Impact of a Neonatal Early-onset Sepsis Risk Calculator on Antibiotic Use in a Regional Australian Hospital

1. INTRODUCTION

Early onset sepsis (EOS) is a potentially fatal condition [1, 2]. However, it is also rare, affecting between 0.3 and 0.8 neonates per 1000 births in developed countries [1-4]. EOS remains a diagnostic challenge [5-7]. Despite evidence against the use of non-specific infection biomarkers in neonates, worldwide many neonatal facilities, including tertiary NICUs, currently continue to use these to investigate and guide management of neonatal EOS [6-11]. Blood cultures, while offering excellent sensitivity and specificity when obtained correctly, yield results within 24-48 hours, necessitating empiric antibiotic use while awaiting definitive diagnosis [6, 10, 12]. The number of neonates receiving parenteral antibiotics greatly exceeds those with culture-positive sepsis [1, 2, 5]. Antibiotic use is associated with infant-parent separation, delayed initiation of breastfeeding and parental anxiety [1, 13, 14]. Early antibiotic use and its impact on infant microbiome is the topic of current research and has potential associations with asthma, obesity, and type 1 diabetes [15-18]. Reducing unnecessary antibiotic use in neonates is therefore a vital initiative.

The Neonatal EOS Calculator is an evidence-based algorithm which provides risk estimates based upon objective clinical data which are readily available in most circumstances [1, 2, 19]. Multiple international studies have reported safe reductions in antibiotic use between 42-59% when using the EOS Calculator, with the Calculator being endorsed by the American Academy of Paediatrics [2, 5, 14, 20-22]. Two recent meta-analyses found the Calculator was associated with significant reductions in antibiotic use with no evidence of inferiority regarding safety when compared to conventional approaches. A third meta-analysis raised safety concerns regarding missed or delayed treatment, especially in the setting of maternal chorioamnionitis [2, 20, 22].

Studies in the Australian context are limited and there is no regional data to date regarding the safety and efficacy of the neonatal EOS Calculator [1, 20, 23]. Variations in antenatal Group-B Streptococcus (GBS) screening protocols and intrapartum antibiotic guidelines limit the applicability of international studies in the Australian context, which necessitates the need for local evidence. A single prospective study in an Australian tertiary centre yielded similar results to those observed in American and European research [1]. Although similar studies are being undertaken at large metropolitan tertiary centres in Australia, we are unaware of any current research in regional Australia.

Grampians Health is a health service that encompasses a number of health services and encompasses the main campus (Grampians Health Ballarat) servicing the Ballarat region and numerous smaller health services across the Grampians region. The paediatric service at the Grampians Health Ballarat supports neonates born at ≥32 weeks gestation and has a Special Care Nursery of 12 bed capacity. During the study period, there were 2822 births at the Grampians Health Ballarat and GBS screening is standard antenatal care. The evaluation and management of early onset sepsis was derived from a non-standardised assessment of risk factors (generally requiring 2 or more risk factors, such as prematurity, CTG abnormalities, maternal temperature, GBS status, antenatal antibiotics, prolonged rupture of membranes) along with clinical condition of the baby and screening blood tests. Rigid guidelines were not routinely used during this period, as was the case throughout most equivalent institutions the investigators are aware of. The purpose of this study was to establish whether using the Neonatal EOS Calculator in this health service would lead to a reduction in neonatal antibiotic use whilst maintaining patient safety. Adding to the limited body of research in the application of the Calculator in Australia, specifically, the regional context, it was anticipated that our findings will support a prospective trial of the Calculator including local and wider policy changes in the management of neonatal EOS.

As a secondary outcome measure, was to establish the relationship, if any, of commonly used biochemical tests used in neonates with suspected EOS to determine their utility in management decisions. The sensitivity and specificity of biochemical tests for EOS, in particular C-reactive protein (CRP), varies widely between studies and data suggest they are of limited value in this context [6, 7, 11]. We hypothesise that our findings will align with previous data and support the health service to reduce its reliance upon biochemical markers as screening tools, while encouraging the transition to the EOS calculator as an objective assessment tool.

2. METHODS

3. RESULTS

A total of 2822 births were identified in the study period and 402 birth admissions were assigned a diagnostic code relevant to sepsis or parenteral antibiotic use. After exclusion criteria were applied, 296 were included in the final analysis. Demographic and clinical features of the cohort are outlined in Table 2. Of these, 292 received parenteral antibiotics (Fig. 1). Three neonates were investigated for presumed EOS, however, did not receive antibiotics. Of these, two did not have blood cultures obtained. Lastly, EOS was suspected in one neonate; however, investigations were not initiated due to parental choice. No neonates were excluded based on insufficient available data. Maternal temperature was not recorded for 21 mothers and a temperature of 37.0°C was used in Calculator analysis as in previous studies [23].

