Outcomes of Nosocomial Viral Respiratory Infections in High-Risk Neonates

BACKGROUND AND OBJECTIVE: Neonatal respiratory disease, particularly bronchopulmonary dysplasia, remains one of the leading causes of morbidity and mortality in newborn infants. Recent evidence suggests nosocomially acquired viral respiratory tract infections (VRTIs) are not uncommon in the NICU. The goal of this study was to assess the association between nosocomial VRTIs, neonatal respiratory disease, and the health care related costs. METHODS: A matched case–control study was conducted in 2 tertiary NICUs during a 6-year period in Nottingham, United Kingdom. Case subjects were symptomatic neonatal patients with a confirmed real-time polymerase chain reaction diagnosis of a VRTI. Matched controls had never tested positive for a VRTI. Multivariable logistic regression was used to test for associations with key respiratory outcomes. RESULTS: There were 7995 admissions during the study period, with 92 case subjects matched to 183 control subjects. Baseline characteristics were similar, with a median gestation of 29 weeks. Rhinovirus was found in 74% of VRTIs. During VRTIs, 51% of infants needed escalation of respiratory support, and case subjects required significantly more respiratory pressure support overall (25 vs 7 days; P < .001). Case subjects spent longer in the hospital (76 vs 41 days; P < .001), twice as many required home oxygen (37%; odds ratio: 3.94 [95% confidence interval: 1.92–8.06]; P < .001), and in-hospital care costs were significantly higher (£49 664 [$71 861] vs £22 155 [$32 057]; P < .001). CONCLUSIONS: Nosocomial VRTIs in neonatal patients are associated with significant greater respiratory morbidity and health care costs. Prevention efforts must be explored.

WHAT'S KNOWN ON THIS SUBJECT: Viral respiratory tract infections (VRTIs) cause severe respiratory morbidity in ex-preterm infants after NICU discharge. VRTIs are now recognized to be more prevalent in the NICU, but their impact, before discharge, during this early period of life is unclear.

WHAT THIS STUDY ADDS:
This study identifi es the adverse impact that VRTIs, particularly rhinovirus, have on newborn infants during their initial NICU admission. Identifi cation of the associated signifi cant respiratory morbidity and health care costs should focus efforts on reducing these nosocomially acquired infections.
Neonatal bronchopulmonary dysplasia (BDP) is associated with long-term respiratory morbidity, including asthma, chronic obstructive airways disease, 1 -3 and poor neurodevelopmental outcomes. 4 Despite improvements in the respiratory management of preterm infants, and subsequent increased survival, many countries have not reported reductions in BPD. In the United Kingdom, 68% of extremely preterm survivors born in 2006 had BPD, unchanged from 1995 rates. 5 Current US estimates suggest there are between 5000 and 10 000 new cases of BPD annually, 2 with respiratory causes of death second only to immaturity in extremely preterm infants. 6 The pathogenesis of BPD is multifactorial; the presence of inflammation and prematurity are key factors. The inflammatory changes observed are usually a result of infection, ventilatory trauma, and hyperoxia. 7 Few studies have explored the impact of viral respiratory tract infections (VRTIs) in this process, although animal studies have highlighted their potential adverse role. 7,8 After discharge from the NICU, surviving preterm infants with subsequent VRTIs are more likely to develop severe respiratory disease or experience recurrent wheeze and asthma. [9][10][11] Nosocomial VRTIs are often unrecognized or clinically underdiagnosed in the NICU, 12,13 with up to 52% of admissions testing positive for viral DNA according to polymerase chain reaction (PCR). 14 It is therefore plausible that VRTIs acquired during the NICU period could worsen BPD in preterm infants, although previous studies have mostly been small observational cohorts. 12, 13, 15 -19 The goal of the present study was to assess the association between nosocomial VRTIs acquired in the NICU and respiratory morbidity up to hospital discharge.  20 Two control subjects were matched per case subject according to the following: (1) gestation (±1 week); (2) center in which they received the majority of their care; and (3) birth within 6 months of the case subject to minimize variations in care over time. All infants studied were admitted to the NICU after birth.

Study Design
Data were collected on a standardized pro forma, from both the clinical notes and hospital electronic records. Gestational age was determined by ultrasound scanning or, if this option was not available, last menstrual period was used. Clinical data included birth weight, mode of delivery, number of siblings, number of courses of maternal steroids, maternal infections in pregnancy, maternal smoking, intrauterine growth retardation, and surfactant doses. Postnatal data included time in the hospital, days of respiratory support, and other comorbidities such as congenital heart disease, retinopathy of prematurity (requiring treatment), and a clinical diagnosis of necrotizing enterocolitis.
For case subjects, the escalation of respiratory care was defined according to set criteria by at least 1 step up the support pathway from their immediate preinfection status: no support → oxygen only → continuous positive airway pressure → bilevel positive airway pressure → invasive ventilation → high-frequency ventilation or extracorporeal membrane oxygenation. BPD was defined as the requirement in infants <32 weeks' gestation of oxygen for >28 days and at 36 weeks' corrected gestation. 21 Infants' requirements for home oxygen were based on the criteria defined by the British Thoracic Society 22 using an agreed hospitalwide guideline.
Ethical approval was given by the University of Nottingham Medical School Ethics Committee (reference LTd10042014).

