Introduction
Available literature indicates that 25–50% of hospitalized patients receive antibiotics, of which 20–50% are either unnecessary or inappropriate [
1‐
5]. Antimicrobial stewardship programs (ASPs) are coordinated programs designed to improve the appropriateness of antibiotic use [
6‐
8]. One of the cornerstones of an ASP is to monitor the total amount of local antibiotic use and use this information to guide and evaluate targeted ASP interventions [
7,
9].
Several units of measurement are available to standardize total antibiotic use. Recommended metrics are defined daily dose (DDD) and days of therapy (DOT) [
10‐
13]. DDD is defined by the World Health Organization (WHO) as the assumed average maintenance dose per day for a drug used for its main indication in adults [
14,
15]. One DOT represents the administration of a single agent on a given day regardless of the number of doses administered or dosage strength [
10].
Advantages and disadvantages of both metrics have been recognized. DDDs allow for standardized comparison of aggregate antibiotic use between hospitals and are usually extracted from hospital billing or hospital dispensing records, which makes the metric applicable even in countries with limited access to computerized pharmacy or prescription data. However, there are substantial limitations to this metric. For example, DDDs are influenced by dose adjustment and will therefore underestimate antibiotic use in patients in whom dose adjustment is required, for instance children or patients with renal impairment. Also, DDD is a unit of measurement and does not necessarily reflect the recommended or prescribed daily dose [
7,
10]. Using DOT is recommended by the IDSA guidelines, as it is usually based on patient-level prescription data and therefore not influenced by dose adjustment. However, prescription data are difficult to assess without computerized physician order entry (CPOE) of individual patients [
7,
10].
In Europe and in the USA, surveillance reports on antibiotic use and resistance rates are issued annually. The European Surveillance of Antimicrobial Consumption Network (ESAC-NET), which is managed and coordinated by the European Centre for Disease Prevention and Control (ECDC), and the Dutch Working Party on Antibiotic Policy (SWAB) provide reference data on hospital antibiotic consumption using DDD [
16,
17]. In contrast, the US Center for Disease Control and Prevention (CDC) uses DOT [
18]. These reports are used for surveillance purposes, but they have not shown to be useful in ASPs, as they do not provide a detailed assessment of quantitative antibiotic use, e.g., per diagnosis or medical specialty, which is needed to guide and evaluate targeted ASP interventions.
A recent literature review stated that electronic assessment of antibiotic use data is potentially useful for the purpose of antimicrobial stewardship; however, the best approach to retrieve reliable quantitative data is not yet clear and might also be determined by local hospital settings, procedures, and budget [
19]. The objectives of the present study were (1) to explore for the hospital setting the possibilities of quantitative data retrieval on the level of medical specialty and (2) to describe factors affecting the usability and interpretation of these quantitative metrics.
Discussion
In this observational multicenter study, we observed a large variation in antibiotic use between and within hospitals and a low correlation between DDD and DOT as metrics of total antibiotic use in hospitalized adult patients. We explored several factors potentially affecting the retrieved data and found that part of the variation in quantitative antibiotic use is likely caused by differences in organizational factors, data sources, data registration, and data extraction. Also, we showed that for measuring quantitative antibiotic use for ASP purposes at the level of medical specialty, it is currently preferable to use patient-level prescription data.
Previous literature showed that differences in antibiotic use between hospitals can be partially explained by patient mix or hospital characteristics. For example, vancomycin use is significantly higher in university hospitals as compared to large teaching or general hospitals, due to differences in patient mix [
22]. However, in our study, the variation in mean monthly antibiotic use between and within the hospitals was more extensive than the differences in antibiotic use presented in the annual antibiotic consumption reports [
22], and therefore unlikely to result exclusively from differences in patient mix and hospital type.
