Article Text

Download PDFPDF

Modelling the population-level impact of tai-chi on falls and fall-related injury among community-dwelling older people
  1. Lesley Day1,
  2. Caroline F Finch2,
  3. James E Harrison3,
  4. Effie Hoareau1,
  5. Leonie Segal4,
  6. Shahid Ullah2
  1. 1Monash University Accident Research Centre, Melbourne, Australia
  2. 2School of Human Movement and Sport Sciences, University of Ballarat, Ballarat, Australia
  3. 3Research Centre for Injury Studies, Flinders University, Adelaide, Australia
  4. 4University of South Australia, School of Nursing and Midwifery, Adelaide, Australia
  1. Correspondence to Dr Lesley Day, Monash University Accident Research Centre, Clayton, Melbourne, Victoria 3800, Australia;{at}


Objective To model the population level impact of tai-chi on future rates of falls and fall-related injury in older people as a tool for policy development.

Design An epidemiological and economic model for estimating population-level effectiveness of tai-chi.

Setting Australia, 2009.

Patients or subjects Australian community-dwelling population aged 70+ years, ambulatory and without debilitating conditions or profound visual defects.

Intervention Group-based tai-chi, for 1 h twice weekly for 26 weeks, assuming no sustained effect beyond the intervention period.

Main outcome measure Total falls and fall-related hospitalisation prevented in 2009.

Results Population-wide tai-chi delivery would prevent an estimated 5440 falls and 109 fall-related hospitalisations, resulting in a 0.18% reduction in the fall-related hospital admission rate for community-dwelling older people. The gross costs per fall and per fall-related hospital admission prevented were $A4414 (€3013) and $A220 712 (€150 684), respectively. A total investment of $A24.01 million (€16.39 million), equivalent to 4.2% of the cost of fall-related episodes of hospital care in 2003/4, would be required to provide tai-chi for 31 998 people and achieve this effect.

Conclusions Substantial investment in, and high population uptake of, tai-chi would be required to have a large effect on falls and fall-related hospitalisation rates. Although not accounted for in this study, investment in tai-chi is likely to be associated with additional significant health benefits beyond falls prevention. This approach could be applied to other interventions to assist selection of the most cost-effective falls-prevention portfolio for Australia and other countries.

  • Accidental falls
  • tai-chi
  • aged
  • evidence-based practice
  • epidemiological modelling
  • elderly
  • falls
  • methods
  • policy

Statistics from

Falls pose a considerable threat to the safety, health and independence of older people, particularly in countries with ageing populations.1 Deleterious effects include injury, loss of confidence, mobility and independence, decline in health status, and transfer to institutional care.2–6 Falls are relatively common, with 28–39% of people aged 65+ years falling at least once a year, and up to 50% of these experiencing multiple falls.7 8 More than 50% of falls among older people occur at home.2

Evidence from randomised controlled trials (RCTs) shows that falls among community-dwelling older people can be reduced.3 However, RCT results cannot directly convey the effect that specific interventions might have on population-level falls incidence, nor is it possible from such results to gauge the effect of differing levels of uptake and compliance, or predict the effect in a different population.9 10 Such limitations in other areas have been addressed by epidemiological and health economic modelling.9 These approaches enable extrapolation of health and injury outcomes to the longer term, generalisation between locations and population subgroups, and variation of key parameters through systematic sensitivity analysis. Modelling thereby facilitates better understanding of the comparative outcomes expected from different interventions and the ultimate translation of trials evidence into policy and practice.10

This paper describes a modelling approach and results for a group-based tai-chi intervention for community-dwelling older people if delivered at the population level in Australia. Given that commitments have been made to invest resources in falls prevention in Australia,11 our objective was to estimate the population-level impact on falls, and the required resources, if a proven falls intervention was offered to the eligible population. This forms part of a larger study to estimate the potential population-wide impact of six proven falls interventions.12


Description of relevant factors and outcomes

We used the updated Cochrane review3 to source current best efficacy evidence on minimising falls among community-dwelling older people. Group-based tai-chi, as a multi-component exercise programme, was shown to be effective in reducing fall rates. The Cochrane review drew on four trials, which on average delivered 2 h of tai-chi each week for 26 weeks.13–16 Inclusion and exclusion criteria, as well as the efficacy results from these trials, were incorporated into an epidemiological model applied to the Australian population.

