Background.Antimicrobial stewardship programs aim to reduce inappropriate hospital antimicrobial use. At the Johns Hopkins Children's Medical and Surgical Center (Baltimore, MD), we implemented a World Wide Web–based antimicrobial restriction program to address problems with the existing restriction program.
Methods.A user survey identified opportunities for improvement of an existing antimicrobial restriction program and resulted in subsequent design, implementation, and evaluation of a World Wide Web–based antimicrobial restriction program at a 175-bed, tertiary care pediatric teaching hospital. The program provided automated clinical decision support, facilitated approval, and enhanced real-time communication among prescribers, pharmacists, and pediatric infectious diseases fellows. Approval status, duration, and rationale; missing request notifications; and expiring approvals were stored in a database that is accessible via a secure Intranet site. Before and after implementation of the program, user satisfaction, reports of missed and/or delayed doses, antimicrobial dispensing times, and cost were evaluated.
Results.After implementation of the program, there was a $370,069 reduction in projected annual cost associated with restricted antimicrobial use and an 11.6% reduction in the number of dispensed doses. User satisfaction increased from 22% to 68% and from 13% to 69% among prescribers and pharmacists, respectively. There were 21% and 32% reductions in the number of prescriber reports of missed and delayed doses, respectively, and there was a 37% reduction in the number of pharmacist reports of delayed approvals; measured dispensing times were unchanged (P=.24). In addition, 40% fewer restricted antimicrobial–related phone calls were noted by the pharmacy.
Conclusion.The World Wide Web–based antimicrobial approval program led to improved communication, more-efficient antimicrobial administration, increased user satisfaction, and significant cost savings. Integrated tools, such as this World Wide Web–based antimicrobial approval program, will effectively enhance antimicrobial stewardship programs.
Antimicrobials, the most commonly prescribed therapeutic agents, are administered to 30%–50% of hospitalized patients [1, 2]. As much as 50% of hospital antimicrobial use is unnecessary because of treatment of contaminants, overly broad antimicrobial coverage, or prolonged treatment, all of which result in selective pressure and emergence of antibiotic-resistant bacteria [1, 3–8]. The Infectious Disease Society of America and the Society for Healthcare Epidemiology recommend restriction of antimicrobial use to prevent the emergence of drug resistance [9, 10]. A survey of 48 participating pediatric hospitals in the Pediatric Prevention Network revealed that 32 had antimicrobial restriction policies and 21 had an antimicrobial restriction task force [11, 12].
There are 2 major varieties of antimicrobial stewardship programs (ASPs): (1) prior approval programs, in which authorization precedes release of restricted antimicrobials, and (2) prospective audit with feedback, in which prescribed and released antimicrobials are reviewed for appropriate dose, route, duration of treatment, streamlining and/or de-escalation, or discontinuation. Although ASPs may reduce antimicrobial use [13–16] (in part by focusing on expensive antimicrobials), they can be challenged by conflicting interests of the stakeholders (i.e., the primary physician vs. the infectious diseases practitioner), which may lead to limits on prescriber freedom and to dissatisfaction. Regardless of the approach, ASPs can face logistical challenges that can compromise patient care through miscommunications (particularly during telephone interactions), prolonged approval waiting times, or missed doses .
In 1983, the Johns Hopkins Children's Medical and Surgery Center (JHCMSC) implemented a prior approval ASP using pager and telephone communication. Anecdotally, the JHCMSC program was plagued with miscommunications, delayed and missed doses, user dissatisfaction, and an increasing workload attributable to increasing numbers of approval requests. We hypothesized that a World Wide Web (WWW)–based program that efficiently documents rationales for approval requests, decisions, and recommended durations and that transmits information to all relevant parties in real time would reduce redundancy and workload, would improve communication and user satisfaction, and would be cost neutral.
Setting and details of the existing approval program.The intervention and study were conducted at the JHCMSC, a 175-bed tertiary care children's hospital in Baltimore, Maryland. At JHCMSC, a prior-approval ASP restricts the prescribing of certain broad-spectrum and high-cost antimicrobials, as mandated by the Pharmacology and Therapeutics Committee (table 1). The policy is enforced by the pediatric infectious diseases service represented by the pediatric infectious diseases fellows (PIDFs) and the pediatric pharmacy.
