Retrospective review of antimicrobial use for gastroschisis patients in Kigali, Rwanda: can improved stewardship reduce late inpatient deaths?

Gastroschisis mortality is 75–100% in low-resource settings. In Rwanda, late deaths are often due to sepsis. We aimed to understand the effect of antimicrobial use on survival. We conducted a retrospective review of gastroschisis patients at a tertiary hospital in Kigali, Rwanda between January 2016–June 2019. Demographics, antimicrobial use, microbiology, and outcomes were abstracted. Descriptive and univariate analyses were conducted to assess factors associated with improved survival. Among 92 gastroschisis patients, mortality was 77%(n = 71); 23%(n = 21) died within 48 h. 98%(n = 90) of patients received antibiotics on arrival. Positive blood cultures were obtained in 41%(n = 38). Patients spent 86%(SD = 20%) of their hospital stay on antibiotics and 38%(n = 35) received second-line agents. There was no difference in age at arrival, birth weight, gestational age, silo complications, or antimicrobial selection between survivors and non-survivors. Late death patients spent more total hospital days and post-abdominal closure days on antibiotics (p < 0.001) compared to survivors. There was no difference in the proportion of hospital stay on second-line antibiotics (p = 0.1). Conclusion: We identified frequent late deaths, prolonged antibiotic courses, and regular use of second-line antibiotic agents in this retrospective cohort of Rwandan gastroschisis patients. Future studies are needed to evaluate antimicrobial resistance in pediatric surgical patients in Rwanda. What is Known: • Global disparities in gastroschisis outcomes are extreme, with <4% mortality in high-income settings and 75-100% mortality in low-income settings. • Antimicrobial surveillance data is sparse across Africa, but existing evidence suggests high levels of resistance to first-line antibiotics in Rwanda. What is New: • In-hospital survival for gastroschisis was 23% from 2016-2019 and most deaths occurred late (>48hrs after admission) due to sepsis. • Rwandan gastroschisis patients received prolonged courses of antibiotics and second-line antibiotics were frequently used without culture data, raising concern for antimicrobial resistance. What is Known: • Global disparities in gastroschisis outcomes are extreme, with <4% mortality in high-income settings and 75-100% mortality in low-income settings. • Antimicrobial surveillance data is sparse across Africa, but existing evidence suggests high levels of resistance to first-line antibiotics in Rwanda. What is New: • In-hospital survival for gastroschisis was 23% from 2016-2019 and most deaths occurred late (>48hrs after admission) due to sepsis. • Rwandan gastroschisis patients received prolonged courses of antibiotics and second-line antibiotics were frequently used without culture data, raising concern for antimicrobial resistance.


Introduction
Global under-five mortality decreased by more than 50% between 1990 and 2015 as a result of efforts spurred by the Millennium Development Goals-a set of public health priorities and targets established by the United Nations [1]. Unfortunately, neonates-infants less than 28 days old-have not seen the same proportional decrease in mortality during the same timeframe. Neonatal deaths now account for nearly half (47%) of deaths among children under age five [2]. While infection, prematurity and birth asphyxia represent the leading causes of neonatal mortality, surgically treatable conditions contribute significantly [2]. In fact, congenital anomalies (which often require surgical management) are estimated to be the 5 th leading cause of under-five mortality, up from the 6 th leading cause in 2005 [3,4]. Furthermore, disparities in global surgical outcomes are extreme in the neonatal population.
Gastroschisis, a common congenital abdominal wall defect which occurs in one in 2000-3000 births live births, has a less than 4% mortality rate in high-income countries (HIC), but 75-100% mortality rate in low-and middle-income countries (LMICs) [5]. This disparity can be attributed to HICs' advances in antenatal ultrasound, prenatal diagnosis, durable intravenous access, total parenteral nutrition (TPN), and neonatal anesthesia and surgical care capacity, which remain largely unavailable in LMICs. In most HIC hospital systems, mothers undergo antenatal ultrasounds, which allows for gastroschisis to be identified prenatally. Infants with gastroschisis are then born at centers with neonatologists, pediatric surgeons and pediatric anesthesiologists. While exact protocols vary by institution, postnatal management in HICs typically includes covering the bowels with a preformed silo, reducing the bowels back into the abdominal cavity over several days and then either surgically closing the abdominal wall defect or using the sutureless closure technique [6,7]. Antibiotic therapy is often initiated at birth and continued only until the bowels are covered unless longer duration antibiotics are indicated due to maternal and neonatal risk factors. Without many of the human and material resources used for gastroschisis care in HICs, pediatric providers in LMICs are forced to adapt.
Previously published work concerning gastroschisis outcomes in Rwanda demonstrated an improvement in survival from 0 to 22% within the last five years [8]. Such a marked improvement in survival was closely linked to the establishment of a functional pediatric surgery service in 2017 and rapid development of capacity for pediatric critical care at The University Teaching Hospital of Kigali (Centre Hospitalier Universitaire de Kigali [CHUK]). In a retrospective analysis of all gastroschisis patients admitted to CHUK from 2016-2019, most neonates arrived within 24 h of birth and the majority (80%) survived the initial 48 h after admission. However, more than half of patients who survived the initial 48 h died prior to hospital discharge (late deaths) [8]. The precise etiologies for late deaths are not completely understood, but clinical experience among Rwandan experts suggest that sepsis and antimicrobial resistance (AMR) are the driving factors.
Globally, sepsis is known to be the most common cause of death in surgical neonates and this problem is exacerbated by AMR [9][10][11]. Across all age groups in sub-Saharan Africa, 23.7 per 100,000 deaths are attributable to AMR [12]. In Rwanda specifically, data available to guide antimicrobial stewardship is sparse, but has identified increasing rates of multidrug resistance among gram negative bacteria [13,14]. Gram negative bacteria are native to the gastrointestinal tract and are commonly encountered in cases of sepsis due to gastrointestinal pathologies, such as gastroschisis [13,[15][16][17]. Sepsis diagnosis and management, therefore, not only timely transfer and care at a tertiary care center, are critical to further improvement in gastroschisis outcomes. In this study, we seek to understand patterns of antimicrobial use for gastroschisis patients at a single tertiary hospital in Kigali, Rwanda. In doing so, we aim to identify modifiable factors to minimize late deaths in gastroschisis patients in Rwanda.

