Family Practice Vol. 16, No. 5, 495-500
© Oxford University Press 1999
Influencing antibiotic prescribing in general practice: a trial of prescriber feedback and management guidelines
Royal Australian College of General Practitioners Training Program, PO Box 197, North Ryde NSW 2113, Australia,
a School of Community Medicine, University of New South Wales,
b Faculty of Medicine, University of Sydney and
c Faculty of Medicine, University of Newcastle, NSW, Australia.
N Zwar, Wolk J, Gordon J, Sanson-Fisher R and Kehoe L. Influencing antibiotic prescribing in general practice: a trial of prescriber feedback and management guidelines.Family Practice 1999; 16:495–500.
Received 18 September 1998; Revised 23 April 1999; Accepted 13 May 1999.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodResultsDiscussionReferences |
|---|
Background. The extent of use of antibiotics to treat upper respiratory infections in general practice is an area for concern due to the increasing problem of bacterial resistance. Effective educational strategies to promote rational prescribing are needed.
Objectives. We aimed to examine the effectiveness of prescriber feedback and management guidelines in reducing antibiotics prescribing by GP trainees for undifferentiated upper respiratory tract infection, and in improving the choice of antibiotic for tonsillitis/streptococcal pharyngitis. The research tested a stepwise approach to targeting educational input to high prescribers.
Method. General Practice trainees in New South Wales (n = 157) were randomly allocated to a treatment group (n = 78) which received an education intervention on antibiotic use, or to a control group (n = 79) which received an intervention on an unrelated topic. Trainees completed three practice activity surveys, each of 110 consecutive patient encounters, with 6-month intervals between surveys. Prescriber feedback and management guidelines on use of antibiotics for URTI and choice of antibiotic for tonsillitis/streptococcal pharyngitis were delivered in a written form between surveys 1 and 2. An educational outreach visit to high prescribers occurred between surveys 2 and 3. Outcome measures were the rate of antibiotic prescribing for all indications, for URTI and prescribing of select antibiotics for tonsillitis/streptococcal pharyngitis.
Results. Antibiotic prescribing by the intervention group declined over three occasions from 25.0 to 23.3 to 19.7 per 100 URTI problems, while the control group increased from 22.0 to 25.0 to 31.7 per 100 URTI problems (P = 0.002). Prescribing in agreement with accepted guidelines for tonsillitis/streptococcal pharyngitis increased over time in the intervention group from 55.6 to 69.8 to 73.0 per 100 problems, but decreased in the control group from 59.6 to 57.5 to 58.5 (P = 0.05).
Conclusion. Prescriber feedback and management guidelines were shown to influence antibiotic prescribing for URTI and choice of antibiotic for tonsillitis/streptococcal pharyngitis. This study provides a model for targeting educational input to those prescribers who most need to change their behaviour.
Keywords. Antibiotics, feedback, guidelines, prescribing, respiratory tract infections.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionReferences |
|---|
Approximately 80% of the Australian population visit a GP in any 12 month period1 and between 60 and 80% of general practice consultations involve a prescription.2 Antibiotics are the group of drugs most commonly prescribed in general practice, accounting for 20% of all prescriptions. There is evidence that the Australian population has a higher rate of antibiotic consumption than comparable countries such as Britain, the United States and Sweden.3
A major and continuing criticism of antibiotic use in community practice is frequent prescribing for self-limited viral respiratory illness.4,5 In Australia there is also a problem with inappropriate choice of antibiotic agent, with the widespread use of broad spectrum drugs for acute sore throat.6,7 There is evidence of rapidly increasing rates of resistance to antibiotics in common bacterial pathogens and that the widespread use of antibiotics, plus the breadth of spectrum of the drugs used, is contributing to this increase.8
Soumerai et al.9 have reviewed educational methods for improving prescribing in primary care. Feedback to prescribers, mailed-out educational material, group education and educational outreach (or ‘academic detailing’10) were four of the approaches examined with the most convincing evidence found for the effectiveness of educational outreach. However, subsequent studies of educational outreach have been less convincing.11–13 Those studies that have shown an effect have moved away from using exclusively the educational outreach approach14,15 towards more complex interventions involving elements of a number of educational and administrative strategies.
