Quality improvement report

Kaiser Permanente early-onset sepsis calculator as a safe tool for reducing antibiotic use among chorioamnionitis-exposed term neonates: Qatar experience

Abstract

Being an important cause of early-onset neonatal sepsis, clinical chorioamnionitis in the mother results in frequent laboratory workup and antibiotic use for the neonate. Neonatal intensive care units (NICUs) in Qatar follow the categorical approach recommended by the Centers for Disease Control and Prevention, USA, and all chorioamnionitis-exposed neonates receive antibiotics.

Our project aimed to reduce antibiotic use among chorioamnionitis-exposed, asymptomatic term babies by adopting the early-onset sepsis calculator (EOSCAL). Reduction of blood culture and NICU stay duration were added as secondary objectives later.

The Institute of Healthcare Improvement Model of Improvement was used. Antibiotic use rate was the primary outcome measure. Blood culture rate and early transfer to the postnatal ward were added after 1 year. The process measures included the EOSCAL use rate and calculation error rate. The rate of positive culture among untreated babies within the first week was taken as a balancing measure. Monthly data were collected from February 2020 and entered as run charts. Calculation errors were dealt by multiple PDSAs. Additional outcome measures were added in January 2021. Data collection and monitoring continued till December 2022.

Among 3837 inborn NICU admissions, 464 (12 %) were chorioamnionitis-exposed babies. Of them, 341 (74%) cases were eligible for inclusion. Among eligible cases, 270 (79%) did not receive antibiotics. Blood culture could be avoided among 106 (97% of low-risk babies) and NICU stay was reduced among 45 (92% of eligible low-risk babies). None of the untreated babies developed sepsis during the first week.

Implementation of this project effectively and safely reduced the antibiotic use and blood culture rate among term, well-appearing babies exposed to chorioamnionitis. The project resulted in enhanced patient safety, experience and flow and reduced cost. It is recommendable to other NICU settings in Qatar.

What is already known on this topic

  • The early-onset sepsis calculator (EOSCAL), recommended by researchers from the Kaiser Permanente Hospitals, USA, is a widely accepted tool for risk stratification and management of neonates at risk for early-onset sepsis. However, some researchers have questioned the safety of the tool, especially when applied to babies exposed to chorioamnionitis.

What this study adds

  • When symptomatic and preterm babies are excluded, EOSCAL is a safe and effective tool for reducing antibiotic use among babies exposed to chorioamnionitis.

How this study might affect research, practice or policy

  • Incorporation of EOSCAL to routine clinical care of asymptomatic term neonates exposed to chorioamnionitis will result in reduced antibiotic use, cost reduction and improved patient flow.

Problem

Chorioamnionitis (CA) refers to the inflammation, often of infective origin, that affects the placenta, fetal membranes and amniotic fluid and complicates up to 3.9% of deliveries.1 CA is a significant risk factor for early-onset neonatal sepsis.2 3

The Centers for Disease Control and Prevention (CDC) and the Committee on Fetus and Newborns (COFN) recommended empiric antibiotic treatment for all infants born to mothers with CA.4

Al Wakra Hospital (AWH) is a level-2 hospital under the Hamad Medical Corporation, Qatar, catering to 5000–6000 deliveries every year, and our neonatal intensive care unit (NICU) admits 1100–1300 neonates annually.

An audit conducted in our NICU in 2019 showed that the rate of suspected CA among mothers delivering at AWH increased from 1% in 2016 to 2.8% in 2019. In parallel, the proportion of CA-exposed neonates admitted to our NICU increased from 9.3% of admissions in 2016 to 11.5% in 2019. Based on the CDC/COFN recommendation, all these babies received intravenous antibiotics for a minimum of 48 hours, although up to 75% were well-appearing term babies. Among all term neonates admitted to our NICU during 2019, 42% received first-line antibiotics.

Our team reviewed the background data and explored the current evidence for managing neonates exposed to CA. The updated American Academy of Pediatrics (AAP) guideline5 recommended the use of the early-onset sepsis calculator (EOSCAL) as a reliable and safe option for risk stratification and management of neonates at risk for EOS.

We decided to use the EOSCAL to reduce antibiotic use among well-appearing term neonates exposed to suspected maternal CA in our unit.

