Modification of general practitioner prescribing of antibiotics by use of a therapeutics adviser (academic detailer)

Abstract

Aims

This was a pilot study of the use of a clinical pharmacist as a therapeutics adviser (academic detailer) to modify antibiotic prescribing by general practitioners.

Methods

Following a visit by the adviser (March-May), 112 general practitioners were recruited and randomised to control or active groups. A panel of experts prepared a best practice chart of recommended drugs for upper and lower respiratory tract infections, otitis media and urinary tract infections. The adviser made a 10–15 min visit to each prescriber in the active group (June–July), gave them the chart and discussed its recommendations briefly. Doctors in the control group were not visited nor given the chart. Prescription numbers for all prescribers were obtained from the Commonwealth Health Insurance Commission for the pre(March-May) and postdetailing (August–September) periods using a three month lag time for data collection. Data for total numbers of prescriptions and for selected individual antibiotics used in these two periods were analysed using nonparametric statistics.

Results

Prescribing patterns were similar for the control and active groups in the predetailing period. For both groups, there were significant (P < 0.03) increases (45% for control and 40% for active) in total number of antibiotic prescriptions in the post compared with the predetailing period. This trend was anticipated on the basis of the winter seasonal increase in respiratory infections. In line with the chart recommendations for first-line treatment, doctors in the active group prescribed significantly more amoxycillin (P < 0.02) and doxycycline (P < 0.001) in the post vs predetailing periods. By contrast, doctors in the control group prescribed significantly more cefaclor (P < 0.03) and roxithromycin (P < 0.03), drugs that were not recommended. The total cost of antibiotics prescribed by doctors in the control group increased by 48% ($37 150) from the preto postdetailing periods. In the same time period, the costs for the active group increased by only 35% ($21 020).

Conclusions

We conclude that the academic detailing process was successful in modifying prescribing patterns and that it also decreased prescription numbers and costs. Application of the scheme on a nationwide basis could not only improve prescriber choice of the most appropriate antibiotic but also result in a significant saving of health care dollars.

Introduction

In recent years there has been considerable concern about appropriate prescribing practices and costs of the Pharmaceutical Benefits Scheme (PBS) in Australia [1]. For example, inappropriate prescribing of antibiotics may lead to the emergence of resistant bacterial strains [2, 3], as well as to increased costs, particularly where new expensive drugs are heavily promoted. Methods of addressing these problems for prescribed pharmaceuticals include the use of academic detailing [4, 5], feedback of cost information [6], provision of ‘practice guidelines’ [7], and provision of continuing medical educational information [8]. Of these methods, academic detailing has been the most successfully and widely used method for influencing prescribing [9–14].

A fax poll of General Practitioners (GPs) in the Osborne Division of General Practice in early 1996 indicated that 85% of them considered that prescribing practice needed to be reviewed and 71% agreed that visits by a therapeutics adviser (academic detailer) would be an acceptable method for dissemination of unbiased educational information. Prescribing of antibiotics has been one area where significant concerns have been raised about inappropriate prescription, development of resistance, choice of drug and cost [2, 3, 15–18]. In the present study we have used an academic detailing approach to modify prescribing behaviour for antibiotics used in the treatment of upper and lower respiratory tract infections, otitis media and urinary tract infections. Numbers of prescriptions for GPs who received an educational intervention and controls who did not were compared for 3 month periods before and after the educational intervention.

Methods

Development of best practice guidelines

Guidelines for prescribing were developed by an expert panel and summarized in the form of a one page chart of first-, second-and third-line drug treatment recommendations for urinary tract infections, bacterial tonsilitis, otitis media, acute bacterial bronchitis and mild pneumonia (Appendix 1). Recommendations were in line with published Australian therapeutic guidelines for antibiotics [19]. All antibiotics chosen were available on the Australian PBS.

