Background
Antimicrobial resistance (AMR) is an urgent public health threat globally and in the US. 1,Reference Fitzpatrick, Suda and Poggensee2 In 2013, the Centers for Diseases Control and Prevention estimated that infections with AMR were associated with at least 23,000 excessive deaths annually in the US. 3 Antimicrobial stewardship programs (ASPs) play vital roles in reducing unnecessary antimicrobial use and combating AMR. One of the key recommendations for ASPs is to measure antimicrobial consumption as a metric. Reference Barlam, Cosgrove and Abbo4
To be effective, ASPs need practical methods for monitoring antimicrobial use. Reference Barlam, Cosgrove and Abbo4–Reference Berrevoets, Ten Oever and Sprong6 In 2014, the Veterans Health Administration (VHA) mandated all its facilities develop and maintain an ASP. Reference Kelly, Jones and Echevarria7,8 However, according to a 2015 mandatory survey of 140 VA facilities, 64% reported that limited IT support and/or data tools presented “substantive challenges to achieving optimal antimicrobial use.” 9
To address this critical resource gap, we developed a multicenter, automated electronic dashboard for ASPs that displays risk-adjusted benchmarking metrics for different categories of antimicrobial consumption. This dashboard was deployed in the acute care, intensive care, and long-term care units of 10 VHA hospitals that participated as pilot users. This study aimed to evaluate the impact of these metrics on ASP activities and the acceptance of these metrics among ASP members.
Methods
Ethics
The Institutional Review Board of the University of Iowa and Iowa City Veterans’ Health Care System approved this study. A waiver for written informed consent was granted.
Setting
VA Midwest Health Care Network (VISN 23) serves over 440,000 enrolled Veterans through an integrated system of 10 acute-care medical centers and eight long-term care facilities, ie, community living centers (CLCs). Additionally, VISN 23 has 69 community-based outpatient or outreach clinics and four domiciliary residential rehabilitation treatment programs, but these were not covered by the implemented dashboard. Hospital capacity at each site ranges from 15 to 229 beds, while CLC size is up to 225 beds. VISN 23 also includes special programs such as a spinal cord injury (SCI) program, cardiac surgery, polytrauma, and transplant.
The dashboard development team is located at the Iowa City VA Health Care System and collaborated with antimicrobial stewards at all 10 medical centers. All facilities designated physician and pharmacist champions for ASPs according to the VHA internal guideline, but the availability of local expertise, especially infectious diseases specialists and informatics support, varied. Specifically, four medical centers did not have any local infectious diseases specialist, and only two facilities had pharmacists who specialize in infectious diseases. No facility had informatics support specifically allocated for ASPs.
Dashboard tool
We built a system to extract patient-level data for antimicrobial consumption and demographics for acute inpatient and long-term care units at all VHA hospitals each month, utilizing the VHA’s Corporate Data Warehouse. We also collected data for underlying diagnoses prior to hospital admissions and procedures 90 days prior to or during admission. Underlying diagnoses were obtained from inpatient and outpatient diagnoses (International Classification of Diseases 10th edition [ICD-10]) and were classified into 86 categories based on Hierarchical Condition Categories. 10 Procedures were obtained from inpatient and outpatient procedure records recorded as ICD-10 procedure codes or Current Procedural Terminology codes. Those codes were classified into 224 categories based on Clinical Classifications Software developed by the Agency for Healthcare Research and Quality. 11 We performed risk adjustments based on negative binomial regression models with patient- and unit-level factors by calculating observed-to-expected ratios of antimicrobial use for each hospital and for specific units within each hospital.
Risk adjustment models included month, unit types (eg, intensive care unit [ICU] vs non-ICU for acute care), specialty, age, gender, comorbidities (50 and 30 factors for acute care and long-term care, respectively), and preceding procedures (45 and 24 procedures for acute care and long-term care, respectively). We created additional models for each antimicrobial category based on National Healthcare Safety Network (NHSN) definitions. For each hospital, risk-adjusted benchmarking metrics and a monthly ranking within the VHA system were visualized and presented interactively to end users through the dashboard (Supplementary Material).
