In the majority of studies on hospital readmissions, a readmission is deemed to have occurred if a patient is admitted within a time window of the previous discharge date. However, these time windows have rarely been objectively justified. This paper develops a framework to determine the optimal time window in defining patient readmissions. First, we capture the readmission process by fitting a special case of a Coxian phase-type distribution, which is expressed as a mixture of two generalized Erlang distributions. Second, we apply the minimum classification error approach to compute the optimal time window. Using the English national hospital episodes statistic (HES) dataset, we demonstrate the usefulness of the approach in the case of chronic obstructive pulmonary disease (COPD), stroke, congestive heart failure (CHF), and clinical conditions associated with hip and thigh fractured patients.
|Title of host publication||Procs 22nd IEEE Int Symposium on Computer Based Medical Systems|
|Subtitle of host publication||CBMS 2009|
|Number of pages||7|
|Publication status||Published - 2009|