2.1 Study design and program registration
This design was registered in the international prospective register of systematic reviews (https://www.crd.york.ac.uk/PROSPERO/ID = CRD42020146932). The detail of this protocol was conducted follow the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-analyses Protocols (PRISMA-P) statement [22]. Since this study is a secondary study of the literature, formal ethical approval is not applicable.
2.2. Including criteria
2.2.1. Types of studies
Only the clinical randomized controlled trials (RCT) of oral care interventions will be included in this study. The quasi-RCTs and non-RCTs will be excluded. If there are cross-cover randomized controlled trials, the first period will be included as a parallel group trial.
2.2.2. Types of patients
ICUs patients did not receive mechanical ventilation, without infection of the lower respiratory tract at admission will be included. Hospital-acquired pneumonia must be diagnosed as following: temperature >37.8°C, and chest radiograph, cough or subjective dyspnea. If ICUs patients with and without mechanical ventilation are involved in a study, only these RCTs with over 50% of patients without mechanical ventilation will be included in this review. If different setting patients such as rehabilitation unit, nursing home, community were included in RCTs, only these RCTs with over 50% of patients without mechanical ventilation from ICUs will be included in this study.
2.2.3. Interventions
Patients in the experimental group received clearly-defined oral care procedures, including decontamination of oropharyngeal cavities with antiseptics, oral and pharyngeal cavity rinse, nurse-assisted tooth brushing. Patients in the control group received no treatment, ‘usual care’, or placebo. Studies which compared different types of oral care will not be included. These studies in which oral care used as one part of whole treatment protocols will be excluded.
2.2.4. Outcomes
2.2.4.1. Primary outcomes
The incidence of nosocomial pneumonia defined as the primary outcome of this study. Nosocomial pneumonia was defined as an infection of the lower respiratory tract that is diagnosed at least 48 h after the patient has been admitted to hospital, which is not present or incubating at the time of hospital admission [23].
2.2.4.2. Secondary outcomes
Other outcomes such as mortality, 30-day mortality, duration of ICU stay, oral health indices (including periodontal index, plaque index, bleeding index, gingival index, etc.), the usage of antibiotic, adverse effects of the intervention, economic data were defined as secondary outcomes.
2.2.5. Search strategy
2.2.5.1. Online electronic databases
Four English online electronic databases including Embase (via embase.com), Medline (via PubMed), CINAHL (via EBSCOhost), and Cochrane Central Register of Controlled Trials (CENTRAL) will be systematic searched without language restrictions from their inception to December 31, 2020. Four Chinese-language databases include WanFang Database, Sino-Med Database, Chinese Science and Technology Periodical (VIP) Database, and China National Knowledge Infrastructure (CNKI) database will be searched from their inception to December 31, 2020. The English terms were used individually or combined “intensive care” “nosocomial infection” “oral care” “mouth care” and the Chinese searching terms were “zhong zheng jian hu (intensive care),” “yi yuan huo de xing fei yan(nosocomial infection),” “ kou qiang hu li oral care)”. The search strategy we have built for Medline via PubMed was showed in Table 1, which was performed according to Cochrane Handbook for Systematic Reviews of Interventions [24]. Some adaptive changes will be made when searching in other databases.
2.2.5.2. Additional resources
Although there are no full-text articles in Chinese or English, articles with Chinese or English titles or abstracts will also be screened.
2.3. Data collection and analysis
2.3.1. Research management and screening.
The literature management software, EndNoteX7, will be used in this study to manage all records collected. Before two independent reviewers read the title and abstract of the trials, duplicate records will be removed from the literature database. Then obviously unqualified records will be identified. For potentially eligible records, full text of these records will be read by two independent reviewers. Finally, qualified literature was selected according to the pre-determined inclusion criteria. All records included must fit the type of study, type of patients, intervention, design (PIS) strategy of this review. Original author will be contacted via email when there is more information is needed to make a decision. If there is any disagreement, the final consensus is generated through discussion with a third reviewer. Details of the literature screening will be reported following the PRISMA flow diagram (Fig. 1).
