The Effect of Single-Room Care Versus Open-Bay Care on the Incidence of Bacterial Nosocomial Infections in Pre-Term Neonates: A Retrospective Cohort Study

The LUMC is a tertiary university-affiliated hospital with a 25-bed level-III NICU constructed as a single-room facility with an average annual admission rate of approximately 500–600 in- and out-born neonates. The ward contains 16 single-patient rooms and four twin-rooms (Supplementary Figure S1). In addition, a separate negative-pressure isolation room designed to prevent the spread of airborne pathogens is located adjacent to the main entrance of the ward. Each infant is hospitalised in a private room equipped with 1 hand-washing facility with touch taps and a dispenser with alcohol-based hand rub. Additional hand rub dispensers are available adjacent to each incubator. The average nurse-to-patient ratio is 1:1 for infants with high-dependency care, and 1:2 for infants requiring intermediate care. Prior to the transition to the new single-room facility, the NICU consisted of three open-bay rooms, two which served as intensive care (IC) units and one as a high-care (HC) unit. The two IC units were configured with, respectively, nine and seven beds, and the HC unit with nine beds. Bed occupancy rates depended on volatility in demand and staffing levels. Each unit was equipped with one sink (Supplementary Figure S2). Hand rub dispensers were located next to each incubator and at the central nursing station. Depending on the level of care required, the standard nurse-to-patient ratio was 1:2 and 1:3 in the IC and HC units, respectively. The average area surrounding each incubator was approximately 2 m².

We conducted a retrospective cohort study to evaluate the rate of nosocomial infections and antibiotic consumption in premature infants before and after transition to the new single-room NICU. Two study periods of 2 years each were compared: one prior to (May 15, 2015–May 14, 2017) and one following (May 16, 2017–May 15, 2019) unit transition. Approval and a waiver from the need to provide written informed consent were obtained from the institutional review board of LUMC (G19.070).

Neonates born at < 32 weeks gestational age and admitted to the NICU between May 15, 2015 and May 15, 2019 were included in this study. Infants born on the day of the unit transition (May 15, 2017) and those who were admitted to both unit types were excluded. Only data from first admissions were analysed.

For all study participants, the following data were extracted from electronic medical records for the period between admission and hospital discharge: birth date, gestational age, birth weight, sex, delivery mode (i.e. caesarian section or vaginal delivery), 5-min Apgar score, presence of major congenital anomalies, presence of surgical pathology, exposure to invasive mechanical ventilation, duration of invasive mechanical ventilation, exposure to a central-line, exposure to intravenous antibiotic therapy ≤ 24 h postpartum and ≥ 72 h after admission, total length of antibiotic treatment, prophylactic fluconazole use, length of hospital stay, in-hospital neonatal mortality and infection-related mortality (i.e. death attributed to infection as either the immediate or underlying cause). Information on NIs was obtained by reviewing patients’ medical records, including laboratory (i.e. C-reactive protein), microbiology (i.e. organisms isolated from clinical cultures) and radiology (i.e. ultrasound, chest and abdominal radiography) results. Data pertaining to CLABSI included central-line type (i.e. umbilical venous, umbilical arterial or central-venous catheter), number of central-lines placed per infant, total duration of central-line use and isolated pathogenic microorganism(s) in case of a confirmed CLABSI.

Evaluation for sepsis was performed at the discretion of the attending physician in accordance with hospital protocol and included blood-culture analysis and at least one quantitative C-reactive protein (CRP) measurement 18 h after the initial sepsis work-up. Only culture results for specimens collected ≥ 72 h after NICU admission were considered. In case an infant was evaluated for sepsis more than once, all new unrelated episodes – i.e. absence of clinical signs or positive cultures between two episodes – were regarded as separate infection episodes.