In the original ASP (figure 1), the prescriber contacted the PIDF by pager; the PIDF gave verbal approval or disapproval and notified the pharmacy of approved antimicrobials by phone. Pharmacy notification of denied antimicrobial requests was not required. Restricted antimicrobial agents required verbal approval by the PIDF, who was contacted via pager between 8 a.m. and 11 p.m. During overnight hours, the initial doses of newly ordered restricted antimicrobials did not require approval (for convenience purposes), with subsequent doses requiring approval. After approving a request, the PIDF notified the pharmacy by phone (voicemail messages were not allowed), and the pharmacist logged the approval in the pharmacy's computer system (RxTFC, version 4.00; GE Healthcare). The PIDFs did not keep a written record of requests or approval status. Once the pharmacist was notified of the approval by the PIDF and an order on paper was sent and reviewed by the pharmacy, the restricted antimicrobial could be dispensed. Approval duration was indefinite, and there was no functionality to limit antibiotic therapy duration. Program challenges were increasing numbers of requests (600 requests per year during 2000–2001 and 1400 requests per year during 2000–2001), inconsistent documentation, prolonged antimicrobial use (unpublished data), and user dissatisfaction. The flow diagram (figure 1) outlines the original ASP and potential system failures.
Intervention.The WWW-based ASP was developed by a team of pediatric infectious diseases physicians, pharmacists, and an informatician. The system was implemented on a secure server and integrated into the institution's existing text-paging system. It includes an automatically generated inpatient census, lists of restricted and unrestricted antimicrobials, a notification system with text paging and e-mail reminders about requests (approved, disapproved, expired, pending, and missing), and “automatic” approvals for specific drug indication combinations (e.g., vancomycin for empirical treatment of suspected drug-resistant gram-positive meningitis). The number of “automatic” approvals was limited to facilitate discussion with the PIDF and education about appropriate antimicrobial use. The system does not disclose which drug indication combinations are eligible for automatic approval, to avoid inappropriate use of this feature by prescribers. A clinical decision support tool provides users with limited antibiotic-specific information on precautions, interactions, adverse events, and therapeutic drug monitoring. Users select patients from lists sorted by name, medical record number, or location and are prompted to select the requested antibiotic. The prescriber must provide the indication for antibiotic use (from an antibiotic-specific list or through free text), with supporting culture data. Once a request is generated, the PIDF is paged with the approval request. Using the WWW-based system, the PIDF enters an approval, indicating the duration of treatment, or rejects the request with rationale. The response is automatically transmitted simultaneously to the prescriber and to the pharmacy via pager. The duration of approval is recorded in the pharmacy computer system, which generates a stop date and time for antimicrobial treatment. Once the paper order is processed by the pharmacy, the restricted antimicrobial is dispensed. The ASP automatically pages prescribers with missing requests and expiring approvals and pages the PIDF with pending approval requests. The algorithm in figure 2 displays the flow of the system.
In preparation for implementation, pharmacists were inserviced by the participating pharmacy staff; the pediatric house officers attended a lecture and a practical session, and rotating surgical services were educated by the clinical pharmacist for each hospital floor. Instructional guides were posted at all workstations. The PIDFs were provided a wireless broadband laptop computer to facilitate ubiquitous access. The online antimicrobial approval program was implemented on 1 June 2005.
Evaluation.We evaluated the approval process before and after implementation of the WWW-based ASP. The evaluation period before implementation was from 1 July 2004 through 30 June 2005, and the evaluation period after implementation was from 1 July 2005 through 30 June 2006. The study was approved by the Johns Hopkins Institutional Review Board.
Measures.The assessment tools used before and after implementation included (1) satisfaction of prescribers and pharmacists, (2) delivery times for restricted and unrestricted antimicrobials, (3) evaluation of the number of dispensed restricted and unrestricted antimicrobial doses, and (4) evaluation of the cost of dispensed restricted and unrestricted antimicrobials.
Survey.An online survey was designed to capture prescriber knowledge of antimicrobial restriction policies, satisfaction with the approval process, and perception of missed and delayed doses. Pharmacists were surveyed separately regarding satisfaction with the approval process, delayed approvals, and the volume of restricted antimicrobial–related telephone calls. The preimplementation surveys were administered in the fall of 2004, and the postimplementation surveys were administered from November 2005 through January 2006.
