Statistical approaches for identifying heavy users of inpatient mental health services

A lack of consensus exists concerning how to identify "heavy users" of inpatient mental health services.

To identify a statistical approach that captures, in a clinically meaningful way, "heavy" users of inpatient services using number of admissions and total time spent in hospital.

"Simple" statistical methods (e.g. top 2%) and data driven methods (e.g. the Poisson mixture distribution) were applied to admissions made to adult acute services of a London mental health trust.

The Poisson mixture distribution distinguished "frequent users" of inpatient services, defined as having 4 + admissions in the study period. It also distinguished "high users" of inpatient services, defined as having 52 + occupied bed days. Together "frequent" and "high" users were classified as "heavy users".

Data driven criteria such as the Poisson mixture distribution can identify "heavy" users of inpatient services. The needs of this group require particular attention.