Remote Stewardship for Medically Underserved Nurseries: A Stepped-Wedge, Cluster Randomized Study Free

We evaluated the efficacy of a level 1 nursery-specific remote ASP using a stepped-wedge cluster randomized design (Fig 1) over a 36-month period (July 2017 to June 2020).²¹  This included a 15-month baseline period in which no nursery was receiving the intervention; a 9-month “step-in” period during which time the nurseries began receiving the remote ASP in a randomly generated order, and a 12-month postintervention period during which time all nurseries were receiving the intervention. We hypothesized that our nursery ASP would reduce both the proportion of infants exposed to antibiotics and the total volume of antibiotic use by 25% from preintervention rates. This study was approved as nonregulated research (program evaluation) by the University of Texas Health San Antonio Institutional Review Board.

Eight level 1 nurseries participated in the study. These nurseries were selected from a convenience sample of nurseries in our region, many of which refer infants requiring a higher level of neonatal care to the primary investigator’s center. There was no cost to participating centers. Each nursery was in a Health Resources & Services Administration medically underserved area with a rural designation; 3 were designated critical access hospitals.²²  The median distance from each center to San Antonio was 113 miles (range, 61–173 miles). These centers deliver a median of 380 infants annually (range, 120–710) and are staffed by a median of 2 physicians (range, 1–4; 11 family medicine providers, and 5 pediatricians). Characteristics of the participating centers are shown in Table 1. All liveborn infants ≥35 0/7 weeks gestation were included; infants who required transfer to a higher level of care at the time of delivery (eg, ≤34 weeks gestation, severe respiratory disease, or congenital malformation) were excluded.

The primary intervention was delivery of a nursery-specific remote ASP. The program included 24/7 provider-to-provider phone consultation with a neonatal infectious disease specialist (J.B.C.); on-site or virtual continuing medical and nursing education lectures, including such topics as early-onset sepsis, the neonatal sepsis calculator, chorioamnionitis, congenital syphilis, neonatal antibiotic pharmacodynamics and pharmacokinetics, and the risks and benefits of antibiotic exposure in newborns; and prospective audit and feedback of prescribing practices. The provider-to-provider consultations were available at any time and to any provider, including physicians, nurses, infection preventionists, quality officers, pharmacists, etc. Consultations could be by phone, video call, text, or e-mail, although phone calls were encouraged. Continuing education lectures were delivered on-site whenever possible and were also recorded to allow asynchronous viewing at each site. The neonatal sepsis calculator was discussed with each site and reinforced as a useful stewardship tool. Audit and feedback were based on the study by Gerber et al.²³  “Report cards” were tailored to each center and included their 6-month rolling average of antibiotic use (proportion exposed and volume of antibiotic use), the center’s degree of improvement from baseline level, and how their center was performing relative to other deidentified participating centers.

The 2 primary outcomes were the proportion of infants exposed to any antibiotic and the volume of antibiotic use, measured as days of therapy (DOT) per 1000 patient-days.²⁴  Secondary outcomes included several balancing measures: infant length of stay, need for infant transfer to a higher level of care, early-onset sepsis, and death. Participating centers collected a 1-page data abstraction sheet for each infant, including gestational age, birth weight, mode of delivery, maternal group B streptococcal colonization and intrapartum antibiotic prophylaxis, infant antibiotic dosing, duration and indication, any positive culture results, and length of stay. These sheets did not contain any protected health information and were transmitted in batches via e-mail or fax at the end of each month.

Based on an estimated antibiotic exposure prevalence of 6%, an α of 0.025, and no infants lost to follow up, an estimated 8404 infants would provide 80% power to detect a 25% reduction in antibiotic exposure. Infants were designated as preintervention or postintervention depending on whether the ASP was in effect in their nursery at the time of their birth. Patient- and center-specific variables were compared in the pre and postintervention periods using χ² or Wilcoxon rank testing as appropriate. For the primary analysis of antibiotic exposure, a multivariate mixed-effects logistic regression model was used to investigate changes at the infant level after adjustment for the study center. For the primary analysis of antibiotic consumption (DOT per 1000 patient-days), a similar model was used to investigate changes at the center level, with intervention period forced in to the second model a priori. Bonferroni correction was used for the 2 primary outcomes; therefore, a 2-tailed P value of .025 was considered statistically significant for each of the 2 primary outcomes. All calculations were performed using Stata v16.1 (Stata Corporation, College Station, TX).

