Background

While significant progress has been made in preventing device and procedure-related healthcare-associated infections (HAI), the threat of antimicrobial resistant organisms (ARO) and Clostridium difficile continues. In the USA, the prevalence rate of HAI was 4 % in 2011 [1], and it has been estimated that there are at least two million ARO-related infections and 23,000 deaths each year [2], resulting in $26–$33 billion additional medical costs [3]. An estimated 220,000 HAI and 8000 related deaths occur in Canada per year [4]. Healthcare-associated C. difficile and vancomycin-resistant Enterococci infections increased from 2007 to 2012, and carbapenemase-producing organisms appeared in 2010 [5]. The cost of readmissions alone due to nosocomial C. difficile-associated diarrhea is estimated to be at least $128,200 CDN per year per facility [6]. These observations highlight the need for more effective prevention and control practices and better therapy.

Outbreaks of HAI in healthcare facilities are not only serious clinical events when affecting vulnerable patient populations but are highly disruptive to care delivery. Closure of affected clinical areas typically involves suspending new patient admissions and has been used as a means of controlling HAI outbreaks [7]. However, ward closures restrict patient access to necessary care, may lead to detrimental outcomes, and can be extremely expensive to implement. Consequently, the role of ward closure in outbreak control should be better understood.

Complete ward closures are typically exercised when other outbreak measures have failed, or in the setting of highly virulent organisms, or those known to spread rapidly [8]. However, whether ward closure is a necessary control intervention is not clearly established in the literature.

A number of studies have described the use of ward closure for the purpose of outbreak control. One systematic review of worldwide HAI epidemics published in the Worldwide Database of Nosocomial Outbreaks between 1965 and 2005 found that some level of ward closure was used in 194 outbreaks, with a median closure time of 14 days and closure rate of 12.4 % [8]. Geriatric units were significantly more likely to be closed due to outbreaks compared to pediatric wards, and infectious pathogens were significantly more likely to lead to ward closure compared to contaminated medical equipment. Two specific groups of pathogens were most often associated with ward closure: norovirus (for 44.1 % of ward closures) and influenza/parainfluenza virus (for 38.5 % of ward closures).

The literature generally suggests that ward closure is a necessary control intervention as part of a bundle [9–11] versus a bundle that does not include ward closure [12, 13]. However, an analysis of a large standardized data set from 2009–2012 from the Hospital Norovirus Outbreak Reporting Scheme in the UK found that in instances where no ward closure was used, the length of outbreaks was similar to those where wards were closed but with fewer patients and healthcare workers (HCW) affected (in total and per day of outbreak) [14].

To gain a better understanding of the role of ward closure in controlling outbreaks, we systematically reviewed the published academic literature examining the use and impact of ward closure for controlling outbreaks in the acute care hospital setting. In addition to this review, we developed a web-based environmental scan survey that was distributed to IP&C practitioners and physicians at acute care sites across Canada. The present systematic review had two objectives: (1) to identify studies that describe ward closure as an outbreak control measure in sufficient detail to determine how ward closure was defined and what was done and (2) to determine the impact of ward closure on outbreak control by answering the following question: In hospitalized patients of all ages, does the use of partial or complete hospital ward closure have a significant impact on the control of an outbreak due to invasive infection or colonization by pathogenic microbes with the potential for spread, as compared to not using hospital ward closure, with or without the use of other infection control interventions and/or practices? These two questions guided the protocol development, which then informed the screening and selection process.

Methods

This review is not registered with PROSPERO.

Search strategy and selection criteria

To identify relevant references for this review, we searched the following databases with no language restrictions or other limits: Ovid MEDLINE, including In-Process & Other Non-Indexed Citations, 1946 to 7 July 2014; Ovid EMBASE, 1974 to 7 July 2014; CINAHL Plus with Full Text, 1937 to 8 July 2014; and Cochrane Database of Systematic Reviews, 2005 to May 2014. Our search consisted of selected subject headings and keywords related to the use of ward closure, combined with terms for outbreaks of infectious diseases (see Additional file 1). We also searched IndMED, using the same keywords, and LILACS, using a combination of the keywords in English and some of their Spanish and Portuguese equivalents. In addition, we searched reference lists from retrieved articles and journals, conference proceedings, and the websites of the Centers for Disease Control and Prevention, the International Centre for Infectious Diseases, and the World Health Organization.

