Project

Back to overview

Robust negative binomial regression with application to the analysis of hospital length of stay

English title Robust negative binomial regression with application to the analysis of hospital length of stay
Applicant Marazzi Alfio
Number 141266
Funding scheme Project funding (Div. I-III)
Research institution Institut Universitaire de Médecine Sociale et Préventive - IUMSP CHUV et Université de Lausanne
Institution of higher education University of Lausanne - LA
Main discipline Mathematics
Start/End 01.07.2012 - 31.08.2014
Approved amount 90'941.00
Show all

All Disciplines (4)

Discipline
Mathematics
Science of management
Medical Statistics
Public Health and Health Services

Keywords (3)

Robust estimation; Negative Exponential Model; Regression

Lay Summary (English)

Lead
Lay summary

Length of stay (LOS) is the most widely used indicator of hospital cost of stay and LOS means, percentiles, and other statistics are essential for the pricing of hospital procedures, hospital budgeting, and reimbursement. Since extremely long or unusually short stays are often observed, advanced robust statistical techniques (not affected by the presence of atypical data) are vitally needed to assist in the computation of reliable LOS statistics.

In this project we will propose and develop a family of new robust estimates of negative binomial regression, which has often been recommended in the literature as a convenient model for LOS. We will focus on the weighted likelihood approach which provides highly robust and efficient estimates.

The application of these methods could help provide more reliable LOS statistics and identify misclassification problems. This is an essential process for the efficient application of patient classification systems in hospital management both in Switzerland and internationally. In addition, this project will include a new (currently lacking) specific product for the analysis of asymmetric count distributions into the easy-to-use statistical environment R in public domain. This tool will therefore be made available to a broad community of researchers in different fields, both those with extensive statistical expertise and those with a relatively limited statistical background.

Direct link to Lay Summary Last update: 21.02.2013

Responsible applicant and co-applicants

Employees

Collaboration

Group / person Country
Types of collaboration
University of Buenos Aires Argentina (South America)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Ca' Foscari University of Venice Italy (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication

Scientific events

Active participation

Title Type of contribution Title of article or contribution Date Place Persons involved
ICORS 2013 Talk given at a conference WEIGHTED LIKELIHOOD NEGATIVE BINOMIAL REGRESSION 08.07.2013 Saint Petersburg, Russie, Russia Marazzi Alfio; Amiguet Michael;
ERCIM 2012 Talk given at a conference A generalized weighted likelihood estimator 01.12.2012 Oviedo, Espagne, Spain Amiguet Michael;


Associated projects

Number Title Start Funding scheme
66895 Régression robuste pour réponses à distribution asymétrique et application à l'analyse du coût de séjour hospitalier 01.09.2002 Project funding (Div. I-III)
116357 Robust regression with censored data and application to the analysis of hospital cost of stay 01.07.2007 Project funding (Div. I-III)
54146 Statistical tool for modeling length of stay distributions with the help of covariates: robust regression for asymmetrically distributed responses 01.03.2000 Project funding (Div. I-III)

Abstract

This is a proposal to continue a series of projects on the development of robust statistical methods for the analysis of positive and asymmetrically distributed random variables. We are considering methods for modelling positive discrete random variables as a function of a number of covariates and in the presence of outliers, i.e., atypical extreme observations. Our main practical interest is the analysis of the length of hospital stay (LOS) as a function of covariates available on administrative and medical files. LOS is the most widely used indicator of hospital cost of stay and LOS means, percentiles, and other statistics are essential for the pricing of hospital procedures, hospital budgeting, and reimbursement. Since extremely long or unusually short stays are often observed, advanced robust statistical techniques are vitally needed to assist in the computation of reliable LOS statistics.
-