2.4. Data collection and analysis

The search terms on PubMed are as follows: “Health-Preserving Sports” (and such as “Six-Character Tactic” OR “Tai Chi” OR “Five-Animal Exercise” OR “Eight-Section Brocade” OR “Muscle-Bone Strengthening Exercise”); “COVID-19” OR “Corona Virus Disease 2019” OR “Novel Corona Virus” OR “2019-nCoV”; “convalescence” OR “rehabilitation” OR “recovery”; “randomized controlled trial” OR “randomized” OR “randomly” OR “clinical trial.” According to the characteristics of the database, the comprehensive retrieval of the combined of Medical Subject Headings (MeSH) and text words was carried out. The full search strategy for PubMed is provided in Table ​Table1,1, the same strategies are used in other electronic databases.

Search strategy for the PubMed database.

Before searching the literature, all reviewers will discuss and determine the screening criteria. After the screening requirements are clearly defined, the 2 reviewers will independently review and screen the titles and abstracts yielded by the search against the inclusion criteria, and then excluded some duplicate studies or studies with incomplete information. After literature retrieval, the literature records will be imported into EndNoteX9 software for management. Any inconsistency is resolved by discussing with the third reviewer. If the full text is not available, we will try to contact the appropriate author. We chose the PRISMA flow chart to show the process of selecting literature for the entire study (Fig. ​(Fig.11).

Flow chart of the study. Adapted from Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocols (PRISMA-P). CBM = Chinese Biomedical Literature Database, CENTRAL = Cochrane Central Register of Controlled Trials, CNKI = China National Knowledge Infrastructure, CSTJ = China Science and Technology Journal Database, EMBASE = Excerpta Medica Database, MMA = Mixed Martial Arts, TCM = traditional Chinese medicine, WF = Wan-Fang Database, WOS = Web of Science.

In this protocol, we will use 95% confidence interval (CI) risk ratio (RR) to rigorously analyze the dichotomous data. And for the continuous data, mean difference (MD) or standard MD (SMD) is used to measure the efficacy of 95% CI.

The data will be extracted and recorded onto an Excel file, will include at least the following items: The title of the first author, publication journal name, year of publication, study sample size, intervention methods, intervention results, bias risk assessment, and findings. The result will be cross-checked by two reviewers, any disagreement will be resolved by consensus, and any persistent disagreement will be arbitrated by a third reviewer. We will contact the corresponding authors by telephone or email for additional information if necessary. All data will be analyzed by the Review Manager software (RevMan V.5.3).

Two reviewers will use the Cochrane Handbook for Systematic Reviews of Interventions to assess the methodological quality of each trial. The risk of bias was evaluated for each study by random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other sources of bias. A judgment as to the possible risk of bias on each of the domains will be made from the extracted information; the risk of bias is evaluated at 3 levels: low risk, high risk, and unclear risk.

We will try our best to ensure the integrity of the data. When there are missing data, we will attempt to obtain missing data by contacting the corresponding author. If the corresponding author cannot be contacted, we will remove the experiment with incomplete data.

Each outcome will be calculated and combined using the RevMan 5.3. Specific implementation was based on the current version of the Cochrane Handbook for Systematic Reviews of Interventions. If tests of heterogeneity are not significant, the Mantel-Haenszel method will be chosen for fixed effect model, and if statistical heterogeneity is observed (I² ≥ 50% or P < .1), the random effects model will be used. If heterogeneity is substantial, we will perform a narrative, qualitative summary.

If data are available, we will conduct a subgroup analysis according to different Patients characteristics, intervention method (Six-Character Tactic, Tai Chi, Five-Animal Exercise, Eight-Section Brocade, Muscle-Bone Strengthening Exercise), duration of treatment, and outcome measures.

Sensitivity analysis is used to analyze research quality, intervention method, publishing type, and so on. The trials with quality defects will be excluded to ensure the stability of the analysis results.

We will use the evidence quality rating method to evaluate the results obtained from this analysis. Five factors: bias, inconsistent, inaccurate, indirect, and publication bias; 4 evaluation levels: high, medium, low, and very low.