2.3.2. Data extraction and management.
First, all data will be extracted according to the data table previously prepared by one researcher. Then all data extracted will be verified by another researcher. The information we will extract is multifaceted. The information will be extracted from each included trial as following: (1) Basic information of studies including the journal, publish year, author, author institutions, title; (2) The characteristics of participants of RCTs including sex, age, inclusion/exclusion criteria; (3) Intervention used in groups, including the intensity, frequency, and period; (4) The methodological information of each trial including random sequence generation, allocation concealment, blinding of participants, and blinding of outcomes assessment; (5) Outcomes including instruments, drop-out, follow-up, and adverse events. If there is any missing data, original author will be emailed to ask for it. The measurement data will be described using mean and standard deviation (SD) or standard error (SE). The count data will be described using the number of events. The two arms’ data that most fit our aims of this review will be extracted rather than all arms if there are more than two arms in study. Before entering the data into the RevMan5.3 for analysis, necessary data conversion will be performed via spreadsheet software (Microsoft Excel). If there is any disagreement, the final consensus is generated through discussion with a third review author.
2.4. The risk of bias assessment for included studies
The quality of included studies will be assessed using “Risk of bias” (ROB) table, which was recommended by Cochrane Handbook for Systematic Reviews of Interventions[24]by two reviewers independently. This tool is a two-part tool, addressing the following six specific domains when defining the quality of an RCT: (1) sequence generation; (2) allocation concealment; (3) blinding of participants, personnel and outcome assessors; (4) incomplete outcome data; (5) selective outcome reporting; (6) Other sources of bias. Each domain consists of one or more specific entries. Within each entry, the first part is to describe what happened in the study. The second part is a judgment of the risk of bias for that entry. This is achieved by answering a pre-specified question for each entry. In all cases, if there is insufficient information about the part description, the judgment as will usually be “unclear” risk of bias. An answer “No” indicates high risk of bias and an answer“Yes” indicates low risk of bias. If there is any disagreement, the final consensus is generated through discussion with a third review author.
2.5. Handling of missing data
If there is any missing data, original authors will be contacted to request it. If the missing data is not obtained, the analysis will be performed only using the available data. Effect of missing data on the final results will be discussed in the discussion section.
2.6. Assessment of heterogeneity
Heterogeneity across trials will be detected by χ2 test with a 0.10 level as the cut-off value (P<0.1). Heterogeneity across trials will be quantified using I2 statistic. Studies with an I2 value of more than 75% will be considered to have high degree heterogeneity. If the included studies have good homogeneity, the overall effect will be synthesized [25]. Otherwise, the sources of heterogeneity will be explored via subgroup and analysis meta-regression [25].
2.7. Evaluation of publication bias
If there are more than ten studies in an outcome, a funnel plots will be structured to funnel plot used to investigate publication bias [25]. Asymmetry in the funnel plot indicates possible publication bias.
2.8. Data synthesis
If there are sufficient studies focusing on similar comparison and the same outcomes, meta-analysis will be undertaken using fixed-effect model or the random-effect model dependent on heterogeneity results. Otherwise, only the qualitative analysis will be carried out.
2.9. Subgroup analyzes
Subgroup analysis will be undertaken according to whether toothbrush was used or not. Subgroup analysis will also be undertaken according to the different types of mouthwash.
2.10. Sensitivity analysis
The studies that with a low methodological quality adversely affect the strength of the evidence, so sensitivity analysis will be performed to investigate the effect of these trials on the evidence. It will be performed via excluding a study and comparing the results changes. The results will be reported and discussed the discussion part.
2.11. Grading the quality of evidence
The quality of evidence for outcomes will be evaluated according to the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system, which rates the quality the quality of evidence into four levels (high, moderate, low, very low levels) [26].