Antibiotic delivery time.Restricted and unrestricted antimicrobial dispensing times were evaluated during 2 separate 13-day periods in the pre- and postimplementation phases (October 2004 and February 2006, respectively). The time from the documentation of the written antimicrobial order to transcription into the pharmacy computer system was used as a surrogate marker for dispensing time.
Dose and cost assessment.Data on dispensed doses of restricted and unrestricted antimicrobials for both periods were obtained from the inpatient financial database. Use of antimicrobial therapy was calculated as antimicrobial days per 1000 patient-days for both periods. Because of changes in antimicrobial policy during the study period, respiratory syncytial virus monoclonal antibody (Synagis) and the liposomal amphotericin agents were excluded from the cost and dose analyses.
Combining the inpatient financial database with administrative and financial data on antimicrobial cost in the children's center, we calculated costs of restricted and unrestricted antimicrobials during the pre- and postimplementation periods. Data included purchase costs, but not administration costs, of the restricted antimicrobials. Although the launch date of the program was 1 June 2005, we analyzed the postintervention period beginning on 1 July 2005, to allow for a brief transition period. It was anticipated that the purchasing cost would increase for almost all restricted antimicrobials during the study period. Assuming equal antimicrobial use patterns, we expected that purchasing costs would be higher during the postimplementation period because of the escalation in costs of antimicrobials. Furthermore, changes in contracts and generic status would have to be considered. To simplify the estimated cost savings, we calculated adjusted preimplementation cost, which we compared with actual postimplementation costs. Adjusted preimplementation cost was calculated as the number of doses during the preimplementation period multiplied by the number of doses and price during the postimplementation period, and actual postimplementation cost was calculated as the number of doses during the postimplementation period multiplied by the number of doses and prices during the postimplementation period. Cost savings were calculated by subtracting adjusted preimplementation costs from actual postimplementation costs.
Complexity and duration of hospital stay.Data on patient complexity and mean duration of hospital stay for all patients who received antimicrobials were analyzed to detect potential differences in severity of illness between the pre- and postimplementation periods. The complexity score was derived using the All-Patient Refined Diagnosis Related Group expressed on a 4-point scale .
Statistical analyses.Dispensing times of restricted and unrestricted antimicrobials were compared with use of the nonparametric Wilcoxon rank sum test. Exempt antibiotics and those that were prescribed during the overnight hours were not included in this analysis, because they are automatically dispensed without the need for PIDF approval. Comparison of duration of hospital stay and acuity were analyzed with use of a 2-tailed Student's t test.
Cost analysis.The annual cost of restricted antimicrobials, excluding respiratory syncytial virus monoclonal antibody and the liposomal amphotericin agents, decreased from the projected cost by $370,069 (decrease of 21.6%) within 1 year after implementation of the online ASP. Cost of unrestricted antimicrobials was similar (∼$570,000) during both periods.
Requests.During the preimplementation phase, a mean of 220 approval requests per month for restricted antimicrobials was recorded. During the postimplementation phase, the mean number of requests per month increased to 342 (total no. of annual requests, 4104; no. of requests per day, 11.2). Of the 4104 requests, 3600 (87.7%) were initial requests and 504 (12.3%) were renewal requests. The proportion of requests that were approved was 89.4% (3669 requests), and 446 (12.2%) of these requests were approved automatically by the program for specific drug indication combinations. Of the 3669 approved requests, 1950 (53.1%) were granted limited approval (for ⩽3 days) and, therefore, required renewed approval if the provider wanted to continue treatment. Of the 1950 limited approvals, only 256 (13.1%) had subsequent renewal requests, and of those, only 101 (5.2% of the 1950 limited approvals) were subsequently approved for additional treatment durations.
