Vancomycin remains one of the most commonly prescribed antibiotics in NICUs despite recommendations to limit its use for known resistant infections. Baseline data revealing substantially higher vancomycin use in our NICU compared to peer institutions informed our quality improvement initiative. Our aim was to reduce the vancomycin prescribing rate in neonates hospitalized in our NICU by 50% within 1 year and sustain for 1 year.
In the 60-bed level IV NICU of an academic referral center, we used a quality improvement framework to develop key drivers and interventions including (1) physician education with benchmarking antibiotic prescribing rates; (2) pharmacy-initiated 48-hour antibiotic time-outs on rounds; (3) development of clinical pathways to standardize empirical antibiotic choices for early-onset sepsis, late-onset sepsis, and necrotizing enterocolitis; coupled with (4) daily prospective audit with feedback from the antimicrobial stewardship program.
We used statistical process u-charts to show vancomycin use declined from 112 to 38 days of therapy per 1000 patient-days. After education, pharmacy-initiated 48-hour time-outs, and development of clinical pathways, vancomycin use declined by 29%, and by an additional 52% after implementation of prospective audit with feedback. Vancomycin-associated acute kidney injury also declined from 1.4 to 0.1 events per 1000 patient-days.
Through a sequential implementation approach of education, standardization of care with clinical pathways, pharmacist-initiated 48-hour time-outs, and prospective audit with feedback, vancomycin days of therapy declined by 66% over a 1-year period and has been sustained for 1 year.
Antibiotics are the most commonly prescribed class of medications in NICUs1,2 and are often prescribed with great variability between providers and institutions.2,3 Although antibiotic use has proven successful in reducing infant mortality from bacterial infections,4 overuse is common.2 Overuse of antibiotics in newborns is associated with an increased risk of mortality5,6 and morbidities including necrotizing enterocolitis (NEC),5 candidiasis,7 resistant pathogens,8,9 and adverse drug effects such as nephrotoxicity.10 Antibiotic exposure in neonates is also associated with additional long-term morbidities such as asthma,11 inflammatory bowel disease,12 and obesity.13
Vancomycin is one of the most commonly prescribed antibiotics in NICUs.14 It is primarily indicated for treatment of methicillin-resistant staphylococcal infections because its use for other indications is often unwarranted because less toxic alternatives are preferable.14–17 Vancomycin is associated with acute kidney injury in up to 9% of neonates.10 Because of the aforementioned risk, the American Academy of Pediatrics and the Pediatric Infectious Disease Society, through the Choosing Wisely campaign, recommend judicious use of vancomycin and avoiding empirical vancomycin in newborns unless there is a known risk for resistance to narrower-spectrum agents.18
When compared to peer hospitals, vancomycin usage in the Children’s National Hospital (CNH) NICU in 2016 was 3 times higher (95th percentile) than the mean use in children’s hospitals contributing to the Pediatric Health Information System database (112 vs 34 days of therapy [DOTs] per 1000 patient-days). Informed by this benchmarking data and by preliminary data revealing that approximately half of vancomycin use was for indications for which our team believed vancomycin was unnecessary (empirical use for NEC and culture-negative sepsis), our specific aim was to reduce vancomycin DOT per 1000 patient-days by 50% in 1 year and sustain this reduction for 1 year.
This prospective single-center quality improvement (QI) project was conducted in a 60-bed level IV NICU in an urban academic freestanding children’s hospital. All infants admitted to the CNH NICU were outborn and transferred for subspecialty care. After admission, all infants receive a nasal methicillin-resistant Staphylococcus aureus (MRSA) polymerase chain reaction surveillance test at the time of admission and weekly thereafter. The incidence rate of MRSA colonization during the study period was 0.92 per 1000 patient-days.19 A hospital-wide antimicrobial stewardship program (ASP) had been developed within 1 year of the onset of this project and focused on the development of clinical pathways but had not yet incorporated NICU stewardship interventions. This project was undertaken as a QI initiative at CNH; it did not constitute human subjects research and was not under the oversight of the institutional review board.