Modelling approach

The approach began with an estimate of the number of eligible people in 2009, and compared two pathways—one in which there is a coordinated effort to offer tai-chi to all members of this population, and one in which there is no such effort (figure 1). We defined the eligible population as those who most closely matched the combined trial inclusion criteria. The falls rate applied to those who are not offered tai-chi, and those who are offered tai-chi but do not take it up, was the general falls rate for this age group in Australia, derived from published sources.17 18 The pooled falls rate reduction from the Cochrane review3 was applied to this baseline falls rate to estimate the number of expected falls among those who take up tai-chi. In two of the four trials contributing to the pooled rate reduction, falls were monitored for 6 months, and in a third, for 5.5 months.14–16 In the fourth trial, the period was 48 weeks.13 We have therefore estimated the effect on falls rate for a 6-month period as the base case scenario.

Figure 1

Structure of falls intervention model.

Data sources and definitions

Population data and factors used to select eligible population

In three of the four trials, the targeted age group was 70+ years,13 14 16 while the remaining trial targeted those aged 60+ years.15 An estimate of the proportion of community-dwelling people aged 70+ years who would meet the combined remaining eligibility criteria of the four trials was obtained from the most recent Disability, Ageing and Carers Survey, which provides prevalence estimates stratified according to degrees of limitation in three core activities: communication, mobility or self-care.19 By selecting community-dwelling people aged 70+ years and excluding those who were profoundly limited in, and required assistance with, their core activities, the proportion that would be eligible was approximated (80.4% rounded). This proportion was applied to the 2009 projected resident population from the Australian Bureau of Statistics (n=2 095 579) giving an estimate of 1 684 091 Australians eligible for participation.

Intervention uptake

The uptake rate was defined as the proportion of older people who are likely to respond to an invitation to participate in tai-chi. This rate was estimated from an Australian tai-chi trial among older people,15 which reported an average 1.9% uptake by the eligible population in 21 suburbs in response to local paper advertisements (A Voukelatos, unpublished data, 2007). The 25th and 75th centiles (table 1) were selected to assess the effect of higher or lower uptake rates.

Table 1

Base case parameters for the tai-chi intervention model, Australia

Fall rates

We calculated the rate in the absence of the tai-chi intervention, by pooling results from two prospective Australian studies17 18 of older people with similar inclusion and exclusion criteria, and fall definition, to the Cochrane trials.13–16 The pooled average from these studies was 0.89 falls per person per year, or 0.45 per person per 6 months. We applied the 37% reduction calculated in the Cochrane review3 to this fall rate, to obtain the base case expected falls rate among people taking up the tai-chi intervention (table 1).

A generalised linear model with an underlying negative binomial family20 21 was applied to published Australian falls distributions15 to simulate the falls frequency data and produce the standard errors required to construct the 95% confidence intervals shown in table 1. The software R (version 2.8.1) was used.22 Dispersion parameters were estimated from the mean and variance of the published distributions, and the group size (n=297) was obtained by pooling the sample sizes of the prospective Australian studies 17 18 used to calculate the baseline fall rate.

