Journal of Antimicrobial Chemotherapy (2003) 51, 379 384 DOI: 10.1093/jac/dkg032 Advance Access publication 6 January 2003 Antimicrobial practice Laboratory antibiotic susceptibility reporting and antibiotic prescribing in general practice Thean Yen Tan 1 *, Cliodna McNulty 2, Andre Charlett 3, Nazma Nessa 3, Clare Kelly 4 and Trevor Beswick 4 1 Public Health Laboratory, University Hospital Wales, Cardiff, South Glamorgan CF14 4XW; 2 Public Health Laboratory, Gloucester; 3 Statistics Department, Communicable Disease Surveillance Centre, London; 4 NHS Executive South West, Department of Health, Bristol, UK Received 25 June 2002, returned 26 August 2002, revised 9 October 2002; accepted 9 October 2002 This study set out to investigate whether there was an association between antibiotic susceptibility reporting from microbiology laboratories and antibiotic prescribing for urinary tract infections in the community. Data were collected over a 3 month period using a prospective questionnaire survey of general practitioners, who submitted and received a mid-stream urine (MSU) result from selected microbiology laboratories in England and Wales. In addition, prescribing analyses and cost (PACT) data were requested from the Prescription Prescribing Authority. The study demonstrated an association between laboratory reporting of antibiotic susceptibilities and antibiotic prescribing for treatment of urinary tract infections. The reporting of susceptibilities to oral cephalosporins and nitrofurantoin from microbiology laboratories was associated with increased prescribing of each antibiotic. This association was demonstrated for the choice of empirical antibiotic therapy and the choice of antibiotic prescribed for each studied episode of urinary tract infection. PACT data demonstrated a consistently greater use of antibiotics that were reported by the servicing laboratory, although this was only statistically significant for nitrofurantoin. This study demonstrates that there is an association between antibiotic susceptibility reporting from microbiology laboratories and antibiotic prescribing for the treatment of urinary tract infections. Keywords: antibiotic reporting, prescribing, microbiology laboratory Introduction Infections constitute up to 40% of consultations in general practice. 1 General practice surgeries submit 40 400 urine samples per 1000 patients and laboratory reports result in a change in antibiotic therapy in 28% of cases. 2 However, there is little information to document the influence of microbiology reports on antibiotic selection and prescribing. In the UK, microbiology laboratories invest significant efforts in the process of quality assurance 3,4 and the standardization of antimicrobial susceptibility testing methodology 5 but much less attention has been given to the choice of antibiotics to test and report. This study aimed to determine whether different antibiotic susceptibility reporting protocols for urinary isolates from microbiology laboratories was associated with differences in antibiotic prescribing in primary care. We hypothesized that general practitioners (GPs) who were served by laboratories that routinely reported particular antimicrobials (e.g. nitrofurantoin, quinolones and cephalosporins) would prescribe... *Corresponding author. Tel: +44-29-2074-4825; Fax: +44-29-2074-2161; E-mail: theanyen.tan@phls.wales.nhs.uk... 379 2003 The British Society for Antimicrobial Chemotherapy
T. Y. Tan et al. these agents more frequently for the treatment of urinary tract infections (UTIs) than doctors who were served by laboratories that did not report these antimicrobials. Materials and methods A questionnaire-based study was developed to investigate antibiotic prescribing by GPs for the treatment of UTIs that were served by laboratories with differing antibiotic reporting protocols. In addition, antibiotic prescribing for the treatment of all infections from each participating primary care group was measured by obtaining prescribing analyses and cost (PACT) data. PACT data represent the main source of information on the prescribing of GPs in England, 6 and a similar scheme exists in Wales. Questionnaire survey of antibiotic prescribing practices from general practice In total, 100 questionnaires were distributed to GPs in each primary care area serviced by 13 Public Health Laboratories over a 4 month period in the year 2000. The primary care area was delineated by obtaining a complete list of GP practices serviced by each participating laboratory. The geographical area served by these laboratories included areas throughout England and Wales. Laboratory selection was based on susceptibility reporting determined by an earlier questionnaire survey of all Public Health Laboratories in England and Wales. 