Planning of Key Interventions
To understand the indications for vancomycin, a chart review was completed from July to September of 2017, with 514 days of vancomycin therapy identified. Physician notes were reviewed to identify indications for vancomycin. An indication of “rule out sepsis” was defined as vancomycin use <72 hours for “suspected sepsis” pending culture results, whereas “culture-negative sepsis” was defined when vancomycin use was continued ≥72 hours despite negative culture results when the terms “suspected sepsis” or “culture-negative sepsis” were documented. For clinical indications for which culture results would not drive antibiotic choices, the result of the patient’s routine MRSA surveillance swab was noted. The most common indications for vancomycin use were NEC, rule out sepsis, and culture-negative late-onset sepsis (LOS) (Fig 1). Baseline data informed the development of clinical pathways for common conditions resulting in empirical vancomycin courses.
The Institute for Healthcare Improvement Model for Improvement was used as the framework for our QI initiative. The key driver diagram (Fig 2) provided a guide for project implementation. This project had 4 major interventions, described below.
Intervention 1: Development of an Interdisciplinary Team and Provider Education
The interdisciplinary QI team included practitioners from neonatology, infectious diseases (ID), pharmacy, nursing, and QI. The team leader was an ID physician.
An education session codelivered by an ID and neonatology physician summarized institutional antibiotic prescribing practices compared to peer hospitals and the rationale for stewardship. During this session, neonatology providers were polled about their prescribing practices. A postcourse evaluation form sent to all participants indicated that all respondents were likely to change practice after the education session.
Intervention 2: Pharmacist-Initiated 48-Hour Time-Out
The division of pharmacy implemented a formal policy requiring clinical pharmacy specialists and unit-based pharmacists to perform a “48-hour antibiotic time-out.”21 Using Sentri7 (Pharmacy OneSource, Inc, Madison, WI), all clinical pharmacists’ personalized dashboard identified patients who had been on antibiotics for ≥48 hours. Monday through Friday, the clinical pharmacist evaluated the patient list and prompted discussion of the antibiotic plan on rounds. Providers were encouraged to either discontinue vancomycin orders or to switch to a narrower-spectrum antibiotic as indicated.
Intervention 3: Clinical Pathway Development
Derived from this initiative, new clinical pathways for LOS and NEC were developed with consensus from neonatology, ID, and surgery faculty (Supplemental Figs 7 and 8). The clinical pathways were developed after reviewing existing literature, national consensus guidelines, local antibiotic susceptibility data from clinical cultures of hospitalized neonates over the past several years, and guidelines implemented at other institutions.22–26 At our institution, it is standard to perform universal MRSA surveillance for all infants after admission to the NICU and weekly thereafter. The clinical pathways incorporated these MRSA screen results to identify infants at low risk for MRSA infection. The clinical pathways recommended discontinuing vancomycin in infants without MRSA colonization if no pathogens warranting vancomycin therapy were isolated from cultures after 48 hours. The clinical pathways were revised on the basis of stakeholder feedback. Final approval was provided by the stakeholders and the institutional antimicrobial subcommittee of the pharmacy and therapeutics committee. The pathways were posted in the NICU at workstations, mobile computers, and on the intranet and disseminated via e-mail to neonatology staff and reviewed in educational sessions for neonatology practitioners and trainees.
Intervention 4: ASP Prospective Audit and Feedback
A member of the ASP (ID physician or pharmacist) reviewed all vancomycin orders in the NICU daily (Monday through Friday). If opportunities to de-escalate were identified, recommendations were communicated either via the NICU pharmacist or directly with phone calls to the medical team. If the recommendation was not followed within 24 hours, the ASP medical director contacted the neonatology attending directly.
The primary outcome measure was vancomycin DOTs per 1000 patient-days. A DOT was defined as ≥1 doses of intravenous vancomycin administered to the patient on a calendar day. Antibiotic data were obtained from the hospital pharmacy dispense database. NICU patient-days were obtained from the hospital daily census.
A secondary outcome measure was vancomycin-associated acute kidney injury (AKI), defined according to the neonatal modified Kidney Disease Improving Global Outcomes criteria,27 as either ≥50% increase in serum creatinine above the baseline (lowest in the 6 months before initiating vancomycin), with the increased value above a threshold of 0.5 mg/dL or an absolute increase of 0.3 mg/dL that occurred during the course of vancomycin or up to 48 hours after its discontinuation.
