Disinvestment is the removal or reduction of previously provided practices or services, and has typically been undertaken where a practice or service has been clearly shown to be ineffective, inefficient and/or harmful. However, practices and services that have uncertain evidence of effectiveness, efficiency and safety can also be considered as candidates for disinvestment. Disinvestment from these practices and services is risky as they may yet prove to be beneficial if further evidence becomes available. A novel research approach has previously been described for this situation, allowing disinvestment to take place while simultaneously generating evidence previously missing from consideration. In this paper, we describe how this approach can be expanded to situations where three or more conditions are of relevance, and describe the protocol for a trial examining the reduction and elimination of use of mobilisation alarms on hospital wards to prevent patient falls. Our approach utilises a 3-group, concurrent, non-inferiority, stepped wedge, randomised design with an embedded parallel, cluster randomised design. Eighteen hospital wards with high rates of alarm use (≥3%) will be paired within their health service and randomly allocated to a calendar month when they will transition to a “Reduced” (<3%) or “Eliminated” (0%) mobilisation alarm condition. Dynamic randomisation will be used to determine which ward in each pair will be allocated to either the reduced or eliminated condition to promote equivalence between wards for the embedded parallel, cluster randomised component of the design. A project governance committee will set non-inferiority margins. The primary outcome will be rates of falls. Secondary clinical, process, safety, and economic outcomes will be collected and a concurrent economic evaluation undertaken.