Two authors independently reviewed the title and abstract of all articles resulting from the searches and the retrieved full texts of the relevant articles. The reviewers appraised the published full-text articles for inclusion according to the five criteria described below; articles were rejected if they did not meet all of the criteria. Disagreements during title and abstract screening and full-text review were resolved through third-party adjudication.

Only those articles that were outbreak investigation studies of hospitalized patients at acute care hospitals/facilities, including teaching and specialized institutions, were included. Studies set in a long-term acute care hospital were also included; however, studies set in a long-term care facility, rehabilitative setting, or outpatient clinic at a tertiary acute care hospital/facility were excluded. To be included, studies needed to identify ward closure (complete or partial) for at least 48 h (or length not specified) as an intervention to help control outbreaks. We defined “complete ward closure” as the application of ward closure across all beds on a ward/unit and “partial ward closure” as the application of ward closure to some, but not all, of the beds on a ward/unit. “Ward closure” included any or all of the following: no new patients admitted to the area; no transfers to other units within the healthcare facility allowed unless required for ongoing care; and no transfers to other healthcare facilities, including long-term care, with no restrictions on discharge home [14]. “Ward closure” was also assumed if the following synonyms and word variants were used: “unit closure,” “wing closure,” “partial hospital closure,” “halt new admissions,” “partial hospital closure,” “no new admission,” “closure,” “limited admissions,” “delayed admissions,” and “department closure.” Studies were also included only if a comparison intervention or another infection control intervention other than ward closure was applied and if they discussed control of the outbreak(s) under investigation as an outcome. We adopted the Alberta Health Services definition of outbreak: “the perceived, or true occurrence of more cases of a communicable disease than expected in a given area, or among a specific group of people over a defined period of time” [15]. Measures of this outcome included narrative accounts of outbreak control, number of cases of illnesses, number of colonized or infected inpatients, attack rates, relapse rates, and number of deaths attributable to the causative pathogen. Only original research studies were included, but conference abstracts were reviewed for relevance; if an abstract was deemed relevant, the corresponding author was contacted by one of the librarians for the published full text. We also excluded studies that used surveys, secondary data analysis, non-original reports, grey literature, editorials, letters, cost analyses, and reviews.

Data extraction and analysis

The included studies were systematically reviewed and relevant data was extracted from each article on the following parameters: study design, setting and population characteristics, causative pathogen(s), details of ward closure, details of other outbreak control interventions, outcomes relevant to the review, including the number of patients colonized and/or infected, and the role of ward closure for controlling the outbreak were extracted and recorded by one of the authors. Data from non-English full-text articles were extracted by a researcher who was a native or fluent speaker of the language and had knowledge of data extraction for systematic reviews. Relevant extracted data were collated in a descriptive summary and tabular format based on the findings from the parameters listed above.

We adopted Juni and colleagues’ [16] component approach to assess the quality of each study included in this review. Six evaluative criteria were adapted from components of the GRADE approach [17] and the Downs and Black checklist [18] to develop an aggregate measure for “confidence in the estimate of effect of the body of evidence,” as done by Hsu and colleagues [19] using GRADE. The first five criteria were taken from the Downs and Black’s checklist for measuring study quality and the sixth criterion was developed by the authors to assess the accuracy (reliability and validity) of the outcome measures: (1) Are the characteristics of the patients included in the study clearly described? (2) Is the intervention of interest clearly described? (3) Are the main findings of the study clearly described? (4) Were the main outcome measures used accurate (valid and reliable)? (5) Did the authors address the issue of confounding in the analyses from which the main findings were drawn? (6) Did the authors confirm cases using acceptable diagnostic methods? For each criteria, a score of “0” was assigned if the criteria was not met and “1” if the criteria was met, providing a summated score between 0 and 6 for each article, where 0–1 indicates very low quality; 2–3 indicates low quality; 4–5 moderate quality; and 6 indicates high quality.

Results

From the 2095 references gathered from all the sources searched, a total of 97 English and non-English articles in Dutch, French, German, Japanese, and Spanish were accepted for inclusion (Fig. 1).

Fig. 1
figure 1

PRISMA flow diagram

Full size image

Of the 97 included studies, 67 were case series, 14 were case–control studies, 5 were cohort studies, 5 were before-and-after studies, 5 were interrupted time series studies, and 1 was a time series study. As there were no studies that included a controlled comparison between ward closure and other interventions, the studies included in this review only allowed us to fulfill our first objective. Thus, this review purely focused on studies that described how ward closure was used as an outbreak control intervention and its impact on the outbreak.