Compliance with the clinical pathway for NEC and LOS pathways was determined by measuring vancomycin DOTs beyond 48 hours for (1) NEC and (2) culture-negative sepsis in patients with a negative MRSA screen result. These were considered subgroup outcome measures. Three 3-month audits were performed through chart review by reviewing the medical records of all infants prescribed vancomycin. For each audit, all vancomycin courses for that period were reviewed to identify the vancomycin indication. Each vancomycin DOT was associated with a specific indication. To provide context for possible changes in antibiotic prescribing practices for NEC, rates of NEC per 1000 patient-days were collected from the Children’s Hospital Neonatal Consortium.
A third process measure, the number of vancomycin orders per month for which a 48-hour time-out was performed and documented by the unit-based pharmacist (and whether it was followed), was used to monitor pharmacist compliance with the 48-hour time-out policy.
To ensure no patient harm resulted from reduction of vancomycin use, the incidence of S aureus bloodstream infections and Gram-positive sepsis-related mortality per 1000 patient-days was monitored. Gram-positive sepsis-related mortality was defined as a death in the setting of a blood culture with growth of a Gram-positive bacterial pathogen within 14 days. Positive blood culture results were extracted from the microbiology laboratory database, and mortality data were collected from our institution’s database maintained as part of the Children’s Hospital Neonatal Consortium.
To monitor the effect of these interventions on overall and alternate antibiotic use, additional balancing measures included total antibiotic DOTs per 1000 NICU patient-days and clindamycin, oxacillin, and ampicillin DOTs per 1000 NICU patient-days.
Statistical process control charts tracked outcome measures over time and were created by using Process Improvement Products software (QI Charts) and QI Macros for Excel. Statistical process control charts rules were applied for differentiating special- versus common-cause variation.28 A u-chart was used for the outcome measures vancomycin DOTs per 1000 patient-days and vancomycin-associated AKI events per 1000 patient-days.
Modified control charts were used for the first 2 process measures, evaluating vancomycin DOTs beyond the first 48 hours of therapy for patients with NEC with negative MRSA screen results and culture-negative sepsis with negative MRSA screen results. NEC rates per 1000 patient-days with 95% confidence intervals (CIs) were calculated for a Poisson distribution and compared between the pre- and postintervention periods. Because the date of event (NEC) was not available but rather linked to the date of patient admission, 3 months (representing >95th percentile for NICU length of stay at our institution) was used as a washout period. For the third process measure, a p-chart was used to evaluate the proportion of pharmacist-initiated 48-hour time-outs for vancomycin documented each month, using all vancomycin courses prescribed for >48 hours as the denominator, excluding those opportunities that occurred on a weekend when a pharmacist would not be available.
Gram-positive sepsis-related mortality rates and incidence rates of S aureus bacteremia episodes per 1000 patient-days with 95% CIs estimated for a Poisson distribution were compared between the baseline (September 2016–September 2017) and the periods after implementation (October 2017–September 2019). Antibiotic DOTs per 1000 patient-days were monitored by using a u-chart.
Primary Outcome Measure
During the 1-year baseline period (September 2016–September 2017), mean vancomycin use was 112 DOTs per 1000 patient-days. After establishing the QI team and implementing the first 3 interventions (provider education, pharmacist-initiated 48-hour time-outs, and clinical pathway rollouts), special-cause variation was detected, resulting in a 29% decrease in vancomycin use to 80 DOTs per 1000 patient-days. After implementation of ASP team daily review, a second centerline shift was detected, with further decline in vancomycin use to 38 DOTs per 1000 patient-days. Over a period of 12 months (October 2017–September 2018), these interventions resulted in a 66% reduction in overall vancomycin use from baseline. These results were sustained for 1 year (Fig 3).
For the baseline period, the rate of vancomycin-associated AKI was 1.4 events per 1000 patient-days. After the interventions, special-cause variation was noted with a decline to 0.1 events per 1000 patient-days (Fig 4).