Nine thousand, two hundred and seventy-seven infants were born during the study period (4586 preintervention, 4691 postintervention). There were no differences among demographics and clinical features between the infants born pre and postintervention (Table 2). The proportion of infants exposed to antibiotics declined from 6.2% pre-ASP to 4.2% post-ASP (absolute risk reduction 2%, relative risk 0.68 [95% confidence interval (CI), 0.63% to 0.75%]; Table 3). Similarly, total antibiotic consumption declined from 117 to 84.1 DOT per 1000 patient-days (−28% [95% CI, −22% to −34%], Fig 2). No safety signals were observed; the mean length of stay was unchanged (1.84 days preintervention, and 1.83 days postintervention) as was the frequency of infant transport to a higher level of care (12, 0.26% preintervention; and 11, 0.23% postintervention). Three infants had early-onset sepsis, 2 in the preintervention period (0.44 per 1000 live births, 1 group B Streptococcus, and 1 Escherichia coli) and 1 in the post-intervention period (0.21 per 1000 live births, 1 group B Streptococcus). There were no infant deaths in the pre or postintervention periods.

Combined continuing medical and nursing education presentations were attended by an average of 8 participants (interquartile range [IQR] 5–10, range 2–12). Nurses accounted for 53% of attendance, followed by physicians (41%), advanced practice providers (4%), and pharmacists (2%). There were 451 provider-to-provider telephone consultations during the study period, for an average of 3.3 calls per participating center per month. Physicians accounted for the majority of calls (427, 95%) followed by nurses (16, 4%), and pharmacists (6, 1%). Indications for consultation are shown in Table 4. Provider-to-provider consultations addressed 38 distinct topics; congenital syphilis was the most common question (149 calls, 33%) followed by suspected early-onset sepsis (81 calls, 18%), and despite all coronavirus-related questions occurring only in the final 4 months of the study, SARS-CoV-2 questions were the third most common (42 calls, 9%). The median response time was <1 minute (IQR, 0–3 minutes, range, 0–30 minutes). The median length of the call was 4 minutes (IQR, 2–6 minutes). Eighty eight percent of the phone consultations occurred during work hours (ie, 8:00 am to 5:00 pm, Monday through Friday). Three hundred and twenty (71%) of the phone consultations resulted in a management change, including 6 infants and 2 mothers transferred to a higher level of care; and 11 unnecessary infant transfers prevented. For example, on 1 occasion a center anticipated transferring an infant for congenital syphilis evaluation including lumbar puncture, but provider-to-provider consultation determined that a lumbar puncture was not indicated due to adequate maternal treatment, allowing the infant to remain in couplet care with the mother.

A nursery-specific remote antimicrobial stewardship program reduced the proportion of infants exposed to any antibiotic by 32% and the overall volume of antibiotic use by 28% in 8 medically underserved, rural level 1 nurseries. Encouragingly, no adverse safety signals were seen despite the relatively large cohort size; however, the study was not powered for individual safety signals. Specifically, no increase in length of stay, early-onset sepsis, or unplanned transfers to a higher level of neonatal care were identified after the ASP implementation. Importantly, the study also established that the data collection process used for this remote ASP is simple and feasible for personnel at participating sites. Finally, the on-call obligation for the infectious disease specialist was manageable; the majority of the calls took <5 minutes and came during normal business hours. This is presumably because level 1 nurseries are not routinely staffed by physicians overnight. In summary, this study provides preliminary data in support of the efficacy and feasibility of remote stewardship to minimize unnecessary or prolonged antibiotic use for infants born in rural or medically underserved areas that do not have direct access to pediatric infectious diseases consultation or pediatric-specific ASPs. Remote strategies may be particularly beneficial in larger, low-population density states like Texas, which averages 30% fewer primary care physicians per capita than the United States average.²⁵  Of 254 Texas counties, 211 (83%) are designated by the Health Resources and Services Administration as medically underserved, and 33 counties (13%) have no primary care physicians at all.

Our experience adds to the growing literature in support of telestewardship programs for low-resource, rural, or medically underserved settings. Telehealth ASP has generally been successful in adult acute care settings. Vento et al²⁶  demonstrated the effectiveness of a combined on-site and telehealth-supported ASP, including telehealth consultation, for rural hospitals in Utah. Similarly, Klatt et al²⁷  used a telehealth ASP to address antibiotic utilization for urinary tract infections, skin and soft tissue infections, and pneumonia at a community hospital in Wisconsin that lacked on-site ASP support. Similar programs in Australia and Pennsylvania have also been successful.^28,29  However, pediatric data are still extremely limited, including a paucity of data supporting telehealth ASP in the nursery setting. There is high-quality evidence for telehealth strategies for other neonatal conditions.³⁰  For example, the use of telemedicine for diagnosis of neonatal congenital heart disease can reduce the time to diagnosis and treatment while also minimizing unnecessary transports for infants without significant malformations.³¹  Telemedicine for infants with cleft lip or palate is extremely cost-effective and can minimize time to surgery and miles traveled for care coordination.³²  Telemedicine has also been used to alleviate the effect of a growing shortage of pediatric ophthalmologists on retinopathy of prematurity screening and treatment.³³  The use of telemedicine to reduce costs, minimize care delays, prevent unnecessary transports or travel, and mitigate workforce issues can all be applied to neonatal infectious diseases and antimicrobial stewardship efforts as well.³⁴  Further investigation into pediatric and neonatal telehealth ASPs may help close a significant knowledge gap and reduce health disparities between urban and rural centers.

