Drug utilization information can be used to draw conclusions about something other than drug use per se. For example, anti-epileptic drug prescriptions have been used to estimate the prevalence of epilepsy, and patient drug use records may inform us about their health status. In this way, drug utilization measures are used as an indicator for concepts more dicult to measure, such as prevalence of disease, treatment of adverse events, health status, quality of care. The objective of this article is to describe how drug utilization can be used as indicators, how the use of an indicator impacts the methodology and ยฎnally to suggest some methods to create a more robust indicator.
First, we need to deยฎne an indicator. According to Webster, the verb to indicate' means to point out or to point to' or to be a sign, symptom, or index' while the noun indicator' means one that indicates' 1 . According to Last's dictionary a health indicator may be deยฎned as A variable, susceptible to direct measurement, that reยฏects the state of health of persons in a community. Examples include infant mortality rates, disability days, etc.' 2 . This is a case where a precise measurement, such as a mortality rate, is used to suggest something about a much vaguer concept as the health status of a country. Mortality is not exactly what we want to use, but as an indicator it provides a quantity that can be easily measured and analysed. More directly relevant for our purposes is the term indicator variables', a which are . . . variables hypothetically linked to the variable [studied], which itself cannot be directly observed' 3 .
This kind of surrogate measure has much potential in pharmacoepidemiologic methodology and, in fact, is being used. Other familiar names used for the same idea are markers, proxies, or surrogate measures. It is especially important to recognize that we are measuring something else than what we really want to know. Five situations in which drug utilization measures have been used as an indicator will be presented.