pMineR

R libraries for Process Mining for Healthcare

The pMineR project was born in 2016 and presented the first time at AIME 2016, in Vienna. We are glad for the opportunity to be here now to present the current state of this never-ending development, share visions and perspective and collects all the potential contributes of users and interested colleagues.

pMineR is an R library specifically designed to support Process Analysts to work in the clinical domain, providing Process Discovery algorithms, tools for Conformance Checking, Trace and Event Log analysis and methods for representing and working with given institutional processes, Consensus flow, Clinical Protocols, etc.. The library implements those features considering the most frequent statistical tools used in healthcare, such as non-parametric tests (Chi-Square, Fisher's Exact test), Surivival Analysis (Kaplan Meier curves, log-rank tests) and also Machine Learning methods and predictors to estimate the most relevant clinical outcomes in terms of events or times. It also implements an easy to use internal language for the interpretable representation of clinical guidelines and subsequent compliance analysis (crossing the domains of Process Mining in Healthcare and Clinical Computer Interpretable Guidelines)
For more info, visit : https://github.com/PMLiquidLab/pMineR.v046

The TUTORIAL

This Tutorial is designed to provide a general overwiew on pMineR, and introduce the partecipants to the main modules, by short hands-on sessions, to cope with Process Discovery, Conformance Checking and Statistical Analysis of paths, clinical guidelines, events and timing. The Tutorial will also exploit invited speakers to present their experience on real-world data analysis and open projects based on pMineR. Finally, an open round-table will be the opportunity to share opinions, ideas about the topic or a further evolution of the libraries. An indicative program is shonw below :

Frontal Lesson Modules ( 1h 30', Roberto Gatta ):

• General introduction
• Event Log data Loading, internal structure, events, time ( pMineR::dataLoader() )
• Performing queries and Quality-of-data oriented trace inspection ( pMineR::QOD() )
• Process Discovery with First Order Markov Models and CareFlow Miner ( pMineR::firstOrderMarkovModel() and pMineR::CareFlowMiner() )
• Clinical Guideline Conformance Checking with Pseudo Workflow language ( pMineR::confCheck_easy() )

Short lectures ( 1h ):

• Real-world Process Discovery
  • Michel Cuendet. A differential process mining analysis of COVID-19 management for cancer patients. Centre Hospitalier Universitaire Vaudoise (CHUV), Lausanne
  • Federico Mastroleo. Facing the "what went wrong?" in treatments organization/delivery in Radiotherapy
• Real-world Conformance Checking
  • Stefania Orini and Claudio Singh Solorzano. Clinical Guideline modeling with PWL for Mild Cognitive Impairment
  • Mariachiara Savino. Conformance Checking metrics for Clinicall Guidelines

Round Table ( 30' ):

• Freewheeling discussion about the current issues or future opportunities of the discipline and/or the project. For esample: empowering the PD with more ML? PD-oriented scheduling and planning DSS, are they possible? Engaging MDs: winning and losing strategies? PM4HC and Semantic web technologies: which possible marriage? Big Data = Big Problems: can we scale PM analysis? Federated PM to cope with Patient's privacy and data ownership, in multicentric clinical studies? .... and any other Wild and Crazy ideas!

pMineR/pMinShiny frontal lectures

Real-world use cases, presented by invited speakers

Round Table, interactive open discussion