7 reporting practices for each participating laboratory were confirmed by the investigators before distribution of questionnaires. All laboratories reported susceptibilities to trimethoprim and ampicillin. Two laboratories did not report susceptibilities to nitrofurantoin, five laboratories did not report susceptibilities to cephalosporins and seven did not report susceptibilities to quinolones. Questionnaires were attached to 100 consecutive positive mid-stream urine (MSU) reports at each participating laboratory. Wherever possible, laboratories were instructed to refrain from sending more than one questionnaire to an individual GP. A positive report was defined as any MSU specimen with significant growth of an isolate for which antibiotic susceptibility testing was carried out. Each questionnaire contained details of the urinary isolate identified with reported antibiotic susceptibilities, but contained no patient or practice identifiers. Completed questionnaires were initially returned to each local laboratory, and subsequently forwarded to the study investigators. GPs were asked to provide information on their antibiotic prescribing practices for uncomplicated UTIs, including the antibiotic that they had prescribed for the current investigated episode of UTI. The antibiotic choices provided for the questionnaire included amoxicillin, oral cephalosporins, quinolones, co-amoxiclav, nitrofurantoin and trimethoprim (questionnaire available from the authors). Analysis of antibiotic prescribing results from questionnaire survey prescribing of GPs was examined by two separate methods. GPs were asked to nominate their first choice of antibiotic for the treatment of uncomplicated UTIs in general. These choices were examined in order to detect any difference in empirical antibiotic choices between GPs serviced by laboratories with different antibiotic susceptibility reporting practices. GPs were then asked to name the actual antibiotic prescribed for the UTIs surveyed in each questionnaire received. The antibiotic prescribing for these episodes could be subdivided into two groups: antibiotics prescribed before receipt of the laboratory report, and antibiotics prescribed after receipt of the laboratory report. From these data, it was possible to compare the actual antibiotic prescribing for the surveyed UTIs for GPs serviced by separate laboratories. Nitrofurantoin, quinolones and cephalosporins were chosen as the indicator antimicrobials for analysis because these were the only antibiotics where differences in reporting existed between the various laboratories. It was not possible to analyse differences in antibiotic prescribing between reporting and non-reporting laboratories for ampicillin and trimethoprim, because all laboratories reported these two antibiotics. PACT data PACT data provide information about prescribing in general practice in England, and include all items prescribed by GPs and nurse prescribers. The volumes and prices of antibiotic s can be measured, but not the specific infection for which a particular antibiotic was prescribed. However, nitrofurantoin and norfloxacin are almost entirely prescribed only for the treatment of UTIs. Any regional differences in antibiotic prescribing for UTIs may therefore be measurable by PACT data for these two antibiotics. Primary care groups served solely by each participating laboratory were identified. PACT data for oral cephalosporins, co-amoxiclav, nitrofurantoin, norfloxacin, quinolones and total antibiotic prescribing were requested for every participating primary care group for the year 1999. Analysis of PACT data Statistical analysis of the PACT data was carried out on Microsoft Excel and Stata 7. Due to an observed positive skew in the antibiotic prescribing units, the data were logtransformed before the analysis. A t-test was then carried out to assess whether there was a significant difference in prescribing in the areas where an antibiotic is reported compared with where the antibiotic is not reported. As well as the reported s, the estimated ratio of prescribing in the 380