There is evidence that prescribing feedback can be an effective intervention,16,17 particularly where individual prescribing profiles are accompanied by therapeutic recommendations.18 Both Sanzarro19 and Eisenberg20 have suggested that principles of effective feedback are that it is individualized and gives more information on behaviour compared with specific standards. Eisenberg further suggests that feedback is more effective if delivered in an interpersonal encounter, although research evidence to support this view is limited.
This study set out to evaluate an innovative educational intervention using individual feedback and comparison to a set of standards delivered in a face-to-face encounter. The intervention incorporated the concept of stepwise care in behavioural change,21 where more attention is given to those individuals who are not responding to a low-intensity intervention. The objectives of the intervention were to reduce antibiotic prescribing by general practice trainees for undifferentiated upper respiratory tract infection (URTI) and to improve the choice of antibiotic for tonsillitis and streptococcal pharyngitis.
|
Method |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionReferences |
|---|
A randomized trial was conducted with reciprocal controls where general practice GP trainees were allocated to an education intervention on antibiotic use or a control group which received an intervention on an unrelated topic (benzodiazepine use). GP trainees enrolled with the Royal Australian College of General Practitioners Training Program in New South Wales were approached to participate. Of a total of 707 trainees, 237 eligible trainees were identified.
Excluded were those trainees in hospital rotations, those working less than 12 hours per week and those planning more than 3 months' absence from general practice during the study. Of the 237 eligible trainees, 193 consented to take part in the study (eligible consent rate 81.4%) and were randomized into groups. There were 36 losses to follow-up over the study and prescribing data from these doctors were excluded from the analyses. The most common reason for withdrawal from the study was taking time off from general practice for travel or for maternity leave (18 trainees). This left 157 who completed all three practice activity surveys, 78 in the intervention group and 79 in the control group.
Trainees in the study completed three practice activity surveys of 110 consecutive patient encounters each with 6-month intervals between surveys. Trainees were told that the study would examine methods of improving the management of common general practice conditions but were not told that antibiotic prescribing was a key issue before completing the pre-intervention survey. As the intervention involved feedback on antibiotic use for URTI and tonsillitis, the trainees were aware that this was of interest to the researchers when they completed the subsequent morbidity and treatment surveys. Data was coded using the International Classification of Primary Care.22 The survey instrument and coding accuracy checks followed the methods used in the Survey of Morbidity and Treatment in General Practice in Australia 1990–1992.
Intervention
The education intervention on antibiotic prescribing was delivered as a two-step process. The first step consisted of individual prescriber feedback for URTI and tonsillitis/streptococcal pharyngitis, specific management guidelines on acute sore throat, and patient handouts on symptomatic management of sore throat. These materials were mailed to the GP trainees following the first practice activity survey. The second step followed the second practice activity survey (survey 2). Prescriber feedback and management guidelines were again mailed to GP trainees in the antibiotic intervention group. An educational visit was undertaken with those trainees who at survey 2 were prescribing an antibiotic on more than one occasion for every ten URTI problems managed or who were using other than the recommended antibiotic for tonsillitis or streptococcal pharyngitis (n = 48). Trainees located in rural areas received the educational intervention by telephone (n = 11).
The prescribing threshold was based on the outcome of a consensus panel held specifically for the purposes of the research project. The panel consisted of GP trainees, GP educators and infectious disease specialists who were asked to define appropriate levels of prescribing in general practice. Nineteen trainees were identified as prescribing below the threshold for antibiotic prescribing for URTI and in accordance with guidelines for tonsillitis, and therefore were defined as not needing the educational visit or telephone call.
