Parenteral Antibiotic Use in Acute-Care Hospitals: A Standardized Analysis of Fourteen Institutions
- Infectious Diseases Section, Department of Medicine, Carney Hospital and Boston University School of Medicine, Boston, Massachusetts
- Reprints or correspondence: Dr. Philip Carling, Infectious Diseases Section, Carney Hospital, 2100 Dorchester Avenue, Boston, MA 02124.
-
Presented in part: 5th annual meeting of the Society for Health Care Epidemiology, San Diego, 2 April 1995 (abstract M-34).
Abstract
Despite increasing concerns regarding the need to optimize appropriate antibiotic use in hospitals, a standardized method for evaluating interinstitutional antibiotic use has not been developed. To address this issue, antibiotic use was analyzed by means of a uniform methodology among 14 acute-care hospitals. Data were standardized by use of a defined daily dose for each antibiotic while adjusting for patient volume by calculating use per 1000 patient-days. Within the group, there was a 68% range in total parenteral antibiotic expenditures and wide variability in the use of individual agents. Analysis of these differences indicated that only the use of active antibiotic-management programs clearly correlated with antibiotic cost per 1000 patient-days (P<.001). Given these results, we believe that wider comparative analysis of antibiotic use with a standardized methodology in conjunction with standardized analysis of nosocomial infection rates and antibiotic resistance data may enhance the stewardship of antibiotics in acute-care hospitals.
It is widely recognized that large portions of pharmacy expenditures in acute-care hospitals represent acquisition costs for parenteral antibiotics, accounting for 20%–40% of a hospital's total drug budget [1–3]. Beginning in 1973, hospitals have been encouraged to develop policies and procedures to maximize appropriate antibiotic use while containing costs [4]. In the 1980s, Kunin [5, 6] recommended a wide range of both active and passive interventions to optimize the appropriate use of antibiotics while controlling costs and limiting selective pressure on the development of antibiotic resistance. These issues were most recently evaluated by the Joint Committee on the Prevention of Antimicrobial Resistance, who provided a broad range of evidence-based recommendations and strategic goals to optimize antibiotic use and limit the emergence of antimicrobial resistance within hospitals [7].
Although many hospitals monitor antibiotic expenditures for budgetary purposes, an analysis of such expenditures as part of an assessment of various forms of management interventions has been reported for only a relatively small number of institutions [8–17]. Although these reports have documented various aspects of antibiotic use and management interventions within individual institutions, lack of a standardized method for conducting and reporting such evaluations has prevented the analysis of interinstitutional use and expenditures.
The development of systems by hospitals to measure the use of antimicrobials is an essential element of an optimal antibiotic-control program, as defined by the Joint Committee on the Prevention of Antimicrobial Resistance [7]. Despite this, specific methods for standardizing the analysis of antimicrobial use have yet to be developed that could optimize longitudinal monitoring within a given institution while facilitating comparison of interinstitutional trends and variances in use. The present study was designed to assess the applicability of a standardized analysis model, which had been previously used to prospectively evaluate parenteral antibiotic use in an individual institution [12], to compare and analyze parenteral antibiotic use among a group of acute-care hospitals.
Methods
On the basis of a shared interest in developing a better understanding of the epidemiology of parenteral antibiotic use and expenditures, pharmacists and infectious disease clinicians at 14 acute-care hospitals provided data for the present analysis. The hospitals included in this study were self-selected for their interest in sharing information on antibiotic use from a much larger group of hospitals who were invited to participate in the analysis but were unable to do so, primarily because of limitations of data collection due to the retrospective nature of the project. None of the hospitals or the individuals supplying the data received any form of compensation for their participation in the project.
We collected information about total grams of parenteral antibiotics used during 12 consecutive months and about total medical and surgical patient-days, as well as descriptions of the types of antibiotic management policies and procedures being used by the individual hospitals by means of a standardized questionnaire (available from the author). After analysis of the questions revealed that several of the hospitals used prospective antibiotic-management programs such as that first described by Woodward et al. [8] and later by others [9–17], the 14 hospitals were divided into 2 groups for the purpose of comparison. For the purpose of comparison, we grouped together those hospitals that used only a variety of passive control policies and procedures, such as automatic stop orders, antimicrobial order forms, limited formularies, measures to control contact between pharmaceutical representatives and prescribers, educational interventions (e.g., institutional guidelines for antibiotic use), and restricted antibiotic susceptibility reporting by the microbiology laboratory; these we labelled hospitals with “passive antibiotic-management programs.” Hospitals that used such procedures along with an active, prospective intervention program that involved a clinical pharmacist and staff-level infectious diseases—trained physician team, such as that initially described by Woodward et al. [8], we grouped as hospitals with “active antibiotic-management programs.”
