A key question policy makers are interested in is whether public health interventions that can be regarded as best practices could be successful implemented in countries other than the country where the policy was originally implemented. Public health interventions that present best practice characteristics, such as Let Food be Your Medicine (LFYM), Multimodal Training Intervention (MTI) and the StopDia intervention, are being assessed as part of the OECD project on best practices. However, while these interventions have been successful in one context, they may not be successful in another for multiple reasons, including population, economic and political factors.
This paper presents a data-driven transferability assessment using cluster analysis, to identify groups of countries that have the greatest potential for the successful transfer of a specific interventions. For each of the three best practice interventions mentioned above, key success factors are identified and country-level data on these factors collected from public sources. Then, countries are clustered into groups with similar characteristics. Based on these characteristics, tailored recommendations are made for each cluster of countries regarding the potential transfer of the best practice intervention. This analysis helps policy makers decide whether or not to transfer a public health intervention, and what factors to pay particular attention to when doing so.
Four clustering methods are compared (k-means, k-medoids, hierarchical and DBSCAN), using two different methods for preparing the data (Gower distance matrix and aggregated context scores). On balance, k-medoids using Gower distance is found to be the most effective method for clustering countries into groups, taking into account validation statistics, data characteristics, interpretability of the results and flexibility to use with other datasets. The resulting clusters successfully separate the countries into interpretable groups depending on their potential for transferring each intervention.
