Detecting anomalous care pathways in long-term care network

Hambisa Keno, Nan Kong, Steven Landry, Wanzhu Tu, Christopher M. Callahan

Research output: Chapter in Book/Report/Conference proceedingConference contribution


Expensive facilities in the long-term care network are identified both by federal and state payers as potential targets to reduce care cost. Achieving such targets requires having a comprehensive insight into care utilization dynamics across the whole spectrum of the long-term care continuum. Interrelationships between care sites, dependency in length of stay at care sites, recurring care patterns and anomalous care patterns are among these comprehensive insights. In this paper, we present a study on one of these issues namely, identification of anomalous care pathways in a long-term care network. The aim is to identify patients with expensive care costs such as those diagnosed with dementia from their potentially anomalous care pathway captured from spatio-termporal utilization data. We characterized long-term care pathway for patients with and without dementia diagnosis based on utilization patterns from simulated data. A pathway is represented using 17 features, which include the total number of transitions among different long-term care facilities in five years, the total numbers of institutionalizations and hospitalizations, and the associated length of stay among others. We used a distance-based anomaly detection algorithm to identify anomalous pathways within the transition data set for patients without dementia diagnosis. Among the 17 features considered, we found only two features, namely the total numbers of transitions and institutionalizations, to be useful in detecting anomaly. The reduction in the number of features useful for anomaly detection limits the applicability of our work in making early intervention recommendations since the comprehensive transition history will be hidden.

Original languageEnglish (US)
Title of host publicationIIE Annual Conference and Expo 2015
PublisherInstitute of Industrial Engineers
Number of pages7
ISBN (Electronic)9780983762447
StatePublished - 2015
EventIIE Annual Conference and Expo 2015 - Nashville, United States
Duration: May 30 2015Jun 2 2015

Publication series

NameIIE Annual Conference and Expo 2015


OtherIIE Annual Conference and Expo 2015
Country/TerritoryUnited States

All Science Journal Classification (ASJC) codes

  • Control and Systems Engineering
  • Industrial and Manufacturing Engineering


Dive into the research topics of 'Detecting anomalous care pathways in long-term care network'. Together they form a unique fingerprint.

Cite this