Educational visits comprised one 20-minute appointment, and were conducted with trainees in their practices by senior GP trainees trained in the technique of educational outreach.10 The trainee's prescribing feedback, the assessment and management of sore throats and URTI, and the barriers for that individual doctor in changing prescribing behaviour were explored. The telephone interview with rural trainees covered the same content.
Statistical analyses
GP trainees were the unit of randomization and the unit of analysis. Outcome measures were (i) the rate of antibiotic prescribing for all indications, (ii) for URTI and (iii) the rate of prescribing of select antibiotics for tonsillitis/streptococcal pharyngitis. For all outcome measures and for diagnosis rates of common respiratory infections per 100 encounters, the statistical test used was a repeat measures ANOVA with planned orthogonal contrasts comparing (a) surveys 1 and 2; and (b) the average of surveys 1 and 2 with survey 3. Across the three surveys, standard diagnostic tests for the assumptions underlying ANOVA, particularly homogeneity of variance, were performed and found satisfactory. In the analysis of prescribing for all indications and for URTI (outcomes 1 and 2), less than 10% of doctors had no opportunity to prescribe in any one practice activity survey. These doctors had the group mean applied to their prescribing for that occasion.
For the third outcome measure, greater than 30% of doctors for each survey had no opportunity to prescribe a select antibiotic for tonsillitis/streptococcal pharyngitis because no cases presented with these conditions. Consequently, about 70% of GP trainees had incomplete data over the three occasions, too many to be replaced with group means. An alternative method for handling these data whilst retaining the repeat measures ANOVA model was to randomly assign all GPs to eight subgroups: four intervention subgroups and four control subgroups. The mean rate of prescribing of select antibiotics was computed for each subgroup, and these subgroups were used in the ANOVA model as though they were individual prescribers. Use of ANOVA in this way provides an extremely conservative significance test, due to the loss of power as the effective n drops to 8. This choice was preferred over an encounter-based analysis, which would move away from use of a consistent statistical model and the unit of randomization (GP trainees).
|
Results |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionReferences |
|---|
The characteristics of the trainees in the intervention, control, losses to follow-up and non-consenting groups were examined. There were no significant differences between these groups with respect to mean age, sex, length of general practice experience or University of graduation.
Diagnostic behaviour between groups over the three surveys is shown in Table 1
. An unintended effect of the survey or intervention could be that the GP trainees changed their diagnostic behaviour, possibly to justify giving an antibiotic rather than actually changing prescribing behaviour. Although there was an overall wintertime increase in the rate of diagnosis of URTI (survey 2), no differences in diagnostic behaviour were found between the two groups (all F values for contrasts of interest were less than 1).
|
The results of total antibiotic prescriptions per 100 encounters is shown in Table 2
. There was a greater decrease in total antibiotic prescribing in the intervention group compared with the control group from surveys 1 and 2 to survey 3, which approached statistical significance (P = 0.08).
|
Antibiotic prescriptions for URTI are shown in Table 3
. Viral rhinitis, rhinitis, coryza, cold, viral pharyngitis, pharyngitis (not specified) were all coded as URTI. Results showed a significant decline in antibiotic prescribing from surveys 1 and 2 to survey 3 in the intervention group when compared with the control group (P = 0.002).
|
Table 4
shows the mean rate at which the GP trainees prescribed an antibiotic consistent with Australian guidelines in the treatment of tonsillitis/streptococcal pharyngitis. Antibiotic Guidelines23 specify narrow spectrum penicillin (penicillin, phenethicillin, penicillin IM, penicillin V, procaine penicillin, benzathine penicillin, benzylpenicillin) or erythromycin. The results showed an increase in adherence to the guidelines in the antibiotic intervention group compared with the control group from survey 1 to survey 2 (P = 0.05).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionReferences |
|---|
The present study demonstrates that a stepped approach to training in the form of management guidelines plus personalized prescriber feedback can improve the prescribing behaviour of GP trainees over time. This was particularly true for reducing prescribing of antibiotics for URTI, supported by a trend towards reduced prescribing of antibiotics for all indications and an increase in prescribing of select antibiotics for tonsillitis/streptococcal pharyngitis.
