The intervention was based on theories and experience usually referred to as social marketing or academic detailing.4,5,6,10
It is a framework for dissemination and implementation of activities to improve prescribing. A combination of adult learning theories and the marketing experience of the pharmaceutical industry are directed at improving the rationality of prescribing.
All doctors and pharmacists from groups assigned to the individual visit intervention arm were individually contacted by telephone. They were told of the aim of the study (to improve antidepressant prescribing in elderly people and measure the effectiveness of an educational programme) and invited to participate in the programme. For those who agreed, an appointment was made for a 20 minute visit with the lead investigator (MvE), who is a doctor. This session emphasised the unique therapeutic difficulties of treating older people and the problems of anticholinergic side effects. Participants were given a leaflet containing an evidence based summary of the most important information.
All sessions were based on a priority list for issues to be discussed. Depending on the length of the visit and the responses of the professionals, the following items were discussed (in order): altered pharmacodynamics and kinetics in elderly people,18,19
increased vulnerability of elderly people to side effects,20,21
the need to avoid anticholinergic antidepressants in elderly people,22
and difficulties in diagnosing depression, especially in elderly people.17
Participants were shown the overall data on prescribing of antidepressants in the past year to illustrate that most anticholinergic antidepressants are prescribed to people aged over 60.14
The initial visits included no further comment on personal performance. At the end of each visit another appointment was made for about four months later. During the second visit a graph was provided showing personal performance and the proportion of prescriptions for anticholinergic antidepressant versus less anticholinergic antidepressants in three age categories: under 60, 60-70, and over 70 years old.
For the group intervention arm, all group coordinators were contacted to ask permission to use one full meeting for the educational programme. The content of these presentations was essentially the same as in the individual contacts. At the end of the first visit, permission to use part of another meeting was requested. In this second meeting, a graph of accumulated prescribing in the group was shown and personal graphs were handed out. All contacts for both intervention arms were performed by MvE. The control arm was not contacted.
The effectiveness of this intervention is best reflected in the choice of antidepressant for patients starting treatment. To define incident users of antidepressants, we used the prescription reimbursement records described above. For each prescription we calculated the number of days the prescription would cover, using the prescribed daily dose and the package size.14
In this way, a time window of probable use was created. We assessed all antidepressant prescriptions from July 1995 onwards. If the patient had not previously been prescribed antidepressants or if the interval since the last prescription was over 180 days, the patient was considered an incident user.14
In the Netherlands all people with national health insurance are allocated to a general practitioner. We determined the periods before the educational visits, between the visits, and after the visits for each general practitioner in the region to allocate each incident patient to the right period for each general practitioner. In order to calculate incidence rates (number of incident users/1000 person years), the number of patient days per period was calculated to determine the denominator. For general practitioners in the control arm and those who declined to participate in the intervention arms we assigned average visit dates calculated from data on the doctors in the intervention arms who were visited. Incident use of highly anticholinergic antidepressants and less anticholinergic antidepressants was calculated for each general practitioner per period.
We used a Poisson regression model to estimate rate ratios of starting highly anticholinergic antidepressants and less anticholinergic antidepressants in elderly people in both intervention arms in relation to the control arm. The evaluation was done on an intention to treat basis in order not to overestimate the effect of the intervention by including only the most responsive doctors. Since randomisation was performed at a group level and correlated outcomes within a group can influence precision (95% confidence intervals),25
we studied rate ratios with and without correction for correlated outcomes (exchangeable correlation matrix) using longitudinal data analysis (Spida). This did not materially influence outcome. Point estimates were virtually identical and 95% confidence intervals changed less than 3% (there was no change in significance of effects estimates).
Since it was not possible to correct for baseline rates with Spida, we used Egret, although in this program it is not possible to analyse correlated Poisson outcomes. In Egret, rate ratios were estimated after correcting for sex and baseline rates, with baseline rates as an offset variable. We estimated the effects of the first and second visits and of both visits together. The effects in each intervention arm and of both interventions together were also measured.