[ comparison of statistical modeling strategies for analyzing length of stay in Shiraz education hospitals]

ABSTRACT

The modeling of inpatient length of stay [LOS] has important implications in various aspects of health care management, such as hospital management and planning, the management of health care resources and the performance in hospital care. For many health care management systems, it is important to develop a comprehensive analysis of LOS and to identify hospital- and patient-related characteristics influencing LOS variations. By targeting relevant factors, appropriate policies can be developed to manage the hospital care and the health care resources effectively. Except a specific and limited study, a few studies have carried out on LOS modeling in Iran. This study has been carried out to compare different models for explaining LOS. A sample of 3546 patients was selected using randomized clustered sampling scheme from different wards of Shiraz educational hospitals in 2005 and their hospitalization [LOS] rate registered as response variable and other demographic variables. Advance statistical models such as Poisson regression, ZTP, ZTNB and percentile regression were applied to analyze and modeling the data. Mean +/- S.D of LOS were calculated as 8.4 +/- 13.7days. Over-dispersion for LOS was observed. Therefore, negative binomial, gamma, zero truncated Poisson [ZTP], zero truncated negative binomial [ZTNB] and percentile regression were applied. The results showed that ZTNB is an efficient model for explaining LOS in Shiraz's hospitals. Given heterogeneity in LOS in different wards in Shiraz hospitals, negative binomial model explains adequately the LOS values. Also, percentile regression model seems appropriate for LOS which was applied to our data. Application of the cited models here to other hospitals scattered around the country is suggested