susceptibility reporting and prescribing in general practice Table 1. Comparison of stated empirical antibiotic of first choice by GPs Number of GPs (%) who selected this antibiotic as first-choice empirical therapy for uncomplicated UTIs Oral cephalosporins 52/592 (8) 6/313 (2) <0.01 4.9 (2.1 14.1) Nitrofurantoin 13/869 (2) 0/169 (0) 0.1 undefined Quinolones 1/410 (0) 6/474 (1) 0.09 0.2 (0.0 1.6) Figures in parentheses are reported as a percentage of the denominator. Table 2. Actual antibiotic prescribing by GPs for the episode of UTI investigated Number of GPs who prescribed this antibiotic for the investigated episodes of UTI (%) Oral cephalosporins 100/616 (16) 31/300 (10) 0.04 1.6 (1.1 2.5) Nitrofurantoin 35/835 (4) 0/149 (0) 0.01 undefined Quinolones 45/459 (10) 24/383 (6) 0.06 1.6 (0.95 2.9) Figures in parentheses are reported as a percentage of the denominator. reporting to non-reporting areas was calculated together with the 95% confidence intervals. Statistical analysis Data from the questionnaires and PACT data were entered into Epi-Info, and analysed in Stata 7 and Epi-Info. The χ 2 test and Fisher s exact test were used to test for statistical significance for any associations noted. Results In total, 1069 of 1300 (82%, range 65 90%) distributed questionnaires were returned. GPs stated empirical antibiotic choices Overall, 90% of GPs selected trimethoprim as antibiotic of first choice in uncomplicated UTIs. The other empirical antibiotic choices selected were an oral cephalosporin (6%) and a quinolone (1%). However, GPs in areas serviced by a laboratory that routinely reported cephalosporin susceptibilities on urinary isolates were four times more likely to select an oral cephalosporin as their first choice of empirical therapy (Table 1). No statistically significant differences in the choice of either quinolones or nitrofurantoin were noted for GPs serviced by laboratories that reported or did not report each antibiotic. GPs prescribed antibiotic therapy for the investigated episode of UTI With regard to antibiotic prescribing for the reported cases of UTI in this study, GPs who were serviced by laboratories that reported cephalosporins or nitrofurantoin were significantly more likely to prescribe these antibiotics (Table 2). In laboratory areas where cephalosporin susceptibilities were reported, 16% of GPs prescribed this antibiotic as opposed to 10% in non-reporting areas (P < 0.05). The corresponding figures for quinolone prescribing were 10% in reporting areas, compared with 6% in non-reporting areas (P = 0.06); 4% of GPs serviced by laboratories that reported antibiotic susceptibilities to nitrofurantoin prescribed this antibiotic, compared with 0% in the corresponding group serviced by a non-reporting laboratory (P < 0.05). Subgroup analysis was carried out to compare antibiotic both before and after receipt of the laboratory result. 381
T. Y. Tan et al. Table 3. therapy prescribed by GPs before receipt of laboratory MSU report prescribed for investigated episode of UTI before receipt of microbiology report (%) Nitrofurantoin 16/501 (3) 0/87 (0) 0.07 undefined Oral cephalosporins 50/364 (14) 16/181 (9) 0.10 1.6 (0.9 3.2) Quinolones 32/277 (12) 11/226 (5) <0.01 2.6 (1.2 5.8) Table 4. therapy prescribed by GPs after receipt of laboratory MSU report prescribed for investigated episode of UTI after receipt of microbiology report (%) Nitrofurantoin 19/334 (6) 0/62 (0) 0.04 undefined Oral cephalosporins 50/252 (20) 15/119 (13) 0.09 1.7 (0.9 3.5) Quinolones 13/182 (7) 13/157 (8) 0.70 0.85 (0.4 2.1) Table 5. Comparison of antibiotic prescribing as measured by PACT data from reporting and non-reporting laboratory areas Estimated ratio of antibiotic prescribing (laboratories that report antibiotic/laboratories that do not report antibiotic) 95% CI Cephalosporins 1.40 0.75 2.64 0.26 Quinolones 1.35 0.98 1.86 0.06 Norfloxacin 1.29 (laboratories that report quinolones/laboratories 0.29 5.80 0.72 that do not report quinolones) Norfloxacin 2.08 (laboratories that report norfloxacin/laboratories 0.10 44.37 0.54 that do not report norfloxacin) Nitrofurantoin 2.78 1.36 5.69 0.01 Co-amoxiclav 1.61 0.86 3.02 0.10 Combined analysis 1.54 1.06 2.23 0.02 Before receipt of the laboratory result, GPs served by any laboratory that reported quinolone susceptibilities were more than twice as likely to have initially prescribed this antibiotic (12% versus 5%, P < 0.01) (Table 3). The strongest association between antibiotic reporting and prescribing following receipt of the laboratory report was for nitrofurantoin (6% versus 0%, P = 0.04) (Table 4). Although more GPs who were served by laboratories that reported cephalosporin susceptibilities prescribed cephalosporins following receipt of the microbiology report, this increase was not statistically significant. PACT data Analysis of the PACT data was carried out to investigate whether these reported differences in antibiotic selection and prescribing were quantifiable as differences in antibiotic and cost data (Table 5). Prescribing volumes for nitrofurantoin were nearly three times higher in primary care areas served by a reporting laboratory (P < 0.05). For all the other antibiotics concerned, prescribing of an antibiotic was higher in areas served by the laboratories that reported the antibiotic, although the results failed to reach statistical sig- 382
