2026-11-26 –, Holme Room
Background: Qualitative Comparative Analysis (QCA) is a case-based method that examines combinations of causal conditions that lead to an outcome. Unlike variance-based methods, which average the effects of independent variables to explain an outcome and assume that the effect of each independent variable remains constant regardless of the values of other variables, case-based methods seek to explain outliers and unexplained variance by producing open-ended, descriptive results. Grounded in set theory, QCA focuses on the effects of cases (rows) on an outcome by considering observations as sets, rather than focusing on the effects of variables (columns). Case-based methods can synthesise evidence and provide contextual information about cases that remain unexplained in variance-based analyses. Traditional QCA has primarily included patient and clinical characteristics as causal conditions, with less attention given to incorporating process conditions into case-based analyses.
Objective: In this project, we assess whether process features captured in electronic medical records are sufficient or necessary conditions leading to 30-day unplanned readmission among patients with heart failure. We aim to extend causal conditions to include process conditions and use QCA to examine whether process features are sufficient or necessary for discovering configurations associated with the outcome.
Methods: We extracted eligible patients from the de-identified, publicly available Medical Information Mart for Intensive Care (MIMIC-IV) database. Patients were included if they were aged ≥18 years and had a first heart failure hospitalisation between 2011 and 2022. Patients deceased during the index hospitalisation were excluded. In total, we extracted 12 tables from MIMIC-IV, including hospital transfers, patient demographics, laboratory measurements, and procedures to create an event log capturing patient trajectories from hospital admission to discharge. Local process mining was then applied to identify the most common events observed in the dataset and were used to generate process features. The events and their order were included in QCA to assess whether process conditions were sufficient or necessary in configurational pathways leading to 30-day unplanned readmission.
Results: A total of 285 patients with a first heart failure hospitalisation between 2011 and 2022 were identified. Of these, 95 (33.3%) experienced a 30-day unplanned readmission. The average age was 67 years, 94 (34.0%) patients were female, 69.0% were Caucasian, and the mean Charlson Comorbidity Index was 5.0. Local process mining identified three events that were associated with the outcome: transfer to the operating room (OR), multi-lumen catheter insertion (ML), and arterial line insertion (AL). The relationships between these events were also derived, including AL→ML and AL→OR. By logically eliminating and synthesising the conditions and analysing the truth table, we identified that no single condition was sufficient to produce the outcome. However, the combination of AL and ML was associated with readmission (N = 4) and demonstrated perfect consistency (consistency = 1.0). Incorporating the sequential process condition AL→ML also identified four readmission cases with a consistency score of 1.0, indicating that all patients who followed this pathway were readmitted within 30 days.
Conclusion: By applying process mining to extract process conditions and integrating these conditions into a set-theoretic, case-based analysis, we found that process features were able to discover the cases leading to 30-day unplanned readmission. The inclusion of process features in QCA offers more precise and meaningful insights into patient care pathways and demonstrates the potential value of combining process mining and case-based analysis in medicine.
Haeri Min is a second-year PhD candidate in the School of Computer Science at the University of Sydney, supervised by Associate Professor Simon Poon and Dr Angus Ritchie from the Sydney Medical School. Prior to commencing her PhD, she worked as a full-time Statistical Programmer at the Westmead Applied Research Centre for 3.5 years, where she analysed clinical data, including NSW linked data and randomised controlled trials, to support research aimed at reducing the burden of cardiometabolic and other chronic conditions. Her doctoral research focuses on integrating process mining and case-based analysis to enhance evidence-based medicine. At this conference, she will present her recent work exploring whether local process models derived from clinical event logs can provide additional insights into patient outcomes beyond traditional patient-specific factors.