Determining readmission time window using mixture of generalised Erlang distribution

Eren Demir, Thierry J. Chaussalet, Haifeng Xie

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

3 Citations (Scopus)


The absence of a unified definition of readmissions has motivated the development of a modelling approach, to systematically tackle the issue surrounding the appropriate choice of a time window which defines readmission. The population of discharged patients can be broadly divided in two groups - a group at high risk of readmission and a group at low risk. This approach extends previous work by the authors, without restricting the number of stages, that patients may experience in the community. Using the national data (UK), we demonstrate its usefulness in the case of chronic obstructive pulmonary disease (COPD) which is known to be one of the leading causes of readmission. We further investigate variability in the definition of readmission among 10 strategic health authorities (SHAs) in England and observe that there are differences in the estimated time window across SHAs. The novelty of this modelling approach is the ability of capturing time to readmission that exhibit a non-zero mode and to estimate an appropriate time window based on evidence objectively derived from operational data.
Original languageEnglish
Title of host publicationProcs 20th IEEE Int Symposium on Computer-Based Medical Systems
Subtitle of host publicationCBMS 2007
PublisherInstitute of Electrical and Electronics Engineers (IEEE)
ISBN (Print)0-7695-2905-4 , 1063-7125
Publication statusPublished - 25 Jun 2007
Event20th IEEE International Symposium on Computer-Based Medical Systems - Maribor, Slovenia
Duration: 20 Jun 200725 Jun 2007
Conference number: 20th


Conference20th IEEE International Symposium on Computer-Based Medical Systems


Dive into the research topics of 'Determining readmission time window using mixture of generalised Erlang distribution'. Together they form a unique fingerprint.

Cite this