From the details provided within the context of the setting and population, the studies were organized firstly by the organ system(s) affected and secondly by the genus of the causative pathogen within each of these organ system categories. The organ system and genus categorization lent itself very well to an additional categorization by the mode of transmission, which is the basis for infection prevention and control precautions. The organ system categories included: “gastrointestinal,” consisting of 17 studies; “respiratory,” consisting of 11 studies; and “multiple/mixed,” which includes the central nervous system, skin/soft tissue, urinary tract, eye, abdominal, and vascular or when more than one system is affected simultaneously and consisting of 63 studies. Of the studies in the third category, eight studies described predominant colonization, 12 studies described predominant infection, and 43 studies described a combination of colonization and infection. A sixth category included six studies that described the impact of infection control policies and of specific interventions on outbreak control. The modes of transmission relevant to our study included contact (both direct person-to-person and indirect via fomites and inanimate objects), droplet (via large droplets within a 1–2 m radius of the individual), and airborne (via small droplet nuclei capable of spreading over distance of greater than 2 m through the air [20]).

Gastrointestinal system (Table 1)

Table 1 Summary table for accepted studies—gastrointestinal system
Full size table

We identified 17 studies [21–37] on outbreaks involving C. difficile, norovirus, rotavirus, Salmonella panama, small round structured virus, or small round structured virus and small round featureless virus. The primary mode of transmission for all these pathogens is direct person-to-person contact and indirect contact with contaminated surfaces [20]. The outbreaks occurred at single facilities, of which four occurred at the facility-wide level, and resulted in gastrointestinal system colonization and/or infection among 3–116 inpatients. Between two and ten intervention strategies were used in conjunction with ward closure to control the outbreaks.

The definition of ward closure varied across the studies, and ward closure lasted between 3 days and 1 month among the studies that reported length of closure. Six studies defined ward closure as prohibiting new admissions to the affected clinical area (i.e., unit/ward/bay) [21–26]. Widdowson and colleagues reported on a study that utilized a phased approach, first halting new admissions and discharging all cases, then halting all admissions and discharging all patients from the area [27]. Three studies described completely stopping both admissions and transfers [28–30]. New admissions and transfers were stopped and transfers were limited in the studies by McCall and Smithson and by Stevenson and colleagues [31, 32]. In addition to stopping new admissions, transfers were limited in two studies [33, 34] and discharges were limited in one study [35]. In Hoffman and colleagues’ study, only transfers were limited [36]. One study did not specifically describe their definition of closure [37].

Of nine studies that reported achieving outbreak containment, six attributed it to all the measures used [21, 25, 26, 28, 31, 33], two studies attributed it to multiple, but not all the measures used [27, 37], and one study did not specify which measures contributed to the outcome [36]. In two studies, the reduced number of new cases of colonization and infection was attributed to all the measures used [30, 32]. In two other studies, the authors were uncertain which measures contributed to the reduced number of new cases in one [35], while authors of the other did not report which measures contributed to the reduced number of new cases [29]. Kienitz and colleagues reported that new cases continued to be identified until the pediatric ward was closed; however, newly admitted patients became infected until commercial milk was found to be the source of the outbreak [24]. In three studies, the authors did not report whether the outbreaks were controlled; however, they reported that either all or a number of the measures that were used could be effective at achieving outbreak control [22, 23, 34].

Respiratory system (Table 2)

Table 2 Summary table for accepted studies—respiratory system
Full size table

Eleven studies examined outbreaks of influenza A, parainfluenza, parainfluenza and respiratory syncytial virus, severe acute respiratory syndrome, Streptococcus pneumoniae, or Streptococcus pneumoniae and Streptococcus spp. [38–48] The primary mode of transmission for these pathogens is via the combination of both droplet through respiratory secretions and direct and indirect contact [20]. The outbreaks occurred in one to multiple wards/units at single facilities, of which two were at the facility-wide level, and one was at multiple hospitals. The outbreaks resulted in respiratory system infection and/or colonization among 7–30 inpatients; the number of affected patients was not reported in one study [38]. In addition to ward closure, one to nine other interventions were used to control the outbreaks.