Proportion of falls resulting in hospitalisation

Few published prospective falls studies have reported the proportion of falls among community-dwelling older people that result in hospitalisation. We obtained an estimate, and the 95% CI (table 1), by pooling data from two Australian prospective studies.23 24

Costs of tai-chi implementation

Costs for tai-chi leaders were obtained from Arthritis Victoria, a major provider of group-based tai-chi classes ($A81–152 per class; mid range $A116.50). Average venue hire costs were obtained from state-based falls-prevention coordinators during a consultative process.11 Music licence fee costs were obtained from the Australasian Performing Rights Association and Phonographic Performance Company of Australia websites.25 26 We assumed that exercise class leaders and hired venues would carry their own professional indemnity and public liability insurance, and the costs would be included in their respective hourly rates. Advertising cost estimates were obtained from local newspapers in both capital cities and major regional centres, and we estimated that two advertisements would be required per class group. An allowance was made for 16 h per class for coordination (arranging venue, leader and advertising; registering class participants; follow-up of those missing classes). All costs are shown in 2009 Australian dollars. Using the class time and programme length data reported in the four tai-chi trials,13–16 we estimated that the average dose of tai-chi required to produce the estimated rate reduction was two classes of 1 h each, for 26 weeks. We estimated the total cost of providing tai-chi for those who take it up, cost per fall prevented, and cost per hospital admission averted.

Sensitivity analysis

Sensitivity analyses were performed using the lower and upper bounds of the base case point estimates. The more or less effective intervention scenarios were used with the intervention fall rates set at the upper and lower bounds of the confidence intervals, respectively, and the control fall rate set at the base case scenario point estimate. We also included three variations on the base case scenario in which tai-chi participants contributed $A3, $A5 or $A10 per class to assess the effect on cost to government, and also recognising the direct benefits of participation.

Predicting the effect on fall-related hospital admissions

We used the tai-chi model outputs and national hospital records from 2002–6 (the most recent national data) to estimate the percentage reduction in the fall-related hospitalisation rate that could be expected in 2009. Using previously published methods,27 average annual age-specific fall-related hospitalisation rates for 2002–3 to 2005–6 were applied to the matching age groups in the 2009 population projections to obtain the 2009 expected fall-related hospitalisation count for people 65+ years. Although the tai-chi model output related to people 70+ years (the group to whom the trial results can be generalised), we calculated the impact on the hospitalisation rates for 65+ years, as this is the age group most often used in Australian setting policy targets and is therefore more relevant to local policy makers when assessing the potential for falls interventions to meet their targets. The number of hospitalisations prevented predicted by the tai-chi intervention model was subtracted from the total count, and the resulting percentage reduction in the falls hospitalisation rate was calculated.


In the base case scenario with 1.9% of the eligible population (n=31 998) taking up tai-chi, 5440 falls and 109 fall-related hospital admissions would be prevented (table 2). This equates to one fall prevented for approximately every six participants, one fall-related hospital admission averted for every 294 participants, and one fall-related hospital admission avoided per 50 falls prevented.

Table 2

Estimation of number of falls prevented and hospital admissions averted, tai-chi intervention model, Australian population 2009

Percentage changes relative to the base case scenario generated by changing the values of the input variables are shown in table 2. A negative value means that fewer falls and falls hospitalisations would be prevented compared with the base case. The variables with the greatest influence on the estimated number of falls prevented are the uptake rate and level of intervention effectiveness. The estimated proportion of falls resulting in hospital admission is an additional influence on the number of hospital admissions prevented.

The unit cost for the tai-chi intervention was $A750 per participant for the 26-week programme (table 3). This translated to costs of $A4414 (€3.013) and $A220 712 (€150 684) per prevented fall and hospital admission, respectively, in the base case scenario (table 4). Cost per fall prevented was determined by intervention effectiveness and intervention cost. Cost per hospital admission prevented was more sensitive to intervention effectiveness and the proportion of fallers admitted to hospital than the variation in intervention cost. The main influences on the total cost were the uptake rate and the variation in the intervention cost.

Table 3

Implementation costs associated with twice weekly classes for 26 weeks, tai-chi intervention

Table 4

Cost effectiveness, tai-chi intervention model, Australia 2009

The total cost of a 6-month tai-chi programme for community-dwelling older persons aged 70+ years, capable of participation, in Australia was estimated at $A24.01 million (€16.39 million), based on 31 998 participants (1.9% of all eligible persons). If participants contributed to class costs, government costs would be reduced by 21%, 35% and 70% for participant contributions per class of $A3, $A5 and $A10, respectively.

