JAC Advance Access originally published online on October 24, 2006
Journal of Antimicrobial Chemotherapy 2007 59(1):162-163; doi:10.1093/jac/dkl440
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Correspondence |
Hospital antibiotic prescribing data require careful local interpretation
Department of Microbiology, City Hospitals Sunderland, UK
*Tel: +44-191-5656256; Fax: +44-191-5410531; E-mail: andrew.berrington{at}chs.northy.nhs.uk
Keywords: drug utilization , antimicrobial agents , data collection
Sir,
The recent publications of hospital antibiotic consumption data from the European Surveillance of Antibiotic Consumption (ESAC) project,1 and subsequent correspondence,2 make particularly interesting reading in the UK. Eight years ago the House of Lords Science and Technology Committee reported that: ‘Those responsible for the NHS Information Technology Strategy should consider the contrast between the excellent data on GP prescribing ... and the lack of data on antimicrobial use in hospitals. All hospitals should install computer systems for patient-specific prescribing information at ward level.’3 This problem is being belatedly addressed in Scotland, where acute Trusts are to be required to analyse and report use of key antimicrobials in defined daily doses (DDDs) per 1000 bed days,4 but as yet there has been no coordinated response for the rest of the UK. However the idea at last seems to have momentum and it would not be at all surprising if the Scottish recommendations were replicated elsewhere.
There is no doubt that at a national or regional level, antimicrobial prescribing data can provide insights into the spread of resistance and can facilitate comparison between different healthcare systems.5 It would clearly be preferable to have access to both ambulatory and hospital prescribing figures. However, there are numerous caveats to the interpretation of such data that become particularly important at the level of individual hospitals, if this is the direction in which we are headed. In the UK we are becoming used to hospital surveillance data, for instance rates of methicillin-resistant Staphylococcus aureus (MRSA) bacteraemia and Clostridium difficile disease, being presented to the public in the form of hospital league tables, and it is easy to anticipate the same happening with prescribing data.
In Sunderland we have recently begun feeding back prescribing data to clinicians. Most prescribing is done electronically, and it is not too challenging to manipulate the information in a database of drug use that can be stratified according to agent, clinician and/or clinical area. By converting total amounts into DDDs, and linking the data to bed day figures, we can inform individuals or groups of clinicians about their antimicrobial usage rates in DDDs per 1000 bed days. We can also point out significant differences between clinicians whose case mix is apparently equivalent, as these might reflect variation in practice and perhaps opportunities to reduce unnecessary prescribing.
Our initial perception is that this has been useful in reminding colleagues of the importance of prudent prescribing, promoting adherence to policies and informing audit questions. At a community level we are also reassured that the antimicrobial consumption of inpatients is a comparatively low (albeit slightly underestimated) 1.0 DDDs per 1000 population served per day. However, it has become clear that the data require careful interpretation, and I have three broad areas of concern about how the figures might be used if they became routinely collected.
(1) Are antibiotic prescribing data genuinely comparable between local healthcare organizations, for instance if a central agency were to publish hospital ‘league tables’ of antimicrobial prescribing?
Ongoing data collection at the level of prescribing (rather than supply for instance) is only feasible if prescribing is electronic. The electronic prescribing module that we use covers most of the organization, but specifically it does not capture prescribing in the critical care unit, in operating theatres, or in some long-stay units outside the main hospital. It is likely that other systems in other organizations would face equivalent constraints, and consequently would collect a subset of prescribing data different in subtle but perhaps important ways. Our system also allows us to restrict our analysis to administered doses, which we choose to do in order to ignore the surprisingly high proportion of prescribed doses that are not actually given, but others might be forced to include these. Finally, as is the case with the MRSA and C. difficile figures, hospital case mix is crucial. Patients trigger prescribing thresholds at different points (e.g. acute haematology patients versus long-stay rehabilitation patients) and consume antimicrobial DDDs at different rates (e.g. renal patients or children versus otherwise healthy adults). At a national level these might even themselves out, but at the level of an individual healthcare organization the measured antimicrobial consumption will be subject to various biases that will act to prevent meaningful comparison.
(2) Is the ‘DDD per 1000 bed days’ an appropriate unit of measurement of antimicrobial prescribing?
