at Tuskegee University

College of Veterinary Medicine







  Level II ›



 Chapter 7:  Epidemiologic Methods



Topic 1: Causal Associations
Topic 2: Risk Analysis
Topic 3: Cohort (Prospective) Studies
Topic 4: Case Control (Retrospective) Studies

Assignment: Ch. 11, 12 Text

Ch. 6, 8 Susser: Causal Thinking in the Health Sciences

The objective of many epidemiologic studies is to identify factors that are causally related to the occurrence of a disease and, once such a factor is found, to assess how much it contributes to the occurrence of the disease. If a causal association between a factor and a disease is shown to exist, the factor will be called a determinant. Given this association, one may expect that the higher the level of the factor, the more frequent the disease in the population. Conversely, one may also expect that manipulation, more specifically elimination of the factor would lower the incidence of the disease, i.e. the occurrence of new cases. This of course is prevention and is one of the major goals of epidemiology.

Epidemiologic studies that aim at identifying associations between a factor and disease frequency can be observational or experimental; they can be cross-sectional or longitudinal. In an observational study events such as factor prevalence, disease prevalence or disease incidence are observed and recorded as they present themselves in nature, either at a point in time (cross-sectional) or over a period of time (longitudinal). In an experimental study the exposure factors are manipulated, i.e. introduced in a controlled manner, at varying levels, over differing lengths of time; the resulting disease frequencies are recorded.

Although a causal relationship between a factor and a disease must be unidirectional in time, i.e. from cause to effect, epidemiologic studies can proceed in either direction in time, from cause to effect (cohort studies), or from effect to cause (case-control studies). In a cross-sectional study of factor prevalence and disease prevalence the time relationship must be assumed; in a longitudinal study this time relationship is respected.

Most surveys, particularly prevalence surveys, are cross-sectional studies done to estimate a population parameter such as the prevalence of a disease (infection): P(D); the prevalence of a risk factor: P(F); and to search for an association between the two.

It is convenient to exhibit the results of a cross-sectional study in a contingency table according to the following model:

Disease Factor Present Absent

For (the) ketosis survey referred to previously this would be:

Factor + - Total
No Stanchions


One can say that a factor and a disease are associated when the disease occurs at a "much" higher rate in the presence of that factor than in its absence, i.e.,

pdf >>pdf_small

One can also say that a factor and a disease are associated when the factor occurs at a "much" higher rate in the diseased animals than in the non-diseased group, i.e.

pfd >> pfd_small

From the above it is obvious that there is a risk of disease for both the exposed and the non-exposed groups of animals. These risks expressed as incidence or prevalence are:

DFF and DF_small , respectively.

There is, however, an excess risk for the exposed group. This excess risk can be measured by computing either the ratio of the two risks (relative risk) or the difference between the two (attributable risk).

A. Association between factor and disease

Three levels of association between a factor and a disease are generally recognized: an apparent association, a statistical association, and a causal association. The following hypothetical example serves to illustrate these types of association. Let us assume that the incidence of ketosis in fresh cows was 160/1000 per year in group A and 480/1000 per year in group B, and that the majority of cows in group A are housed in free stall barns while in group B they are housed in barns with stanchions (Table 1).

Type of Housing Incidence of Ketosis
Free Stall

1. Apparent association

From the above table one could easily be tempted to say that there might be an association between the type of housing and ketosis, more specifically that there is an apparent association between stanchion type housing and the incidence of ketosis.

This apparent association is the lowest level of association, illustrated below.


Many associations of this type are derived from descriptive epidemiologic studies, e.g. prevalence surveys: the morbidity rate in one sub-population is found to be higher than the morbidity rate for the whole population, or than the morbidity rate for another similar sub-population. The question is asked: what factor can account for the difference, specifically, what factor accounts for the higher morbidity rate. In an attempt to answer, one seeks to identify a number of candidate factors and begins to collect frequency data for these factors in the respective populations.

2. Statistically significant associations

Once an apparent association is shown, one will ask: is the association statistically significant?


In the case of the example given above, where the risk of disease is three times greater for cows in stanchions than for cows in free stalls, a chi-square test would soon reveal a statistically significant association. This does, however, not mean that the association is causal? If it is not causal, what else can it be?

A statistically significant association may be spurious (false); it may be due to bias in the study method; or to chance occurrence, i.e. when the outcome of an event is declared statistically significant even though it results from random variation.

A statistically significant association may also be secondary.


The disease and the factor are associated with only because both are related to some other underlying conditions. The association is non-causal since manipulation of the factor will not affect the disease frequency.

    forwar_arrow Next page














Tuskegee University is accredited by the Commission on College of the Southern Association of Colleges and Schools  (1866 Southern Lane, Decatur Georgia 30033, (404) 679-4500 to award baccalaureate, master's professional and doctoral degrees

All content Copyright © 2009 SVM, Tuskegee University. All Rights Reserved.