The growing digitalisation of the healthcare system has transformed electronic health records (EHRs) into a valuable source of information for understanding and improving care processes. However, their effective management remains a challenge.
In this context, a study led by Carlos Fernández Llatas, researcher at the SABIEN group of the ITACA Institute at the Universitat Politècnica de València, in collaboration with the Universitat Jaume I, has developed an innovative Process Mining (PM) approach specifically adapted to the clinical field. It enables care pathways recorded in electronic health records to be represented in a clearer and more comprehensible way.
The work, recently published in the Journal of Biomedical Informatics, proposes a new methodology based on declarative techniques that allows expert knowledge to be incorporated into the analysis of clinical data. This overcomes one of the historical limitations of process mining: the complexity of the models generated when they deal with repeated activities. The methodology aims to facilitate interactive exploration of real processes from data recorded in EHRs.
“This approach makes it possible, for the first time, to interactively integrate clinical knowledge into the discovery process, generating models that better reflect the reality of treatments and improve interpretation by medical and healthcare management staff,” highlights Carlos Fernández-Llatas, principal investigator of the study.
What is Process Mining?
Process Mining is a well-established discipline in industrial and administrative sectors, but its application in healthcare presents significant limitations. The particular characteristics of the clinical environment —such as the wide variability among patients, repeated activities (tests, treatment cycles, visits), and the complexity of recording care— generate excessively dense and unintelligible models, known as the “spaghetti effect,” which hinder their practical use by clinical and managerial staff.
“The models produced with standard techniques do not faithfully reflect the logic of clinical algorithms or actual therapeutic decisions. This prevents professionals themselves from interpreting or using these models for care improvement,” notes Fernández-Llatas.
Contribution of the new approach: integration of clinical knowledge through declarative expressions
In this context, the new approach introduces the possibility for healthcare professionals to actively participate in the model development process, in a field where the continuous presence of repeated activities has limited the effective use of Process Mining techniques. To achieve this, it relies on declarative expressions —a way of representing knowledge through flexible and understandable rules that define key conditions of the process.
The study proposes a set of innovative functionalities —such as milestones, circuits, and protected regions— that make it possible to differentiate critical stages (for example, before and after surgery), identify specific therapeutic pathways, or isolate relevant regions of the process for independent analysis.
The methodology has been validated in a use case centred on the treatment of prostate cancer patients, using synthetic data from Simulacrum —a database generated from real oncology records of the UK National Health System. The study focused on patients treated with Docetaxel, one of the most common drugs for this type of tumour.
In fact, the results show that the models generated after incorporating expert knowledge are much more comprehensible, structured, and useful for analysing patient pathways, identifying deviations, detecting opportunities for improvement, and supporting clinical decision-making.
“We have demonstrated that it is possible to generate models that accurately reflect different therapeutic patterns, such as the administration of chemotherapy before or after surgery, and that professionals can interact with the system to refine these models according to their clinical experience,” explains Fernández-Llatas.
A step forward in advancing healthcare based on real data and processes
The impact of this methodology is particularly relevant in the context of the digital transformation of the healthcare system, as it enables more effective use of data available in electronic health records, helping advance towards a more personalised, efficient, and value-based healthcare model.
“The healthcare system holds millions of records that, with the right tools, can be turned into useful knowledge for redesigning processes, reducing inefficiencies, and improving patient care. This research is a key step in that direction,” concludes Fernández-Llatas.
The research was funded by the Ministry of Science sand Innovation through project PID2020-113723RB-C21, and its results open new lines of work aimed at clinical validation with healthcare professionals and extension to other conditions and care contexts.