Doses dispensed and antibiotic-days.The number of doses of restricted antimicrobials dispensed decreased by 11%, from 125.8 doses per day (during the preimplementation period) to 111.8 doses per day (during the postimplementation period; a reduction of 5085 doses annually). The numbers of antibiotic-days (defined as the number of days that antibiotics were used) per 1000 patient-days were 485.4 days per 1000 patient-days (during the preimplementation period) and 417.6 days per 1000 patient-days (during the postimplementation period; a reduction of 14%). Figure 3 shows the most commonly requested and dispensed restricted antimicrobials. In addition, there was a 12% reduction in the number of doses of unrestricted antimicrobials dispensed, from 227.5 to 201.0 doses per day (9603 doses annually). Of note, the number of patient-days was nearly equivalent during the 2 periods (49,671 during the preimplementation period and 49,434 during the postimplementation period).
Severity of illness and duration of hospital stay.Assessment of demographic data on patients hospitalized at JHCMSC who received antibiotics during the pre- and postimplementation periods revealed similar mean complexity scores (2.17 vs. 2.22; P<.001). The mean duration of hospital stay (±SD) was also similar between the 2 periods (6.78±14.3 patient-hospital days vs. 6.67±14.1 patient-hospital days; P=.65).
Dispensing time.During the pre- and postimplementation periods, 80 and 84 orders for restricted antimicrobials, respectively, were evaluated for effects of the WWW-based ASP on dispensing time. Mean dispensing times (±SD) for restricted antimicrobials decreased from 2.59±3.73 h to 2.44±2.06 h (P=.24).
Assessment of 175 and 317 orders for unrestricted antimicrobials during the pre- and postimplementation periods, respectively, revealed that the mean dispensing time (±SD) decreased from 2.87±3.70 h to 1.93±2.20 h (P<.001). Although there were no differences in the dispensing times between restricted and unrestricted antimicrobials during the preimplementation period (P=.11), during the postimplementation period, the dispensing time for unrestricted antimicrobials was significantly shorter than that for restricted antimicrobials (P<.05).
Provider assessment.One hundred twenty-one prescribers responded to the preimplementation survey, compared with 78 prescribers who responded to the postimplementation survey. Of the 81 total pediatric house staff, 71 (88%) during the preimplementation period and 50 (62%) during the postimplementation period responded to the survey. Providers reported a 21% reduction in the number of missed antimicrobial doses and a 32% reduction in the number of delays of restricted antimicrobial doses during the postimplementation period, compared with the preimplementation period (table 2). Prescriber satisfaction with the ASP was 22% during the preimplementation period, compared with 68% during the postimplementation period (table 3). Pharmacists reported a 40% decrease in the number of telephone calls related to restricted antimicrobial use and a 37% decrease in the number of delayed approvals of restricted antimicrobials during the postimplementation period, compared with the preimplementation period. Of 15 surveyed pharmacists (60% response rate), 13% were satisfied with the ASP during the preimplementation period, compared with 69% during the postimplementation period. The preimplementation surveys generated comments focused on major shortcomings, such as miscommunications resulting in missed doses and overall dissatisfaction, and the postimplementation surveys generated comments focused on technical matters (e.g., expired approval pages being sent after a patient was discharged from the hospital) rather than on the approval process itself. Survey comments included that the WWW-based ASP enhanced communication, streamlined approvals, facilitated discussion, led to fewer workflow disruptions, and provided time savings for residents and PIDFs (data not shown).
The implementation of a WWW-based ASP at a tertiary pediatric hospital center led to (1) improved user satisfaction, (2) decreased numbers of missed and delayed doses, (3) decreased numbers of doses of restricted antimicrobials dispensed, and (4) significant antimicrobial cost savings. The WWW-based ASP offered a repository of approval status for antimicrobial requests and rationales for approval or denial. Furthermore, the system allowed automatic approvals, providing time savings for PIDFs without compromising the clinical decision support provided by the PIDFs for more-complicated requests. The database of requests provided reinforcement of the approval process by allowing all parties to review the status of their requests. In addition, this database provided a mechanism for identifying breaches in the approval process (i.e., order entry before submission of an approval request). Most importantly, the program and the database improved multidisciplinary communication.
