Page 50 - Clinical Small Animal Internal Medicine

18 Section 1 Evaluation and Management of the Patient

called the “prevalence” (see Chapter 2 for examples), and An alternative approach to obtaining prevalence
VetBooks.ir is contextually tied to a specific point in time. Knowing ratios when there is a health outcome is to sample sub-
sets of a population defined by certain characteristics.
such information may be valuable in planning preven-
tive initiatives and having an appropriate inventory of
sample a fixed number of sexually intact and sexually
pharmaceuticals available. Hospitals, however, may not For example, an investigator may choose to deliberately
have an adequate number of patients at a single time to altered male outdoor cats, and subsequently ascertain
estimate prevalence, so instead patients fitting inclusion the prevalence of FeLV in these two groups. A preva-
criteria may be captured over a period of time, and the lence ratio relating the prevalence of FeLV in one sex-
frequency of the factor of interest is then obtained as ual status group versus the other is estimable, but
patients are enrolled. This special type of prevalence is not the converse due to the sampling being based on
called “period prevalence” and is contextually defined by sexual status.
the beginning and end dates of patient accrual (e.g., the Another sampling variant of cross‐sectional studies is
proportion of patients entering a hospital over one the two‐stage prevalence case–control study. This
month present with clinical signs of upper respiratory approach is indicated when a relatively rare disease in a
disease). It is important to note that no assumptions clinical population is studied, and the goal is to estimate
need be made about when individuals with the outcome not the prevalence of the disease but instead factors that
actually developed it; it is sufficient to note that at the may be associated with it but cannot be readily meas-
time of measurement, they either had it or did not (the ured without additional data gathering and considera-
latter from either never having developed it or from tion of cost/time expenditures. Initially, a cross‐sectional
recovering from it). study is undertaken in a clinical population to ascertain
Prevalences measured from cross‐sectional studies are disease status. When the cost and effort of obtaining
not restricted to being crude (unconditional) measures. prevalence factor information prohibit procurement of
Factor‐specific (conditional) prevalences are also esti- it for every individual in the population, those with the
mable, as in sex‐specific or age‐specific prevalences. rare disease (prevalent cases) can have it measured,
Moreover, such specific measures can be jointly used for along with only a sample of the remaining nondiseased
comparative purposes, as in relative prevalences (preva- individuals (nonprevalent controls). Thus, prevalence
lence ratios). For example, it may be of interest to esti- factor information is efficiently obtained only on a sub-
mate the proportionate change in the prevalence of feline set of the entire population – either all or some of the
leukemia virus (FeLV) infections in intact outdoor males diseased individuals and a sample of the nondiseased
compared to sterilized outdoor males. individuals. The ratio of individuals in the two groups is
Sampling in cross‐sectional studies can be accom- at the discretion of the investigator, although a 1:1 ratio
plished in several ways. The most straightforward is to is statistically the most efficient for a fixed study size.
either study an entire population of individuals or take a The prevalence case–control study is the cross‐sectional
random sample from one. With either approach, indi- counterpart of the (longitudinal) incidence case–control
viduals can be cross‐classified by their health outcome study discussed later.
status as well as other factors to assess joint associations.
Using the previous example, the prevalence of FeLV can
be measured in intact and sterilized outdoor male cats, Longitudinal Observational Studies
or conversely the prevalence of male sexual status can be
measured in cats with and without FeLV. Prevalence The distinguishing feature of longitudinal medical stud-
ratios provide an objective measure of how relatively ies, compared to cross‐sectional studies, is that individu-
common or rare an outcome of interest is between als are monitored over time, typically being followed
groups, but should not be used for causal inference (such from a state absent a particular health outcome to a later
as making statements about likelihood, risk, or probabil- state that may or may not have experienced a change in
ity of a health outcome) because when measured simul- outcome status. The two principal types of such studies
taneously, it is often not possible to know if prevalence are cohort studies and case–control studies, which differ
factors, as putative risk factors, occurred prior to or in how individuals in them are sampled for study inclu-
after the health outcome occurrence. In addition, such sion. Longitudinal studies may also be characterized by
factors may not potentially alter only disease incidence their temporality; relative to the study onset, they can be
but also disease duration and recovery, all of which affect retrospective, with historical data being used during the
prevalence. Prevalence ratios, as measures of associa- follow‐up period, or prospective, with future data being
tion, therefore have utility for hypothesis generation but collected following study onset. Less commonly, studies
lack a foundation for making inferences about causal can be ambispective, with both retrospective and pro-
associations. spective components.