We estimated that 75 754 fall-related hospitalisations among people aged 65+ would be expected in Australia in 2009. On the basis of our base case prediction of 109 fall-related hospital admissions averted, the annual fall-related hospital admission rate in 2009 could be expected to be reduced by 0.18%. Under the high hospitalisation rate assumption combined with the more effective intervention scenario, a 0.24% reduction in fall-related hospital admissions is predicted, while the less effective intervention scenario would produce a 0.10% reduction.


This is the first published study to apply epidemiological and economic modelling to assess the population-level impact of a falls-prevention intervention among older people. It provides extensive methodological detail which could be readily applied to estimate the effect of falls interventions in other countries. The approach could also be applied to other injury issues where there is sufficient intervention evidence.

We estimated that, if implemented in the eligible Australian population in 2009, a group-based tai-chi programme would prevent 5440 falls and 109 fall-related hospital admissions for a total cost of $A24.01 million. This equates to $A4414 and $A220 712 per fall and per fall-related hospital admission prevented, respectively. The estimated effect on the fall-related hospital admission rate for persons aged 65+ years is a 0.18% reduction.

Our study clearly shows that an intervention can be highly effective in a randomised trial, but have very little population-level impact. Group-based tai-chi is a relatively effective falls intervention in a controlled setting delivering a fall reduction of 37%.3 However, unless population-level uptake is considerably higher than the evidence-based rate used in our tai-chi model, this intervention will have only limited impact on population falls and fall-related hospital admission rates. Some among the eligible population would never take up tai-chi. This highlights the importance for policy of looking beyond RCT results to the population health context.

We have taken the conservative position of generalising the trial results to community-dwelling people aged 70+ years with similar fitness and mobility levels to the trial participants. The younger age group (60–69 years) may experience similar levels of falls reduction, but there is no firm evidence to support this contention. Certainly, tai-chi may be less effective among frailer older people.13

In defining the eligible population, we relied on estimates from a 2003 population survey.19 If the proportion of community-dwelling older people with restrictions in their core activities changed between 2003 and 2009, the number of eligible people would change accordingly, affecting the number of falls and falls hospitalisations prevented in the base case scenario and the total cost of programme delivery. The cost per fall and per fall hospital admission prevented would remain unchanged.

We have assumed that the size of the falls reduction reported in the Cochrane review,3 based on one Australian and three American studies, is generalisable to community-dwelling older Australians with similar fitness and mobility characteristics. In addition, we assumed that intervention compliance (ie, the proportion of classes attended by the participants, and the extent of practice between class sessions) would equal that achieved in the trials.13–16 Population-level compliance may actually be lower than that achieved in a research trial, resulting in a less effective intervention, with fewer falls and fall hospitalisations prevented.

We assumed that all willing eligible persons would take up the programme in the first year, and that it would not be repeated in subsequent years for the new cohort who would become eligible as they reach 70 years of age. We conservatively selected 6 months as the intervention effect period for the base case scenario, consistent with the follow-up periods in most of the trials.13–16 However, there is some evidence of effect beyond the 6-month period. In one trial, 66% of participants maintained tai-chi practice and sustained the falls reduction over 12 months.14 If this level of sustained effect was achieved in the base case scenario, ∼1.7 times as many falls and fall-related hospital admissions would be prevented for the same investment, with a flow-on effect on the hospital admission rate. The cost per fall and per fall-related hospital admission would be reduced by 30%.

There is a high degree of confidence in the other key assumptions. There is some evidence to support the assumption that falls reductions translate to the same level of effect on fall hospitalisations.28 In theory, an exercise programme might further reduce fall hospitalisations by improving bone strength to the point where fractures would be reduced in the event of a fall, so that fall hospitalisations would be reduced via two different mechanisms. However, it is unlikely that participation in a 26-week tai-chi programme could achieve such an improvement in bone strength. It is also unlikely that any control group activity in the trials would have had an effect on falls, although if it had, the effect on fall rates would be greater than our estimates.