Setting and context
CHUK is a 520-bed public, tertiary care hospital in Rwanda's capital city. It is one of four referral hospitals in the country and the only hospital offering comprehensive pediatric surgical and anesthesia services for Rwanda's 12.2 million population [18]. In addition, CHUK has a 3-bed PICU with limited neonatal ventilation and a high-dependency unit with oxygen available. Due to CHUK's large catchment area, the hospital receives a significant number of outborn neonates (born outside of CHUK's facility), who are transferred in by ambulance for surgical evaluation. Patients transferred to CHUK with gastroschisis are received in the pediatric emergency room and co-managed by a team of pediatricians, pediatric surgeons, pediatrics residents, and general surgery residents. Most infants arrive with no coverage of the bowels. Cloth or gauze wrap and rarely plastic covering (foley bag) is applied prior to arrival. Cleansed and re-used preformed silos (PFS) are applied in the CHUK emergency room, and simple gastroschisis patients routinely undergo staged reduction and bedside closure [19]. Most gastroschisis patients are admitted to CHUK within one day of birth [8]. Those admitted with complex gastroschisis (atresia, ischemia, volvulus, or perforation) received palliative care in this sample.
In 2011, Rwanda's Ministry of Health published National Neonatal Protocols, which included empiric antibiotic guidelines for several clinical diagnoses [20]. The protocol recommends using ampicillin and gentamicin for first-line, empiric treatment of neonatal infection. Per the guidelines, a 7-14 day course is recommended. In our study, secondline antibiotics are those with broader coverage than empiric antimicrobial suggested in the National Neonatal Protocols (e.g., vancomycin, meropenem, etc.).

Study population and outcomes
We completed a single-center, retrospective review of all gastroschisis patients admitted to the CHUK between January 2016 and June 2019 [8]. Reporting of this study was done in accordance with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) checklist [21]. Demographic data included birth location, sex, birth weight, gestational age, and age at admission. The outcomes of interest were 1) survival to hospital discharge, 2) associations between blood culture results and survival outcomes, and 3) comparison of patterns of antimicrobial usage between survivors and non-survivors, including specific antimicrobial regimens, proportion of hospital stay on antibiotics, and proportion of post-abdominal closure stay on antibiotics.

Data collection
Patients were identified from hospital registries and their medical charts were obtained from hospital archives. Data was extracted and entered into a secure REDCap database. Abstracted data included demographics, referral history, admission interventions, daily antimicrobial usage, operative details, in-hospital complications, length of stay, and mortality. All data were de-identified and stored on an encrypted University of Florida server. Further methodological details are available in previously published work [8]. This project was approved by the CHUK Ethics Committee and University of Florida Institutional Review Board.

Data analysis
Descriptive and univariate analyses were conducted with primary outcome being survival to hospital discharge. Fisher's exact test was used to compare groups on categorical outcomes. All analyses were performed using R statistical software package (V.4.0.2, The R Foundation for Statistical Computing).