To allow for accurate interinstitutional comparison of expenditures for parenteral antibiotics, confounding differences related to acquisition costs and variations in the amount of each antibiotic used per day to treat individual patients were standardized for the group. Acquisition costs per gram were standardized by use of a 1994 catalog of prescription drug costs [18]. In addition, each parenteral antibiotic was assigned a fixed or defined daily dose (DDD) in the manner recently supported by the World Health Organization [19], by means of the recommended daily adult dose found in the 1994 edition of the American Hospital Formulary System compendium of prescribing information [20].
For each antibiotic, treatment days were calculated by dividing the total grams used by the DDD. The annual expenditure for each antibiotic was then calculated by multiplying the cost per gram by the total grams used. The total treatment days and annual expenditures were divided by patient-days for each institution. To avoid fractional numbers, these figures were multiplied by 1000 to arrive at DDD and expenditures per 1000 patient days and days of treatment. Using a computerized spreadsheet program we have described [12], we collated data for each hospital on individual spreadsheets, an example of which is illustrated in table 1.
Figure 1
Total parenteral antibiotic costs expressed as percentage above and below the mean total cost per 1000 patient-days for group. Hospitals A—E had active management programs (group A), whereas hospitals F—N used only passive management interventions (group B). Calculated mean cost per 1000 patient-days was $14,014.00 for the group.
Figure 2
Examples of variability in use of 3 antibiotics whose use showed greatest variability (ciprofloxacin, ceftriaxone, and cefuroxime) and least variability (vancomycin, oxacillin/nafcillin, and imipenem), expressed as defined daily dose (DDD) per 1000 patient-days (PTD) for group A hospitals (those with active management programs) and group B hospitals (those with only passive management programs).
Table 1
An example showing the format of the data analysis spreadsheet used to calculate defined daily dose (DDD) and costs per 1000 patient-days (PDs) for each of the 14 participating hospitals.
In analyzing the data from the 14 hospitals, we first performed a descriptive analysis to evaluate the number of medical/surgical patient-days, the number of hospitals with passive or active antibiotic-management programs, and the types and quantities of antibiotics used. We then used univariate linear regression techniques to evaluate the relationship between these types of controls (active and passive) on use of parenteral antibiotics and all other variables. Finally, we performed multiple linear regression techniques to evaluate the effect of policies regarding parenteral antibiotic use on 3 different outcomes in each hospital: annual expenditures, total cost per patient-day, and the percentage of difference between the hospital's cost and the average cost per 1000 patient days while controlling for the Medicare case mix index of each hospital, whether or not the hospital was a teaching institution, and the number of patient-days.
In an effort to understand how the use of specific antibiotics may have accounted for either the differences or similarities between hospitals, we sorted all antibiotics for each hospital according to the cost per day of treatment from the most to the least expensive. The average cost per day of treatment was then computed. Antibiotics were then classified according to whether they fell above or below the mean, and the pattern of use of the more and less-expensive antibiotics was explored. Additionally, antibiotics were analyzed by grouping certain antibiotics into a broadspectrum group (cefotaxime, ceftazidime, ceftriaxone, ciprofloxacin, imipenem, and ticarcillin/clavulanate) to compare them with the relatively narrow-spectrum remaining group, and the patterns of use according to these 2 categories were evaluated across hospitals and hospital characteristics. For both categorizations, comparisons of means and medians were evaluated by use of the t test and the nonparametric Wilcoxon ranked sum test, respectively. Significance was defined as P<.05 by use of a two-tailed test. All descriptive and statistical analyses were done with the Prophet statistical software program (BBN Systems and Technologies, Cambridge, MA).
Results
The 14 hospitals compared in this study were all acute-care, private, nonprofit institutions providing primary and secondary care in urban or suburban population centers in the northeastern (10), southeastern (3), and far western (1) United States. They ranged widely in size, having between 19,000–105,000 medical/surgical patient-days (mean, 58,787) during the 12-month period analyzed. Four were medical school-affiliated teaching hospitals with 2–5 accredited residency programs. Overall expenditures for parenteral antibiotics other than antiviral and antifungal agents ranged from $119,000–$1,740,200 (mean, $738,086). Despite the wide range in size of the hospitals and their expenditures for parenteral antibiotics, acuity of patient care (as defined by Medicare case mix indices) reflected a narrow range, from 1.13–1.37.
