Determining the Factors Affecting the Length of Hospital Stay Using Hurdle Model
Sajadi Farkhondeh Alsadat1, Hasanzadeh Fatemeh2, Alizadeh Mojtaba3*
1. Assistant Professor, Department of Statistics, Faculty of Mathematics and Statistics, University of Isfahan, Isfahan, Iran
2. Assistant Professor, Department of Statistics, Khansar Campus, University of Isfahan, Isfahan, Iran
3. PhD Student in Public Management, Faculty of Management, Islamic Azad University, Khorasgan Branch, Isfahan, Iran
*Corresponding Author: Mojtaba Alizadeh
Address: Khorasgan University, Daneshgah Blvd., Arqavanieh, Jey St., Isfahan, Iran
Tel: 009831-35354001 Email: firstname.lastname@example.org
Background & Objectives: There are different statistical methods to identify the factors effective in increasing and decreasing the patients’ length of stay (LOS), each of which having its own advantages and disadvantages. This study aimed to determine the factors affecting LOS in a selected hospital using the available count models and comparing them with the Hurdle model as a new model.
Methods: This descriptive-analytical study was conducted using a retrospective cross-sectional design in a selected hospital in Isfahan in 2019. Using Poisson and Hurdle (binomial and Poisson) count models, the effect of demographic factors of the hospitalized patients and the reasons for visiting the hospital on LOS was analyzed. The goodness of fit of the models and their comparison were performed using deviance statistic.
Results: The mean of LOS was 8.89 days and the median of LOS was 3 days. From among the three models, the binomial Hurdle model was determined appropriate in which, the constant and variables such as age, social security insurance and uninsurance, hospital emergency visits and 115 emergencies, and reasons for visiting psychiatric, orthopedic, and ophthalmic wards were significant. By changing the type of insurance from the reference of other insurances to social security insurance, the mean of LOS decreased by 0.088 units and for uninsured patients, this decrease was by 0.539 units.
Conclusion: Considering the values of the deviance statistic for Poisson, Hurdle Poisson and negative binomial Hurdle models, the negative binomial Hurdle model is the most appropriate for analyzing these data.
Keywords: Length of stay (LOS), Inpatients, Hospital, Count Models, Hurdle Models
Citation: Sajadi FS, Hasanzadeh F, Alizadeh M. Determining the Factors Affecting the Length of Hospital Stay Using Hurdle Model. Journal of Health Based Research 2020; 6(2): 139-50. [In Persian]