Nosocomial infection was defined as isolation of a pathogenic organism from a blood or other sterile fluid culture obtained ≥ 72 h after admission in combination with clinical symptoms of infection (temperature instability (< 36.5 or ≥ 38 °C), abdominal distension, feeding intolerance, respiratory distress, increase in ventilation requirements, apnea, hypotension, hyper-/hypoglycemia, metabolic acidosis, abnormal laboratory signs (i.e. leukocytosis, thrombocytopenia) or a CRP ≥ 10 mg/L). CLABSI was defined as a laboratory-confirmed bloodstream infection occurring ≥ 72 h after birth and associated with an indwelling central-line and not related to an infection elsewhere. In addition, the definition requires that the central-line had been in place for > 2 calendar days, up until the day of removal or the following day. Confirmed CLABSIs were subsequently categorised into one of three criteria based on the isolated pathogen (Supplementary Table S1). A positive culture with coagulase-negative staphylococci (CoNS) was considered an infection when a CRP ≥ 10 mg/L was measured within 36 h of infection onset. Cultures yielding other low virulence organisms (Corynebacterium spp., Micrococcus spp.) without elevated inflammatory markers were regarded as contamination and removed from further analyses. Given their contribution to short- and long-term neonatal morbidity, and in order to provide a comprehensive measure of the burden of NIs among preterm neonates, additional non-CLABSI NI types were also evaluated. As standardised definitions for NIs within the neonatal population are lacking, definitions for the non-CLABSI NI types were developed using well-established reference sources and adjusted according to specific neonatal symptomatology and diagnostic procedures. The diagnosis of a NEC was based exclusively on clinical parameters, as data on the prevalence of concurrent positive blood cultures in infants diagnosed with NEC are inconclusive [11]. Only grades IIA–IIIB according to Bell’s staging criteria were collected [12]. Remaining definitions pertaining to each NI type are listed in Supplementary Tables S1 and S2.

Data are reported as mean and standard deviation (SD), median and interquartile range (IQR) or absolute number and percentages, as appropriate. In univariate analysis, categorical data were analysed using the Chi-square test or Fisher’s exact test, and continuous data using the independent t test or Mann–Whitney U test, as appropriate. Rates for NIs were calculated as the number of infection episodes per 1000 patient-days (incidence density) and number of infection episodes per 100 infants (cumulative incidence rate). CLABSI rates were calculated per 1000 central-line days. Considering infections as rare events, differences in infection rates in Poisson distribution were calculated using the exact conditional test. Shewhart U charts were used to evaluate the presence of special-cause variation with regards to quarterly CLABSI and NI rates [13]. To evaluate change in antibiotic use between the two study periods, total days of treatment (DOT) per antibiotic agent and length of treatment (LOT) were calculated. If an infant received two or three different antibiotic agents within 1 day, each was counted as a separate DOT. Finally, to identify predictors for NI, univariate and multivariate Fine-Gray Competing Risk Regression models were created in which death without infection was modelled as a single competing outcome. Given that the outcome of death may modify the probability that an infection occurs, the competing risk modelling approach was chosen to allow for direct assessment of the association between a covariate and the cumulative incidence of infection. Unit type was included as the sole variable in the univariate model as the main outcome of interest. For the multivariate model, predictors were chosen based on clinical relevance. Retainment in or elimination from the model was based on maximum log-likelihood estimates using the stepwise selection approach. Subdistribution hazard (sDH) ratios comparing the probability of NI for each predictor value, conditional upon survival or occurrence of the competing risk (death), were computed along with corresponding 95% confidence intervals (CI). All p values were two-tailed and p < 0.05 was regarded as statistically significant. All statistical analyses were performed using R (v.3.6.2) [14].

Derivation of the study sample is shown in Fig. 1. Totals of 367 and 345 premature infants were admitted to the open-bay and single-room units, respectively, whose demographic characteristics are described in Table 1. There were significantly more infants in whom a central-line was inserted (61% vs. 48%, p = 0.005) during the open-bay period compared to after the conversion to single-room units. No other significant differences in neonatal characteristics between the two study periods were identified.

A total of 200 NI episodes occurred in 164 (23%) infants (Table 2). No differences were found in the incidence density per 1000 patient-days (13.68 vs. 12.62 p = 0.62) nor cumulative incidence rate per 100 neonates (23.97 vs. 22.02, p = 0.59) between the historic open-ward and new single-room unit.

Apart from a slight increase in the number of primary blood stream infections and upper respiratory tract infections, an overall decrease was seen in the number of episodes for the remaining infection types after transition to the new unit, despite not being statistically significant (Table 3). No infant developed a meningitis during either study period.

During both study periods, three neonates died of a hospital-acquired infection, with mortality directly attributable to infection being 0.4% (Table 1). Of these three neonates, two were hospitalised in the new SRU. The cause of death of these three neonates were, respectively, E.coli sepsis-related multiorgan failure, methicillin-sensitive S. aureus-related respiratory and circulatory failure, and fulminant sepsis after NEC.