Sampling and Real-Time PCR Analysis
All respiratory samples, either nasopharyngeal and/or lung aspirates if ventilated, were collected at the time of the presumed infection and analyzed by using the multiplex PCR Respiratory Viral Panel assay according to a standard operating procedure in a certified National Health Service laboratory (Supplemental Table 3). Data were also collected on any other positive microbiologic samples found ±24 hours of the PCR diagnosis of a VRTI.

In-hospital Cost of Care
Costs of care were calculated based on the 2011-2012 published tariffs for neonatal care in the UK National Health Service. 23 Costs for each level of care (intensive, high dependency, and special care) were calculated to the point of discharge or death.

Data Analysis
Data were analyzed by using Stata version 12 (Stata Corp, College Station, TX). Continuous data are presented as means and SDs or medians and interquartile ranges (IQRs) for parametric and nonparametric data, respectively. Conditional logistic regression was used to compare the baseline characteristics of case subjects and matched control subjects, and then to compare respiratory outcomes; these outcomes included number of days of ventilation required, need for continuous positive airway pressure, and number of days spent in intensive care and high dependency. Univariable and multivariable logistic regression, with a priori adjustment for gestation and center, was used to calculate odds ratios (ORs) for the association between several potential risk factors and escalation in respiratory support, as well as discharge on home oxygen. Potential predictors of respiratory morbidity studied included the following: sex, birth weight, mode of delivery, antenatal steroids, maternal infection, chorioamnionitis, intrauterine growth restriction, maternal smoking, surfactant, Apgar score, cardiac anomalies (including patent ductus arteriosus), and necrotizing enterocolitis. Variables significant at the 10% significance level (P < .1) in a univariable logistic regression model were included in a multivariable model. Data were summarized by using ORs with 95% confidence intervals (CIs). Analyses were 2-sided and considered significant if P < .05.

Baseline Characteristics
During the study period, there were 7995 admissions to the NICUs where the study was conducted (1332 per year) with a total of 79 642 beddays. A total of 275 infants (92 case subjects and 183 control subjects) met the inclusion criteria. All case subjects were matched with 2 control subjects, with the exception of 1 infant who only had a single control subject meeting the inclusion criteria. Case and control subjects were evenly matched, with a median gestation of 29.4 weeks (IQR: 26.9-33.9 weeks) for the case subjects and 29.6 weeks (IQR: 26.9-34.0 weeks) for the control subjects. Both groups had similar baseline characteristics ( Table 1).

Viral Infections
There were a total of 95 nosocomial VRTIs diagnosed by using PCR in 92 symptomatic infants ( Table 2). Of these, 84 (91%) were preterm, with 56 (61%) at ≤30 weeks' gestation. These findings equate to a symptomatic nosocomial VRTI incidence of 3.8% in infants admitted at ≤30 weeks' gestation. The median day of life of infection was 43 days (IQR: 25-76 days). Rhinovirus was the dominant pathogen and was detected in 70 case subjects (74%). There were 5 cases of influenza A H1N1 pandemic strain with escalation of respiratory support 3

Respiratory Support
Overall, 51% (n = 47) of infants required escalation of respiratory support during a VRTI episode ( Fig  1). A greater proportion of infants <28 weeks' gestation required new respiratory pressure support during the VRTI (48%) compared with those ≥28 weeks' gestation (20%; P < .01). For case subjects, there were no associations identified with the need for escalation of respiratory support and key patient characteristics (Supplemental Table 4).

Outcomes: Case Subjects Versus Control Subjects
Case subjects required significantly more ventilation, with a median of 7 days (IQR: 2.5-27.5 days) compared with 2 days (IQR: 0-8 days) for control subjects (P < .001). Similarly, case subjects had a greater requirement for continuous positive airway pressure at a median of 18 days (IQR: 1-37 days) compared with 5 days (IQR: 0-33 days) for control subjects (P = .026). Case subjects spent significantly more care days at the level of intensive care and high dependency; they also remained in the hospital longer (76 vs 41 days; P < .001) ( Fig 2).
Among case subjects <32 weeks' gestation, 65% (n = 40) had a diagnosis of BPD at 28 days compared with 52% (n = 71) of control subjects (P = .09), which remained not significant at 36 corrected weeks' gestation (case subjects, 52%; control subjects, 39% [P = .81]). Univariable conditional logistic regression did not identify any clinical variables associated with VRTI at the predefined statistical level (P < .1) to include in a multivariable model.

