Risk of nosocomial respiratory syncytial virus infection and effectiveness of control measures to prevent transmission events: a systematic review

Respiratory syncytial virus (RSV) causes a significant public health burden, and outbreaks among vulnerable patients in hospital settings are of particular concern. We reviewed published and unpublished literature from hospital settings to assess: (i) nosocomial RSV transmission risk (attack rate) during outbreaks, (ii) effectiveness of infection control measures. We searched the following databases: MEDLINE, EMBASE, CINAHL, Cochrane Library, together with key websites, journals and grey literature, to end of 2012. Risk of bias was assessed using the Cochrane risk of bias tool or Newcastle–Ottawa scale. A narrative synthesis was conducted. Forty studies were included (19 addressing research question one, 21 addressing question two). RSV transmission risk varied by hospital setting; 6–56% (median: 28·5%) in neonatal/paediatric settings (n = 14), 6–12% (median: 7%) in adult haematology and transplant units (n = 3), and 30–32% in other adult settings (n = 2). For question two, most studies (n = 13) employed multi‐component interventions (e.g. cohort nursing, personal protective equipment (PPE), isolation), and these were largely reported to be effective in reducing nosocomial transmission. Four studies examined staff PPE; eye protection appeared more effective than gowns and masks. One study reported on RSV prophylaxis for patients (RSV‐Ig/palivizumab); there was no statistical evidence of effectiveness although the sample size was small. Overall, risk of bias for included studies tended to be high. We conclude that RSV transmission risk varies widely during hospital outbreaks. Although multi‐component control strategies appear broadly successful, further research is required to disaggregate the effectiveness of individual components including the potential role of palivizumab prophylaxis.


Introduction
Respiratory syncytial virus (RSV) causes a significant public health burden; a systematic review and meta-analysis estimated that globally the infection caused 33Á8 million (95% confidence interval [CI] 19Á3-46Á2 million) new episodes of acute lower respiratory tract infections in children <5 years old in 2005. 1 It is an important cause of severe respiratory disease in children, particularly those at high risk of acute lower respiratory tract infections. 2,3 RSV is also common in adults, especially the elderly and other highrisk groups such as those who are immunocompromised. [4][5][6] RSV outbreaks among vulnerable hospitalised patients are of particular concern as affected patients are more likely to experience longer hospital stays, with increased risk of morbidity and mortality. 4,7,8 Numerous hospital outbreaks have been reported in multiple age groups and settings including neonatal intensive care, haematology, transplant and oncology units. [9][10][11][12] RSV infection does not lead to long-term immunity. 13 There is currently no specific treatment for RSV nor a licensed vaccine, 14 so controlling transmission is crucial. RSV is transmitted via large nasopharyngeal secretion droplets from infected individuals. 15 These droplets enter via the mucus membranes of the eyes, nose and mouth following close contact, or self-inoculation after touching contaminated surfaces. 15 Standard (routine) respiratory infection control procedures such as isolation of cases, high standards of hand hygiene, cohort nursing and use of personal protective equipment (PPE) have been reported as effective, to varying degrees, in the prevention and control of RSV outbreaks in nosocomial settings. [16][17][18] In more recent years, immunoprophylaxis with the monoclonal antibody palivizumab has been used during hospital RSV outbreaks for patients at high risk of severe complications (e.g. preterm infants). [10][11][12]19,20 To our knowledge, the effectiveness of RSV-specific infection control measures in the hospital setting have not been subject to a high-quality systematic review. A Cochrane review of physical interventions to prevent respiratory virus infections was published in 2011, but this was not specific to RSV or acute settings and did not seek to identify studies reporting on the effectiveness of palivizumab. 21 We aimed to address the aforementioned gaps in the evidence base through a systematic review of the published and unpublished international literature. The specific research questions were as follows: (i) What is the risk of nosocomial RSV transmission where patients may have been potentially exposed to the infection (epidemiologically suspected or microbiologically confirmed) during an outbreak in a hospital ward or unit? and (ii) What is the effectiveness of infection prevention and control measures to minimise nosocomial transmission of RSV in the hospital setting?

