Interventions:
The interventions appeared to be appropriate comparators as they were identified through a comprehensive review. This review was not up to date and so the authors included more recent studies, but it was not clear whether they also looked for new interventions. The interventions were not extensively described, but references were given. Usual care was also not well defined, which makes it difficult to assess whether the results would be applicable to other settings.
Effectiveness/benefits:
The data were appropriate to the study setting and their sources appeared to be robust and were referenced. The authors did not report the literature review methods, such as the inclusion criteria and databases searched, so it was not clear if the review was systematic. The key risk reductions were provided in a table, but their derivation was not described. The approach used to determine the utility data was only partly reported; one study used the Rosser matrix, but the methods of other studies were not reported. QALYs were an appropriate benefit measure given the impact of falls on survival and quality of life. Future benefits were discounted.
Costs:
The perspective was clearly defined and the key costs were included. These costs were reported as totals for the interventions and a full breakdown of the cost items was not given, which limits the ability to replicate the analysis for other settings. Other details, such as the price year, sources of data, and use of discounting, were given. Where possible it seems that medical care inflation was used to adjust the costs and this was appropriate.
Analysis and results:
The model structure was not described in detail and no diagram was provided. Medical management was more effective and cheaper than all other treatments, but the authors did a second analysis excluding this intervention and Tai Chi on the grounds that they were unlikely to be sustainable in the elderly population. It would have been preferable, if the data were available, to have included the sustainability of the interventions in the model and this would have eliminated the need for a second analysis. The incremental approach was appropriate for comparing the relative cost-effectiveness of the different interventions with that of usual care. The probabilistic sensitivity analysis was useful for assessing the impact of the parameter uncertainty. The results were reported reasonably well, in a graph, but the costs and effectiveness figures were not reported and only the one non-dominated cost-effectiveness figure was clearly reported. The authors also discussed the strengths and limitations of their approach.
Concluding remarks:
This cost-effectiveness analysis was satisfactory in the selection of the clinical and economic data and in the reporting of the results, but there were a few limitations. The results lacked some detail, but generally the authors’ conclusions appear to be appropriate.