we can compare mortality between 2 counties in New York State and simultaneously examine the average level of an air pollutant for those two counties for the same time period. An important source of bias in ecologic design is something called the ecologic fallacy. This is when an association observed between variables on a group level does not necessarily represent the association that exists on the individual level. For example, if we take an average air pollution measure for an entire county, this may not actually be the level of pollution that every individual is exposed to. Some individuals may be exposed to much less, others to much more. The advantages of conducting ecologic studies are:

• The types of data commonly used are routinely collected for purposes other than conducting research, thus reducing the cost of the study since the epidemiologist does not have to collect the data.

• Some exposures are inherently ecologic. For example in a community with a lot of air pollution sources (factories, traffic, etc.), it may be easier to assign exposure on the group-level than for an individual.

• You can study many different types of exposures on one type of health outcome. • You can study what we refer to as “contextual effects” on health, such as looking at the impact of

community-level poverty on health.

The major disadvantages of the ecologic design include: • The quality of the data may vary. Since the data is typically not collected for conducting an

epidemiologic study but instead for other purposes, information may be missing or not as complete as we would like.

• There is difficulty in identifying complex relationships. Since we are averaging information for a group, we may “wash out” some relationships and not observe them.

Excelsior College PBH 321

Page 2

• Temporality is difficult to establish. When we measure both exposure and outcome at the same time, we cannot be sure that exposure preceded the outcome. (Logically, if exposure does not precede disease, it cannot be a causal factor for that disease.)

• The ecologic fallacy. Cross-sectional Studies Cross-sectional studies also examine a health outcome in relation to exposure in a defined population at a specific period of time. These studies can be envisioned as a “snapshot” of exposure and disease - where current disease status at a particular time is examined in relation to current exposure level during the same time period. In other words, there is no follow-up over time. A common method of conducting a cross- sectional study is via surveys, where questions regarding exposure and disease are asked of an identified population. Cross-sectional designs are useful for public health planning purposes, such as identifying the number of beds needed in a hospital or determining a community’s need for family planning services. They are often useful for hypothesis generation, since they may provide clues as to possible risk factors for a condition that can be better examined with a more rigorous study design. The advantages of conducting cross-sectional studies are:

• Can typically be conducted over a short time period, reducing study cost. • Good for hypothesis generation. • If conducted among a sample of the population, we may say that study findings are generalizable.

The major disadvantages of the cross-sectional design include:

• The quality of the data may vary. Since the data is often not collected for conducting an epidemiologic study but instead for other purposes, information may be missing or not detailed adequately.

• Cross-sectional studies usually are generally limited to identifying prevalent disease, because they measure disease status at only one time point.

Just as in the ecologic design, temporality between exposure and disease is difficult to establish.

E

O

Measured during the same time

period

E

O

Measured during the same time

period