Protocol registration and study conduct
The review protocol was registered with the PROSPERO International prospective register of systematic reviews, registration number: CRD42013003835. 22 It was conducted following the general principles of the Cochrane Handbook for Systematic Reviews of Interventions, 23 and is reported according to the requirements of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). 24

Search strategy
We searched the following databases using MeSH and freetext terms: MEDLINE, EMBASE, CINAHL and the Cochrane Library (CENTRAL). Bandolier and the Cochrane Library (CSDR, DARE, NHS HTA databases) were searched for evidence-based reviews, and NHS Evidence (NHS Clinical Knowledge Summaries and the National Library of Guidelines) to identify guidelines containing relevant data. Two of the most relevant journals (Influenza and Other Respiratory Viruses and Eurosurveillance) were hand-searched. Additional searches were conducted via Google, the Health Protection Agency website (now Public Health England [PHE]), the World Health Organization and the US Centers for Disease Control and Prevention. Experts in the field were consulted. Grey literature was sought via the Web of Science, NHS Evidence and OpenSIGLE. Reference lists of the most relevant records (~100) were searched, and Web of Science (Science Citation Index) and Google Scholar were used for citation tracking. Unpublished epidemiological data were sought from the PHE respiratory outbreaks database.
Our search strategy was designed to identify studies providing data addressing either or both research questions. The generic list of search terms is available in the protocol, 22 and the full electronic search strategy for MEDLINE in Appendix 1. Searches were executed in February/March 2013 and included publications from the inception of each database to the end of 2012, in the English language. Searches were limited to humans. No restriction was placed on either the publication type (e.g. abstracts, unpublished works etc. were eligible) or study design. Review papers were not eligible for inclusion but were obtained for reference list searching.

Study selection (inclusion and exclusion criteria)
Search records were imported into Endnote. After removal of duplicates, records were assessed for eligibility using a threestage sifting process sequentially reviewing titles, abstracts and full texts. Each record was independently assessed by two reviewers with the involvement of a third reviewer to resolve disagreements. To be eligible, studies had to address at least one of the two research questions. Only studies conducted in hospital settings were eligible. We were interested in clinically suspected RSV or bronchiolitis, or microbiologically confirmed RSV, epidemiologically suspected to be nosocomial in origin. We accepted any description of a nosocomial (rather than community) transmission, according to the authors' own definition, as eligible. No restriction was placed on laboratory technique for identifying RSV.
To be eligible for inclusion for question one, studies had to (i) provide data on the risk of nosocomial RSV transmission to patients (attack rate), defined as follows: number of nosocomial RSV or bronchiolitis cases/number of patients potentially exposed to RSV in the ward or unit; and (ii) be conducted during an outbreak in a hospital ward or unit, defined as two or more cases of RSV infection linked epidemiologically in time and place or microbiologically confirmed. Studies providing data for a whole RSV season or routine surveillance data were not eligible. Research question two was defined as follows: Population: patients, staff or visitors at risk of RSV infection in the hospital setting; Intervention: RSV infection control measures; Comparator: infection control measures which differ from the intervention, or no intervention; and Outcome: nosocomial RSV transmission in the intervention versus comparator group. For question two, studies had to (i) state one or more RSV infection control interventions, (ii) utilise a comparator group and (iii) provide data on nosocomial RSV transmission in the intervention versus comparator groups, with no restriction placed on the type of data that were reported (e.g. risks or rates, risk or rate ratios, the ratio of RSV cases that were nosocomial). For question two, studies were not restricted to those conducted in the context of a specific outbreak (e.g. routine surveillance data comparing two RSV seasons were eligible). Studies which assessed the use of palivizumab to prevent RSV outbreaks were eligible for inclusion. Assessing the effectiveness of season-long palivizumab prophylaxis for individual high-risk patients or severity of RSV infection was beyond the scope of this systematic review.

Data extraction
Two reviewers independently extracted data using a predefined template. Disagreements were resolved by discussion or by a third reviewer. For research question one, we extracted the following: country and year of outbreak, hospital setting, study objective and nosocomial transmission risk (attack rate, number of nosocomial cases, number of patients at risk). For research question two, we extracted the following: country and year of outbreak, study design, hospital setting, infection control measures for intervention and comparator groups, and information on effectiveness of control measures.