Blood culture volumes were not recorded; however, in keeping with available evidence, 0.5-1.0ml for all neonatal blood cultures is practiced at the institution. Five blood cultures returned positive results, all of which were deemed to be contaminants (Table 3). One blood culture result was unavailable. No positive blood cultures with pathogenic organisms were recorded. In cases where Cerebral Spinal Fluid (CSF) was obtained, 5 showed elevated CSF white cell count, but there were no positive CSF cultures. One case of viral meningitis was identified (Parechovirus). A total of 12.2% (36) of neonates required to transfer to tertiary centres, and there were no recorded deaths. Return correspondence was unavailable for 4 neonates transferred to tertiary centres. These neonates had an unknown total duration of antibiotics; however, no other critical data were missing from their local records. Maternal temperatures of ≥38.0°C were recorded for 11 cases, only one of which met the clinical criteria for chorioamnionitis.

4. DISCUSSION

The findings of the retrospective review support the EOS calculator as an effective measure to reduce empiric antibiotic use in the neonatal population within the regional health service. These findings, coupled with previous studies, justify a prospective study to be conducted within the health facility’s neonatal unit to examine the safety and efficacy of the Calculator. However, due to an absence of blood cultures and proven EOS cases, we cannot comment on the safety of the calculator in this study.

It is unfortunate, though not unexpected, that no positive blood cultures with pathogenic organisms were recorded. During the study period, 2822 births were recorded at the health service. The absence of positive blood cultures may reflect a local incidence of 0.3 per 1000 live births, a figure that is consistent with previous Australian studies. Blood culture volumes were not recorded, although staff are instructed to collect 0.5-1mL in keeping with current evidence.

The absence of local incidence data presents a challenge when investigating a rare condition, a difficulty amplified in a regional setting with smaller birthing numbers. Added to this, high-risk deliveries are transferred to larger tertiary facilities for birthing, while unwell neonates transferred to larger tertiary facilities may reduce the overall number of blood cultures obtained. We postulate that considering the birth numbers, retrospective data from the previous 10 years will be required to establish the local incidence of EOS. Though sought, this data was unavailable at the time of publication. Determining the incidence figure to use presents an ongoing difficulty for regional and rural health services that use the Neonatal EOS Calculator. Future research could address this further.

The findings related to the secondary outcome support the hypothesis that biochemical markers (initial and peak CRP, WCC and neutrophil count) correlate poorly with risk estimates from the EOS calculator. In this case, they are of limited utility in the decision to commence antibiotics. We conclude that CRP is not useful as a screening test in infants with low-risk EOS Calculations. A high CRP may influence a clinician to commence IV antibiotics, despite EOS calculator recommendations and an examination indicating a well-child. This would negatively impact the efforts to reduce antibiotic use.

While elevated initial CRP does positively correlate with a high-risk EOS Calculation, these subjects would receive empiric antibiotics regardless of CRP value should the EOS Calculator be introduced. Furthermore, with the introduction of the EOS Calculator, greater numbers of neonates at high risk of EOS may be identified and treated early, avoiding potentially missed cases when CRP has not been obtained. It will also omit treatment delays while awaiting CRP results or false negatives from non-elevated initial CRP. These findings support the decision to move away from using biochemical markers as decision aids in screening for EOS and utilising serial examination to determine clinical status.

Given the ubiquity of the use of CRP and WCC in evaluating infants with suspected sepsis at our facility, this may represent a cultural challenge during the implementation phase, thus careful education of clinicians will be essential. The authors recognise that many neonatal clinicians continue to utilise CRP in other clinical scenarios, for example, to guide the duration of treatment in infants commenced on IV antibiotics. The utility of CRP in this setting was not evaluated during this study and we cannot comment on this. Further studies may be warranted.

New methods to reduce antibiotic exposure among neonates are continuing to evolve – serial physical examination is emerging as a feasible and safe approach [24]. In considering future directions, it is important to consider the context of practice. At Grampians Health Ballarat, as in many regional Australian health services, the paediatric junior medical staff are rotating staff from other health services who have varied paediatric exposure and typically have limited neonatal experience. In these settings, it is valuable to have an objective assessment tool that is clearly understood by senior and junior staff alike, as afforded by the neonatal EOS Calculator.

CONCLUSION

The EOS Calculator may effectively reduce empiric antibiotic use in our regional health care facility. Future directions, including a prospective trial require careful consideration of safety and clinician discretion based upon the clinical scenario. The secondary outcome findings support reducing reliance on biochemical markers such as CRP and WCC as screening tools in neonates suspected of EOS, and instead utilising the EOS Calculator and serial clinical evaluation.

REFERENCES