Requirement for Home Oxygen
At time of discharge, 37% (n = 34) of case subjects required home oxygen compared with 17.5% (n = 32) of control subjects (P = .002). Case subjects were also more mature,

FIGURE 1
Percentage of case subjects requiring escalation of respiratory support at the time of VRTI diagnosis grouped according to birth gestation category; n = 95 episodes of infection. BiPAP, bilevel positive airway pressure; CPAP, continuous positive airway pressure.

FIGURE 2
Median (95% CI error bars) days spent at each level of care and total hospital stay for case subjects (shaded boxes) compared with control subjects (clear boxes). ***P < .001.

In-hospital Cost of Care
The 2011 National Health Service tariff costs 23

DISCUSSION
Until recently, nosocomially acquired VRTIs in the NICU were believed to be uncommon. The present study is the largest to date and demonstrates that VRTIs in the NICU are associated with significant respiratory morbidity and increased requirement for home oxygen, and they result in more than double the in-hospital care costs before first discharge. Using a rigorous matched case-control method, the baseline characteristics of the infants were well balanced, supporting our belief of the validity of the results.
Neonatal patients often develop respiratory symptoms during septic episodes and are usually screened for bacterial infection before initiation of antibiotics. Recent studies have highlighted that either parallel screening at the time of sepsis concerns, or routine surveillance monitoring, identify VRTI in 8% to 52% of infants in the NICU. 12 -14 Our study focused on symptomatic infants with a suspected VRTI and identified that the most dominant pathogen was rhinovirus. Although other studies 13,14,19 have not found this pathogen to be the dominant finding, many of these studies were conducted before the widespread use of PCR, and their methods may not reliably identify rhinovirus. After respiratory syncytial virus infection, rhinovirus is a common pathogen and responsible for many community respiratory infections, 24 -26 and it can cause severe respiratory disease in older children requiring hospitalization. 11,27 As the dominant pathogen in our cases, rhinovirus was associated with a need for increased respiratory support in one-half of the infants, with many of the premature infants requiring additional pressure support ventilation, similar to other smaller studies. 12,14,17 This scenario results in an increase in the level of care and barrier nursing, and it requires additional nursing and medical input to manage.
Rhinovirus infection was seen throughout the year, with many of the other viruses only observed during the colder months in keeping with seasonal trends in the United Kingdom. 25 33 In addition, animal models offer evidence that viral infection, in combination with oxygen supplementation, has an adverse impact on the immune response and potential subsequent lung recovery. 34 It is interesting that, in the present

FIGURE 3
Median (95% CI error bars) health care cost according to level of care and total hospital stay for case subjects (shaded boxes) and control subjects (clear boxes). ***P < .001.
study, case subjects requiring home oxygen were significantly more mature (by ∼2 weeks) compared with control subjects requiring home oxygen. The reasons for this outcome are unclear but could reflect a more mature immune system mounting a greater inflammatory response to the viral pathogen, resulting in more lung injury.
There are few published reports on neonatal nosocomial-acquired H1N1, and our study adds to the data available highlighting the significant respiratory disease this strain causes, also evident in other outbreaks. 35 The main limitations of the present study include its retrospective nature, although the data were collected prospectively. In addition, only those infants the team believed were symptomatic were screened for VRTI, and thus the overall incidence may have been greater if asymptomatic infants were also screened. Because symptom data were not collected, the clinical diagnosis used a pragmatic approach rather than a clearly defined definition. This study is based on infants in only 2 tertiary NICUs, and the generalizability to other centers therefore must be established. However, the Nottingham NICUs during the period of study had a relatively low rate of BPD in ventilated infants 24 to 32 weeks' gestation according to the Trent Neonatal Survey, 40 with rates significantly lower than similar units during the 2010-2012 period, perhaps suggesting our study may underestimate the impact of VRTI in other centers. More detailed lung function data in infancy and childhood would also provide valuable long-term information on the effects of VRTIs during early lung development.

CONCLUSIONS
This large, well-matched, casecontrol study reported the significant respiratory morbidity and resource implications associated with nosocomial VRTIs in the NICU. Rhinovirus was the dominant pathogen, and more than onethird of the infants required home oxygen. One could hypothesize that nosocomially acquired VRTIs not only contribute to BPD in symptomatic infants but also play a role in those infants with unrecognized infection. With little change in the incidence of BPD, despite advances in respiratory management, and increasing evidence that many infants have unrecognized VRTI, this hypothesis is plausible. Large, prospective studies could incorporate surveillance programs to test our theory. Strategies such as restricting visiting could reduce the burden of nosocomially acquired VRTIs, 17 especially when the viral community load is high during pandemics.

POTENTIAL CONFLICT OF INTEREST:
The authors have indicated they have no potential confl icts of interest to disclose.