Risk of bias assessments
We assessed risk of bias, by domain, for all studies providing comparative data on the effectiveness of infection control interventions (i.e. addressing research question two). Experimental and prospective cohort studies were assessed using the Cochrane risk of bias tool, 23 and retrospective cohort studies using the Newcastle-Ottawa scale. 25 Abstracts, conference posters or proceedings were not assessed formally due to the limited information available.

Data synthesis
A narrative approach was used to synthesise the extracted data and quality assessments according to the framework described by the Economic and Social Research Council and recommended by the University of York Centre for Reviews and Dissemination. 26 Due to weak study designs and heterogeneity between studies, including the range of different control measures applied with most studies implement-ing multicomponent measures, it was not considered appropriate to carry out a meta-analysis.

Included studies
The searches returned 16 558 records, 6913 after removal of duplicates, with an additional six studies obtained through reference list scanning. Forty studies were eligible for inclusion, 19 addressing research question one and 21 addressing research question two (none addressed both) ( Figure 1). One outbreak recorded in the database held by the PHE Respiratory Diseases Department met the eligibility criteria for research question two. Table 1 summarises the 19 studies providing data on the risk of nosocomial RSV transmission. Eight were from Europe, six from the United States and five from elsewhere. Most (n = 14) were in neonatal/paediatric units (13 of which were neonatal units), 19,20,[27][28][29][30][31][32][33][34][35][36][37][38] three in adult units for haematological cancers and/or bone marrow/stem cell transplant recipients (hereafter referred to as immunocompromised adults), [39][40][41] and two in other adult units (a psychiatric ward and a continuing-care ward for the elderly). 42,43 In all outbreaks, either all or the majority of diagnosed RSV cases were laboratory confirmed. The extent of case searching varied. The number of persons at risk ranged from 9 to 60 in neonatal/paediatric settings,  in adult units for immunocompromised adults and 25-27 in other adult units. RSV transmission risk varied by hospital setting: from 6% to 56% (median: 28Á5%) in neonatal/paediatric settings, 6-12% (median: 7%) in units housing immunocompromised adults and 30-32% in other adult care settings. All studies utilised at least some type of infection control measures (either in place prior to the outbreak or implemented in response to it). The outbreak reported in the PHE database was in an adult haematology unit that included bone marrow transplant recipients. There were 20 persons at risk and the RSV transmission risk was 30%. Table 2 summarises characteristics of the 21 studies addressing research question two. Four were from Europe, 15 from the United States and two from Canada. There were 13 experimental or prospective cohort studies 15,16,18,[44][45][46][47][48][49][50][51][52][53] and seven retrospective cohorts 54-60 (one was an abstract only and there was not enough information to identify the study type). 61 Most (n = 18) were conducted in neonatal/paediatric settings, 15,16,18,[44][45][46][47][48][49][50][51][52][53][55][56][57]60,61 with three in units housing immunocompromised adults. 54,58,59 All studies employed laboratory confirmation of RSV diagnoses.

Effectiveness of control measures to prevent transmission events
Studies on the effectiveness of interventions to prevent nosocomial RSV transmission to patients and staff are summarised in Tables 3 and 4, respectively. We found no eligible studies on interventions to prevent transmission to visitors. A range of different outcome measures were reported in eligible studies including the nosocomial transmission risk before and after the intervention, the rate of transmissions (e.g. per number of patient-days at risk) in the intervention versus control group, and risk or rate ratios.