Vancomycin DOTs for NEC beyond the first 48 hours of therapy declined from 66 per month during the baseline period (July–September 2017) to 20 DOTs per month (February–April 2018) and subsequently to 8 DOTs per month (August–October 2018) with an overall reduction of 88%. Excluding the first 2 DOTs, vancomycin DOTs for culture-negative sepsis decreased from 22 DOTs per month in the baseline period to 13 DOTs per month (February–April 2018) and subsequently to 8 DOTs per month (August–October 2018) (overall 63% reduction) (Fig 5). The incidence rate of NEC in the preintervention period (September 2016–July 2017) was 4.31 per 1000 patient-days (95% CI: 3.24–5.73) and was 2.63 per 1000 patient-days (95% CI: 1.88–3.67) in the postintervention period (December 2017–December 2018). The incidence rate ratio of the postintervention period compared to the preintervention period was 0.61 (95% CI: 0.39–0.95).
After implementation, the proportion of documented pharmacist-initiated 48-hour time-out interventions out of all possible opportunities averaged 49% with wide variation month to month (Supplemental Figs 9 and 10).
S aureus bacteremia incidence ranged from 0 to 1 episode per month, with no significant difference in incidence from the preintervention period (0.42 episodes per 1000 patient-days) to the postintervention period (0.31 episodes per 1000 patient-days) (P = .57). The Gram-positive sepsis-related mortality rate was 0.089 deaths per 1000 patient-days in the preintervention period (September 2016–September 2017) and 0.099 deaths per 1000 patient-days in the postintervention period (October 2017–January 2019), with an incident rate ratio of 1.10 (95% CI: 0.12–10.1).
Antibiotic prescribing for all antibiotics in the NICU declined by 20% from 600 in the baseline period to 480 DOTs per 1000 patient-days in the postintervention period. Ampicillin use remained the same, whereas the clindamycin and oxacillin prescribing rates increased (Fig 6).
In the level IV NICU of an urban academic freestanding children’s hospital with a high baseline vancomycin prescribing rate, a sequential approach of education (level 1 reliability), standardization of care with clinical pathways (level 2 reliability), pharmacist-initiated 48-hour time-outs (level 2 reliability), and prospective audit with feedback (level 2 reliability) lowered the rate of vancomycin DOTs per 1000 patient-days by 66% over a 1-year period. This exceeded the target and has been sustained for 1 year. The patient safety implications of this achievement are highlighted by decrease in vancomycin-associated AKI without a change in Gram-positive sepsis-related mortality or S aureus infections. Although vancomycin DOTs per 1000 patient-days decreased by 66%, an overall 20% reduction in total antibiotic DOTs per 1000 patient-days was noted, suggesting that not all vancomycin reduction was achieved through replacement with an alternate antibiotic.
The relevance of these results are highlighted by a recent large retrospective review of >300 NICUs across the United States identifying vancomycin as the fourth most common medication used.14 We found an acute need to decrease the vancomycin within our NICU on the basis of benchmarking data, but because vancomycin continues to be 1 of the most commonly prescribed antibiotics in NICUs across the United States, the interventions we have described are likely to be of interest to other centers.
Authors of previous studies evaluating judicious vancomycin use in the NICU setting have used strategies focusing on both limiting empirical starts of vancomycin29,30 and de-escalation from vancomycin,31 which was the primary focus in our study. To our knowledge, this is the first work to reveal a significant change in vancomycin-associated AKI in neonates. Admittedly, this is not a surprising result given the high baseline vancomycin prescribing rate and substantial decrease in opportunities to develop vancomycin-associated AKI. Two retrospective chart reviews of blood culture positive for LOS revealed that not using vancomycin empirically for LOS did not negatively impact clinical outcomes including sepsis duration and mortality rate.29,30 Chiu et al32 implemented a guideline in 2 tertiary-care NICUs with low rates of MRSA infection and found both a significant decrease in the number of vancomycin starts as well as a reduction in the number of infants treated with vancomycin per 1000 patient-days. Compared to their study, our initiative resulted in a greater reduction in vancomycin DOTs while focusing primarily on de-escalation of vancomycin rather than limiting initiation of empirical vancomycin. Holzmann-Pazgal et al31 evaluated the effectiveness of a combination of education, guideline development, and prospective audit with feedback in a single-center tertiary-care NICU. They found significant decreases in vancomycin use after education and guideline development but no additional decreases after adding audit with feedback. Our study, on the other hand, revealed initial 30% reduction in vancomycin use after education and clinical pathway implementation, but it was only after implementing prospective audit with feedback that our target was achieved with a decline in vancomycin use of >50%. One notable difference is that in their study, prospective audit with feedback was performed by NICU staff physicians, whereas in our model, the antimicrobial stewardship team (ID physician and/or pharmacist) provided the feedback to NICU physicians to de-escalate vancomycin when indicated.