The affected clinical area was closed to new admissions in six studies [39–44]. New admissions were stopped in addition to discharges in two studies [45, 46] and transfers in another [47]. Liu and colleagues reported that construction work was undertaken during closure; however, they did not specify the details and length of closure [48]. In the study reported by Owolabi and Kwolek, admissions were initially limited then completely stopped [38]. Ward closure lasted from 1 week to 2 months in six studies; the length was not clear in two studies.

In the studies that achieved outbreak containment, this outcome was attributed to all the measures used in four studies [40, 46–48], and multiple, but not all the measures used in two others [41, 43]. Outbreak containment was attributed specifically to closure in one study [39] and hand hygiene in another [42]. In three studies, the authors did not report which measures contributed to outbreak control [38, 44, 45].

Other and multiple/mixed systems: predominant colonization (Table 3)

Table 3 Summary table for accepted studies—other and multiple/mixed systems with predominant colonization
Full size table

The mode of transmission for the microbes described within this category is via contact [20]. Eight studies reported on outbreaks of Enterococcus, Escherichia coli, Klebsiella pneumoniae, or Staphylococcus aureus that resulted predominantly in colonization and involved 3–59 patients at single facilities [49–56]. Between 4 and 11 other interventions were used in addition to ward closure to control the outbreaks.

In six studies, no new admissions were permitted to the affected clinical area. Ward closure entailed limiting transfers and partial closure of four beds in the outbreak described by Delmare and colleagues [49] and restricting admissions and limiting transfers in the outbreak described by Rettedal and colleagues [50]. The length of closure ranged from approximately 3 days to 3 months. Barrett and colleagues did not specify their use of closure [51].

The authors of four studies attributed outbreak containment to all the measures implemented [50, 52–54]. Delamare and colleagues attributed control to multiple measures [49]. Barrett and colleagues attributed control to treating nasal carriers with nasal mupirocin [51], and van der Zwet and colleagues attributed control to cohorting of colonized patients [55]. Additional patients became colonized after control measures were implemented in one study [55].

Other and multiple/mixed systems: predominant infection (Table 4)

Table 4 Summary table for accepted studies—other and multiple/mixed systems with predominant infection
Full size table

The major mode of transmission for the microbes described within this category is via contact with the exception of adenovirus and the echovirus where both contact and droplet transmission occur [20]. We identified 12 studies that reported on outbreaks of Acinetobacter baumannii, adenovirus, echo 19 virus, Enterobacter cloacae, Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa, or Staphylococcus aureus that resulted predominantly in infection among 4–48 patients [7, 57–67]. The authors reported using 1–11 interventions in addition to ward closure.

Ward closure involved closing the affected clinical area to new admissions in eight studies [7, 57–63]. Admissions to the affected clinical area were limited in two studies [64]. In the study reported by Fujiwara and colleagues [65], admissions were first limited and then completely stopped. Two studies did not define the closure used during the outbreak [66, 67]. Ward closure lasted from 1 to 10 weeks in seven studies that reported the length of closure.

Authors reported achieving outbreak control in 11 studies. This outcome was attributed to all the measures used in three studies [57, 65, 67]. Successful containment was attributed to multiple measures, excluding ward closure, in two studies [57, 60] and the treatment of HCW carriers in another [63]. How outbreak containment was achieved was unknown in three studies [7, 61, 62]. Gupta and colleagues attributed the reduction of new cases to all measures instituted [59]. Kaneko and colleagues attributed control specifically to environmental disinfection [64]. One study did not identify which measure(s) contributed to control [66].

Other and multiple/mixed systems: combination of colonization and infection (Table 5)

Table 5 Summary table for accepted studies—other and multiple/mixed systems with combination of colonization and infection
Full size table

Of the remaining studies, 43 reported outbreaks that affected other or multiple organ systems and resulted in both of infection and colonization [68–110]. The major mode of transmission for the microbes described within this category is via contact with the exception of Coxsackie virus and parvovirus where both contact and droplet transmission occur [20]. The studies reported on outbreaks of the following: Acinetobacter baumannii, Coxsackie virus, Enterobacter aerogenes, Enterobacter cloacae, Enterococcus faecium, Escherichia coli, Klebsiella pneumoniae, parvovirus, Salmonella, Serratia marcescens, Staphylococcus aureus, or Streptococcus. For all but one study that involved a total of seven hospitals, the outbreaks occurred at one facility and affected a total of 3–245 patients.