There were differences in the effect of the educational intervention between the decision not to prescribe antibiotic for URTI and the decision of which antibiotics to prescribe for tonsillitis and streptococcal pharyngitis. The reduction in antibiotic prescribing for URTI was more pronounced following the face-to-face educational visits than with the written feedback alone. The continued effect occurred despite the fact that the educational visits were directed only to a subset of the higher prescribing GP trainees.
There were no significant differences in diagnostic behaviour between intervention and control groups over the three surveys. This, combined with the trend towards decreased prescribing of antibiotics for all indications in the intervention group, suggests that the GP trainees did not change their diagnosis to justify an antibiotic prescription, e.g. diagnose bronchitis or tonsillitis rather than URTI. While a decrease in prescribing occurred in the intervention group, in the absence of an education intervention, an increase in antibiotic prescribing was noted in the control group.
The study suggests that achieving behaviour change in prescribing of antibiotics for URTI requires an intensive educational intervention. The face-to-face encounter uses the interpersonal aspects of the encounter to interest the doctor and motivate change, and ensures that attention has been paid to the message, whereas written communication can easily go unread. The use of a stepwise intervention strategy was helpful in this study in that it identified 19 (24%) doctors who, according to our criteria for appropriate prescribing for URTI and tonsillitis/streptococcal pharyngitis, did not need the costly face-to-face intervention.
The significant improvement in antibiotic choice for tonsillitis and streptococcal pharyngitis appeared following the written feedback and was only slightly enhanced by the educational visits. This prescribing change involves a substitution behaviour rather than a decision not to prescribe, and appears to be more suitable for a written feedback intervention. Such a change in prescribing may be more readily accepted by patients than a decision not to prescribe at all and therefore may be perceived by the doctor as less threatening.
Although the practice activity survey data collected in this study were not externally validated, this type of doctor-recorded data has been shown previously to have a good correlation in assessment of patients' problems between independent observers24 and a high level of recording and coding accuracy,25 and has been used in the defining survey of morbidity and treatment in Australian general practice.2 However, as there is some concern that GPs may at times formulate the diagnosis to justify treatment26 we did examine for, but did not find, systematic differences in diagnostic behaviour between the intervention and control groups.
The study used the individual GP trainee or groups of GP trainees as the unit of analysis. De Burgh et al.,13 in a similar study, adopted an encounter-based analysis on the basis that patient factors may be a confounder of an intervention effect and an encounter-based analysis allows for adjustment for patient, doctor and practice characteristics. Patient factors are a confounder only if they change over time in some way related to the intervention. Otherwise the process of randomization should distribute patient and practice factors equally between groups. In this study it was decided that analysis of the unit of randomization and use of parametric statistical tests in accordance with the study design was a more important issue.
Another statistical issue for the analysis was the issue of clustering. Patients who attend a particular GP may be more like each other than a random sample of the general population, so there is a dependency in errors related to the cluster. Adjustment for clustering is a field still in development and there is no generally accepted means of dealing with it.27 If a simple adjustment had been made, that is applying a more stringent probability standard, P = 0.01, then the changes between groups of antibiotic prescribing for URTI would still be significant.
The effect on prescribing costs was examined for the intervention on antibiotic prescribing for URTI. A benefit to cost ratio of 1.3 was calculated if the effect of the intervention only lasted for the duration of the study (prescribing cost reduction of $A273 per doctor). If the effect continues for another 3 months, the ratio increases to 1.6 ($A319 per doctor) and if the effect continues for another 6 months then the ratio is 1.8 ($A364 per doctor), while the costs of providing the intervention were $A205 per doctor.