All hospitals used a variety of passive antibiotic-management policies and procedures, including formulary restrictions and the use of a pharmacy committee to determine which antibiotics were to be maintained on the hospital formulary. Most hospitals also used additional passive antibiotic controls, such as selective antibiotic susceptibility reporting, weekly parenteral antibiotic renewal order policies, educational efforts aimed at optimizing prescribing practices of physicians, and the restriction of pharmaceutical promotion activities. Five hospitals also used prospective interventional types of approval or recommendation programs carried out by teams of clinical pharmacists and infectious diseases staff physicians. Accordingly, these 5 hospitals were defined as having active antibiotic-management programs (group A), whereas the remaining 9 hospitals, which used only passive management policies and procedures, were evaluated as a separate group (group B). None of the hospitals had active programs to facilitate changing patients' therapy from parenteral to oral during the time frame studied.
Using multiple linear regression methodology, we found that total yearly expenditures for parenteral antibiotics were highly correlated with the size of the hospital, as measured bymedical/surgical patient-days (P<.001), although such a correlation failed to be maintained when analyzed on the basis of cost per 1000 patient-days. Total cost per 1000 patient-days also failed to correlate with either the teaching status of the hospital (P=.25) or the Medicare case-mix indices (P=.13) for the 14 hospitals studied. When hospitals were stratified according to the types of antibiotic-management programs being used, the 5 hospitals with active antibiotic-management programs (group A) had expenditures that were clearly lower than those in hospitals with only passive controls (group B) (P<.001), despite the fact that overall use of parenteral antibiotics was 36.7% greater for group A hospitals (670 DDD per 1000 patient-days) than for group B hospitals (490 DDD per 1000 patient-days). By expressing expenditures in terms of the percentage of difference above and below a mean value for the 14 hospitals, we found that the least costly hospital was 38% below the mean and the most costly 30% above the mean for the group (figure 1).
In an effort to understand how the use of specific antibiotics may have accounted for differences in expenditures, the 29 parenteral antibiotics evaluated were organized according to cost per DDD from the most expensive to the least expensive. By calculation of the average cost per DDD for all parenteral antibiotics ($37.56), the antibiotics were divided into 2 groups. Use of the 10 antibiotics that were above average in cost varied quite widely among the hospitals (table 2). The use of these 10 antibiotics was not different for teaching versus nonteaching hospitals, did not correlate with the Medicare case mix index, and was not found to be different between group A and group B hospitals. In contrast, hospitals with active management programs used significantly more of the less-expensive group of antibiotics per patient-day than did group B hospitals (P < .005).
Table 2
Comparison of the use of the 10 antibiotics that were above average in acquisition cost per defined daily dose (DDD), showing the range in use among the 14 hospitals analyzed.
By ranking the parenteral antibiotics according to their overall use among the 14 hospitals and then comparing use between group A and group B with respect to individual antibiotics, we found both similarities and differences (table 3). Although analysis of individual antibiotics disclosed particularly wide variation in use for antibiotics such as cefuroxime, ceftriaxone, and ciprofloxacin, even those antibiotics that were used less variably, such as vancomycin, oxacillin/nafcillin, and imipenem, showed substantial differences in use (figure 2).
Table 3
Parenteral antibiotic use per 1000 patient-days (PDs) for all antibiotics evaluated, from most to least used.
Discussion
Our analysis of parenteral antibiotic use among 14 similar acute-care hospitals disclosed wide variations both in overall expenditures per patient-day and in the use of many individual antimicrobial agents. Although expenditures did correlate with the size of the institutions, as measured by total patient-days per year, there was no correlation when costs were expressed in terms of expenditures per 1000 patient-days or with their status as teaching hospitals. In contrast, there was a strong correlation between cost per 1000 patient-days when hospitals with active antibiotic-management programs were compared with hospitals that used only passive controls (P < .001). Although there may have been a trend toward increasing expenditures on the basis of increasing acuity, as measured by the Medicare case mix index (P = .13), the size of the study along with the relatively narrow range of acuity represented precluded ideal analysis of this issue.