Among these studies, the definition of ward closure varied widely and lasted from 1 week to 2 months in 19 studies that reported the length of closure. Ward closure was defined as limiting and then not accepting new admissions to the affected clinical area in 30 studies [68–97], limiting admissions in three studies [98–100] and limiting transfers in two studies [101, 102]. Ward closure entailed both stopping new admissions to the affected clinical area(s) and limiting transfers or discharges in four studies [103–106]. Boyce and colleagues reported that permanent closure of a burn unit was necessary to control a MRSA (methicillin-resistant Staphylococcus aureus) outbreak that could not be controlled by the use of other measures, including temporary closure on three occasions [107]. Three studies did not provide specifics of their use of ward closure [108–110].

Successful outbreak containment was reported in the vast majority of the studies. This outcome was attributed to multiple measures in five of the studies [70, 73, 92, 103, 104], multiple measures, excluding ward closure in one study [77], and to all the measures used in 13 of the studies [71, 75, 80, 81, 84, 88, 89, 95, 99–102, 105]. Other studies attributed outbreak control specifically to the closure of the affected ward(s) [69, 78, 82, 85, 109, 110], provision of dedicated and disposable equipment [72], disinfection of equipment [94, 97], construction of a cohort isolation ward outside of the affected hospital [108], disinfection of the affected clinical area(s) during closure [76, 87, 90, 98, 106], cohorting enabled by ward closure [79], and treatment of healthcare workers for carriage [93], as well as death of the infected inpatients [86]. Seng and colleagues reported not knowing which measure(s) contributed to outbreak containment [91]. While the authors of three studies reported unsuccessful containment [68, 74, 96]. Boyce and colleagues reported that permanent closure of the burn unit, the source unit, was necessary to control a MRSA outbreak on other units [107]. Moretti and colleagues reported that a combination of measures contributed to a statistically significant reduction (p < 0.001) in the number of cases of colonization and infection [83].

Studies on infection prevention and control policies or specific interventions (Table 6)

Table 6 Summary table for accepted studies—infection prevention and control policies and specific interventions
Full size table

We identified six studies that focused on the impact of specific infection prevention and control policies or a control intervention [12, 13, 111–114]. The mode of transmission for the microbes described within this category is via contact [20]. All the studies involved new policies and/or interventions that influenced ward closure prerequisites, ward re-opening criteria, and impact of alternate measures to that of ward closure on outbreak control. Recorded outcomes of the new policies and interventions include duration of closure in two studies [12, 13], bed-days lost in two studies [12, 13], and rate of new infection cases in four studies [111–114].

In two studies reporting on norovirus outbreak(s), bay closures supplemented with other measures were reported to have a greater impact on the reduction of closure length and bed-days lost than ward closure as a primary intervention [12, 13]. Although a number of other interventions were used, Garcia and colleagues attributed a reduction in the episodes and incidence density of infections to cleaning and disinfection during sequential closure of affected clinical areas [114]. In two other studies, the authors indicated that successful containment could not be achieved when ward closure was used as part of the control strategy. In their 11-year study, Selkon and colleagues found that a dedicated isolation unit with controlled ventilation was crucial to reducing the incidence rate of nosocomial MRSA infections [112]. Stone and colleagues observed a significant decrease in the incidence rates of C. difficile infection and MRSA when a new policy entailing hand hygiene, education, and restriction on antimicrobial treatment was implemented [113]. Lastly, Farrington and colleagues reported on the incidence of MRSA during the application of a MRSA control policy aimed at eradication over 10.5 years and relaxation of the same policy for the next 1.5 years [111]. The authors reported a notable increase in MRSA incidence following the relaxation period; however, the authors noted that the increase could not be solely attributed to the relaxation of the policy as there was also an increase in admission of MRSA carriers.