Our study showed that investment of $A24.01 million in group-based tai-chi programmes in Australia would have a small effect on total falls and falls hospital admissions, reducing the fall-related admission rate by less than 0.5%. This would not, on its own, even curb the current average annual fall-related hospital admission rate increase of 2.5% for men and 1.1% for women.29 The annual cost of fall-related episodes of hospital care in Australia provides some context for the required level of investment. The most recent available estimate of this cost is $A566 million for 2003/4.27 An investment of $A24.01 million would be equivalent to 4.2% of the cost of fall-related episodes of hospital care in 2003/4. Comparison with other proven falls interventions would be useful in guiding policy towards the most cost-effective falls intervention for the population level.

Additional relevant factors in preparing for widespread implementation of tai-chi for community-dwelling older people include sufficient appropriately trained tai-chi leaders and accessible venues. Implementation may be more difficult in rural and more remote areas because of a limited trained workforce, lack of critical mass required to make class groups viable, and fewer transport options.

We estimated the effect of tai-chi on one health outcome: falls. Other health benefits include improvements in musculoskeletal function (strength, balance, flexibility), cardiorespiratory function (including blood pressure), glucose metabolism, body mass index, immune capacity, mental health and quality of life.30–33 Tai-chi could therefore affect key risk factors (physical inactivity, high body mass, high blood pressure) that explain considerable proportions of attributable health loss associated with cardiovascular disease, diabetes mellitus and, to a lesser extent, cancer.34 Anecdotal evidence suggested that most participants enjoyed tai-chi, generating direct ‘consumption’ benefits. Our study shows that, if participants were to contribute even $A5 towards the cost of a class, government costs would be reduced by 35%. Investment in tai-chi for older people is expected to deliver a range of health benefits and reduced health-related costs beyond those associated with falls prevention, thereby making the intervention more cost-effective at a population health level. Ideally the wider benefits would be incorporated into policy decisions.


The translation of RCT results to the population level is possible and provides useful supplementary information for decision makers. Tai-chi programmes may present good value for falls-prevention resources, if the cost per participant can be substantially reduced compared with the cost estimates used here. We have found that, despite tai-chi representing an effective intervention, considerably higher population uptake than that observed in a recent trial would be required for a national roll-out of tai-chi to have a large effect on fall-related hospitalisation rates. This is not to say that investment in tai-chi is not worthwhile, given the potential for other health benefits and overall cost-effectiveness. This modelling approach, applied to other proven falls interventions, may assist in selecting the most cost-effective population-level falls-prevention portfolio for Australia and other countries.

What is already known on this subject

  • Reliable evidence shows that interventions, such as group-based tai-chi, can reduce falls among community-dwelling older people.

  • Trial results rarely convey the effect that interventions may have at the population level.

  • Modelling enables extrapolation of trial results to broader populations and assessment of the effect of variations in different key input variables on cost-effectiveness estimates.

What this study adds

  • An approach for modelling the population-level effectiveness and cost-effectiveness of proven falls interventions to better guide decision-making with respect to policy and practice.

  • Demonstration of the utility of an approach to estimating the population-level effect of one proven falls intervention—group-based tai-chi.

  • Significant investment in, and high levels of population uptake of, tai-chi programmes would be required to produce meaningful reductions in the fall-related hospitalisation rates in Australia.


We thank the following people who generously provided data for use in this project: Dr Alex Voukelatos, Sydney South West Area Health Service, Australia; Rob Maris, and Professor Steve Wolf, Emory University, USA. Soufiane Boufous, George Institute, Australia, provided advice on working with falls hospitalisation data. Clare Bradley, Flinders University, Australia, undertook the hospitalisation rate analyses.



  • Funding This project was funded through the Australian Government Department for Health and Ageing (Falls and Community Injury Grants Program). LD and CF were supported by National Health and Medical Research Council (NHMRC) Senior Research and Principal Research Fellowships, respectively. SU was supported by an Injury Trauma and Rehabilitation (ITR) Research Fellowship funded through a NHMRC Capacity Building Grant in Population Health.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.