Article Text

Download PDFPDF

The impact of a sepsis quality improvement project on neurodisability rates in very low birthweight infants
  1. Jonathan W Davis1,
  2. David Odd2,3,
  3. Sally Jary1,3,
  4. Karen Luyt1,3
  1. 1Regional Neonatal Unit, St. Michael's Hospital, University Hospitals Bristol Foundation NHS Trust, Bristol, UK
  2. 2Regional Neonatal Unit, Southmead Hospital, North Bristol NHS Trust, Bristol, UK
  3. 3School of Clinical Sciences, University of Bristol, Bristol, UK
  1. Correspondence to Dr Jonathan W Davis, Regional Neonatal Unit, St. Michael's Hospital, University Hospitals Bristol Foundation NHS Trust, Bristol BS28EG, UK; j_davis84{at}


Objective Very low birthweight (VLBW; <1500 g) infants with late-onset sepsis (LOS) have an increased risk of neurodisability. Care bundles to reduce bloodstream infections in neonatal intensive care unit (NICU) are effective in reducing LOS. Our aim was to determine if a sepsis reduction bundle introduced through a quality improvement project would impact neurodevelopmental outcomes in VLBW infants.

Design Cohort study.

Setting Level 3 regional NICU in the South West of England.

Patients VLBW infants born between 2002 and 2011.

Interventions A sepsis reduction care bundle implemented between July 2006 and December 2007.

Main outcome measures The primary outcome was risk of coagulase-negative Staphylococcus (CONS) infection diagnosed >3 days of age. Secondary outcomes were death and moderate cognitive impairment. A logistic regression model was derived using the birth era as the independent variable with adjustment for typical confounders.

Results In total, 379 infants were born in the preintervention cohort and 378 in the postintervention cohort. The CONS infection rate was reduced after the intervention (26.7% vs 14.1% p<0.001). Death prior to discharge reduced without reaching statistical significance (14.1% vs10.9%, p=0.195). The rate of cognitive disability reduced in the postintervention cohort (18.8% vs 6.1%, p=0.042). The adjusted ORs (95% CI) for CONS infection, death and cognitive impairment were 0.46 (0.29 to 0.72), 0.73 (0.43 to 1.24) and 0.3 (0.07 to 1.33), respectively.

Conclusions There appears to be an association between reduced cognitive disability and the implementation of a sepsis reduction bundle. Further study in larger series is required to confirm these findings.

  • Neonatal infection
  • Quality improvement
  • Neurodisability

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What is already known on this topic?

  • Nosocomial infection in very low birthweight (VLBW) infants remains problematic.

  • Infection in VLBW infants substantially increases the risk of neurodisability.

  • Infection reduction care bundles are effective in reducing late-onset sepsis.

What this study adds?

  • Infection reduction bundles reduce coagulase-negative Staphylococcus and other late-onset infections in VLBW infants.

  • We describe an association between reduced infection rates and cognitive disability at 2 years of age.


Very low birthweight (VLBW, <1500 g) infants are at high risk of infection during their stay in neonatal intensive care unit (NICU),1 and, importantly, infections in VLBW infants have an additive detrimental effect on neurodevelopmental outcomes.2 ,3

Late-onset sepsis (LOS) is defined as infection commencing after 48–72 hours of age.1 The prevalence of LOS as reported by large data collections such as the Vermont Oxford Network (VON) and the National Institute of Child Health and Development (NICHD) is between 10% and 20%.4–6 VLBW infants are at increased risk of infection because of their immunological immaturity,1 exacerbated by intensive care interventions such as central venous catheterisation and ventilation.7

Care bundles are defined as a limited number of specific practices essential for effective and safe patient care and that implemented together result in additional improvements.8 Studies have demonstrated the effectiveness of ‘care bundles’ on the overall incidence of infection in the NICU population, mainly in state-wide collaborations,8–10 although no study has until now demonstrated improvement in neurodevelopmental outcome or survival.

This work is based in a 31 bedded tertiary-level neonatal unit with approximately 100 VLBW infants born per annum. Between 2001 and 2004, the rate of LOS in VLBW infants was 40%, compared with 20%4 for the VON average and hence an ‘infection care bundle’ was developed and introduced through a quality improvement project (QIP) in an attempt to improve infection rates.

Our aim was to determine the impact of the introduction of an infection reduction bundle in a tertiary neonatal unit on VLBW infants' survival and neurodevelopmental outcomes.


The measures introduced were based on the internet-based newborn improvement collaboration for quality improvement programme from the VON.11 The quality improvement strategy began with a broad assessment of practice by a small group of cross-discipline individuals and led to the identification of a series of potentially better practices (figure 1).

Figure 1

Examples of potentially better practice (PBP) identified and implemented. NICU, neonatal intensive care unit.

The bundle consisted of improving hand washing and cleanliness of incubator environment, targeted aseptic intervention for venous and arterial line insertions, as well as the introduction of the aseptic non-touch technique (ANTT) for the access of central and arterial lines.12 In addition, closed systems were introduced on all arterial access sets and microclaved ports on venous sets. All these measures were underpinned with regular, rapid-cycle audits to identify and rectify areas of improvement. Rapid-cycle audits were performed around hand and environmental cleanliness and ANTT, with immediate display of results to staff. Monthly compliance rates were displayed on run charts, which were visible to all staff on the unit (figure 2).

Figure 2

Example of compliance chart August 2006–January 2007.