Risk of bias

Owing to the nature of the studies included in this review, a number of potential confounders and sources of bias were identified. Firstly, none of the studies controlled for confounding, and the majority of them did not address the confounding factor bias when discussing the impact of the interventions used. All of the studies used ward closure in combination with other interventions, and as such, the impact of each measure on outbreak containment could not be determined. Relatedly, there may also have been a potential for a dose–response relationship in studies that increased the extent of ward closure, for example, from closing the unit to select admissions to closing to all admissions. Secondly, there is the potential for selection bias in studies that did not use epidemiological typing and, subsequently, could not confirm that all affected patients were colonized or infected with the same strain of the causative pathogen. Another source of bias stems from the selection of specific outcomes. Furthermore, some of the case definitions relied on the presence of symptoms and did not confirm cases with any diagnostic method or epidemiological typing, resulting in case finding bias. The studies could have also been subject to recall bias as the vast majority of articles are retrospective. As all the articles included in this study reported on successfully controlled outbreaks, it is highly likely that the reviewed literature is vulnerable to publication bias. Many of the articles failed to address these potential sources of bias that may have contributed to the main findings and, particularly, the impact of ward closure on containing the outbreaks. This failure may be attributed to the retrospective and observational nature of outbreak investigation studies. Fourthly, definitions of ward closure were varied between studies, potentially creating bias in understanding the impact of ward closure and in determining whether the studies used complete or partial closure.

Discussion

We sought to identify studies that describe the use of ward closure as an intervention in outbreak control and determine its importance. Our systematic review expands on existing work by providing an extensive review of the epidemiological literature on the use of ward closure as an intervention to control outbreaks of pathogenic microbes among inpatients hospitalized in acute care settings. We identified 97 studies that described the use of ward closure as part of a bundled approach to their strategy. None of the studies used ward closure in a setting where it was able to be isolated as a singular control measure, limiting our assessment of the direct efficacy of ward closure on outbreak containment, which was one of our primary objectives.

It was not possible to draw any firm conclusions about the impact or effect of ward closure from the studies for a number of reasons. Firstly, the use of “ward closure” varied considerably within the papers included in the review. Our review was unable to identify whether partial or complete closure was instituted in the vast majority of the studies, as precise definitions were not used to describe the type or nature of ward closure. The results suggest that there is not a universal definition of “ward closure”; rather, ward closure refers to restrictions on patient movement into and out of a unit/ward or a facility and could encompass a number of qualities and multiple phases and/or degrees of application. Secondly, with the exception of the prevention and control policy and intervention studies, all of the studies of the included papers were reports or descriptions of outbreak investigations. As investigators could not manipulate exposures (i.e., the outbreak), all outbreak studies were observational in nature and the results were thus susceptible to a number of potential confounders. The vast majority of the included articles did not record these potential confounders or were not adjusted accordingly in any type of additional analysis. The studies were vulnerable to multiple biases, including confounding factor bias, publication bias, and recall bias, and none of them reported taking measures to prevent them or address their source. As Cooper and colleagues [115] noted, these studies generally did not meet standards of planned research as most, if not all, outbreak reports were written retrospectively. Thus, the majority of the included studies were considered to be of poor quality as the nature of outbreak investigation reports rendered the use of high-quality study designs such as randomized controlled trials unfeasible. Thirdly, all of the studies used combinations of measures in an attempt to reduce or terminate transmission. As a result, the relative contribution of each measure, and especially ward closure, could not be determined. The lack of attribution could be due to the reporting style, as many authors listed all the measures used without providing information on whether they were instituted consecutively or concurrently. Overall, ward closure was generally used at a late stage in conjunction with other measures, primarily hygienic and disinfection measures. Finally, considerable variability across the studies limits the generalizability and comparability of the outcomes of the studies. Thus, we considered the conclusions to be very weak when authors stated that the containment of an outbreak could be attributed to any one of the measures used as potential alternative factors accounting for the main findings could not be dismissed.

Our review highlights potential areas for further research on the role of ward closure as an intervention measure for managing and terminating outbreaks. Improving the quality of reporting can be a first step to addressing the difficulties in assessing the applicability and generalizability of these studies [116]. Given the complexities of outbreak investigations and the nature of the studies, clear and detailed reporting enables greater understanding of the context of the outbreak, the outbreak itself, and the control measures used, which may or may not include ward closure. Reports of outbreaks that use ward closure should include a clear definition or description of ward closure, timing of ward closure, and at which point it was used in the investigation. Given the nature of outbreak investigations, an experimental design would not be feasible. However, since the role if any of ward closure in containing outbreaks is unknown, quasi-experimental design is ethically unacceptable. Future research can improve the rigor and internal validity by using study designs of higher quality such as prospective cohort studies and cluster randomized trials. For example, a cluster randomized design study of ward closure, or no ward closure plus a defined bundle of other interventions for specified outbreak organisms, could be conducted.

