Queueing Network Approach to the Analysis of Healthcare Systems H. Xie, T.J. Chaussalet and P.H. Millard Health and Social Care Modelling Group (HSCMG)

Slides:



Advertisements
Similar presentations
Whos in the beds? Experience of using the AEP in the UK Paul Forte ORAHS Working Group Prague, July 2003.
Advertisements

Modeling of Data. Basic Bayes theorem Bayes theorem relates the conditional probabilities of two events A, and B: A might be a hypothesis and B might.
Introduction Queuing is the study of waiting lines, or queues.
Health Innovation Exchange
Burn Injury Jo Myers BSc (hons), RGN, Dip(He)RSCN Lead Nurse
Transportation Problem (TP) and Assignment Problem (AP)
Capacity Setting and Queuing Theory
Experimental Design, Response Surface Analysis, and Optimization
Peter Ward Senior Physiotherapist Acute Medicine Driving Healthcare Change Through HSCP Research February 28 th, 2014 Carole Murphy Senior Occupational.
Discrete Probability Distributions
Modelling Activities at a Neurological Rehabilitation Unit Richard Wood Jeff Griffiths Janet Williams.
Visual Recognition Tutorial
Towards Feasibility Region Calculus: An End-to-end Schedulability Analysis of Real- Time Multistage Execution William Hawkins and Tarek Abdelzaher Presented.
1 Learning Entity Specific Models Stefan Niculescu Carnegie Mellon University November, 2003.
Basics of Statistical Estimation. Learning Probabilities: Classical Approach Simplest case: Flipping a thumbtack tails heads True probability  is unknown.
1 Performance Evaluation of Computer Networks Objectives  Introduction to Queuing Theory  Little’s Theorem  Standard Notation of Queuing Systems  Poisson.
Presenting: Assaf Tzabari
Little’s Theorem Examples Courtesy of: Dr. Abdul Waheed (previous instructor at COE)
Calculating & Reporting Healthcare Statistics
Computer vision: models, learning and inference
The Poisson process and exponentially distributed service time In real life customers don’t arrive at pre-ordained times as specified in the last tutorial…
QUEUING MODELS Queuing theory is the analysis of waiting lines It can be used to: –Determine the # checkout stands to have open at a store –Determine the.
Buffer Management for Shared- Memory ATM Switches Written By: Mutlu Apraci John A.Copelan Georgia Institute of Technology Presented By: Yan Huang.
Queuing Models and Capacity Planning
1 Real-Time Queueing Network Theory Presented by Akramul Azim Department of Electrical and Computer Engineering University of Waterloo, Canada John P.
Intermediate Care a range of integrated services to promote faster recovery from illness, prevent unnecessary acute hospital admission support timely discharge.
 1  Outline  stages and topics in simulation  generation of random variates.
Flows and Networks Plan for today (lecture 5): Last time / Questions? Blocking of transitions Kelly / Whittle network Optimal design of a Kelly / Whittle.
Network Analysis A brief introduction on queues, delays, and tokens Lin Gu, Computer Networking: A Top Down Approach 6 th edition. Jim Kurose.
Manijeh Keshtgary. Queuing Network: model in which jobs departing from one queue arrive at another queue (or possibly the same queue)  Open and Closed.
ECES 741: Stochastic Decision & Control Processes – Chapter 1: The DP Algorithm 31 Alternative System Description If all w k are given initially as Then,
Introduction to Queueing Theory