Risk of bias assessments
Cochrane risk of bias assessments were carried out for the 13 experimental or prospective cohort studies (Table 5). For domains relating to selection and performance bias (random sequence generation, allocation concealment, blinding of participants and personnel), the risk of bias was deemed high for all but one study in which the risk for the first two domains was unclear 50 . No studies reported blinding of participants and personnel (that would largely not have been possible due to the nature of the interventions). Detection bias (blinding of outcome assessors) was considered low risk for all studies because RSV cases were laboratory confirmed. None of the studies sufficiently adjusted for potential confounding. An additional potential bias is the ascertainment of community-acquired rather than nosocomial RSV cases. Eleven of the 12 studies investigating the risk of RSV transmission to patients provided a clear definition of a nosocomial case, with most (n = 8) defining this as a case occurring at least 5 days after hospital admission (some used a higher cut-off). 15,18,[45][46][47][48][49]51 In studies assessing the risk of    25 (Table 6), most scored highly on selection of the study groups with six studies 54-58,60 awarded three or more stars for this domain. However, all studies received zero stars for 'comparability' as none adjusted for confounders or utilised a study design that matched individuals in the intervention and comparator groups. Although all but one study provided a clear definition of a nosocomial RSV case, 59 ascertainment of the outcome was poor overall (generally due to insufficient follow-up or inadequate reporting of this); six studies were awarded one star only for this domain [54][55][56][57][58]60 and one awarded zero stars. 59 Multicomponent interventions (Table 3) Most studies (n = 13) employed multicomponent infection control strategies (two or more measures combined). 16,18,46,48,49,[51][52][53][54]56,[59][60][61] These comprised a wide range of measures including the following: prompt RSV casefinding among symptomatic patients; screening all patients on admission; screening staff and/or visitors; isolation policies and/or staff/patient cohorting; restriction of visitors (e.g. no young children); and staff training and/or compliance monitoring. Most studies made some use of personal protective equipment (PPE) (e.g. gowns, gloves, masks, goggles). Studies of multicomponent control measures essentially used 'standard infection control precautions', that is usual practice (either explicitly stated or assumed, see Table 3), as the comparator group. For example, data for previous RSV seasons prior to the introduction of the intervention were frequently utilised. Nine of the 13 studies presented evidence that nosocomial infections were significantly lower when a multicomponent intervention was implemented and provided supporting statistical data (e.g. P-value or risk ratio with confidence intervals). 16,18,46,48,49,51,54,56,60 Three studies reported data that were suggestive of a beneficial effect but did not present any supporting statistics (such as a P-value). 53,59,61 Relative risk reductions in transmission were variable but tended to be quite substantial and were in excess of 50% for the majority of studies, where calculable. One study using a multicomponent intervention also provided information on transmission to staff, although the risk was actually somewhat higher during the intervention than control period (56% versus 42%, no Pvalue presented). 51 However, Langley et al. 52 compared data for nine different hospitals using different combinations of intervention measures (all included isolation or cohorting) and concluded that RSV transmission to patients was not reduced by any type of isolation policy used, and there was no beneficial effect of a gloving or masking policy. . .'.
**Most of these measures were applied in response to the outbreak (although some may have already been in place prior).