Each hospital had its dedicated dashboard, and the access was restricted only to participating ASP stewards and antimicrobial prescribers. The technical detail of the dashboard has been described elsewhere. Reference Goto, Perencevich and Marra12
Design
The Principal Investigator (M.G.) identified frontline ASP stewards and antimicrobial prescribers from all 10 VISN 23 sites to participate in semi-structured interviews pre and post Center for Antimicrobial Stewardship and Prevention of Antimicrobial Resistance (CASPAR) dashboard implementation with the goal of interviewing at least one stewardship team member from each site. A qualitative trained medical sociologist (D.J.) created both the pre- and post-implementation interview guides informed by the clinical experiences of M.G. and D.L. and structured by the Consolidated Framework for Implementation Research (CFIR). Reference Escoffery, Riehman and Watson13 Two of five CFIR domains were of central focus to this research: the inner setting and intervention characteristics.
After completion of preimplementation interviews, we held a presentation and a live demonstration of the dashboard, then provided participants access to the dashboard. Two months following the initial presentation and the deployment of the dashboard to VISN 23, we recruited ASP champions and antimicrobial prescribers for post-implementation interviews from the live demonstration electronic mailing list. Participants were asked to evaluate several aspects of the dashboard including its ease of use, applicability to ongoing ASP activities, perceived validity and reliability, and advantages compared to other ASP monitoring systems. Additionally, all participants were asked to suggest potential participants in VISN 23 with stewardship experience or knowledge of stewardship metrics who could serve as participants.
Data and analysis
All interviews were conducted by D.J. via audio or videoconference call and were recorded. Interviews were transcribed and deidentified by an internal team of transcriptionists and D.J. reviewed the transcripts for accuracy and completeness.
Cleaned transcripts were uploaded into MaxQDA 2020.4 (VERBI Software, Berlin, Germany). D.J. employed a deductive approach to analysis, using the interview guide and the two CFIR domains of central focus to this study to identify barriers and facilitators associated with the CASPAR dashboard (intervention characteristics) and the participants’ facility (inner setting). An inductive approach to the data followed, identifying central themes across codes.
Results
We completed four preimplementation interviews and 11 post-implementation interviews from nine VISN 23 VA healthcare systems representing diverse perspectives of ASPs: ten pharmacists, four infectious disease physicians, one pharmacy program manager, and one pharmacy executive. Only one participant completed both a preimplementation and post-implementation interview, and two interviews were conducted wherein two ASP members from one facility were interviewed simultaneously. Three facilities were represented by more than one participant.
Four themes were identified throughout pre- and post-implementation interviews related to the adoption of the CASPAR dashboard: (1) the dashboard improved efficiency of established data collection practices, (2) stewards generally accepted the standardized risk-adjusted metrics as potential benchmarks, but facility size and structure shaped perceived utility of the dashboard, (3) additional training and staff was suggested to facilitate dashboard adoption, and (4) some stewards expressed uncertainty surrounding the dashboard’s ability to inform stewardship intervention efforts.
Theme 1. The CASPAR dashboard addressed less efficient established data collection and reporting systems
During pre- and post- interviews, participants were asked to describe how they obtained antimicrobial prescription data, with whom they shared the data, and how they used this data to inform stewardship activities. Most reported using data from the NHSN healthcare-associated infection tracking system and expressed difficulty finding essential data within this system, often relying on other staff to locate information. In contrast, the interface of the CASPAR dashboard was described as “easier to get around in,” “responsive,” and “user-friendly” (Table 1). The dashboard’s “useful data” and graphics were described as easier to incorporate into reports and presentations which facilitated communication of ASP metrics across facilities.
Table 1. CASPAR dashboard improves data collection
In addition to its user-friendly interface, the CASPAR dashboard reportedly reduced the amount of time spent working with multiple systems and staff. Concerning established data collection systems, participants described querying several systems, making stewardship “a little bit messier.” In the words of one physician, the ASP team felt they were, “…not able to focus on actually doing interventions with that data, we’re just trying to get the data” (site 4). In contrast, a pharmacist remarked that the CASPAR dashboard allowed access to data “immediately, at any point in time” without waiting for other personnel to update or retrieve data (site 8).
