Contribution of Unsupervised Clustering Algorithms to Patient Safety Culture Promotion

Patient safety culture (PSC) is a linked to improved organizational performances and fewer adverse events in healthcare. Mainly, PSC is assessed using self-administered questionnaires that define PSC through a number of dimensions. Nevertheless, the latter suffers from a major limit that is dimensional scores aggregation to obtain PSC overall level. To overcome this constraint, the proposed approach consists of applying unsupervised clustering algorithms (i.e. K-means & HAC) to group HEs sharing similar PSC profiles. This approach was applied on PSC assessment data in 10 Algerian pubic HEs. Results showed that HEs can be grouped into 3 clusters having different PSC profiles. The proposed approach has the merit of guiding decision makers to propose adapted improvement actions based on shared inter and intra cluster problematic PSC dimensions.