275
Staff personal protective equipment (Tables 3 and 4) Five studies examined the use of staff PPE in addition to standard precautions, all conducted in neonatal/paediatric settings. 15,44,45,47,50 Two examined the effect on RSV transmission to both staff and patients, 15,45 two examined the risk of transmission to staff only, 44,50 and the remaining study looked only at transmission to patients. 47 Of the three studies providing data on patients (Table 3), 15,45,47 two reported PPE to be effective. Gala et al. 45 implemented an eye-nose goggle for all staff and reported a transmission risk of 43% in the control and 6% in the intervention periods (v 2 : P = 0Á04). Leclair et al. 47 examined the effect of monitoring staff compliance with PPE where it was hospital policy for staff to wear gloves and gowns when in direct contact with RSV patients. This study reported a lower risk of nosocomial transmission during the period in which intensive staff compliance monitoring was implemented compared to the pre-intervention period (relative risk adjusted for intensity of exposure: 2Á9 [95% CI: 1Á5-5Á7]). However, authors of the third study reported that they found no significant difference in the rate of nosocomial RSV transmission to patients when gowns and masks were used (32%) compared with standard procedures alone (41%) (no P-value provided). 15 Of the four studies examining the effectiveness of measures to prevent RSV transmission to staff (Table 4), 15,44,45,50 two found evidence of effectiveness, both of which utilised goggles (one used goggles and masks 44 and the other an eye-nose goggle 45 ).
In the two studies reporting no statistically significant benefit, neither used goggles (just gowns and masks), although it should be noted that in both these studies the risk of transmission (based on the point estimates) was still lower in the intervention than the comparator groups. 15,50 RSV prophylaxis (Table 3) Only one eligible study, in a neonatal unit, reported on postadmission RSV prophylaxis for patients. 57 Standard infection control procedures in the unit involved placing infected infants in single rooms or cohorting them and using droplet/ contact isolation measures. The authors reported no significant difference in nosocomial RSV infection rates during the RSV seasons in which RSV prophylaxis (RSV-Ig or palivizumab) was given monthly to all high-risk infants in the unit in addition to standard infection control procedures: rate ratio for period 1 (no prophylaxis) versus period 2 (RSV-Ig): 0Á67 (95% CI: 0Á03-14, P = 0Á76); and for period 1 (no prophylaxis) versus period 3 (palivizumab): 3Á3 (95% CI: 0Á16-68, P = 0Á37). However, the point estimates indicated a potential beneficial effect of palivizumab, and it should be noted that the power to detect a statistically significant difference was likely low due to the very small number of cases. 57 Other interventions (Table 3) One study compared the RSV transmission risk in wards comprising mainly of individual cubicles with the risk in open wards combined with a smaller number of cubicles. Although the numbers of nosocomial infections were too small to make statistical comparisons, the rate of nosocomial RSV infections was somewhat lower in wards composed largely of individual cubicles (7Á1 versus 4Á2 infections per million susceptible days per infective day). 55 Meanwhile, in a haematology-oncology ward, isolating all patients hospitalised during an RSV season resulted in statistically significantly lower RSV transmission than the previous policy of only isolating patients with severe neutropenia or symptoms of upper and/or lower respiratory tract infection (relative risk in intervention versus control period: 0Á09 [95% CI: 0Á02-0Á38]). 58

Key findings
To our knowledge, this is the first systematic review of nosocomial RSV transmission risk and the effectiveness of infection control measures to prevent transmission in acute care settings. Nosocomial RSV transmission risk is