The starting aim of our project was to reduce antibiotic use among term (37 weeks or more) well-appearing neonates exposed to CA to <50% by 31 July 2020. We adopted the Institute of Healthcare Improvement (IHI) model of improvement for this project. The IHI model has been extensively used by different healthcare organisations worldwide, including ours, and is a proven model for accelerating improvement.6

Background

Early-onset sepsis (EOS), defined as the isolation of pathogenic bacteria from a neonate’s blood or cerebrospinal fluid within 72 hours of birth, is an important cause of neonatal morbidity and mortality.7 EOS is usually due to vertical transmission by contaminated amniotic fluid or during vaginal delivery from bacteria in the mother’s lower genital tract.5 The mortality is approximately 2% to 3% among infants with EOS born at ≥35 weeks gestation.8 9 The important risk factors for EOS include prematurity, Premature rupture of membranes, group B Streptococcus (GBS) carrier state and CA.5 10 CA is considered the most critical risk factor for EOS, and most units, including ours, treated all CA-exposed neonates with intravenous antibiotics. The most recent update11 defines CA as a single spike of intrapartum temperature ≥39.0°C or two spikes ≥38.0°C with fetal tachycardia (>160), leucocytosis (>15 000 cells/mm3) or purulent cervical discharge. However, the identification of CA is challenging. The obstetric providers often take other maternal risk factors also into consideration, and many mothers are suspected to have CA even in the absence of typical clinical criteria. Based on the COFN 20144 recommendation, all these neonates required intravenous antibiotics.

Due to the universal maternal screening and intrapartum antibiotic prophylaxis, the incidence of EOS has decreased considerably over the past two decades.5

Researchers have observed increased risks of later childhood health problems, such as wheezing, asthma, food allergy, inflammatory bowel disease and childhood obesity among children who received antibiotic therapy during the neonatal period.12 13 In addition, neonatal antibiotic administration can also alter the developing gut microbiome.14

In 2018, the AAP clinical report5 recommended the EOSCAL as a safe alternative strategy for risk stratification and management of neonates at risk for EOS.

The EOSCAL is a multivariate risk model developed by Kuzniewicz et al15 in a large prospective study conducted at the Kaiser Permanente hospitals USA.

The EOSCAL is available at https://neonatalsepsiscalculator.kaiserpermanente.org/InfectionProbabilityCalculator.aspx.

Based on six input parameters (baseline risk of EOS, gestational age, peak intrapartum temperature, GBS status, duration of ruptured membranes and timing of intrapartum antibiotics), the calculator can predict the risk (per 1000 live births) at birth and until 12 hours of age, according to the clinical status.

Several subsequent studies have validated the tool, which is widely used worldwide. Many researchers, including the original authors,16–18 observed up to a 50% decrease in early empirical antibiotic use without increasing the number of missed EOS cases or readmission with sepsis during the first week.

However, some researchers questioned the safety of EOSCAL use among the specific population of neonates exposed to CA. Rajbhandari and La Gamma,19 Aghai,20 Ayrapetyan and Carola21 and Carola et al22 observed that the calculator missed or delayed treatment for some neonates with EOS. Subsequently, Sloane et al23 showed that using a baseline risk of 4/1000 did not miss any case of proven EOS among 896 neonates exposed to CA.

The baseline risk of EOS among babies born in AWH is around 1/1000 live birth. At the start of the project, the risk of EOS among CA-exposed babies in our unit was not known. Hence, we decided to use a baseline risk of 4/1000. Since premature babies have an additional risk for sepsis, they were excluded from the project. Based on the evidence that most babies with EOS had clinical signs,24 symptomatic babies were also excluded from the project.

Measurement

Inclusion criteria: all the following

  1. Babies born in AWH.

  2. Mother with suspected CA, as judged by the obstetrician.

  3. Gestational age 37 weeks and more.

  4. Asymptomatic.

Exclusion criteria: any of the following

  1. Babies born outside AWH.

  2. Those with no clinical suspicion of maternal CA.

  3. Gestational age less than 37 weeks.

  4. Symptomatic baby.

Objectives

The primary objective was to reduce antibiotic use among term, CA-exposed, well-appearing neonates to less than 50% by 31 July 2020. Secondary objectives were to reduce blood culture rate and NICU stay among babies with EOSCAL score <1.