In analyzing the basis for the striking difference in expenditures between group A and group B hospitals, we found that group A hospitals used significantly more (P < .005) of the 19 less-expensive antibiotics than did group B hospitals. Some of this difference was most probably due to the greater use of combinations of less-expensive antibiotics (such as ampicillin/ gentamicin/metronidazole among group A hospitals in clinical situations) in which group B hospitals used broad-spectrum monotherapy with antibiotics such as imipenem/cilastatin or timentin/clavulanate. However, this issue was not able to be directly analyzed because of the limitations inherent in pharmacy acquisition cost-based data collection. Although the use of more narrow-spectrum, less costly antibiotics was greater in group A hospitals with active management programs, statistically significant differences were clearly seen only for oxacillin/nafcillin (P < .02), trimethoprim-sulfamethoxazole (P < .05), and gentamicin (P < .05) (table 3). Although analysis of all 3 aminoglycosides as a group failed to show a clearly significant difference between group A and group B hospitals (P < .09), cefazolin, which accounted for 15% of all parenteral antibiotic use, and vancomycin had similar utilization among both group A and group B hospitals.
Although group B hospitals with limited antibiotic controls used more of the 10 most expensive antibiotics than did group A hospitals, this difference was not found to be statistically significant (P = .30), possibly because of the very wide variability in the use of these antibiotics at individual institutions, which ranged from 174%–592% above to 78%–100% below the mean use of each of the antibiotics noted in table 2. Despite there being no significant difference when comparing use of broad-spectrum antibiotics as a group between group A and group B hospitals, the use of each of these antibiotics was greater in group B hospitals, although the difference was clearly significant only for imipenem (P < .01).
One of our most striking findings is that the use of individual parenteral antibiotics varies widely between hospitals, despite the similarities of the group of hospitals analyzed. Although some of these differences were undoubtedly related to the offsetting use of one broad-spectrum antibiotic versus another and possibly the impact of procurement agreements that differed between hospitals, it is of note that the widest range of differences in use was seen among the relatively broad-spectrum antibiotics, such as imipenem (183% above to 90% below mean use), ticarcillin/clavulanate (250% above to 100% below mean use), ceftazidime (240% above to 99% below mean use), and ceftriaxone (198% above to 78% below mean use). In contrast, smaller differences in use were seen for more narrow-spectrum antibiotics, such as vancomycin (55% above to 58% below mean use), oxacillin/nafcillin (112% above to 81% below mean use), and cefazolin (106% above to 58% below mean use). Although some of the differences in individual antibiotic use could reflect variability in the frequency of certain types of infections in different institutions, which was not able to be evaluated in this analysis, the magnitude of the differences in use reported here would seem unlikely to be explained on such a basis, particularly given the narrow range of acuity among the group of hospitals.
Linear regression analysis of overall parenteral antibiotic use strongly suggests that the 68% range in expenditures per patient-day for parenteral antibiotics is due to the presence of active antibiotic-management programs. However, the hospitals within each group demonstrated relatively wide differences in the range of their expenditures: 34% for group A hospitals and 28% for group B hospitals. A definitive analysis of the basis for such variations is beyond the scope of the present study, but it is of note that these variations were substantially greater than variations in the Medicare case-mix index for either group of hospitals (group A, ±15%; group B, ±13%). About half of the variations in expenditures among hospitals within each group could not be explained by the absence or presence of antibiotic-management programs, which suggests that other factors possibly related to practice patterns of physicians within individual hospitals may be playing an important role. Support for such a possibility is found in recent studies of variability in clinical practice, which have identified variances in practice patterns related to cesarean section delivery rates among obstetricians (23.2%) [21], angioplasty rates after myocardial infarction (10%), and coronary artery bypass surgery rates after acute myocardial infarction (3%) [22]. Although none of the hospitals had active programs to facilitate changing parenteral antibiotic therapy to oral therapy or specific programs to optimize the early discharge of patients to receive home iv therapy during the time frames studied, it is possible that variations in such practices between hospitals may have accounted for some of these differences.