Both front- and back-end approval systems [6, 19] may be plagued by the inability to limit the duration of approval of antimicrobials. By providing infrastructure through a database and by improving communication with regard to expiring approvals, our WWW-based program allowed effective implementation and enforcement of approval duration limits and stop dates. Prescribers were strongly encouraged, through repeated pages and reminders, to reassess the patient situation (e.g., culture results and alternative etiology) and to actively submit a subsequent request to continue the use of restricted antimicrobials. Interestingly, subsequent approvals were requested for antimicrobials with initial limited approval duration (⩽3 days) in only 13% of cases, and the majority of these renewal requests were denied. The PIDF determined initial duration of approval on the basis of the clinical scenario and presumptive diagnosis (e.g., pneumonia), with extensions based on review of subsequent approval requests with additional microbiological or clinical data. Of note, no increase in duration of hospital stay was observed during the postimplementation period.
Overall, the number of doses of dispensed restricted antimicrobials and the duration of antimicrobial use decreased after implementation of the program, most likely as a result of limited approvals and the restricted approved durations. It is unclear why the number of doses of unrestricted antimicrobials also decreased, but we speculate that the practice of limiting the duration of therapy with restricted antimicrobials and, possibly, better understanding of antimicrobial use may have prompted providers to limit the duration of therapy with unrestricted antimicrobials also; however, this was not specifically studied. Because the number of patient-days remained unchanged during the postimplementation period, the decrease in the number of administered doses was not a result of decreased patient volume. Furthermore, these findings occurred even with higher patient complexity scores during the postimplementation period.
Interestingly, there was no difference in dispensing times of restricted antimicrobials between the pre- and postimplementation periods, likely reflecting the use of paper ordering, which was used during the entire study period. The units with proactive vigilance in terms of checking orders, particularly the oncology unit, had the shortest dispensing times (data not shown). There was no difference in dispensing times between restricted and unrestricted antimicrobials during the preimplementation period; however, during the postimplementation period, the dispensing time of unrestricted antimicrobials was significantly shorter than that of restricted antimicrobials. This finding requires further exploration.
Computer-assisted programs for prescribing antimicrobials have been linked to decreased numbers of doses of antibiotics, lower drug costs, and lower numbers of drug-associated adverse events  but have been limited by the lack of exportability to other institutions. A unique feature of our WWW-based program is that the PIDFs make decisions without the use of strict clinical algorithms. Prescribers always have the opportunity to communicate directly with the PIDF; this is an appreciated feature in solicited provider feedback. We acknowledge that the use of a PIDF in decision making may provide less standardized decision making and may limit the system's portability to institutions where this service is not available. Standard clinical algorithms, which were being used in 55% of hospitals in a 2006 study, could be easily added to the current WWW-based ASP, which already contains several automatic approval algorithms .
With the increasing cost of antimicrobials, true savings with our program are difficult to assess. As a result, prior studies that have attempted to quantify the impact of ASPs have revealed mixed results . The cost assessment of our system revealed significant projected cost savings, with the program factoring in changes in antimicrobial cost. Several reasons for the reduced number of doses and associated cost savings are possible, but direct cause and effect were not determined, and the study is limited in determining other causes of cost reduction. The effectiveness of the online program must also be weighed by considering time savings, by eliminating the multilevel paging previously required, and by providing a searchable database of approvals and rationales. Furthermore, the impact of limiting excessive antimicrobial use or of limitation of prolonged administration of broad-spectrum antimicrobials on patient outcomes, as well as the potential limitation of antimicrobial resistance, must be considered. These additional outcomes are challenging to measure.
A concern regarding implementation of the online ASP was the potential loss of personal communication and education that occurs during an approval request for antimicrobials. However, postimplementation survey comments revealed that the system had actually facilitated communication and education, while saving time.
Limitations of this study include a lack of specific knowledge about time savings (i.e., the actual reduction in the time that pharmacists spent fielding antimicrobial-related telephone calls). We did not investigate the potential impact of decreased antimicrobial use on resistance rates.
A multifaceted, integrated, and potentially transportable WWW-based ASP improved communication, user satisfaction, and time use, resulting in a safe and cost-effective means of control of antimicrobial drug prescribing practices in a tertiary care pediatric hospital. Integrated tools, such as this WWW-based ASP, effectively enhance antimicrobial stewardship programs.
We thank Dr. Marlene Miller and the Children's Safety Network; Dr. Kwang Sik Kim; Dr. George Dover and the pediatric administration, for their support; and Dr. Sara Cosgrove, for her careful review of the manuscript.
Financial support.Children's Miracle Telethon Fund.
Potential conflicts of interest.All authors: no conflicts.