Networks of Queues Plan for today (lecture 6): Last time / Questions? Product form preserving blocking Interpretation traffic equations Kelly / Whittle.
Queuing Theory Basic properties, Markovian models, Networks of queues, General service time distributions, Finite source models, Multiserver queues Chapter.
Capacity Planning Tool Jeffery K. Cochran, PhD Kevin T. Roche, MS.
CS433 Modeling and Simulation Lecture 12 Queueing Theory Dr. Anis Koubâa 03 May 2008 Al-Imam Mohammad Ibn Saud University.
Tactical Planning in Healthcare with Approximate Dynamic Programming Martijn Mes & Peter Hulshof Department of Industrial Engineering and Business Information.
1 Chapters 8 Overview of Queuing Analysis. Chapter 8 Overview of Queuing Analysis 2 Projected vs. Actual Response Time.
Cancer Trials. Reading instructions 6.1: Introduction 6.2: General Considerations - read 6.3: Single stage phase I designs - read 6.4: Two stage phase.
The Extended Connection-Dependent Threshold Model for Elastic and Adaptive Traffic V. Vassilakis, I. Moscholios and M. Logothetis Wire Communications Laboratory,
Easter 2007 in London. Defining better measures of emergency readmission Eren Demir, Thierry Chaussalet, Haifeng Xie
Flows and Networks Plan for today (lecture 6): Last time / Questions? Kelly / Whittle network Optimal design of a Kelly / Whittle network: optimisation.
OPERATING SYSTEMS CS 3530 Summer 2014 Systems and Models Chapter 03.
Maciej Stasiak, Mariusz Głąbowski Arkadiusz Wiśniewski, Piotr Zwierzykowski Model of the Nodes in the Packet Network Chapter 10.
Queuing Theory.  Queuing Theory deals with systems of the following type:  Typically we are interested in how much queuing occurs or in the delays at.
Flows and Networks Plan for today (lecture 6): Last time / Questions? Kelly / Whittle network Optimal design of a Kelly / Whittle network: optimisation.
R. Kass/W03 P416 Lecture 5 l Suppose we are trying to measure the true value of some quantity (x T ). u We make repeated measurements of this quantity.
Queuing Theory Simulation & Modeling.
‘Environment’ Glossary Administrative categories from UK National Health Service.
Helen Lingham – Chief Operating Officer Gill Adamson – Director of Nursing and Operations.
Exposure Prediction and Measurement Error in Air Pollution and Health Studies Lianne Sheppard Adam A. Szpiro, Sun-Young Kim University of Washington CMAS.
©The McGraw-Hill Companies, Inc. 2008McGraw-Hill/Irwin Probability Distributions Chapter 6.
The Maximum Likelihood Method
OPERATING SYSTEMS CS 3502 Fall 2017
Prepared By : “Mohammad Jawad” Saleh Nedal Jamal Hoso Presented To :
Al-Imam Mohammad Ibn Saud University
Application of a simple analytical model of capacity requirements
The Maximum Likelihood Method
Time-dependent queue modelling
Noa Zychlinski* Avishai Mandelbaum*, Petar Momcilovic**, Izack Cohen*
IV-2 Manufacturing Systems modeling
System Performance: Queuing
Flows and Networks Plan for today (lecture 6):
An Algorithm for Bayesian Network Construction from Data
Buffer Management for Shared-Memory ATM Switches
An Accident and Emergency Case Study
Queueing Theory 2008.
CSE 550 Computer Network Design
Numerical Studies on Braess-like Paradoxes for Non-Cooperative Load Balancing in Distributed Computer Systems By Said Fathy El-Zoghdy, Hisao Kameda, and.
Erlang, Hyper-exponential, and Coxian distributions
Presentation transcript:

Queueing Network Approach to the Analysis of Healthcare Systems H. Xie, T.J. Chaussalet and P.H. Millard Health and Social Care Modelling Group (HSCMG) University of Westminster, London, UK

Outline Patient flow Queueing network models for patient flow Application Conclusion and future works

Introduction Patient flow –how patients move through a healthcare system –a major factor in improving efficiency Patient regarded as progressing through a set of logical stages in the process of care –diagnosis, treatment, rehabilitation, long-stay care, etc. In general, progressig through a set of (conceptual) stages, called phases phase 1 health care system discharge phase 2 phase M

Introduction (cont’d) Compartmental models are developed to explain patient flow through departments (e.g. geriatric medicine) –compartments representing different phases of care –study the long-run behaviour of a system e.g. expected number of patients in system, etc. –usually with no bed capacity constraints e.g. the system has as many beds available as required at any time Clearly, adding bed capacity constraint will provide a more realistic representation of the real world situation.

Queueing network models Queueing network models are widely used to study computer and communication systems –typically, network consists of many service nodes –for a service node service rate: how quickly jobs are dealt with arrival rate: how quickly jobs are arriving We model a healthcare system as a queueing network –a phase (or stage) is treated as a service node –the servers at a service node are all hospital beds performing the corresponding task at the time –the number of servers at a service node varies as patients move through care phases –small number of nodes with a variable number of servers

Queueing network models (cont’d) SSSM SMSM S L hospital ward S M L

Queueing network models (cont’d) Why queueing network? –enable access to a range of established methods dealing with queueing network models –a natural platform to incorporate a bed capacity constraint –direct calculation to stable-state solution (if exists) Bed capacity constraint –in total, K beds available in the system –patients will be refused admission if all beds are occupied Semi-open queueing network model –patients arrive following a Poisson process –system remains “open” when there is bed available –new arrivals are blocked (or lost) when capacity is reached

Queueing network models (cont’d) Basic quantity –For a M phase system, the joint probability distribution of the number of patients in each phase (i.e. a service node) is where, patient arrival rate, and is the normalisation constant. Derived quantities, e.g. –Expected number of patient in system ¼ ( l 1 ; l 2 ;:::; l M ) = 1 H ( K ) Q M i = 1 ½ l i i = l i ! ; f or P M i = 1 l i · K ; ½ i = ¸ e i ¹ i ¸ H ( K )

Queueing network models (cont’d) What is the use of such a model? –calculate the expected (and variance of) number of patients in system, i.e. occupancy level –how the beds are distributed among the phases –calculate the probability that the system is closed to new arrival, i.e. the loss probability –calculate the effective admission rate –how these measurements change with respect to changes in bed capacity, patient management policies

Queueing network models (cont’d) Special cases of the semi-open network –CLOSED network: system is always full the number of patients in each phase jointly follow a mutlinomial distribution –OPEN network: system has infinite capacity to admit patients the number of patients in each phase are independent Poisson

Queueing network models (cont’d) Model parameters –model parameters are arrival rate to system number of phases service rate (or average LOS) at each phase transition probability between phases –Estimate parameters from LOS data LOS follows Coxian distribution Coxian distribution can approximate any distribution with non- negative support arbitrarily well fit Coxian distribution to LOS data with increasing number of phases by maximum likelihood the “best” number of phases suggested by Bayesian Information Criteria (BIC) –Parameters can come from literature

Application Model parameters from literature (Parry, 1996) –analysing data from a geriatric department in England –three phases (stages) were identified acute care, rehabilitation and long-stay care Parry A. (1996) An age-related service revisited. In: Millard PH, McClean SI (editors) Go with the Flow. Royal Society of Medicine Press. pp acute care (9 days) rehabilitation (67 days) geriatric department long-stay care (863 days) new arrival (19 patients/day) 93%83% 7%17% K = 473

Application (cont’d) expected number of patients in system vs. bed capacity expected occupancy level vs. bed capacity %

Application (cont’d) expected occupancy level vs. arrival rate  bed capacity

Application (cont’d) loss probability vs. bed capacity loss probability vs. arrival rate 1.6%

Application (cont’d) loss probability vs. arrival rate  bed capacity

Application (cont’d) loss probability vs. bed capacity expected occupancy level vs. bed capacity

Application (cont’d) effective admission rate to the system vs. bed capacity 18.7

Application (cont’d) expected number of patient in system vs. % change in service rates loss probability vs. % change in service rates

Application (cont’d) expected number of patient in each phase vs. % change in service rates

Conclusion and future works Queueing network models are suitable for modelling patient flow –access to well established methods –natural platform for capacity constraint –quite general phases (conceptual states) are derived from data Future works –different types of patients (mutli-class model) –more than one system queues, blocking