Visualisation of Mental Health Community Service Patient Pathways

This research explores the combined use of Clustering and Process Mining techniques to examine patient pathways following referrals to a Community Mental Health Team (CMHT) under Cardiff and Vale University Health Board (CVUHB) for Secondary Mental Health Care. The primary objective is to cluster cases to gain insights into variations in referral processes and their impact on patient outcomes. A secondary aim is to evaluate the effectiveness of various clustering approaches in forming cohesive groups of pathways that can be modelled simply and precisely.<br/><br/>Using a dataset provided by project partners, an event log was created and analysed for quality and completeness. Several adaptations of clustering algorithms, including Hierarchical Agglomerative Clustering (HAC), DBSCAN, and K-Medoids, were applied to both vectorial and syntactical representations of patient pathways. The results indicate that HAC, with a vectorial representation, was the most effective approach, followed by DBSCAN, while K-Medoids and syntactical approaches were less effective.<br/><br/>Cluster visualisations reveal that the pathway experienced during the Assessment stage significantly influences patient outcomes, including their likelihood of receiving treatment. Notably, over 40% of patients followed the most frequent pathway, which led to less than 1% receiving treatment. However, other clusters demonstrated treatment progression rates as high as 94%. These findings suggest that clustering algorithms can effectively identify behavioural patterns in complex, variable processes, offering potential applications for optimising mental health care pathways.

Name dc:type.qualificationname

Master's Thesis

Level dc:type.qualificationlevel

Student thesis

Year dc:date.issued

2025

Author dc:creator

• Harris, Joel

Advisors dc:contributor.advisor

• Roach, Paul
• Jobbins, Samuel

Language dc:language

eng

Identifier

oai:pure.atira.dk:studenttheses/0ceafb30-6e64-4318-8149-c760d8cf5c4a

OAI identifier oai:identifier

oai:pure.atira.dk:studenttheses/0ceafb30-6e64-4318-8149-c760d8cf5c4a