This study evaluates the feasibility and the preliminary effects of a simplified 10-step Tai-chi programme (a dyadic approach) on the mobility performance of people with mild to moderate dementia.

Four community health centres were recruited and each was randomised to either the intervention group (Tai-chi) or the control group. Tai-chi group received a 16-week 10-step simplified Tai-chi training programme in which additional measures were implemented to enhance participants' engagement. The control group took part in group recreational activities organized by the community centres.

It was hypothesized that the Tai-chi group would outperform the control group regarding their mobility performance.

Four community health centres that provide dementia care services were recruited through convenience sampling. Each was viewed as one cluster and was randomized to either the Tai-chi or the control group, based on computer-generated random numbers prepared by an independent statistician.

Participants allocated to the Tai-chi group took part in the 16-week 10-step simplified Tai-chi programme, which was derived from the traditional Yang style and has been proven to be effective in enhancing older people's balance and mobility. Each week, the dyads attended two 1-hour sessions of centre-based Tai-chi training and practised at least three 30-minute Tai-chi sessions at home. Additional measures targeted cognitively impaired people were implemented to promote engagement, including the adoption of multiple sensory cues, slow and relaxed practice, a dyadic approach, and positive emotional motivation techniques.

Participants allocated to the control group took part in group recreational activities such as watching movies or listening to music, which was organized by the community centres with similar frequency and duration of the Tai-chi sessions that were organized for the Tai-chi group. The control group participants were instructed to continue their usual lifestyles and levels of physical activity.

Inclusion Criteria (for participants with dementia):

• community-dwelling older people aged > 60 years;
• able to walk independently with no walking aid or no more than a single point stick for at least 10 minutes to ensure their mobility was good enough for taking part in the Tai-chi training;
• formally diagnosed with a form of dementia;
• classified with mild to moderate severity of dementia, assessed by the Montreal Cognitive Assessment 5-minute scale with a cut-off score at the 16th percentile according to participants' age and education; and
• able to identify a caregiver who was willing to work as an exercise partner for their Tai-chi practice.

Inclusion Criteria (for caregivers):

• adults aged > 18;
• living with the participants or actively involved in their daily care;
• sufficiently mobile to be able to take part in the Tai-chi training together with the participants; and
• willing to work as an exercise partner with the participants and monitor and encourage them to practice Tai-chi at home.

Exclusion Criteria (for participants and caregivers):

if at the time of and three months before recruitment, they

• had any diseases that might severely affect their balance and coordination, such as Parkinson's disease or myasthenia gravis;
• were hospitalized due to acute illnesses such as myocardial infarction, stroke or hip fracture, or had major surgeries;
• reported that they regularly performed moderately intensive exercise, such as hiking or Tai-chi, for more than 2 hours per week;
• had terminal illnesses such as cancer and were in palliative care; or
• had severe visual or hearing impairment.

For confidentiality, the data will be kept anonymous and the names of all participants including the caregivers will be replaced by reference codes. The data collected will be kept in a locked place and electronic versions will be encrypted, and only be accessible by the researchers. All data will be destroyed within 3 years after the completion of this research.

An independent assessor who was blinded to group allocation assessed the participants' motor performance variables.