Subjects and study protocol

At the time of study there were some 300 GPs in the Osborne Division of General Practice in the Perth Western Australia, metropolitan area. In the preliminary phase of the study, the therapeutics adviser was able to visit 148 GPs and 112 (80%) of them agreed to participate in the study. All gave written informed consent for their PBS prescribing data to be obtained from the Health Insurance Commission. The GPs were allocated randomly to either a control group, or to the active group. Numbers and costs of prescriptions written for each of 49 individual antibiotic PBS items were obtained for a three month (March–May 1997) period. During June and July, GPs in the active group were visited once by the therapeutics adviser. The visit consisted of a brief discussion of the best practice guidelines and each GP was given a laminated copy of the chart for future reference. The control group was not visited, nor given the chart. Prescribing data for both groups also were obtained for a second three month period (August-October, 1997) so that the effects of the therapeutics adviser’s visit could be assessed. A lag time of three months was allowed for counting prescription numbers for each period of interest. GPs were asked to complete a postvisit questionnaire on their immediate reaction to the intervention process. In addition, two months after the project was completed, a focus group was convened and attended by 15 GPs from the active group to assess concerns, acceptance and utility of the project and to suggest improvements.

Statistical analysis

The numbers of prescriptions per prescriber were not distributed normally, therefore between-and within-group comparisons of these distributions were made using Wilcoxon’s 2-sided rank sum test. The 0.05 level was used to test for significance. Demographic proportional data on the two groups of GPs were compared using Chi-square tests.

Results

The 56 GPs in each of the two groups had a similar demographic profile. There were 73% and 68% males in the active and control groups, respectively. Place of graduation was also similar in the active (WA 52%, other States 14%, overseas 34%) and control (WA 55%, other States 11%, overseas 34%) groups. The profile for year of graduation was also similar with some 70% of GPs in both groups having graduated between 1970 and 1989. As assessed by postcode, distribution of the GPs practice location was also generally similar between groups.

Total numbers of prescriptions, median numbers of prescriptions per GP, and total costs for prescriptions written by GPs in the active and control groups for the preand postintervention 3 month periods are summarized in Table 1. Prescribing data were categorized initially as ‘antibiotics’ and ‘all other drugs’. In the pre-intervention period, distributions in both these categories were similar for the control and active groups. Comparison of the pre and postintervention distributions within the active (P = 0.03) and control groups (P = 0.01) showed that antibiotic prescribing increased significantly, while other drugs remained unchanged. Sub-groups of the antibiotic data representing individual drugs recommended in the best practice chart (amoxycillin, amoxycillin with clavulanic acid, cephalexin, doxycycline, erythromycin, penicillin V and trimethoprim) and other commonly prescribed antibiotic products (cefaclor and roxithromycin) were examined in the same way (Table 1). In the preintervention period, prescription count distributions were similar for the active and control groups. Comparison of the preand postintervention data revealed several significant alterations in prescribing. In the active group, doxycycline prescribing increased significantly from a median of 1 to a median of 6 prescriptions/GP (P = 0.001). In addition, in the same group, amoxycillin 250 mg also increased significantly from a median of 3 to a median of 6 prescriptions/GP (P = 0.03). In the control group, there was a signficant increase in prescriptions/GP for cefaclor alone (median of 5.5 increased to 10; P = 0.03).

Table 1.

Numbers and costs for prescriptions written by GPs in the active (n = 56) and control (n = 56) groups for the pre- and postintervention 3-month periods.

Comparing pre and postintervention data (3 months each) for all antibiotics prescribed, the total number of prescriptions increased by 45% in the control group and 40% in the active group. The corresponding increases in cost were 48% and 35%, respectively. Overall, between the two 3 month periods, the GPs in the active group generated antibiotic prescriptions costing $16 130 less than their colleagues in the control group. Lower prescribing rates for cefaclor and roxithromycin (not recommended in the chart) by the active group accounted for 82% of the overall savings. As would be anticipated, for the ‘all other drugs’ category, both prescription numbers and total costs were similar in both periods irrespective of group allocation.

The postvisit GP assessment questionnaire (n = 38) indicated that GPs considered that the prescribing information was delivered in a clear (76%), concise (71%) and unbiased manner (68%), respectively. The focus group of GPs convened at the end of the study generally agreed that the visit by the therapeutics adviser was an acceptable and effective method of providing unbiased information based on evidence-based principles. The 10–15 min time of visit was seen as being appropriate, but some suggested that talking with small groups of practitioners could be an alternative. Both the verbal contact and the chart provided were seen as being valuable.