≥0Á2
RSV, respiratory syncytial virus; PPE, personal protective equipment. *Note that interventions were applied in addition to the 'standard precautions' used for the control group. **Or other measure of nosocomial RSV transmission as specified. ***Note that these two studies also appear in Table 3.
substantial during outbreaks in hospital settings, with notable variations across outbreaks and settings. Most studies in this review were conducted in neonatal/paediatric settings where the median transmission risk was 28Á5%. While all studies utilised at least some infection control measures, the identified transmission risks are concerning and highlight a serious challenge. This review highlights the lack of highquality studies describing the effectiveness of infection control measures to prevent nosocomial RSV.
The majority of studies implemented multicomponent interventions, as appropriate for infection control in hospital settings and in accordance with, for example, UK guidelines. 62 The evidence presented here broadly supports the use of multicomponent measures which tended to achieve relative reductions in transmission risk of over 50%. However, within this context, it was not possible to assess the effectiveness of individual components of control measures, either within individual studies or at the review level. Some components may be more effective than others, and identifying these could result in more efficient use of resources and further reductions in transmission risk. Furthermore, the potential harms of interventions, particularly those without measurable benefit, should not be overlooked.
Personal protective equipment worn by staff may potentially prevent transmission from patients to staff and vice versa. Two studies in which staff eye protection was used (eye-nose goggle 45 or goggles plus masks 44 ) found this to be effective in preventing transmission to staff (the first also reported a reduction in transmission to patients, but this was not investigated in the second study). This finding is consistent with RSV transmission generally occurring through the eye or nose. 63 Evidence for the effectiveness of gowns and masks was lacking in two studies 15,50 although a further study did find high (versus lower) compliance with gloves and gowns to be effective at reducing nosocomial RSV transmission. 47 In relation to the transmission of RSV to   58 may make them impractical to implement in many settings, especially in low-and middle-income countries.
Although palivizumab is considered to be effective in preventing RSV-related hospitalisation in high-risk children (outside the scope of this review), [64][65][66] we uncovered limited evidence regarding its use in hospitalised patients to prevent nosocomial RSV transmission. Our literature search returned 12 studies; however, only one of these met the review eligibility criteria for research question two. 57 Of the 11 other studies (all of which lacked a comparator group), all but one were conducted in NICUs. Eight reported no further RSV cases following palivizumab prophylaxis, 10,11,28,32,33,38,67,68 and one reported no further cases after day 14 of the outbreak having instigated control measures on day nine. 30 Two reported the occurrence of two further cases after palivizumab administration. 19,69 Meanwhile, Silva et al. 20 documented the occurrence of 10 RSV cases following prophylaxis administration to all patients, although the authors noted these infants may have already been in the RSV incubation period when palivizumab was administered. These additional data underscore the need for high-quality studies in hospital settings to generate robust evidence to support clinicians and public health policy, particularly bearing in mind the high cost of palivizumab. 70 Limitations RSV is a significant problem across low-and middle-income countries. 1 However, the majority of evidence on the risk of RSV transmission, and all evidence on interventions to interrupt transmission in this review came from the United States and Europe. Extrapolation of our findings to low-and middle-income countries where resources are lacking may be difficult.
To calculate the nosocomial RSV attack rate, the denominator was the number of persons at risk of infection as few studies provided the person-time at risk. Although this is a crude denominator, it allowed for comparisons across studies. The definition of a nosocomial transmission event varied between studies and we accepted any description of a nosocomial event, as defined by the author, as eligible. This may have resulted in some misclassification between nosocomial and community-acquired RSV. Additionally, patients infected with RSV in hospital who did not develop symptoms until after discharge, were likely not identified by the studies, especially if symptoms were mild and they did not require re-admittance to hospital (most studies did not report following up patients after discharge). Also, we cannot discount the potential for under-reporting of nosocomial RSV outbreaks leading to reduced external validity of our findings.
We did not identify any randomised controlled trials on the effectiveness of RSV infection control measures. Observational studies are subject to inherent biases and furthermore, assessing the risk of bias in non-randomised studies is difficult in itself. 23 On the whole, available studies were assessed as having a relatively high risk of bias. A number of the studies utilised comparator data from a different time period (such as prior RSV seasons) and thus are subject to confounding due to differences between the population groups and levels of exposure to RSV. Studies typically did not clearly report the population characteristics of the two groups or control any potential differences, thus making comparisons difficult and potentially subject to bias. Few studies monitored compliance with infection control measures, which has been reported to frequently be suboptimal. 71 High levels of compliance may be necessary for certain infection control measures to be effective. Finally, it should be noted that a number of the studies were poorly reported with a lack of clarity, for example, with regard to the study population and type/timing of interventions. Reporting of future studies in line with the ORION (Outbreak Reports and Intervention Studies Of Nosocomial infection) statement would improve their usefulness. 72 Conclusion RSV transmission risk varies widely during hospital outbreaks. Although there is a lack of high-quality evidence, multicomponent control strategies appear broadly successful, while PPE interventions using eye protection appear more effective than those using gowns and masks. Further research is required, especially in low-and middle-income countries, to identify the most effective and cost-effective individual control measures including the potential role of palivizumab prophylaxis during nosocomial outbreaks.
The study was supported by the NIHR HPRU in Evaluation of Interventions at University of Bristol and also by the University of Nottingham. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, the Department of Health or Public Health England. BCM, SR and CRB designed and executed the searches. CEF, BCM, CC, SR, KP, the Noso-RSV study group, JPTH and CRB contributed to study selection, data extraction and risk of bias assessments. CEF, CC, JPTH and CRB were involved in data analysis. CEF drafted the manuscript. All authors contributed to the interpretation of the data and writing of the final manuscript and approved it for submission.

Disclosure and competing interest statement
CEF, BCM, CC, SR, KP, RP, JPTH: none to declare. JSN-V-T and CRB are respectively Editor-in-Chief and Associate Editor for Influenza and Other Respiratory Viruses; however they played no role whatsoever in the editorial process for this paper, including decisions to send the manuscript for independent peer-review or about final acceptance of a revised version. All of the above functions were handled alone by Dr Alan Hampson.