We found a low correlation between DDD and DOT for three out of four hospitals that used corresponding extraction procedures for numerators and denominators, in contrary to a large US study by Polk et al. who found an overall linear association between DDD and DOT. In the USA, billing records are used as a data source to measure antibiotic use, and both metrics were calculated from the same data source [
10]. In the Netherlands, costs for in-patient antibiotics are integrated in the overall hospital budget; thus, billing data cannot be used as a data source. Data sources that can be used to measure antibiotic use in the Netherlands are dispensing data, to calculate DDD, and prescription data, to calculate DOT. The low correlation we found between DDD and DOT might therefore be partly explained by the use of different data sources for each metric. In addition, the low correlation could be explained by errors in data registration and extraction procedures, e.g., hospitals extracted prescription data based on the prescribers’ specialty instead of the patients’ specialty of admission (see Table
3).
The Dutch healthcare system at present stimulates reorganization of hospital wards into “mixed wards” (i.e., physical locations with a mix of medical specialties). Basically, all hospitals in our study consisted of mixed wards, and the distribution of medical specialties within mixed wards differed per hospital. The majority of Dutch hospital pharmacies in our study were able to extract prescription data on the level of medical specialty (79%). The main reasons for three hospitals not to be able to extract prescription data were technical difficulties and lack of knowledge of the extraction procedure by the IT specialist at the time of the study. Dispensing data, however, could be extracted on specialty level by only 21% of the pharmacies, as antibiotics are usually dispensed aggregated per ward or unit and are not registered per medical specialty. In addition, antibiotics dispensed to a ward do not provide accurate information whether these antibiotics are actually administered to a patient. Compared to dispensing data, patient-level prescription data are able to give a more adequate estimation of the actually administered antibiotics. In addition, patient-level data can be linked to other data registered in the patient system (such as indication and culture results) making it possible to relate quantitative antibiotic use data to resistance data. As a result, for ASP purposes, prescription data currently provide a more valid metric to compare levels of antibiotic use between medical specialties.
Finally, a study on retrieval of antibiotic use data from computerized pharmacy data on the intensive care unit found that computerized patient-level measures can be derived easily, but the magnitude of discrepancies between computerized antibiotic use data and manual chart review varied, with electronic medication administration records (eMAR) providing maximal accuracy [
23]. In the future, the use of administration registration would be preferable over prescription data; however, calculation of antibiotic use from administration records is not yet possible in most clinical settings.
Our study has several strengths. This is the first study to focus on the process of data registration and extraction, in order to understand variation between hospitals in quantitative antibiotic use and to detect inconsistencies between DDD and DOT as measures of antibiotic use. Also, our study focusses on medical clusters of specialties in the assessment of quantitative antibiotic use, which is highly recommended for the evaluation of ASPs, whereas most studies evaluated antibiotic use on a hospital level or only compared wards with a relatively high antibiotic use, e.g., intensive care units [
10‐
12,
24,
25].
Our study was limited by the relatively small number of hospitals participating in the study. Also, we only included Dutch hospitals. However, as differences between Dutch hospitals likely reflect the (even larger) variability in healthcare organizations throughout Europe, using a variety of electronic patient systems, each with different registration modes and extraction possibilities, the findings of our study are of relevance for other countries as well, including cross-country comparisons.
A recent literature review described the difficulties in secondary use of data from hospital electronic prescribing and pharmacy systems to support ASP, including data access, data accuracy, and completeness, and discussed the complexity of data extraction from multiple electronic systems or hospital sites [
19]. Our study showed that differences between hospitals in organizational factors, data sources, data registration, and data extraction contribute to the variation between hospitals in quantitative use and a low correlation between DDDs and DOTs. A clear understanding of these factors, together with a uniform and transparent approach in defining organizational units within hospitals, and uniform data sources, registration, and extraction procedures are necessary for reliable measurement and valid comparison of antibiotic use.
Acknowledgements
All participating hospitals:
Alrijne Hospital, Leiden
Amphia Hospital, Breda
Amstelland Hospital, Amstelveen
Beatrix Hospital, Gorinchem
Deventer Hospital, Deventer
Diakonessenhuis, Utrecht
HAGA Hospital, Den Haag
Laurentius Hospital, Roermond
Maasstad Hospital, Rotterdam
Maastricht University Medical Center, Maastricht
Meander Medical Center, Amersfoort
Medical Center Haaglanden, Den Haag
Rijnstate Hospital, Arnhem
Tergooi Hospital, Hilversum