Discussion

In addition to discussing best practice guidelines for prescribing antibiotics for upper and lower respiratory tract infections, and urinary tract infections and leaving a laminated chart of these recommendations, our therapeutics adviser also exhorted GPs to resist patient pressure to prescribe antibiotics inappropriately. Although there was a significant winter seasonal increase in antibiotic prescribing between the two 3 month periods, the intervention for the active group resulted in marked changes in prescribing of ‘all antibiotics’ and of selected individual drugs. Prescription numbers and costs for the control group also increased markedly, presumably influenced in part by normal advertising exposure in the industry. In line with expectations, there was no seasonal increase in trimethoprim prescriptions for urinary tract infections. Thus, in agreement with previous investigations [13, 14, 20, 21], our study achieved significant alterations in prescribing by use of academic detailing.

Appendix Recommended antimicrobial therapy for URTI/RTI and UTI; average doses for adults with normal renal function.

Although cost data were included in the chart, drug cost was not discussed by the therapeutic adviser. Nevertheless, total savings achieved for three months ($16 130) for 56 GPs equate to $1152 per GP per year. Decreased prescribing of cefaclor and roxithromycin accounted for the majority of this change. Significant changes in prescribing patterns for amoxycillin (250 mg) and doxycycline were also apparent. While there were also changes for erythromycin and penicillin V, the total numbers of prescriptions for these items were too small to give reliable information. If anything, total cost savings may have been underestimated since the prescribing data we were able to obtain came from eligible pensioners and concession card holders who represent the vast majority of all prescriptions written under the PBS.

The total savings on antibiotics were approximately 1.4% ($16130/$1167 750 as a percentage) of total prescription costs for our active group of 56 GPs. In 1997, the total costs of the prescriptions written under the Pharmaceutical Benefits Scheme was $2,862 746 000 (PBS Statistics 1996–97) and potentially it is therefore possible to save $40 000 000 per annum if the scheme were applied nationally. Even after the cost of program delivery is included, it is likely that a national scheme could deliver substantial cost benefits to the Australian health care system. Moreover, increased adherence to best practice guidelines could lead to other benefits such as better treatment outcomes, less adverse reactions and less need for hospitalization, as well as decreased opportunity for the development of resistant bacterial strains. Soumerai et al. also found that a targeted academic detailing intervention was cost-effective, with greatest savings being achieved in high-volume prescribers [22]. Nevertheless, considerable caution would need to be exercised in the wider application of our scheme since Goldberg et al. [10] have found that implementation of academic detailing is very sensitive to local issues such as organizational culture, particular personnel and the disease conditions involved.

While overall cost savings coincidentally flowed from the present study, this might not always be the case. It is conceivable that appropriate prescribing recommendations might well involve the use of more expensive drugs and thus increase costs. Indeed, this occurred for doxycycline in our study. Nevertheless, if the process results in better patient care, presumably there would be indirect long-term savings in health care dollars.

The postvisit GP assessment of the study indicated general approval of the information delivery process using a clinical pharmacist (GG) as the therapeutics adviser, and suggested that the face-to-face contact visit was essential for success. In retrospect, the study really needed to encompass a longer period and/or a much larger number of GPs, particularly for less frequently prescribed items. We did not attempt to collect prescribing indications for individual prescriptions. This was considered in the design stage, but was rejected because it required the GP to provide a copy of each script, a process that could in itself easily bias the data. In addition, we considered that this additional task was likely to be unpopular with prescribers, and would likely be met with poor compliance. Finally, our study was carried out over a relatively short time (essentially three months) and while significant effects were documented, we did not have the opportunity to document the persistence-time of the changes. Nevertheless, our pilot project successfully altered prescribing patterns and demonstrated significant cost benefits. We conclude that a detailed feasibility study of the wider application of the therapeutics adviser method in the Australian health care system is justified.