Tai chi analyses in community-dwelling older adults suggested fewer falls and fewer people falling, with different certainty ratings. Relative reductions are not personal guarantees or a ranking against other exercise types.
Study and source overview
When a report says tai chi reduced falls, does it mean fewer events or fewer people falling, and in whom?
- Who was reviewed
- Community-dwelling adults aged 60 or older. The full review included 108 trials and 23,407 participants across many exercise types, not only tai chi. The overall mean age was about 76; about 77% were women.
- Tai chi samples
- Rate of falls: seven trials, 2,655 participants. People with at least one fall: eight trials, 2,677 participants. Samples can overlap and must not be added together.
- Comparison and duration
- Control programmes were not expected to reduce falls. Tai chi schedules, course lengths and observation periods varied across trials.
- Date and access
- Published in 2019; the search ended on 2 May 2018. This summary uses the complete public abstract and the corresponding NCCIH explanation, not a new full-text evidence update.
Why are 19% and 20% different results?
| Tai chi outcome | Result and 95% confidence interval | Sample and certainty |
|---|---|---|
| Rate of falls, including repeated events and observation time | Rate ratio 0.81 (0.67–0.99): about 19% lower in relative terms. | Seven trials, 2,655 participants; low certainty. |
| People with one or more falls | Risk ratio 0.80 (0.70–0.91): about 20% lower in relative terms. | Eight trials, 2,677 participants; high certainty. |
A person who falls three times contributes three events to the rate calculation but only one person to the second outcome. The two rows answer different questions.
The 20% is a relative change. As a mathematical example only, if 10 in every 100 people fell in a comparison group, a 20% relative reduction would mean eight in 100: two fewer people, not 20 fewer. This is not the trial's absolute result or a prediction of your risk.
Where do the findings stop?
The review concerned older adults living in the community, not a uniform programme for care-home residents. It excluded trials focused on particular conditions such as stroke. The overall age and sex figures do not independently describe the tai chi subgroup.
Different tai chi programmes and observation periods were pooled. Many included trials had unclear or high risk of bias. Low certainty means the rate estimate may change with better evidence. High certainty for the number of people falling does not mean every participant avoided falls.
The abstract does not provide a complete tai chi-specific safety analysis. Adverse-event information across the review covers multiple exercise types; it cannot be used to calculate a tai chi injury rate. Limited reporting does not establish zero risk.
What questions could I ask about a course?
Ask who the course is designed for, whether it includes balance and functional training, who teaches it and how it is adapted. The 2018 tai chi balance-programme trial describes a particular population, classroom programme and comparison. A general video is not the same programme.
Do not rank separate exercise subgroups by their relative reductions. Participants, controls and observation periods may differ. Comparing two exercise options requires a direct comparison under similar conditions. No course result guarantees that a person will never fall again.
Original source and review coverage
Sherrington C et al. Exercise for preventing falls in older people living in the community. Cochrane Database of Systematic Reviews. 2019;1:CD012424.
DOI：10.1002/14651858.CD012424.pub2
Review coverage: Complete public abstract retrieved through the official Europe PMC API, checked for overall scope, tai chi analyses, comparisons, confidence intervals and search date. The publisher full text was not evaluated.
NCCIH: Tai chi and falls in community-dwelling older adults：Distinguishes the two tai chi outcomes, samples and certainty ratings. It describes the same review, not another set of trials.
By First Chi; source checked .The paper was published in 2019; the website update date does not replace that date.