The ATC/DDD system has various advantages over other units of measurement in that DDDs are internationally defined and independent of things like pack size and price. By dividing by bed days the DDD can be corrected for activity, permitting comparison between different time periods or organizations. However, it is a measure of drug consumption, and while useful for comparative health economics it is not necessarily a good proxy for our principal interest which is resistance pressure and C. difficile-associated diarrhoea (CDAD) risk.
First, the DDD system provides a means of combining prescribing data for different drugs, but it must be remembered that DDDs are not weighted to reflect comparative risk. Three grams of cefuroxime and 100 mg of doxycycline each amount to one DDD of antibiotic. Although these could be easily distinguished, consider a DDD of intravenous co-amoxiclav (3 g of amoxicillin) in comparison to a DDD of the oral preparation (1 g). Once DDDs of different drugs are conflated into overall prescribing quantities they become less useful for describing behaviour and risk at local level.
Second, resistance pressure (and perhaps CDAD risk) is likely to depend on the concentration of active drug at sites of infection and colonization, together with different organisms' repertoires of resistance mechanisms and their means of dissemination. While the bed day might be an appropriate denominator for investigating resistance pressure at the hospital's sewage outfall, it is not necessarily appropriate for standardizing risk to patients on the wards. What probably matters more is the length and intensity of prescribed courses: a patient who receives 250 mg of ciprofloxacin once a day for the duration of his/her 6 week stay is likely to have a greater impact on the bacterial ecology of his surroundings than a patient who receives 750 mg twice a day for a week and 5 weeks without, yet both will contribute equally to the headline rate of consumption. There are alternative units, for instance DDDs per admission (which can give a very different picture of prescribing at the local level),6 but none stands out as optimal for the particular purposes of monitoring antimicrobial usage.
(3) Is it a good idea anyway?
The assumptions underlying feedback of prescribing data are (i) that higher prescribing is worse prescribing, and (ii) that when faced with evidence of high prescribing rates clinicians will tend to become better prescribers. Many readers will have some sympathy with these arguments, but by concentrating on DDDs (with whatever denominator) the message is subverted from ‘prescribe more prudently’ to ‘prescribe less’. Generally we want to raise thresholds for initiation, encourage use of some drugs at the expense of others, and (with some exceptions) shorten courses. Nothing is gained by encouraging a high-prescribing orthopaedic surgeon to reduce their standard operative prophylaxis from 1.5 g of cefuroxime to 375 mg.
To summarize, antibiotic prescribing data, presented in DDDs divided by an index of population or activity, have been shown to be useful at national level. They are often assumed to be equally helpful at local level, but such data require careful interpretation according to knowledge about case mix and prescribing parameters that this unit was never intended to capture. Point prevalence studies probably provide more meaningful information, but are less easy to standardize and more onerous to perform. Feedback to clinicians in the DDDs per 1000 bed days format is probably helpful as a ‘shot across the bows’, but the limitations of the data and particularly of conflated headline prescribing rates should be explicitly acknowledged. Except to inform economic arguments, antibiotic prescribing data in this form should not be used to construct hospital league tables.
Transparency declarations
None to declare.
References
1
Vander Stichele RH, Elseviers MM, Ferech M, et al. (2006) Hospital consumption of antibiotics in 15 European countries: results of the ESAC Retrospective Data Collection (1997–2002). J Antimicrob Chemother 58:159–67.
2
Kern WV, Steib-Bauert M, de With K. (2006) Comment on: Hospital consumption of antibiotics in 15 European countries: results of the ESAC Retrospective Data Collection (1997–2002). J Antimicrob Chemother 58:900–1.
3 House of Lords Select Committee on Science and Technology. (1998) Resistance to Antibiotics and Other Antimicrobial Agents. 7th Report Session 1997–98(The Stationery Office, London, UK).
4
Nathwani D. (2006) Antimicrobial prescribing policy and practice in Scotland: recommendations for good antimicrobial practice in acute hospitals. J Antimicrob Chemother 57:1189–96.
5 Goossens H, Ferech M, Vander Stichele, et al. (2005) Outpatient antibiotic use in Europe and association with resistance: a cross-national database study. Lancet 365:579–87.[Web of Science][Medline]
6
Filius PMG, Liem TBY, van der Linden PD, et al. (2005) An additional measure for quantifying antibiotic use in hospitals. J Antimicrob Chemother 55:805–8.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A. Berrington Antimicrobial prescribing in hospitals: be careful what you measure J. Antimicrob. Chemother., November 1, 2009; (2009) dkp399v1. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||