The most expensive aspect of the intervention was the face-to-face visits, which cost $A110 per doctor. Using the middle ratio of 1.6 and extrapolating the intervention to 10 000 GPs, the savings in drug costs are estimated at $A3.19 million over 2 years and costs at $A2.05 million all in 1 year.
This study has shown that individual prescribing feedback can be an effective educational intervention on the most prevalent problem, with rational drug use in general practice—the prescribing of antibiotics for upper respiratory infections. If applied widely this intervention would have the potential to reduce unnecessary prescribing, slow the development of antibiotic-resistant bacteria and reduce expenditure for the treatment of these conditions. Given the alarming rate with which resistance is increasing and the fact that the development of resistance is thought to be related to the amount of antibiotic consumed in a population,8 there is a pressing need for reduction in antibiotic use.
The study provides research evidence that supports Eisenberg's20 model of effective feedback, that it is individualized, specific and preferably delivered in an interpersonal encounter. The study also demonstrates an approach to targeting educational input at progressive levels of intensity, based on the practitioner's behaviour and response. This approach has the potential to deliver a cost-effective educational intervention and could be used as a model for educational programmes on other prescribing issues in general practice. Further research is needed to determine the duration of effect of change in prescribing brought about by the intervention.
|
Acknowledgments |
|---|
We wish to thank the general practice trainees who took part in the research, the educational outreach visitors (Dr Suzanne Mackenzie, Dr Toni Letton, Dr Lyndall Cooper and Dr Rosa Canalese). Funding was from the Department of Health and Family Services General Practice Evaluation Program.
|
References |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionReferences |
|---|
1 Department of Health and Family Services. General Practice in Australia: 1996. General Practice Branch, Commonwealth Department of Health and Family Services. Commonwealth of Australia. 1996.
2 Bridges-Webb C, Britt H, Miles D et al. Morbidity and treatment in general practice in Australia. Med J Aust 1992; 157: Suppl S1–S56.
3 McManus P, Hammond ML, Whicker SD, Primrose JG, Mant A, Fairall SR. Antibiotic use in the Australian community. Med J Aust 1997; 167: 124–127.[ISI][Medline]
4 Howie JG. Respiratory illness and antibiotic use in general practice. J R Coll Gen Pract 1971; 21: 657–663.[Medline]
5 Mainous AG, Hueston WJ, Clark JR. Antibiotics and upper respiratory infection. J Fam Pract 1996; 42: 357–361.[ISI][Medline]
6 Harvey K. Antibiotic use in Australia. Australian Prescriber 1988; 11: 74–77.
7 Birkett DJ, Mitchell AS, Godeck A, Grigson T, Cully R, Lee C. Profiles of antibacterial drug use in Australia and trends from 1987 to 1989. Med J Aust 1991; 155: 410–415.[ISI][Medline]
8 Turnidge J, Bell J. Antibiotic resistance: the Australian situation. Curr Therapeutics1996; April: 59–64.
9 Soumerai SB, McLaughlin TJ, Avorn J. Improving drug prescribing in primary care: a critical analysis of the experimental literature. Millbank Q 1989; 76: 268–315.
10 Soumerai SB, Avorn J. Principles of educational outreach (‘academic detailing’) to improve clinical decision making JAMA 1990; 263: 549–556.[Abstract]
11
Ray WA, Blazer DG, Schaffner W, Federspiel CF. Reducing antipsychotic drug prescribing for nursing home patients: a controlled trial of an education visit. Am J Public Health 1987; 77: 1448–1450.
12 De Santis G, Harvey KJ, Howard D, Mashford ML, Moulds RFW. Improving the quality of antibiotic prescription patterns in general practice. The role of educational intervention. Med J Aust 1994; 160: 502–505.[ISI][Medline]
13 De Burgh S, Mant A, Mattick R, Donnelly N, Hall W, Bridges-Webb C. A controlled trial of educational visiting to improve benzodiazepine prescribing in general practice. Aust J Pub Health 1995; 19: 142–148.