Target measures

  1. Antibiotic use rate among term well-appearing neonates exposed to clinical CA.

This was the primary outcome measure and was defined as the number of babies who received antibiotics among all term well neonates exposed to clinical CA. The target was set as 50%.

  1. Blood culture rate among babies with a score <1: This was defined as the number of neonates with a score <1 for whom blood CS was sent among all neonates with a score <1.

  2. Percentage of eligible babies with a score <1 who were transferred to the postnatal ward within 36 hours.

The operational definition was the number of neonates admitted with a score <1 and were transferred back to the mother within 36 hours of life among all neonates with a score <1. This was considered only if the mother stayed in the postnatal ward for at least 48 hours.

During the first year, all babies were observed in the NICU until the 48 hours blood culture result was available. After this period, blood CS was performed only when the score was >1. Those babies with scores <1 were transferred to the postnatal ward after 24–36 hours of observation in NICU, provided their mothers stayed at least 48 hours.

Process measures

  1. The use of EOSCAL among terms, well-appearing babies exposed to CA, aiming for 100% compliance.

  2. EOSCAL calculation error, defined as the percentage of target neonates for whom the admitting physician miscalculated the EOSCAL. The recalculation was done by one of the team members. The desired target was 0%.

Balancing measures

  1. Percentage of initially untreated babies who become symptomatic and require antibiotics within 48 hours of age.

  2. Percentage of postdischarge readmission for culture-positive sepsis during the first week among all babies who did not receive antibiotics.

Design

By the end of 2019, a multidisciplinary team, including neonatologists, neonatal nurses and a patient quality coach, was formed. A driver diagram was constructed after two brainstorming sessions. The current evidence about the management of neonates exposed to clinical CA was reviewed thoroughly, and the EOSCAL was finalised as the tool for risk stratification and neonatal management. Based on the existing evidence, we expected at least a 50% reduction in antibiotic use. A critical challenge anticipated was the likelihood of missing a truly septic baby. This possibility was reduced by excluding premature and symptomatic babies, using a baseline risk of 4/1000 for calculation, and lowering the intervention threshold from 4 hours to 2 hours among neonates with equivocal clinical signs.

We ensured sustainability by incorporating the project into the routine care of babies admitted following maternal CA (figure 1).

Figure 1
Figure 1

Flow chart for patient management. EOSCAL, early-onset sepsis risk calculator; GBS, group B Streptococcus; NICU, neonatal intensive care unit.

Strategy

The NICU physicians, nurses from NICU, the labour room and the obstetric theatre were educated about the project through multiple sessions. The project was implemented on 1 February 2020. A pretext template was uploaded to the electronic health records for ease of documentation.

One of the three physician members collected data in turn. Baseline information, parameters needed for the EOSCAL, clinical status blood culture, antibiotic use and data related to readmission during the first week of life were collected and recorded on a predesigned Excel data collection form. The process, outcome and balancing measures were recorded on a separate Excel page. The team also reviewed the EOSCAL score assigned for each case to monitor for any calculation errors.

Daily data were collected during the first 2 months (pilot period), and we observed a high proportion of calculation errors (71%). However, the errors did not affect the management plan in any of the cases. The antibiotic use rate among the target population was consistently below 50% from the first month of implementation (figure 2).

Figure 2
Figure 2

Run chart of calculation error. EOSCAL, early-onset sepsis risk calculator; PDSA, Plan-Do-Study-Act.

During PDSA (Plan-Do-Study-Act) 1, we re-educated admitting physicians through one-to-one sessions, and subsequently, the error was reduced to 6.25% (figure 2). In addition, we continued daily monitoring to ensure that antibiotics were administered for all eligible cases.

As a small percentage of calculation continued to persist, we did a second round of physician education, focusing on those committing frequent errors (PDSA2—figure 3). We observed that most errors were related to inaccurate information on the duration of the rupture of membranes and the timing of intrapartum antibiotics. In PDSA 3 (figure 2), we designed a template inclusive of the input parameters, and printed copies were distributed to the labour room and obstetric operation theatre. The nurses were requested to carry a copy of the completed template while admitting any baby with a history of maternal CA. After PDSA 3, we observed a consistent reduction in the calculation error.

Figure 3
Figure 3

Run chart of antibiotic use rate. PDSA, Plan-Do-Study-Act.