Results
Ninety two patients presented with gastroschisis during our study period. The characteristics of patients are described in Table 1. Sex was evenly distributed with 47 (51.1%) females and 45 (48.9%) males. Mean gestational age was 36.0 weeks (standard deviation (SD) = 2.2). Mean age at arrival was 0.6 days (SD = 0.75), and mean length of stay was 16.1 days (SD = 15.8) overall, 28.3 days (SD = 10.4) for survivors.
Twenty one patients (22.8%) survived to hospital discharge. Among the 71 (77.2%) in-hospital deaths, eight patients (8.7%) died within 24 h of admission, 13 (14.1%) died between 24-48 h, and 48 (52.2%) died more than 48 h after admission. There was no significant difference in age at arrival, birth weight, or gestational age between survivors and non-survivors. Likewise, silo dislodgment prior to final closure was not associated with an increased risk of cultureproven sepsis or in-hospital mortality (p = 0.6). At the time of arrival, 90 (97.8%) patients were treated for sepsis and started on antibiotics. Positive blood cultures were obtained in 38 (41.3%) patients. Patterns of antimicrobial use and regimens are depicted in Table 2. Blood-culture confirmed pathogens and associated outcomes are shown in Table 3.
The most common antibiotic regimen at admission was penicillin and gentamicin (n = 50, 54.3%), followed by penicillin and cephalosporins (n = 23, 25.0%). Patients spent on average 86.2% (SD = 20.1%) of the hospital stay on antimicrobial therapy. There was no difference in antimicrobial usage between survivors and non-survivors. In the subset In comparing late death patients to those who survived to hospital discharge, late death patients spent a greater proportion of hospital stay (90.0% v. 75.0%, p < 0.001) and postabdominal closure stay (79.5% v. 66.9%, p = 0.03) on antibiotics. There was no difference in the proportion of hospital stay on extended spectrum antibiotics (16.5% v. 27.1%, p = 0.1).