14 Avorn J, Soumerai SB, Everitt DE et al.. A randomised trial of a program to reduce the use of psychoactive drugs in nursing homes. N Engl J Med 1992; 327: 168–173.[Abstract]
15 Ray WA, Taylor JA, Meador KG et al. Reducing antipsychotic use in nursing homes. A controlled trial of provider education. Arch Intern Med 1993; 153: 713–721.[Abstract]
16 Tamai IY, Rubenstein LZ, Josephson KR, Yamauchi JA. Impact of computerised drug profiles and a consulting pharmacist on outpatient prescribing patterns: a clinical trial. Drug Intell Clin Pharm 1987; 21: 890–895.[Abstract]
17 Kroenke K, Pinholt EM. Reducing polypharmacy in the elderly: a controlled trial of physician feedback. J Am Geriatric Soc. 1990; 38: 31–36.[ISI][Medline]
18 Rokstad K, Straand J, Fugelli P. Can drug treatment be improved by feedback on prescribing profiles combined with therapeutic recommendations?. A prospective, controlled trial in general practice. J Clin Epidemiol 1995; 48: 1061–1068.[ISI][Medline]
19
Sanazaro PJ. Determining physician performance: continuing medical education and other interacting variables. Eval Health Prof 1983; 6: 197–210.
20 Eisenberg JM. Doctors' Decisions and the Cost of Medical Care: the Reasons for Doctors' Practice Patterns and Ways to Change Them. Ann Arbor, MI: Health Administration Press, 1986.
21 Kanfer FH, Goldstein AP (eds). Helping People Change. A Textbook of Methods. New York: Pergamon Press, 1986.
22 Lamberts H, Woods M (eds). World Organisation of National Colleges, Academies and Academic Associations of General Practitioners/Family Physicians. International Classification of Primary Care. (ICPC). Oxford: Oxford University Press, 1987.
23 Victorian Postgraduate Medical Foundation. Antibiotic Guidelines. 6th edn. Victoria: Victorian Postgraduate Medical Foundation, 1990–1991.
24 Bentsen BG. The accuracy of recording patient problems in family practice. J Med Educ 1976; 51: 311–316.[ISI][Medline]
25 Gelbach S. Comparing methods of data collection in an academic ambulatory practice. J Med Educ 1979; 54: 730–732.[ISI][Medline]
26 Howie JGR. Diagnosis: the Achilles heel. J R Coll Gen Pract 1972; 22: 310.[Medline]
27 Ashby M, Neeuhaus JM, Hauck WW et al. An annotated bibliography of methods for analysing correlated categorical data. Statistics Med 1992; 11: 67–69.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  J. C. van der Wouden and M. M. Kuyvenhoven Clinical decision support systems and antibiotic prescribing. JAMA, April 19, 2006; 295(15): 1771 - 1771. [Full Text] [PDF]
|
|
|
|
 |
|
 I. Welschen, M. M Kuyvenhoven, A. W Hoes, and T. J M Verheij Effectiveness of a multiple intervention to reduce antibiotic prescribing for respiratory tract symptoms in primary care: randomised controlled trial BMJ, August 21, 2004; 329(7463): 431. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. Arroll, T. Kenealy, F. Goodyear-Smith, and N. Kerse Delayed prescriptions BMJ, December 13, 2003; 327(7428): 1361 - 1362. [Full Text] [PDF]
|
|
|
|
 |
|
 T. Y. Tan, C. McNulty, A. Charlett, N. Nessa, C. Kelly, and T. Beswick Laboratory antibiotic susceptibility reporting and antibiotic prescribing in general practice J. Antimicrob. Chemother., February 1, 2003; 51(2): 379 - 384. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. Zwar, J. Henderson, H. Britt, K. McGeechan, and G. Yeo Influencing antibiotic prescribing by prescriber feedback and management guidelines: a 5-year follow-up Fam. Pract., February 1, 2002; 19(1): 12 - 17. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||