The methodology of the scheme has been described in detail elsewhere and is, thus, only outlined here. The general principle is to identify the first 20,000 patients dispensed the new drug and to extract all medical ‘events’ recorded for these patients in the datafiles after the initial prescription while they remain on the medication. An ‘event’ is any diagnosis recorded in the hospital discharge or physician service claim datafiles. Time following the initial prescription was divided into 30-day periods, and events were grouped into the appropriate period by their date of occurrence.

The data used to evaluate the feasibility of the signalling scheme were those produced as a result of linking patients treated for psychiatric disorders and/or dispensed psychoactive drugs between 1976 and 1986 . In this period, a primary discharge diagnosis and, if appropriate, a secondary diagnosis were recorded in each hospital discharge claim using fourdigit International Classification of Diseases (ICD) codes . Only one diagnosis is recorded in each physician service claim using three-digit ICD codes, together with special codes assigned by the physician claims branch either to identify certain conditions more distinctly rather than using the normal ICD code or to identify examinations, procedures and interpretations not associated with a specific diagnosis.

Between 1976 and 1986, Canadian drugs given general coverage under the prescription drug plan were dispensed to Saskatchewan residents for a fee that was a nominal part of the real cost. Both benzodiazepines chosen for the study (alprazolam and lorazepam) are medium-acting products with similar length half-lives (10 to 20 h) that are usually prescribed to alleviate anxiety but are also used for sleep disorders, depression and panic attacks . All benzodiazepines are closely related, but alprazolam’s triazolo-structure makes it a little different from most of the others. Neither drug was selected because of any specific doubts about its safety. Lorazepam was chosen because it is an ‘average’ benzodiazepine that achieved significant market penetration in Saskatchewan in a reasonably short period, while alprazolam was selected because it has been suggested as being more useful in mixed anxiety and depression than other benzodiazepines.

Lorazepam was first given general coverage under the Saskatchewan drug plan in January 1981 and alprazolam in January 1984. The lorazepam patients were the first 20,000 identified as new users and were accumulated from 1981 to 1983, but only 8525 new users of alprazolam were identified between 1984 and 1986.

Because the scheme was designed to signal acute adverse events that may be potential ADRs, rates of events occurring in the first 30 days after starting a new drug (it was assumed that most acute ADRs would occur during this period) were compared with those occurring during the subsequent six 30-day periods. Estimates of the excess rate of occurrence per 1000 patients during the 30 days over those for the succeeding six periods and 99% confidence intervals (CIs) for these estimates, assuming a Poisson model, were calculated.