Further, formal assessment of the frequency and outcomes of unit closure versus no unit closure during an outbreak could be undertaken. This should include gathering information on the type of outbreak where a unit is closed, duration of the outbreak, whether or not the unit is closed, and the impact on patient flow, examining both admission and discharge. While there are some inconsistencies in the quality and format of reporting, there are some metrics that are consistently reported, including number of beds, length of closure, and bed-days lost. This information could inform an economic study using modeling to predict the cost of implementing ward closure. Finally, there are potential ethical and legal considerations in deciding whether to implement closure of care settings during outbreaks that are not addressed in the literature reviewed nor within this review. On the one hand, failure to restrict admissions implies that new and unaffected patients are knowingly admitted to an area known to have ongoing transmission of a potential pathogen; on the other hand, closure of a clinical area may reduce access to care.

While this review was undertaken with rigor and in accordance with the requirements of systematic review methodology, it is important to note its limitations. Firstly, for the majority of articles, data were extracted by a single reviewer; however, initial screening was undertaken rigorously by two reviewers, and disagreements were resolved with a third-party adjudicator. Secondly, the literature available for this review could report a positive effect of ward closure, as it is possible that there are many outbreaks that were controlled without using ward closure and were never published. Similarly, outbreaks where interventions failed to control transmission leading to endemic transmission are less likely to be published. For example, it is common for long-term care facilities to use ward (or facility) closure (along with other interventions) to control gastrointestinal and respiratory outbreaks, and these are seldom published. While the outbreaks are generally controlled and the ward (or facility) is re-opened, the key question is whether ward closure is necessary and effective. Lastly, the review is based on the last electronic search which was completed in July 2014, and as such the review may not be entirely up to date.

It can be concluded that ward closure for containment of outbreaks remains an intervention that is not evidence based in the traditional sense; however, this review demonstrates that ward closure is frequently used and was always used as part of a bundled approach, whether as part of a sequence of, or in parallel with, other interventions, and in this sense, is similar to other public health responses. However, it was interesting to observe that in the majority of the studies in this review, ward closure was applied in the late stages of the overall outbreak response rather than as a first measure. In addition, in 16 studies, despite the use of ward closure, additional cases continued to be reported, suggesting that ward closure was not an effective intervention in these settings. Other than general wards, which were not described well, burn units (n = 3), geriatric wards (n = 3), and neonatal intensive care units (n = 2) were reported more than once (Tables 1, 3, 4, 5, and 6). The most frequently recorded mode of transmission was contact with viral gastrointestinal-associated viruses (four norovirus and one small round structured virus) and bacterial (S. panama, C. difficile, E. faecium, and E. coli), making up 56 % of the pathogenic species. These pathogens are known for their persistence within environmental niches and relative resistance to commonly used disinfectants.

There are also potential ethical considerations in the closure of wards during outbreaks that are not addressed within the context of the reviewed studies and would need to be taken into consideration by infection control personnel and hospital administrators. Admitting new and unaffected individuals to a hospital ward that is known to have ongoing transmission of a potential pathogen, particularly if associated with a high case fatality rate, warrants careful deliberation. The risk of new transmissions needs to be juxtaposed against the failure to contain the outbreak despite closure, the disruption of care delivery, and lack of access to care for other patients and overloading other care units, particularly emergency departments, where the risks of overcrowding and delayed care present other challenges.

With no published controlled studies associated with a benefit from ward closure, infection control practitioners and hospital administrators will need to continue to balance the competing risks, taking into consideration the nature of the outbreak, the type of pathogen and its virulence, mode of transmission, and the setting in which it occurs and take reasonable steps to protect patients, and since ward closure has been used in the past, it will likely continue to be used as an intervention strategy until better quality evidence is available.

Conclusions

The present systematic review could not ascertain the impact of ward closure on outbreak containment for any of the included studies based on our primary objective. Ward closure was commonly reported as an intervention during the course of a wide range of outbreaks, and outbreak control was described in most settings with the use of ward closure, usually in the late stages of the outbreak and was always used in parallel or in sequence with other interventions. Our results highlight that there is no universal definition of ward closure, as it has been defined in various and imprecise ways in the included studies. Since the published literature to date consists of uncontrolled observational study designs that were vulnerable to a number of potential confounders and biases, the actual impact of ward closure could not be determined. Our review has identified a number of research gaps and new opportunities for future investigations. In particular, the ability to determine the generalizability and applicability of ward closure as a control intervention could be improved by standardizing outbreak investigation reporting to include information on the use, role, precision of definition, and timing of ward closure.