The postintervention period ended on June 30th, 2020; the last 4 to 5 months of the study occurred during the opening stages of the SARS-CoV-2 pandemic. The effect on the remote ASP itself was minimal, primarily because most activities including audit and feedback, provider-to-provider consultations, and education were delivered via distance-based strategies from the beginning. The 2 most notable effects of the pandemic were (1) that the minority of centers that preferred in-person continuing medical and nursing education had to transition to a virtual meeting model, and (2) there was a sharp increase in the volume and proportion of phone consultations focused on SARS-CoV-2 issues, particularly infection control and prevention strategies such as infant separation and breastfeeding. A strength of the remote ASP was its ability to rapidly deliver needed information on new or emerging clinical issues without a meaningful change in the structure of the program.

Limitations of this study include those inherent to the stepped-wedge design.²¹  A major consideration with stepped-wedge designs is that because all centers start without the intervention and finish with the intervention over a given time frame, temporal trends that are independent of the intervention can be significant confounders. There was no evidence of a decrease in antibiotic use during the baseline period, so it is probable that the reduction in antibiotic utilization was due to the intervention, but it is impossible to exclude extrinsic confounders in this model. Secondly, there was no wash-out period between preintervention and postintervention periods, which may have resulted in a slight underestimate of the efficacy of the remote ASP during the initial learning period after implementation. However, use of provider-to-provider consultation and decreases in antimicrobial utilization were seen rapidly at every center, suggesting that any learning curve was relatively short. A parallel cluster design was considered in which some centers would have received the remote ASP and others would not. We rejected the parallel cluster design for 2 reasons: (1) the centers were heterogeneous in size, location, and staffing, so we felt that each center would be best served as its own control rather than compared with different centers; and (2) centers in the study may have been less enthusiastic to participate in the data collection if they were not guaranteed of receiving the intervention. Finally, the study was designed to measure reduction in antimicrobial use rather than care quality. There were instances where the optimal care plan was longer therapy. For example, if after provider-to-provider consultation it was determined that an infant required 10 days of penicillin therapy rather than a 1-time intramuscular dose of penicillin, this change represented improved care but worked against the primary outcome of antibiotic use. Future research should include more holistic measures of value as opposed to only reduction in antibiotic use. Finally, as no protected health information was transmitted, it was impossible to distinguish individual treatment decisions as appropriate or inappropriate in some cases.

The relatively small sample size of the study means that only limited generalization assumptions should be made, particularly as the participating nurseries had a preexisting referral relationship with the study center. Larger studies that include more nurseries and a wider variety of rural or medically underserved settings may be needed to determine the scalability of remote nursery ASP. Scalability also will require training additional neonatal infectious diseases content experts to effectively navigate teleconsultations. Furthermore, it is unclear which specific implementations drove improvements in care, as the study was not designed to test individual components of the ASP. We hypothesize that case-by-case or systematic implementation of the neonatal sepsis calculator drove most of the reduction in antibiotic use, whereas provider-to-provider consultation drove most of the improvements in infant transfers. Systematic reduction in antibiotic use may benefit more infants, but individual mother and infant dyads whose care can be improved by provider-to-provider support may have the greatest cost and value benefits, particularly if unnecessary transfers can be prevented. Additional study using dissemination and implementation methodology is needed to tease out the additive value of specific components.

If remote ASP does prove to be a viable strategy for rural or medically underserved nurseries, the impact could be significant. A 2% absolute reduction in antibiotic exposure is a small decrease; however, if applied to the approximately 1.7 million births in rural or medically underserved areas in 2020, this reduction would equate to over 30 000 less antibiotic courses, which in turn could reduce the risk of obesity, asthma, eczema, diabetes, metabolic syndrome, multidrug resistant organisms, and other adverse consequences of early antibiotic exposure. Remote ASP could also increase the quality of care by minimizing unnecessary transfers and reducing the time to diagnosis and treatment of infection.

The number of infants exposed to antibiotics and total antibiotic use declined in medically underserved nurseries after implementing a remote ASP. No adverse safety events were seen, and the remote ASP time demands were manageable. This study provides preliminary data in support of remote ASP as an effective strategy for optimizing antibiotic use in medically underserved newborn nurseries.

ASP

antibiotic stewardship program

DOT

days of therapy