Format

Send to:

Choose Destination
See comment in PubMed Commons below
Psychiatry Res. 2013 Dec 30;210(3):1211-8. doi: 10.1016/j.psychres.2013.09.019. Epub 2013 Oct 4.

Predicting the length of hospital stay of psychiatry patients using signal detection analysis.

Author information

  • 1Kyushu University, Graduate School of Medicine, Department of Health Services Management and Policy, Higashi-ku, Fukuoka 812-8582, Japan. Electronic address: fase_bzm@uinjkt.ac.id.

Abstract

In Japan, the length of hospital stay (LOS) at psychiatric institutions often exceeds a year, and factors related to such stays have been identified. However, we do not know how multiple patient, hospital, and physician factors interact to determine LOS. Patient data were collected from a psychiatric hospital in Osaka, Japan. We developed subgroups, which were determined by interactions related to LOS using signal detection theory. In acute or emergency wards, five factors related to LOS were identified, and subjects were categorized into six subgroups. The indices obtained by the five factors ranged 2.49-3.47 for odds ratio, 0.47-0.84 for sensitivity, 0.40-0.76 for specificity, and 0.52-0.71 for positive predictive value. In general wards, five factors related to LOS were identified, and subjects were categorized into six subgroups. The indices obtained by the five factors ranged 3.02-5.36 for odds ratio, 0.58-0.86 for sensitivity, 0.37-0.68 for specificity, and 0.85-0.92 for positive predictive value. Psychiatrists who have been practicing longer in acute or emergency wards appear to have significantly longer stay of patients, and older or more severe patients tend to be in need of longer inpatient care. Our results provide findings that may be helpful in decreasing LOS at psychiatric hospitals.

© 2013 Elsevier Ireland Ltd. All rights reserved.

KEYWORDS:

Outcome; Prediction; Signal detection analysis

PMID:
24095680
[PubMed - in process]
PubMed Commons home

PubMed Commons

0 comments
How to join PubMed Commons

    Supplemental Content

    Full text links

    Icon for Elsevier Science
    Loading ...
    Write to the Help Desk