susceptibility reporting and prescribing in general practice nificance. Initially an interaction was used to test for heterogeneity between the estimated ratio of antibiotic prescribing, and there was no evidence to suggest that the relative increases in prescribing in the areas served by reporting laboratories were different (P = 0.33). The final row in Table 5 was obtained by pooling these estimates in a regression model. These figures demonstrate that, regardless of which individual antibiotic is examined, laboratory reporting of any antibiotic is associated with a 54% increase in antibiotic prescribing (P = 0.024, 95% CI 6 123%). Discussion There is conflicting evidence from previous literature regarding the link between laboratory antimicrobial susceptibility reporting and antibiotic prescribing. An earlier study that restricted susceptibility reporting on urinary isolates from the community showed increasing use of the antibiotics for which susceptibilities were reported. 8 This study demonstrates that the prescribing of cephalosporins, nitrofurantoin and quinolones for the treatment of UTIs varies across general practices. These reported differences extended from the choice of empirical antibiotics to the actual of antibiotics for the UTIs surveyed in this study, and corresponded with local laboratory antibiotic susceptibility reporting. Laboratory reporting of quinolone or cephalosporin susceptibilities by a microbiology laboratory was associated with a 50% increase in prescribing of each antibiotic by the GPs surveyed in this study, whereas the difference for nitrofurantoin prescribing was substantially larger. Other than for nitrofurantoin, these reported differences in prescribing did not translate into a significant difference in actual antibiotic prescribing as reported by PACT data. There was a trend towards increased antibiotic prescribing linked with antibiotic reporting. However, with such small numbers of laboratories involved, the analysis did not have sufficient statistical power to detect the reported differences as being significant. In addition, PACT data are not specific enough to differentiate between antibiotics prescribed for the treatment of UTIs or other infections. Thus, it is difficult to demonstrate an association between reporting and prescribing of some antibiotics specifically for the treatment of UTIs. Cephalosporins, co-amoxiclav and quinolones can also be prescribed for treatment of other infections. There are several limitations to this study. The sample of GPs in the study was clearly biased towards those who submitted urine samples to their local laboratory. The prescribing practices of doctors who rarely submit urine samples for testing may be different from those surveyed in this study. Some GPs may only submit MSU samples for the treatment of complicated UTIs, which may require different empirical antibiotic treatment. It may not be possible to extrapolate these results to all primary care areas in England and Wales. Only 100 GPs were sampled for each primary care area involved in the study. Furthermore, only primary care areas served by Public Health Laboratories were included in the study. Laboratories were selected for participation in the study by their local antibiotic reporting practices and not by geographical distribution or demographic coverage. Because of these factors, the sample population may not be representative of the general population as a whole. Variation in local antibiotic susceptibility patterns may also account for the differences in antibiotic prescribing. For example, higher rates of resistance to ampicillin, nitrofurantoin or trimethoprim may account for increased of cephalosporins or quinolones. susceptibilities of the reported 100 urinary isolates from each laboratory were analysed for antibiotic resistance rates to ampicillin, nitrofurantoin or trimethoprim (data available on request). The only significant difference in resistance rates noted was increased resistance to ampicillin/amoxicillin in laboratories that reported norfloxacin susceptibilities (resistance rate 59% versus 46%, P < 0.05). Even in this laboratory grouping, there was no significant difference in resistance rates to trimethoprim or