Discussion
We identified frequent late deaths, prolonged antibiotic courses, and regular use of second-line antibiotic agents in this retrospective cohort of Rwandan gastroschisis patients. Although we did not find a survival difference related to antimicrobial selection in this study, our findings and the team's clinical experience highlight the critical need for prospective evaluation of blood culture data and antibiotic sensitivity results to maintain an up to date antibiogram and control AMR in Rwanda.
Initial antibiotic selection in our study was in alignment with the Rwanda Neonatal Protocol, which advocates for using ampicillin and gentamicin for first-line, empiric treatment of neonatal sepsis [20]. The duration of antibiotic therapy, however, is a major concern. Per the guidelines, a 7-14 day course of antibiotics is recommended for treatment of sepsis [20]. The average duration of antimicrobial therapy among survivors in our study population was three weeks. It is important to note that survivors did spend an average of one week off antibiotics prior to hospital discharge, suggesting a satisfactory clinical response. Nevertheless, the overall duration of antimicrobial therapy was considerably longer in our cohort than recommended by the Rwandan National Neonatal Protocols. Additionally, over one-third of our study population received second-line antibiotics, such as vancomycin and meropenem, which were presumably prescribed due to concern for AMR. The true picture of AMR, however, remains unclear due to limited antibiotic sensitivity testing.
AMR surveillance data are sparse across most of Africa, but mounting evidence points to widespread antibacterial resistance among several common culprits in surgical patients, including Staphylococcus aureus, Escherichia coli, and Klebsiella pneumonia [22]. Two systematic reviews on AMR in Africa have identified high levels of resistance to first-line antibiotics and large gaps in diagnostic capacity. In fact, AMR data is missing in 40% of African countries as of 2017 [23,24]. Inadequate AMR surveillance presents a particularly sizeable problem for surgical neonates-a population which dies more frequently of sepsis than any other cause [9,10]. At CHUK, blood cultures are commonplace, but antibiotic sensitivity testing was not routinely available at the time of our study. A recent study on antimicrobial resistance patterns in Rwandan neonatal units found zero percent sensitivity to ampicillin and only 13% sensitivity to gentamicin among 128 positive blood cultures at CHUK [14]. Studies concerning AMR patterns for adult patients at CHUK have identified gram negative bacteria with significant resistance to third generation cephalosporins, and even rising resistance to imipenem [13,25]. In our study population, there was no difference in the type of pathogen identified on blood culture between survivors and Other includes cephalosporin + metronidazole, cephalosporin + vancomycin, penicillin alone, penicillin + aminoglycoside + carbapenem, penicillin + aminoglycoside + metronidazole, penicillin + aminoglycoside + vancomycin non-survivors, except for reduced mortality among patients with staphylococcus-positive blood cultures, likely representing contaminated samples. It is also possible that the lack of statistical difference in blood culture results between survivors and non-survivors is due to small sample size and inadequate power to detect a meaningful difference. Nevertheless, resistance to nationally recommended empirical antibiotic regimens is a major impediment to combatting neonatal infection in Rwanda. Without reliable culture and antibiotic sensitivity data, physicians must rely on clinical suspicion for antimicrobial selection, which delays appropriate treatments, worsens antibiotic resistance to first-line antibiotics, and drives up costs for the healthcare system [12,26]. These issues will only be amplified as neonatal critical care capacity grows in Rwanda. At the patient level, issues with antibiotic use extend beyond AMR. Inappropriate or excessive use of antibiotics are increasingly linked to life-threatening adverse events in neonates, including late-onset sepsis, invasive candidiasis, and death [26][27][28][29]. In our study, survivors received prolonged courses of antibiotics lasting much longer than high-income setting guidelines recommend for simple gastroschisis patients (without evidence of infection) [30][31][32]. Prophylactic antifungal medications are also not routinely provided in our setting, as the only available antifungals are oral medications, which are not suitable for patients who are unable to tolerate enteral feeds. Furthermore, prolonged use of certain second-line agents may be particularly harmful in low-resource settings, where serum concentration monitoring is not feasible. Without the laboratory capacity to measure serum peaks and troughs for antibiotics like vancomycin, we are unable to monitor for nephrotoxic and ototoxic serum drug concentrations [33,34]. Therefore, following guidelines established for high-income settings is challenging in LMICs due to differences in patient environment (open wards), resources, and AMR patterns.
Improving gastroschisis outcomes at CHUK demands much more than timely arrival to a tertiary care center, careful antibiotic management, and better AMR surveillance. Gastroschisis is a complex congenital condition which requires comprehensive neonatal care. There are many infrastructure, personnel, and material resource limitations to overcome in our setting [35]. At CHUK, most neonates treated for surgical diseases are outborn and unable to be admitted to the formal neonatology unit due to infection risk. This lack of centralized neonatal critical care services introduces challenges to specialized nurse training and access to neonatal-specific equipment (radiant warmers, isolettes, small gauge IVs, etc.). Since the time of our study, surgical neonates have been cohorted in a neonatal surgical unit-an evidence-based intervention that has improved surgical outcomes and resource utilization in other LMIC institutions [36,37]. Further improvements in gastroschisis outcomes, and, more broadly, neonatal surgical outcomes, will require careful investigation to optimize multidisciplinary collaboration and daily management (infection control, durable intravenous access, nutrition). This is the focus of an ongoing prospective neonatal surgical outcomes registry, through which we aim to identify modifiable risk factors for in-hospital morbidity and mortality at CHUK.
Our study has certain limitations. First, the study was conducted retrospectively and over a timeframe when neonatal surgical care was evolving at CHUK. While data is presented collectively, there were likely differences in care protocols and outcomes over this timeframe that we were unable to capture. Next, we are unable to screen for congenital heart disease or other congenital abnormalities which may be confounders in our survival outcomes. Additionally, necrotizing enterocolitis (NEC), a potential source of sepsis known to occur in gastroschisis patients, is also difficult to detect in our setting due to lack of bedside x-ray. Finally, antibiotic selection was affected only by blood culture results; there were no urine or cerebrospinal fluid samples sent for culture. Therefore, in cases of culture-negative sepsis, antimicrobial escalation/ de-escalation and discontinuation were dependent on clinical findings alone. Antimicrobial sensitivity testing was likewise not routinely available during our study timeframe. These data will be a critical component of our ongoing neonatal surgical registry and quality improvement initiative at CHUK.
In conclusion, gastroschisis outcomes are improving in Rwanda, but the high proportion of late deaths, prolonged antibiotic courses, and frequent use of second-line agents raise concern for AMR in our population. As overall neonatal care continues to improve at CHUK, AMR surveillance will become increasingly important. Our findings highlight the critical need for routine culture data and antibiotic sensitivity testing at CHUK to maintain an up-to-date hospitalspecific antibiogram, minimize delays in prescribing appropriately targeted antimicrobials, and restrict AMR.
Funding This manuscript was supported by the National Cancer Institute, Fogarty International Center and National Institute of Mental Health of the National Institutes of Health (award numbers T32CA090217 and D43TW010543). The funders had no role in study design or writing of this report.