Participants also felt CASPAR metrics were more complete and accurate than NHSN metrics. The ability to track antibiotic usage in CLCs using the CASPAR dashboard was viewed as a “huge improvement” and addressed participants’ prior difficulties obtaining CLC data (pharmacist, site 5). Furthermore, the ability of dashboard metrics to adjust for “diagnostics, demographics, comorbidities, all different categories” unavailable in NHSN enhanced perceptions of the metrics’ validity (pharmacist, site 4).
Theme 2. Stewards largely accepted standardized and risk-adjusted metrics as potential benchmarks, but facility size and structure shaped perceived utility
Due to a lack of mutually accepted benchmarks within antibiotic stewardship practice, participants felt risk-adjusted metrics were an acceptable way to compare facilities and set benchmarks to assist with goal setting. However, during preimplementation interviews, participants reported NHSN hospital rankings as ineffective for goal setting due to their inability to adjust for significant facility characteristics. After reviewing CASPAR rankings, several participants regarded them as a superior metric. As one physician indicated, the NHSN dashboard “compares hospitals after adjusting for just a small number of factors such as the size of the hospital, whether the hospital has an ICU, and whether it’s a teaching hospital. So far, far fewer factors than [the CASPAR] dashboard is accounting for” (site 7). However, some participants expressed hesitancy to fully adopt the rankings as a benchmarking method due to concerns about the ability of risk adjustments to account for facility differences (Table 2).
Table 2. Perspectives of standardized and risk-adjusted rankings as benchmarks
Participants’ perspectives of their facility’s attributes, including size, structure, and patient population, influenced the perceived value of hospital rankings. Some participants felt CASPAR hospital rankings penalized their facility for its size and patient population. For example, a pharmacy manager from a small facility remarked, “We have to keep in mind that we have a very small sample size that can throw off these [rankings]” (site 6). Similarly, participants from facilities taking in more critically ill patients felt that their rankings would “look worse than they really are” (physician, site 7). Still, other participants felt the hospital rankings did not address a larger theoretical pitfall within the stewardship world:
“I think one of the biggest things that we all struggle with is how do we understand the appropriate level of use? Which I know is a kind of a philosophical part that’s not really a fix the dashboard can give, but I think we all look at this and we wonder, ‘Okay, here we are, we’re, you know, at 25% of facilities here. Is that lower than we want to be? Higher than we want to be?’” (physician, site 3).
Theme 3. Additional training and staff are needed to facilitate CASPAR dashboard adoption
While participants generally reported the CASPAR dashboard increased the efficiency of collecting and reporting data and presented more complete and accurate data, participants felt staff shortages presented challenges to learning how to navigate a new system:
“…it likely starts with staffing where one or two dedicated personnel have to be assigned to this. It’s more like one person a week, a new person picks up the next week. So, there is the lack of continuity and if we can address that I think that would be the first step towards obtaining the data and sharing it” (physician, site 3).
One participant viewed the CASPAR dashboard as a potential tool to advocate for more staff and additional stewardship resources (Table 3).
Table 3. Additional resources needed to fully adopt dashboard
Theme 4: Some stewards expressed uncertainty surrounding how CASPAR informs specific stewardship interventions
When asked what additional functionalities would be useful to incorporate into the dashboard, some respondents reported that rather than adding more metrics, they desired more direction on how to use the metrics to actively improve antimicrobial prescribing at their facility rather than using the dashboard as an evaluative tool for antimicrobial prescribing that already occurred (Table 4). One physician stated, “It just sort of provides a general overall assessment, but it doesn’t tell you exactly how you can improve aside from just decreasing antibiotic use—which is less actionable than some other hospital metrics” (site 7). In a similar vein, a pharmacy manager described the CASPAR dashboard as, “…really interesting—a lot of comparative data that’s very helpful, I just have a hard time trying to use the data and convert that into a measurable process improvement initiative” (site 6).
