Editorial
April 19, 2024
Harnessing the Electronic Health Record to Improve Empiric Antibiotic Prescribing
Anurag N. Malani, Preeti N. Malani
JAMA. Published online April 19, 2024. doi:10.1001/jama.2024.6554
Approximately half of hospitalized adults in the US receive antibiotics, with pneumonia and urinary tract infection (UTI) being the 2 most frequent indications.1 Estimates suggest about 30% of all antibiotics prescribed in US hospitals are either unnecessary or suboptimal,2 and extended-spectrum antibiotic use is common.3Unnecessary antibiotic use contributes to numerous harms, including drug-related adverse effects, increased risk of drug resistance, and higher costs.
Antimicrobial stewardship aims to optimize antimicrobial use, minimize associated harms, and improve clinical outcomes, and over the past 2 decades, the field of stewardship has emerged and evolved. In 2014, the US Centers for Disease Control and Prevention introduced the Core Elements of Hospital Antibiotic Stewardship Programs to help hospitals with implementation.2 Among these Core Elements are hospital leadership commitment, accountability, pharmacy expertise, action, tracking, reporting, and education.
In 2015, the US National Action Plan for Combating Antibiotic-Resistant Bacteria set an implementation goal of the Centers for Disease Control and Prevention’s Core Elements in all hospitals that receive federal funding. By 2022, an impressive 97% of approximately 5100 US acute care hospitals reported meeting all 7 Core Elements of effective antimicrobial stewardship programs compared with only 41% in 2014.4 With much of the low-hanging fruit of stewardship interventions already picked (eg, guideline development and formulary restrictions), the demand for novel, scalable, sustainable strategies remains, especially addressing the empiric use of extended-spectrum antibiotic therapy.
In this issue of JAMA, Gohil and colleagues report the results of 2 related trials examining the use of computerized provider order entry (CPOE) prompts to reduce empiric use of extended-spectrum antibiotics among adults admitted for either pneumonia or UTI.5,6 The INSPIRE Pneumonia and INSPIRE UTI trials were conducted at a network of 59 geographically diverse community hospitals across the US. Using a cluster-randomized design, the investigators compared routine stewardship (ie, education, feedback) with the CPOE bundle composed of routine stewardship plus CPOE entry prompts that recommended standard-spectrum instead of extended-spectrum antibiotics during the first 3 hospital days (empiric period) for patients with a low absolute risk (<10%) of infections associated with multidrug-resistant organisms.
The CPOE algorithm was activated whenever extended-spectrum antibiotics were ordered for pneumonia or UTI in a location outside of an intensive care unit (ICU) within 72 hours of admission, including antibiotics ordered in the emergency department. In these instances, computerized prompts offered clinicians the option of seamlessly replacing orders for extended-spectrum antibiotics with guideline-recommended standard-spectrum antibiotics. The calculation of multidrug-resistant organisms risk was based on a sophisticated model that assessed more than 50 variables using the electronic health record (EHR) and included both hospital-specific (ie, local antibiogram) and patient-specific (ie, demographics, comorbidities, prior health care exposures, and history of multidrug-resistant organisms) factors.
The INSPIRE Pneumonia Trial included 96 451 patients (51 671 during the baseline period and 44 780 during the intervention period). Compared with routine stewardship, the hospitals assigned to the CPOE prompts reported a 28.4% reduction in empiric extended-spectrum days of therapy (rate ratio, 0.72 [95% CI, 0.66-0.78]; P < .001).5The INSPIRE UTI trial included 127 403 patients (71 991 during the baseline period and 55 412 during the intervention period). Hospitals assigned to the CPOE prompts reported a 17.4% reduction in empiric extended-spectrum days of therapy (rate ratio, 0.83 [95% CI, 0.77-0.89]; P < .001) compared with the routine stewardship group.6
Notably, the authors also examined 3 clinically meaningful safety outcomes including (1) days to antibiotic escalation (a switch from standard spectrum to extended spectrum), (2) days to ICU transfer, and (3) hospital length of stay. In addition, 3 sets of post hoc sensitivity analyses addressed the a priori decision to exclude patients initially admitted to a non-ICU location who were subsequently transferred to the ICU on hospital day 1 or 2. Both the safety outcomes and sensitivity analyses offer helpful reassurance that the observed reductions in extended-spectrum antibiotic use did not compromise patient safety or result in adverse clinical outcomes.
