Why modelling a complex intervention is an important precursor to trial design: lessons from studying an intervention to reduce falls-related injuries in older people

Abstract

Objectives: To develop a cost-effectiveness model of a complex intervention from pilot study data in order to inform the viability and design of a subsequent falls prevention trial.

Methods: We used two models; the first estimated the probability of falling over a 12-month period based on a probability tree; the second used Markov simulation to assess the impact of the programme over time.

Results: The first model indicated that our intervention would reduce the proportion falling by only 2.8% over a 12-month period. The major reason for this small effect was that less than a quarter of older people at risk of falling were assessed using our screening tool. Even if policy-makers were willing to spend 30,000 pounds per quality-adjusted life-year gained, there is only a 40% chance that the intervention would be cost-effective. Sensitivity analyses showed that the only scenarios that produced a substantial increase in the effect of the intervention were those in which all older people are assessed.

Conclusions: The model-building approach described in this paper is vital when designing complex trials and where a trial is not possible. Information from the modelling can be used to re-design the intervention. The effectiveness of our proposed intervention appears very small due to its inability to reach those at risk of falling. It is most likely not to be cost-effective. If inability to reach the target group is a weakness common to other similar interventions, this suggests an area for further research.