nitrofurantoin. This study has demonstrated an association between laboratory susceptibility reporting practices and the prescribing of specific antibiotics for the treatment of UTIs. This association was noted at both possible stages of antibiotic prescribing in the choice of empirical antibiotic therapy, and in prescribing for an infection when the susceptibility results were known. It is less clear whether this association would be applicable to all antibiotics prescribed for UTIs, or to antibiotics prescribed for other infections. In a study of this nature, causality is difficult to demonstrate. For example, one possible explanation for the association noted in this study is that microbiology laboratories may selectively choose to report antibiotics that are more commonly used in their primary care area. It would be difficult to ascertain the relative effect of the actual microbiology report itself, and other influences such as local antibiotic reporting practices in general and local prescribing guidelines. In addition, the laboratory report may affect antibiotic prescribing at various stages in the process in the overall decision making process, or at the actual stage of selection and prescribing of the antibiotic. It is well documented that influences on prescribing habits are complex and multi-factorial and include prescribing guidelines, 9 education, 10 the pharmaceutical industry 11 and patient expectations. 12 Nonetheless, the results from this study suggest that laboratory reporting of antibiotic susceptibilities may influence antibiotic prescribing in the community. Further studies will be needed to clarify this association, for example, by altering laboratory reporting practices in the study areas and 383
T. Y. Tan et al. examining the effect of this alteration on subsequent antibiotic prescribing. Acknowledgements We would like to thank the Directors and staff of the following Public Health Laboratories without whom this study would not have been possible: Bangor, Bristol, Cardiff, Chester, Chelmsford, Coventry, Exeter, Gloucester, Ipswich, Lincoln, Norwich, Plymouth and Shrewsbury. We also thank all the GPs who contributed to the questionnaire survey. This study was funded by a grant from the Public Health Laboratory Service. References 1. UK Department of Health. (2002). Getting ahead of the curve. A strategy for combating infectious diseases (including other aspects of health protection) [Online.] http://www.doh.gov.uk/cmo/idstrategy/ (24 July 2002, date last accessed). 2. Tompkins, D. S. & Shannon, A. M. (1993). Clinical value of microbiological investigations in general practice. British Journal of General Practice 43, 155 8. 3. Gray, J. (1999). Quality assurance in a diagnostic microbiology laboratory. Communicable Disease and Public Health 2, 225 6. 4. Bartlett, R. C., Mazens-Sullivan, M., Tetreault, J. Z., Lobel, S. & Nivard, J. (1994). Evolving approaches to management of quality in clinical microbiology. Clinical Microbiology Review 7, 55 88. 5. Andrews, J. M. (2001). BSAC Working Party on Susceptibility Testing. BSAC standardized disc susceptibility testing method. Journal of Antimicrobial Chemotherapy 48, Suppl. 1, 43 57. 6. Majeed, A., Evans, N. & Head, P. (1997). What can PACT tell us about prescribing in general practice? British Medical Journal 315, 1515 9. 7. Tan, T. Y. & McNulty, C. A. (2002). Survey of public health laboratory protocols for reporting the antibiotic susceptibility of urinary isolates submitted from general practice. Communicable Disease and Public Health 5, 33 7. 8. Langdale, P. & Millar, M. R. (1986). Influence of laboratory sensitivity reporting on antibiotic prescribing preferences of general practitioners in the Leeds area. Journal of Clinical Pathology 39, 233 4. 9. Zwar, N., Wolk, J., Gordon, J., Sanson-Fisher, R. & Kehoe, L. (1999). Influencing antibiotic prescribing in general practice: a trial of prescriber feedback and management guidelines. Family Practice 16, 495 500. 10. McNulty, C. A., Kane, A., Foy, C. J., Sykes, J., Saunders, P. & Cartwright, K. A. (2000). Primary care workshops can reduce and rationalize antibiotic prescribing. Journal of Antimicrobial Chemotherapy 46, 493 9. 11. Orlowski, J. P. & Wateska, L. (1992). The effects of pharmaceutical firm enticements on physician prescribing patterns. There s no such thing as a free lunch. Chest 102, 270 3. 12. Macfarlane, J., Holmes, W., Macfarlane, R. & Britten, N. (1997). Influence of patients expectations on antibiotic management of acute lower respiratory tract illness in general practice: questionnaire study. British Medical Journal 315, 1211 4. 384