Building tools to model preference dependence in health settings: Developing a Decision Support System to implement the MACBETH-CI

Context: Multicriteria Decision Analysis (MCDA) methodologies have been increasingly used in evaluations within health and health technology assessment (HTA) settings, with value measurement approaches being the most prevalent. Most studies employ the simple additive model to aggregate the partial impact values of each option, which is only valid if the evaluation dimensions are preference independent. In the health context, this has been shown to be a strong assumption. 2-additive Choquet Integral (CI) models have been pointed out as a versatile aggregation function that allows to model preference dependence relations in multiple contexts. Several studies have already used the 2-additive CI as an aggregation function to model interdependencies, with this being often combined with the MACBETH approach. Nevertheless, several methodological pitfalls can be pointed to some of these studies, and literature in the area does not offer an integrated decision support system (DSS) to enable the implementation of the MACBETH-CI approach to health and HTA settings. Objective: In order to provide a tool to operationalize the MACBETH-CI in a comprehensive and intuitive way, this thesis aims at developing and testing (in HTA context) a DSS and a user-friendly protocol to enable the implementation of the MACBETH-CI approach. The DSS should be designed to address the current challenges identified in the measuring of multidimensional value when preference dependencies occur. Methods: A review of the literature focusing on MCDA studies highlighting current challenges of the simple additive model and on studies using MACBETH-CI approach to model interdependencies was performed. A framework for DSS design was adapted and used to establish the requirements for a new MACBETH-CI DSS. An example pointed out as violating preference independence conditions in the context of HTA was used to verify if having such DSS has the potential to be useful in a real context, and to compare the global scores obtained using CI aggregator with the global scores resulting from the simple additive model (thus helping to understand if adopting an additive model as an approximation is reliable). 28 participants were involved – playing the role of evaluators – in the testing of the DSS with a simple evaluation model in HTA context. Feedback regarding the user experience was collected. Results: The use of the MACBETH-CI DSS evinced the occurrence of preference dependence situations in the tested HTA context. In addition, it was clear that, for the tested example, the CI aggregator model was more precise than the additive model to represent the preferences of participants. Participants provided positive feedback regarding the DSS and the adopted protocol. Conclusions: The proposed DSS and protocol were proven to be a useful tool to identify and model situations of preference dependence. Future research should be performed in order to make the DSS more general and able to model different types of interdependencies in health settings. Additionally, more situations in the context of health settings should be tested.