Although the current study was not able to evaluate the impact of antibiotic-management programs on rates of nosocomial infection or clinical outcomes, these issues have been addressed in reports from individual institutions. Decreases in specific types of nosocomial infections (such as colitis due to Clostridium difficile [23] and vancomycin-resistant enterococcal infections [24–26] caused by individual antibiotic restriction) have been realized without adversely affecting several outcome measures. Although only limited studies to determine whether broad-based active intervention programs adversely affect clinical outcomes have been reported, neither the case-controlled study by Fraser et al. [15] nor a pre- versus postintervention study by Chapnick et al. [27] found any evidence of adverse impacts of their active antibiotic-management programs. Two studies that evaluated the impact of the implementation of active antibiotic-management programs on mortality [8, 11] failed to find any evidence of adverse impact on mortality after implementation of the programs. A recent report by White et al. [17] noted that implementation of an active prior-approval antimicrobial control program did not delay the administration of appropriate antibiotics or reduce survival time among patients with bacteremia due to gram-negative bacilli, did not increase overall mortality or specific mortality associated with different infections, and did not increase the time to discharge from the hospital or intensive care unit in patients with bacteremia due to gram-negative organisms.
Despite the small number of hospitals reported in this study, several factors suggest that the results may have broader relevance than the size of the study would suggest. For example, the mean total patient-days for the 14 hospitals was quite similar to the mean patient-days reported for the cohort of similar acute-care hospitals (those with 100–199 beds) reported by the American Hospital Association in 1994 (58,818 vs. 58,143) [28], the calendar year during which most of our data were collected. Furthermore, the relatively narrow range of the Medicare case mix indices for the group (1.13–1.37) and the fact that the mean Medicare case mix index for the group was quite similar to the 1994 national median Medicare case mix index (1.232 vs. 1.240) [29] suggest that the acuity and complexity of patient care of the 14 hospitals studied was quite similar to those of the average US acute-care hospital in 1994.
This epidemiologic analysis of parenteral antibiotic use among 14 community-based hospitals may not be directly extrapolated to either smaller institutions or larger tertiary-care hospitals. However, the strong correlation between annual expenditures for parenteral antibiotics and the type of antibiotic management programs supported by the hospitals in this report suggests that further analysis of antibiotic use among hospitals is warranted to better understand the cost and effectiveness of such programs, as well as to further assess the basis for the variability in the use of antibiotics that appears to exist even among a sample of similar acute-care hospitals. It is also of note that such initiatives are strongly supported by the Joint Commission on Accreditation of Health Care Organizations, which mandates that institutions develop quality improvement activities and programs, using intrainstitutional data comparisons to establish benchmarks [30].
Unfortunately, the present study was not able to compare and analyze the potential impact of overall antibiotic use or that of broad-spectrum antibiotic use on microbial resistance patterns, because of the lack of standardized methodologies for monitoring such data in the hospitals studied. A recent study by Itokazu et al. [31], which used a standardized epidemiologic model to compare rates of antimicrobial resistance in 369 intensive care units from 1990–1993, suggests that linking such methodology with a standardized model for comparing antimicrobial use may make it possible to quantify the impact of various types of antibiotic-management interventions on the development of antimicrobial resistance over time. Furthermore, the use of a standardized method for comparing antibiotic use between hospitals or within provider groups, such as health management organizations and preferred provider organizations, in a manner similar to that used in this study could represent a useful means for analyzing the costs and effectiveness of the types of antibiotic stewardship programs first advocated by Kunin [4–6], later by Marr et al. [32] and McGowan et al. [33, 34], and most recently by the Joint Committee on the Prevention of Antimicrobial Resistance [7].
Acknowledgements
We wish to express our deep appreciation to the infectious diseases clinicians, pharmacists, and hospital administrators who voluntarily provided the data that made this study possible, as well as to Dr. William R. McCabe, whose advice and guidance throughout the project were greatly valued.
- Received November 16, 1998.
- Revision received June 21, 1999.
- © 1999 by the Infectious Diseases Society of America
References
This Article
- AbstractFree
- » Full Text (HTML)Free
- Full Text (PDF)Free
Classifications
Services
Citing Articles
Google Scholar
PubMed
Related Content
Share
- Email this article
-
CiteULike
Connotea
Delicious
Facebook
Google+
Mendeley
Twitter
Published on behalf of
IDSA Members: For your free access to this journal, log in via the IDSA members area.
The Journal
Impact Factor: 8.186
5-Yr impact factor: 7.898
For The Media
For Authors
- Services for authors
- Self-archiving Policy
- IDSA Conflict of Interest policy
- Instructions for Authors
- ICMJE Form
- Submit Now!
Open access options for authors visit Oxford Open
This journal enables compliance with the NIH Public Access Policy
Corporate Services
Alerting Services
Related IDSA Publications
