Careflow Mining Techniques to Explore Type 2 Diabetes Evolution

Abstract

In this work we describe the application of a careflow mining algorithm to detect the most frequent patterns of care in a type 2 diabetes patients cohort. The applied method enriches the detected patterns with clinical data to define temporal phenotypes across the studied population. Novel phenotypes are discovered from heterogeneous data of 424 Italian patients, and compared in terms of metabolic control and complications. Results show that careflow mining can help to summarize the complex evolution of the disease into meaningful patterns, which are also significant from a clinical point of view.

Using topological data analysis and pseudo time series to infer temporal phenotypes from electronic health records

“…TDA graph nodes can be seen as event data representing a model of progression across states, where each node is identified by a fixed index. Thus we compute Jaccard similarity to assign individual subjects to mined trajectories, as previously exploited in the context of careflow mining [ 22 ]. Jaccard similarity coefficients are computed between each sequence of events that build the individual trajectory and all of the detected trajectories (i.e.…”

“…Data for this study was previously collected for clinical and management purposes during the MOSAIC project funded by the European Commission under the 7th Framework Program, (Theme Virtual Physiological Human, 2013–2016) [ 22 , 27 , 28 ]. Health records were accumulated from 924 pre-diagnosed T2DM patients, which resulted in 13,623 instances in our data set.…”

Improving the In-Hospital Mortality Prediction of Diabetes ICU Patients Using a Process Mining/Deep Learning Architecture

“…Conca et al applied process mining on a DM dataset to investigate if differences in work coordination of caregivers exist and if differences lead to undesired patient outcomes [46]. A further study by Dagliati et al analyzed DM patients in Italy to unveil frequent care patterns that describe the evolution of the disease [47].…”

Mining post-surgical care processes in breast cancer patients

“…This approach can help to identify unmet needs that can cause non-compliances to the guidelines. An interesting and broadly applicable approach to examining careflows in detail, reflecting the temporal nature of the clinical events that compose them, is to use the CAREFLOW Mining (CFM) approach that was recently developed by the authors of this paper, and successfully applied to different clinical settings [4,5]. CFM can be used to extract emerging temporal patterns of clinical diagnoses and procedures from long sequences of events, which cannot be retrieved by resorting to traditional SQL queries.…”