There are several limitations to this study. First, the clinical pathways developed for NEC and LOS were based on best practice and expert opinion because clear evidence to support 1 antibiotic regimen over another is lacking. Interestingly, rates of NEC were lower in the postintervention period compared to the preintervention period, implying that other factors may have contributed to the decline in vancomycin use for NEC (process measure). Second, although the definition of vancomycin-associated AKI is one frequently used in published literature,33,34 additional factors contributing to AKI were not accounted for, such as concomitant nephrotoxic drugs and severity of illness. Additionally, 48-hour time-outs were only documented as having been performed approximately one-half of the time for all potential opportunities; this may have been in part because of pharmacist personnel shortage for several months and represents an opportunity for improvement for the future. Lastly, although balancing measures are important components of any QI project to monitor potential unintended consequences, it should be noted that this was not intended to be a study powered to detect a difference in S aureus infections or Gram-positive sepsis-related mortality. Because these are rare events, a difference may not have been detected by this methodology.
Future directions of our initiative include using clinical decision support within the electronic health record to provide alerts after 48 hours and risk stratifying which patients should be started empirically on vancomycin.32
Through a sequential implementation approach of education, standardization of care with clinical pathways, pharmacist-initiated 48-hour time-outs, and prospective audit with feedback, vancomycin DOTs declined by >60% over a 1-year period. This QI initiative safely decreased unnecessary exposure to a potentially nephrotoxic drug that should be reserved for treatment of known antibiotic resistant infections.
We acknowledge the additional QI team members Marsha Conroy, PharmD, and Anna Espeland, PharmD, as well as Sofia Perazzo, Guillermo Miyashita, and Matthew Tsao for assistance with data collection and Drs Robin Steinhorn, Billie Lou Short, and Roberta DeBiasi for their executive support of this project.
Dr Hamdy conceptualized and designed the study, designed the data collection and management plan, coordinated and supervised data collection, analyzed the data, and drafted the initial manuscript; Mr Bhattarai contributed to the study design and data management plan, performed data interpretation, and drafted sections of and critically revised the manuscript; Dr Basu contributed to the study design and data management plan, supervised data interpretation, and drafted sections of and critically reviewed the manuscript; Dr Hahn contributed to the study design, performed data collection, and drafted portions of and critically reviewed the manuscript; Dr Galiote, Dr Stone, Ms Casto, and Ms Korzuch contributed to the study design, performed data collection, and critically reviewed the manuscript; Dr Sadler performed data collection and data interpretation, drafted portions of the manuscript, implemented interventions, and critically revised the full manuscript; Dr Hammer contributed to the study design, drafted portions of the manuscript, and critically reviewed the manuscript; Dr Slomkowski contributed to the study design, collected data, implemented interventions, and critically reviewed the manuscript; Ms Chase and Mr Chang performed data collection and data analysis and critically reviewed the manuscript; Dr Nzegwu performed data analysis and interpretation and critically revised the manuscript; Ms Greenberg, Ms Ortiz, and Dr Bost performed data analysis and critically reviewed the manuscript; Ms Blake contributed to the study design and conceptualization, performed data collection, and critically reviewed the manuscript; Dr Payne contributed to the study design, performed data interpretation, and critically revised the manuscript; Dr Shah contributed to the study design and critically reviewed the manuscript; Dr Soghier conceptualized and designed the study, designed the data collection and management plan, analyzed and interpreted the data, and critically reviewed the manuscript; and all authors approved the final manuscript as submitted.
FUNDING: No external funding.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.