In this article, we analyzed the concept of the range of drugs prescribed as an alternative or supplement to more usual measurements like the total number of prescriptions. The material in this article shows this to be a useful addition, especially when more precise clinical indicators are absent. Firstly because the range shows clear variations between ATC main groups and subgroups, and between GPs; secondly because it was possible to explain the variation between GPs by factors that have a theoretically interpretable relationship to the range. GPs with larger patient lists, GPs with higher prescribing volumes and GPs who frequently receive representatives from the pharmaceutical industry turned out to have a broader range when controlled for other variables. These statistical relationships could be linked to the hypotheses we formulated on the basis of the literature or theoretical considerations.
A measure which is comparable to the range is DU90% (the number of drugs constituting 90% of the volume expressed in DDDs). This has been used for GP data in Stockholm, Sweden, as a general quality indicator in feedback [32
]. In an intervention project aimed at decreasing the number of inhabitants per GP, no systematic change in DU90% could be detected [33
For measures of prescribing variation, ideally, guideline recommendations should be taken into account. Bergman e.a. [32
] did not find a relationship between DU90% and adherence to a local formulary. The theoretical link between the range of drugs prescribed and quality of drug prescribing is based on the use of restrictive formularies or guidelines. This link was not studied in our article. It would require specific information on the indication to prescribe and on the link between indications and guidelines or formularies. However, in the absence of detailed information on indications, the range of drugs could be used as an overall measure in addition to other drug-specific measures [34
The relationship of the range of drugs prescribed with the number of prescriptions per patient and list size can be easily understood by the fact that the chance that a greater number of different drugs will be prescribed is higher when the number of prescriptions is higher. A second explanation concerns the role of specialist initiated prescription that is subsequently followed by repeat prescriptions by GPs, a well known phenomenon in the Netherlands. This might vary between GPs because GPs who prescribe more usually also refer more to hospital consultants [35
]. Therefore high prescribing GPs might also prescribe more repeat prescriptions initiated by specialists, that often deviate from (GP-)guidelines. An additional explanation for the relationship with prescriptions per patient may be that a broader drug repertory increases the inclination to prescribe because it increases the therapeutic possibilities of the doctor. The causal direction is difficult to infer from cross-sectional data as used by us. This is, of course, a limitation of this study.
A further limitation of our research is that the possibility of multicollinearity cannot completely be ruled out. Although the correlations between the independent variables are below the threshold level of .60 the correlations with non-urbanized areas (.55) and dispensing practice (.43) come close to the threshold. In dispensing practices prescriptions of GPs dispensed by the practice cannot be separated from specialist prescriptions. Removing the 14 dispensing GPs from the analysis lowered the coefficients a bit but did not change our overall conclusion.
GPs working in partnerships did not appear to have a broader range when controlled for other variables. This seems surprising, because we know from previous research that GPs in group practices have a broader range. We expected to see this also on individual level since their patients more often switch between GPs in group practices. Table shows that GPs in group practice more often use oral information sources and receive fewer representatives of the pharmaceutical industry. On the other hand, working in a group practice is often associated with a stronger orientation to esteem by colleagues and to professional guidelines, leading to more rational prescription behaviour and therefore to a lower range[36
]. The more frequent use of oral information (which will often be from colleagues) may point to this (see table ). This might neutralize the range-broadening effects of patients switching between doctors.
Contrary to our expectations, we did not find any statistical relationship with factors associated with the composition of the practice or with the closely linked factor of the location of the practice. These were, in fact, proxy-indicators for differences in morbidity. This brings us to the second limitation of the study, which is that we did not take the diagnosis underlying the prescriptions into account. As a follow-up to this study, we recommend an analysis of the relationship between range and rational prescribing, taking diagnosis and comorbidity into account. This would also provide better opportunities to link the range of drugs prescribed to guidelines. A more specific follow-up study would be to investigate how computerized decision support systems affect the range of drugs prescribed.
Further and more detailed research is necessary into the differences between therapeutic main groups and subgroups with regard to the percentages of available drugs prescribed. We would hypothesize that GPs have a greater inclination to switch to other drugs in therapeutic groups where side-effects play an important role (i.e. hormonal contraceptives). The same could be true of therapeutic drugs with limited or variable effectiveness (i.e. antifungals for dermatological use). This is the grey area where representatives of the pharmaceutical industry play a role, stressing the advantages of their products compared to other products, in the form of fewer side-effects and better results.