Search strategy

We will search MEDLINE, CINAHL, EMBASE, Web of Science, PsychInfo and Global Health, using Medical Subject Headings and keywords, from 1946, 1991, 1947, 1934, 1967 and 1910, respectively, to September 2012. The general search strategy is shown in box 1 and the subject heading will be adjusted for each database. We will use multiple terms to identify culture and intervention. The search will be restricted to English language articles (access to translation services is not available for the review), however we note that a recent systematic review of empirical studies on the effect of English-language restriction on systematic reviews found ‘no evidence overall of a systematic bias from the use of language restrictions in systematic review-based meta-analyses in conventional medicine.’16 In addition to searching the specified databases, to check that the database searches have not missed any studies that may be relevant to our review we will hand search the journals, JAMA, BMJ, BMJ Quality and Safety, Lancet and New England Journal of Medicine and Implementation Science, separately published during the last 5 years (from 2007 to 2012). The topic of hospital-wide interventions is broad and complex, and it is possible that relevant articles may be classified differently to the review search terms. The hand search will serve to check that our search criteria are broad enough, and that an extension of the search criteria is not required. These peer-reviewed journals were chosen as the most likely to publish studies that meet the inclusion criteria, in particular, validated patient outcomes. We will also hand search the reference lists of the relevant Cochrane systematic reviews. Two researchers will conduct the hand search; if disagreement about inclusion of a study occurs a third researcher will arbitrate.

Study selection and exclusion criteria

Under the review's inclusion and exclusion criteria, research must focus on a hospital-wide intervention, that is, mere implementation in the operation theatre, a few general units or intensive care unit is not sufficient. Other inclusion criteria include investigating the organisational factors that may affect the implementation. Studies should also provide patient outcome data before and after the hospital-wide intervention. The review will only include interventions in an acute care setting, that is, rehabilitation centres, primary health cares, ambulatory services and psychiatric facilities will be excluded. Other inclusion criteria include that the study report on empirical research, in peer-reviewed, English language and scholarly journals, as well as the abstract and full text are available. The ‘grey literature’ will be excluded as it is unlikely to yield study designs that meet inclusion criteria.

We will not limit our search to randomised controlled trial studies, since we believe observational studies and controlled before and after studies—with validated data about patient outcomes—can provide useful information to identify the organisational and cultural determinants of hospital-wide interventions.

References identified in the search will be reviewed for inclusion by two researchers. Studies will be excluded only after discussion between at least two reviewers, who will assess and agree on the inclusion and quality rating of the studies. The methodological quality of the reported research will be assessed in accordance with Cochrane Collaboration guidelines.17 The quality of the reporting of the identified studies will be assessed using appropriate critical appraisal tools, such as CONsolidated Standards of Reporting Trials (CONSORT),18 Strengthening the Reporting of Observational Studies in Epidemiology (STROBE)19 or Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA).20 All papers excluded by consensus will be depicted in a document explaining reason for exclusion. Our review will be conducted according to PRISMA guidelines20 ,21 and literature selection will be presented in a PRISMA flow chart.20 The selection criteria may limit the generalisability of study findings, however the scope of the search is appropriate to identify the majority of articles published in the peer-reviewed literature and meeting the study criteria.

Participants

Participating hospitals may include any acute care facility, including metropolitan or rural, and private or public hospitals.

Type of interventions

As noted, we will only include interventions that are hospital-wide and are associated with patient outcomes through validated data collected before and after implementation of the intervention. Also, the organisational elements of the intervention should have been explained in the study to make it qualified for our review.

Comparisons

Comparisons may include acute hospitals with similar nursing–patient ratio, size and region with no intervention.

Types of outcome measures

Patient outcomes may include death rate, the rate of adverse events, patient satisfaction and infection rate.

Assessment of risk of bias

Two reviewers will independently assess risk of bias in eligible studies as outlined in the Cochrane Handbook for Systematic Reviews17: selective outcome reporting and blinding of the research personnel to data collection and analysis. For any non-randomised trials included in the review, the authors will assess any selection bias that may lead to confounding of the outcome. Disagreement regarding assessment of risk of bias will be resolved through discussion between two reviewers. If a consensus is not reached, a third reviewer will be consulted.

Data collection and analysis

Using a standard form created for the review, one author will extract summary descriptive data from these studies. The same author will compile a tabular presentation the study participants and setting, objective, design and method, type of hospital-wide intervention, organisational/cultural factors, patient/process outcomes and findings. The second author will independently review this documentation for accuracy and completeness.

Strategy for data synthesis

If suitable data are available, a meta-analysis will be completed; however, it is likely that included studies will be heterogeneous in nature. Where trial data cannot be combined, two of the authors will conduct a narrative synthesis of the findings in accordance with the review objectives. We will discuss characteristics of the studies and stratify the results according to the type of hospital-wide interventions, organisational factors associated with them and outcomes measured.

Limitations

The review findings will be limited by the number and quality of studies identified by the search strategy. A potential limitation is in selection of the search terms. The concept of a ‘hospital-wide intervention’ is subject to classification, and it is possible that studies could be published that would meet our inclusion criteria, but are not identified by the search engines owing to the use of alternate terms or categorisation. We have attempted to ameliorate this with a hand search over the last 5 years of six prominent general medical journals that we believe are likely to publish studies relevant to our review. The hand searching provides an additional check on the reliability of the search strategy of the electronic databases and will serve to check that an extension of the search criteria is not required. By restricting the search to English language articles we are also potentially eliminating relevant studies from inclusion in our review.