We reviewed the data of CA-exposed neonates who were admitted during the preceding 4 years and observed that none of the term asymptomatic neonates developed positive blood cultures.25

During PDSA 4 (figure 3), we started to use a baseline risk of 2/1000 and antibiotic use was reduced further. Finally, through PDSA 5, we introduced two additional outcome measures for babies with score <1. This included avoidance of blood culture and transferring to the mother within 36 hours.

Results

Among 3837 inborn NICU admissions, 464 (12 %) were CA-exposed babies. Of them, 341 (74%) cases were eligible for inclusion. Among eligible cases, 270 (79%) did not receive antibiotics. Blood culture could be avoided among 106 (97 % of low-risk babies), and NICU stay was reduced among 45 (92% of eligible, low-risk babies).

None of the babies who were assigned for clinical observation without antibiotics developed sepsis during the first week. As another balancing measure, we monitored babies who were initially asymptomatic and later required antibiotics within 48 hours. Of 304 babies who were initially assigned for clinical observation with or without blood culture, 34 developed non-specific symptoms such as vomiting or mild respiratory distress. All these babies received antibiotics soon after the onset of the symptoms. The mean age of starting antibiotics was 3.5 hours. None of them developed culture-positive EOS, and only 1% required antibiotics for more than 2 days.

All three babies with a culture-positive EOS were symptomatic and received antibiotics within the first hour of life.

Lessons and limitations

In our setup, CA is the most common reason for NICU admission and antibiotic use among well-appearing term babies. However, most of these empirically treated babies are later proven unaffected. Using the EOSCAL, we reduced antibiotic use to less than 50% among this population.

The critical challenge we faced was calculation error. The EOSCAL is very sensitive to minor variations in input parameters, and frequent calculation errors were identified during the first few months of implementation. The errors were brought to zero by multiple PDSAs. However, we recognised that constant monitoring of the calculation process is vital for maintaining patient safety.

We observed that some initially asymptomatic babies developed symptoms such as respiratory distress or poor sucking and were then started on antibiotics. However, none of these babies had confirmed EOS, and only 1% received antibiotics beyond 2 days for culture-negative sepsis.

Our centre does not perform routine placental histopathology for confirmation of CA. The clinical diagnosis of CA was purely at the discretion of the obstetricians.

We admit that our data may not confirm the safety of EOSCAL use in this population. Several authors have questioned the efficacy of the calculator in identifying cases with EOS. For example, Scott et al26 reported that using EOSCAL missed 35% of EOS cases, while only 3% was missed when the Australian guideline was used. In a systematic review including 75 EOS cases identified by the National Institute for Healthcare and Excellence (NICE) recommendation, Pettinger et al27 estimated that at least 14 (18%) cases would be missed if EOSCAL was used. Similarly, Snoek et al28 observed that EOSCAL identified only 36% of EOS by 4 hours, compared with NICE (55%) and Dutch (50%) guidelines.

However, our project excluded symptomatic and premature babies who have the highest risk of EOS.

Sharma et al29 applied EOSCAL to 180 CA-exposed babies of gestational age ≥36 weeks and achieved significant reduction in blood culture rate and antibiotic use. However, the study was underpowered to explore the safety in terms of missed or delayed treatment of proven EOS cases. A similar study by Chirvolu et al30 also observed reduced antibiotic use and median hospital stay after implementation of the EOSCAL among CA-exposed late preterm and term neonates. There were no cases of culture proven sepsis.

Conclusion

The multivariate approach for empirical management of EOS (EOSCAL) is not uniformly practised outside the US. NICUs in Qatar used the categorical approach to manage neonates exposed to CA, resulting in 100% of them receiving antibiotics. Our project showed that the use of EOSCAL could safely reduce antibiotic use to 20%. Additionally, we reduced the laboratory evaluation and length of stay among a subgroup of babies with a lower risk of sepsis. As a balancing measure, we monitored all the babies during the first week, and none of the untreated babies developed EOS. Although we did not conduct a formal cost-effectiveness analysis, the project helped in reducing the costs related to NICU stay, laboratory evaluation and antibiotic use. The project is sustainable and is now part of our routine clinical care of CA-exposed neonates. Further work is underway to spread our success to other NICUs in Qatar.