Figure 2 details the quarterly NI rates across the entire 4-year study period. The mean NI rate was 11.2 per 1000 patient-days. U chart analysis did not reveal any occasion in which the variation in NI incidence was higher or lower than expected by random (common-cause) variation, particularly around the time of unit transition.

As the most common type of NIs in neonates, we evaluated the effect of unit transition on the occurrence of CLABSI by assessing changes in central-line use, dwell-time and incidence of central-line infections among neonates with one or more central-lines (Table 4). Over the two study periods, 779 central-lines with 2963 central-line days were placed in 390 infants, with significantly more central-lines placed during the open-bay period compared to the single-room period (430 vs. 349, p = 0.004). No difference was found in average number of central-line days per child, type of inserted central-line, and average age at central-line insertion and removal. Overall, 23 neonates developed at least one CLABSI in the OBU compared to 14 in the SRU, although this difference was not statistically significant (p = 0.51). The 25% reduction in CLABSI rate (14–10.59 per 1,000 central-line days) after unit transition was also not statistically significant (p = 0.51).

Figure 3 shows the quarterly CLABSI rates across the entire 4-year study period, with the mean CLABSI rate being 12.5 infections per 1000 line-days. No special-cause signalling, specifically around unit transition, was identified by U chart analysis. A steep increase in the infection rate in the period May–July 2019 is present, likely resulting from a lower number of line-days (36.37 line-days) compared to prior quarters.

Supplementary Table S3 details the pathogen distribution of the most prevalent NI types, namely, CLABSI and primary bloodstream infection (PBSI). For both infection types, the majority of monomicrobial infections were caused by Gram-positive cocci (CLABSI: 92%, PBSI: 84%), with Gram-negative organisms accounting for 5.4 and 7% of CLABSI and PBSI episodes, respectively. There was a slight decrease in CLABSIs and PBSIs owing to CoNS after unit transition (CLABSI: 87% vs. 71%, PBSI: 70% vs. 68%). Other frequently isolated Gram-positive organisms included S. aureus and Enterococcus spp. (E. faecium), the former increasing in prevalence among PBSIs following transition to the SRU. Two PBSI episodes were found to be polymicrobial, both predominated by Gram-negative organisms. Only a small proportion of CLABSI and PBSI episodes were caused by Gram-negative organisms and fungi throughout the entire study period.

No difference between the two units was found in number of neonates who received antibiotic treatment within the first 24 h postpartum (62% vs. 57%, p = 0.19) (Supplementary Table S4). Likewise, no difference between the units was found in the proportion of neonates who were treated with one or more courses of antibiotics ≥ 72 h after admission (64% vs. 57%, p = 0.06). Total LOTs of 2832 and 2413 days were registered in the OBU and SRU, respectively, with the normalised difference not statistically significant (345 vs. 346 per 1000 patient-days, p = 0.51). Similarly, no significant decrease was found in antibiotic consumption accounting for each agent (DOT, 616 vs. 607 per 1,000 patient-days, p = 0.99).

Supplementary Table S5 shows the subdistribution hazard ratios estimated from the Fine–Gray competing risk regression model. In univariate analysis, unit type was not associated with a change in risk of infection (sDH: 0.95, p = 0.78). In multivariate analysis, 12 variables, namely unit type, gestational age, sex, 5-min Apgar score, birth weight, birth modus, antenatal steroids, duration of invasive mechanical ventilation, exposure to a central-line, central-line dwell-time, antibiotic treatment within 24 h after birth and length of hospital stay were selected to be evaluated as potential risk factors for NI. Only duration of invasive mechanical ventilation was identified as a minor risk factor (sDH: 1.03, p = 0.01). No other variables, including single-room care, were significantly associated with NI in multivariate analysis.

The strengths of our study include the meticulous data collection and clearly defined NI types. However, our data also have several limitations. First, the limited power of our study may have inherently hindered the results from reaching statistical significance. A repeat analysis with a larger study sample and a longer time series is therefore desirable. Second, we were unable to correct for potential prevailing local temporal trends in NIs due to lack of available data. Analysis using segmented interrupted-time series with monthly incidence rates as opposed to a simpler pre- and post-analysis would have allowed for a more refined assessment of the effect of unit transition on infection rates. Lastly, we were unable to identify and model for all other potential modifications in infection-control practices that may have taken place as part of continuous quality improvement efforts.