This was a parallel, randomized controlled trial (RCT) in which participants were randomly allocated into either the Baduanjin group or the control group with an equal allocation ratio. Our study strictly complies with the guidelines of the CONSORT 2010 checklist [15].

The Guangdong Provincial Hospital of Chinese Medicine’s Human Research Ethics Committee approved this study(B2017-139-01). Also, we registered the study on the Chinese Clinical Trial Registry (Registration No.ChiCTR1800015011). It complies with the principles outlined in the Helsinki Declaration and the relevant regulations of the national regulatory agency.

From November 2017 to May 2019, participants were recruited from hospitalized patients in the ICU of The Second Clinical College of Guangzhou University of Chinese Medicine Eligibility criteria included patients who were diagnosed with sepsis, aged 18 to 80 years, and they were given NIV on admission; The delirium assessment (CAM-ICU) was normal. Inhalation oxygen concentration (FiO2) < 0.60; Respiratory rate < 30 bpm; Muscle strength ≥ 4; The primary disease and treatment plan were clear, and endotracheal intubation was not expected; Sepsis diagnostic criteria was a reference to the Third International Consensus Definitions for Sepsis and Septic Shock [16]. Exclusion criteria included participants with prior neuromuscular disease, including severe heart failure (NYHA cardiac function class IV or LVEF ≤ 35%), Guillain-Barre syndrome, muscle weakness, cardiogenic shock, and poor compliance. Written informed consent was obtained before the start of the study.

In a 1:1 allocation ratio, patients were randomly assigned to either the Baduanjin or the control groups. Biostatisticians used the R statistical programme to produce a random sequence to allocate assignments. These assignments were placed in sealed, opaque envelopes with date and signature labels placed over the seals. The trial coordinator unsealed the sequential randomization envelopes and informed the people of their group assignments following the baseline evaluation of patients. In our study, participants could not be blinded, but the researchers who collected the data and statistics were independent and uninformed.

The control group was given routine rehabilitation exercises in the ICU, including the passive movement of limbs, active activities, and functional training. The participants were instructed to do bed activities, including forwarding bending, backward stretching, adduction, abduction, and pronation, for 15–20 min every day.

Participants in the Baduanjin group were given SBE on the base of the control group until they were transferred out of the ICU. The SBE has been developed by our team based on the characteristics of patients with Severe patients in ICU (Supplementary Appendix Video 1). It is composed of 8 types. Each type is done 4–8 times for 12 min each time. Patients exercise daily at 09:00 and 16:00 for 15–20 min each time depending on their physical strength. Two certified instructors were engaged to provide instruction and supervise the participants’ training.

Treatment will be discontinued if any of the following occurs: (1) The patient felt laborious, dizziness, sweating, fatigue and severe dyspnea, SpO2 < 90%; (2) Systolic blood pressure < 90 or > 200mmHg, mean arterial pressure < 65mmHg; (3) The FiO2 of NIV was greater than 60%;(4) The patient’s respiratory rate was greater than 35 times /min.

The primary outcome was the Medical Research Council (MRC) score. Secondary outcomes included changes in Barthel Index (BI), the duration of NIV, length of ICU stay, length of total stay, ICU hospitalization expense, Total hospitalization expense.

Muscle strength was evaluated by the MRC score which was measured at baseline, the 3rd and the day of transfer out of ICU. The MRC score is a 6-level muscle strength assessment method, with a score of 0 to 5 in each level. Muscle strength of bilateral upper limbs (wrist extension, elbow flexion, shoulder joint abduction), and bilateral lower limbs (dorsiflexion, knee extension, hip flexion) were evaluated [17].

The Barthel Index is an interview-based method for assessing to assess participants’ activities of daily living (ADL) [18]. It consists of ten items: feeding, bathing, grooming, clothing, bladder control, bowel control, toilet usage, moving, transferring, and stair climbing [19, 20].The full score is 100 points, < 20 points mean serious functional impairment, life is completely dependent; A score of 20–40 indicates a great need for help in life; A score of 41 ~ 60 indicates life needs; >60 points means basic self-care.

MRC score was used as the primary outcome in this study.According to previous research [21], the control group can increase the average score of MRC by 12 points. Baduanjin can increase the average score of MRC by 24.5 points, and the standard deviation is about 20. According to the formula, \(\text=[(\text^ + \text^)(\text/\Delta)]^2\). Each group’s sample size is around 41 cases for a total of 82 cases in the two groups. Given the dropout and loss of follow-up situation, it is projected that a 15% dropout (loss of follow-up) will necessitate 96 instances.

Data were reported as mean and standard deviation for continuous variables and as percentages for categorical variables. All data were analyzed by a blinded statistician using SPSS 22.0 software with significance set at an alpha level of 0.05. Independent t-tests, chi-square(χ2) tests and Kruskal-Wallis tests were used to examine the homogeneity of demographic and clinical information between the two groups at baseline.Repeated measures analysis of variance (ANOVA) was used to test for change over time (i.e.,baseline, Day3, and Out of ICU) in the MRC for the Baduanjin group and control group.Group comparisons were made by the Mann–Whitney U test for continuous variables and the chi-square test for categorical variables.