Beyond the complex logistics of conducting a trial of this magnitude, the decision to focus on empiric use during the first few days of hospitalization is noteworthy. Hospital-based stewardship efforts tend to emphasize de-escalation of antibiotics after microbiologic testing results return, and few focus on initial empiric prescribing. Empiric antibiotic prescribing for the hospitalized patient has a number of distinctive features. While no one wakes up and says, “I’m going to inappropriately prescribe extended-spectrum antibiotics today,” the decisions surrounding initial antibiotic selection are often subjective and even emotional. Specifically, clinicians may fear that any delay in appropriate antibiotics can increase a patient’s risk for worse outcomes including death, even when the possibility of an antibiotic-resistant pathogen is low.7 Empiric antibiotics in the hospitalized patient are often started in the emergency department and then continued, making inclusion of prescribing in this understudied setting a uniquely pragmatic feature of this study.8 In an effort aimed in part at alleviating potential medicolegal concerns, the CPOE bundle recorded the multidrug-resistant organism risk estimate in the medical record for each patient.
After reviewing the successes reported by Gohil and colleagues, readers may wonder: Could this type of CPOE prompt intervention reduce empiric use of extended-spectrum antibiotics at my institution? And what resources would be required for implementation?
The genius of the CPOE bundle is the elegant integration of CPOE prompts into existing clinical workflows. Although the interventions described may sound simple, the reality is that this system-level initiative would require strong support from leadership and broad collaboration among different groups of front-line clinicians (eg, hospitalists, emergency medicine, pharmacy, infectious diseases). Coordination with information systems technology is also paramount. Audit and feedback with regular reports to clinicians and ongoing coaching requires dedicated stewardship resources. These concerns aside, Gohil and colleagues were able to introduce these interventions at a diverse group of community hospitals—even sustaining the program during the early days of a global pandemic.
Beyond the need for adequate human and technology resources, there are several potential barriers that deserve mention. Other hospitals and health systems use different EHRs, which could be a large hurdle for widespread implementation. Are some EHRs better suited for this type of multidrug-resistant organism risk calculation? Smaller health systems and free-standing hospitals may lack the necessary resources for EHR optimization, potentially exacerbating existing health equity concerns. The authors do not provide specific details on what kinds of additional resources were needed to develop and maintain the CPOE bundle nor do they discuss potential costs or cost savings.
Additionally, patients often seek care at different hospitals and health systems, so key aspects of their microbiologic and antibiotic use histories may not be available. In the future, regional approaches with central repositories of history of multidrug-resistant organisms could help inform empiric antibiotic prescribing.9,10 The use of artificial intelligence may accelerate the ability for EHR vendors to develop more robust models to predict the risk of multidrug-resistant organisms associated with suspected infection. These potential barriers highlight the need for additional studies that focus on the implementation of an intervention using CPOE prompts, especially in settings with a more limited stewardship footprint.
Beyond making sure this approach can work in other settings, there may be opportunities to build on the success of this trial. For example, the CPOE bundle did not include appropriate diagnostic criteria for accurate diagnosis of pneumonia and UTI, which could further improve effectiveness.11,12 Even with the CPOE bundle and significant reductions, extended-spectrum antibiotic use remained high in both the pneumonia and UTI studies, while the rate of multidrug-resistant organisms isolated in cultures was low. Future studies should also consider cost savings, effects on local antibiograms, the nuances surrounding diagnostic certainty (for both pneumonia and UTI), and how to develop policy that incentivizes this type of innovation.
The aptly named INSPIRE trials do just that—provide inspiration and imagination, along with a powerful paradigm to harness the EHR to optimize antibiotic prescribing and improve human health. Rigorous studies that build on the successes reported by Gohil and colleagues are urgently needed.