Table 4. Using the dashboard as an intervention tool
Discussion
ASP leaders desire metrics to assess the impact of stewardship activities in a variety of healthcare settings; Reference Moehring, Anderson and Cochran14,Reference Taber, Weir and Butler15 however, the effort required to extract metrics from electronic health records and translate analysis into tangible interventions are technically challenging and often times impossible with available resources. By combining multidisciplinary expertise (infectious diseases, pharmacy, informatics, and qualitative evaluation), we leveraged electronic medical record data to operationalize a centralized dashboard displaying risk-adjusted, hospital-level antimicrobial use. Through semi-structured interviews conducted before and after dashboard implementation, we identified several attributes that ASP leaders appreciated: a user-friendly interface, antimicrobial surveillance data for long-term care facilities (i.e., CLCs), and more rigorous hospital rankings on antimicrobial use. We demonstrated that there is a relative advantage in our risk-adjusted ASP metrics compared to currently available tools for benchmarking, eg, NHSN. Evaluating performance data facilitates hospital surveillance of defined and consistent metrics to ensure continuous improvement across different settings included in the dashboard. Reference Edwards16
Like other studies focusing on user acceptance of digital interventions for antimicrobial prescribers and stewards, Reference Taber, Weir and Butler15,Reference Van Dort, Carland, Penm, Ritchie and Baysari17,Reference Zetts, Stoesz and Garcia18 we found that although the dashboard released stewards from intensive data gathering, potentially leaving more time for developing appropriate interventions, some stewards expressed distrust of the dashboard metrics. Stewardship team members in facilities larger or smaller than average or caring for more critically ill patients expressed apprehension in accepting dashboard rankings. Additionally, while the dashboard reduced the burden of low staffing during the data collection phase, the difficulty of securing time to learn the new dashboard more thoroughly increased and did not completely resolve the well-documented short staffing of ASPs. Reference Seshardi, Felsen, Sellers and Dumyati19–Reference Greene, Nesbitt and Nelson21 Finally, and perhaps most significantly, some stewards voiced concern of a larger theoretical gap surrounding the interpretation of interfacility rankings and how the dashboard metrics could inform future interventions.
Our work attends to the need to “develop and validate metrics to guide more comprehensive evaluations of antimicrobial-prescribing at the facility-level.” Reference Suda, Livorsi and Goto22 Additional work is needed to evaluate the validity of comparing hospitals using the dashboard’s risk-adjusted version of a standard antimicrobial consumption metric. Furthermore, it remains unclear how the dashboard is being used longitudinally and what the consequences of this use are. Considering these needs, we see the work discussed in this report as an initial step to designing and implementing the dashboard on a broader scale.
Our study has both strengths and limitations. This qualitative study represents one of few studies to assess the acceptance of an automated dashboard tool for evaluating antimicrobial stewardship performance. Since all sites were VA hospitals, our findings may be less relevant to non-VA and non-US facilities. We chose sites with diverse care settings; however, the barriers described were common across sites. Furthermore, while the sample size (n = 15) is relatively small, the purposive sampling strategy and sample size were appropriate choices given the small sampling pool. Additionally, only one participant completed both a pre- and post-implementation interview, and we did not assess perspectives of prescribers not involved in leading ASP activities. Interviewees self-reported their processes, and we did not validate the accuracy of their statements. Furthermore, although interviews were confidential, participants may have been inclined to give socially desirable responses.
Given the limited resources for antimicrobial stewardship personnel, electronic tools in antimicrobial stewardship are an attractive method to facilitate compliance and improve efficiency. Reference Tinker, Foster, Webb, Haydoura, Buckel and Stenehjem23,Reference Rittmann and Stevens24 Emerging information technology is now opening the door to objective assessment of programs by peer-to-peer comparison (benchmarking). This, in turn, allows limited stewardship resources to be allocated for other stewardship activities. Our experience with the development and deployment of a dashboard tool demonstrates a large potential for informatic tools to facilitate antimicrobial surveillance and benchmarking, but the concurrent training and technical supports, as well as transparency for the data collection and risk adjustment, are important to achieve acceptance among ASP leaders.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/ash.2023.203
Acknowledgments
We thank all the ASP champions and antimicrobial prescribers from nine VISN 23 VA healthcare systems that participated from this qualitative study. The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the United States Government.
Financial support
This work was supported by the VHA National Centers for Patient Safety through the Patient Safety Center for Inquiry program and the Agency for Healthcare Research and Quality [grant number K08HS027472 to MG]. The funding source had no role in the design and conduct of the study; collection, management, and analysis of the data; or preparation, review, and approval of the manuscript.
Competing interests
The authors report no potential conflicts of interest. All authors have submitted the ICMJE form for disclosure of potential conflicts of interest.
Open access