The population investigated was all inborn infants less than 1500 g at St. Michael's Hospital, Bristol, UK, between 2002 and 2011. The care bundle was introduced to the unit in stages over 18 months between July 2006 and December 2007. Exposure was defined as before the intervention (January 2002–December 2007) or after the intervention had been fully implemented (July 2008–December 2011). The period between December 2007 and July 2008 was omitted from analysis to allow comparison of full practice implementation. The primary outcome was risk of coagulase-negative Staphylococcus (CONS) infection diagnosed after 3 days of age. Secondary outcomes were any late (>3 days old) bacterial infection, death, cerebral palsy or moderate cognitive impairment. Moderate cognitive disability was defined as a Bayley Scales of Infant Development second edition (BSIDII) mental development index (MDI) below 70 or Bayley Scales of Infant Development third edition (BSIDIII) combined language and cognitive scales below 80 at corrected age of 24 months.13

The VON definition of a CONS infection was used, which is the recovery of that bacterium from blood culture obtained from either a central line or a peripheral blood sample in the presence of one or more signs of generalised infection (such as apnoea, temperature instability, feeding intolerance, worsening respiratory distress or haemodynamic instability) and initiation of intravenous antibiotics treatment for five or more days after blood cultures were obtained. If the infant died, was discharged or transferred prior to the completion of 5 days of intravenous antibiotics, this condition would still be met if the intention were to treat for five or more days. Late bacterial infection was defined as a recognised bacterial pathogen (excluding CONS) recovered from blood culture after day 3 of life.

Initially, the absolute risk of infection was assessed by year of birth. A logistic regression model was derived using the era of birth as the independent variable initially, and then adjusting for the confounders. Confounders were defined a priori as antenatal steroids (completed course), gender, birth weight, multiple birth, gestational age at birth and Apgar scores (at 1 and 5 min). No tests were performed to ‘test’ for confounders. Ordinal variables were checked to ensure that categorised terms did not show better fit than linear terms (all Wald tests >0.05). This was repeated for the primary and secondary outcomes listed above. Univariable comparisons were made using the t-test or Wilcoxon-Mann-Whitney test as appropriate. All analyses were performed using STATA10.

Ethical approval was not specifically requested as this intervention was part of a QIP and was registered as a service evaluation with University Hospitals Bristol NHS Trust.


The results are described first by the baseline characteristics for each group, which were additionally used as confounders to adjust for the ORs later in this section (table 1). In total, 757 patients were born during the study period at birth weight <1500 g. The number of patients in the two cohorts was 379 and 378 in the early (2002–2007) and late (2008–2011) periods, respectively.

Table 1

Characteristics of the study population (n=757)

Although birth weight, gestation and Apgar score were statistically different, there was little numerical difference between the cohorts. The use of antenatal steroids improved significantly between the two time periods (78.1% vs 89.1%, p<0.001).

CONS and any other late infection were significantly reduced between the epochs (26.7% vs 14.1%, p<0.001; 18.8% vs 8.7%, p<0.001, respectively) (table 2, figure 3). On univariate analysis, the rate of cognitive disability was significantly lower in the postintervention cohort (18.8% vs 6.1%, p=0.042) (table 2). The adjusted ORs (95% CI) for CONS infection, death and cognitive impairment were 0.46 (0.29 to 0.72), 0.73 (0.43 to 1.24) and 0.3 (0.07 to 1.33), respectively (table 3).

Table 2

Outcome measures, split by era of birth

Table 3

Association between era of birth and outcome measure

Figure 3

Rates of coagulase-negative Staphylococcus (CONS) infection by year, demonstrating the impact of the intervention.


This is the first description of the long-term impact of a sepsis reduction bundle on neurodevelopmental outcomes in VLBW infants. A reduced rate of cognitive disability will be important if confirmed in larger series. This QIP achieved its primary outcome in reducing coagulase-negative Staphylococcal infection by nearly 50%. In addition, all other late bacterial infections were significantly reduced by the intervention. The intervention has demonstrated sustained impact as the rate of CONS infection in 2013 was 2.1%.

The link between systemic neonatal infection and cerebral white matter injury has been described for more than 40 years.2 ,14 In a population of over 6000 extremely low birthweight infants, Stoll et al,3 reported that one or more episodes of infection significantly increased the risk of cognitive, motor and sensory disabilities. In this study, there was some evidence of a reduction in cognitive disability without significant changes in death or cerebral palsy. After adjustment for potential confounders, the CI around cognitive disability widened, although the point estimate remained similar. While this effect may have been due to chance, it remains consistent with the published literature.

We were limited by the size of the population and the fact that other changes in care may have occurred over the study period, not identified in this work. Although we do know that neuroprotective interventions such as intrapartum magnesium sulfate, delayed cord clamping and enhanced nutrition were not adopted on our unit before 2012 (after the end of the study period). In addition, the neurodevelopmental assessment changed during the study. It is well recognised that the BSIDIII cognitive and language scales overestimate cognitive ability compared with the BSIDII MDI. For this reason, we used a validated13 cut-off of 80 for moderate cognitive disability when using the BSIDIII cognitive and language composites to allow comparison across time.13 There is no validated alignment of cut-offs for BSIDII psychomotor development index and BSIDIII motor score; therefore, BSID motor score was not included in the analysis.

This is the first report where reductions in proven infection and cognitive disability have been associated. This effect needs to be studied and corroborated in larger populations and between institutions before more definite conclusions can be made.


The authors would like to thank Dr Pamela Cairns and Dr David Harding for their curation and administration of the patient database.



  • JWD and DO contributed equally.

  • Twitter Follow Jonathan Davis at @jonathan_davis3

  • Contributors The study was conceived by KL. SJ performed the neurodevelopment assessments. The data were analysed by DO. The significance of the findings was discussed by all four authors. The initial manuscript was written by JWD with input from the other authors. JWD and DO contributed equally to the production of the manuscript.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement The data are extracted from unit-specific data from an international data collaboration.

Linked Articles

  • Editorial
    Roger F Soll William H Edwards
  • Fantoms
    Ben Stenson