International Journal for Research in Applied Science & Engineering Technology (IJRASET)

ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor:6.887

Volume 5 Issue VIII, August 2017- Available at www.ijraset.com

M/M/C: N/FIFO Queueing Analysis for Patient

Flow in Hospital
Rajalakshmi. R 1, Julia Rose Mary. K 2
1
MPhil Research Scholar, Department of Mathematics, Nirmala College for women, Coimbatore, India
2
Associate Professor, Department of Mathematics, Nirmala College for women, Coimbatore, India

Abstract: The increasing population and health-need due to adverse environmental conditions have increased the waiting times
and congestion in hospitals especially in the emergency and accident departments. In such cases, to enhance the level of
admittance to care, optimal beds required in hospital is needed and this can be achieved by adequate knowledge of patient flow.
In this paper, M/M/C: N/FIFO Queueing analysis of Patient flow in hospital is determined. Further their performance measures
and numerical example is analysed. In this model capacity of the system is finite.
Keywords: Mean queue length, M/M/C queue, server utilization, exponential distribution, bed occupancy.

I. INTRODUCTION
Queueing theory is the mathematical study of waiting lines (or) queues. The study of queueing systems finds application in a
variety of real life situations like regulating traffic flow scheduling and facility design. The theory provides models to predict the
behaviour of systems that attempt to render service for randomly arising demands. Thus the Queueing theory had its origin in 1909
when E.K.Erlang published his fundamental paper relating to the study of congestion in telephone traffic.
In- Patient flow one of the vital element is in improving the delivery of health care services. From a clinical perspective, in-patient
flow represents the progression of a patient’s health status. As such, an understanding of patient flow can offer education and insight
to health care providers, administrators, and patients about the health care needs associated with medical concerns like disease
progression or recovery status. Equally important, an understanding of patient flow is also needed to support a health care facility’s
operational activities. From an operational perspective, patient flow can be thought of as the movement of patients through a set of
locations in a health care facility. Then, effective resource allocation and capacity planning are contingent upon patient flow because
patient flow, in the aggregate, is equivalent to the demand for health care services (M. J Cote,[8]). The rising population and health
need due to adverse environmental conditions have led to escalating waiting times and congestion in hospital emergency
departments (ED) Derlet .R.W et al [4]. It is universally acknowledged that a hospital should treat its patients, especially those in
need of critical care, in a timely manner. Incidentally, this is not achieved in practice particularly in government owned health
institutions because of high demand and limited resources in these hospitals.

II. LITERATURE REVIEW

Traffic management, supermarket, health care and many other fields succeed in applying queueing theory for further
progress.Weiss and McCliam [10] used the M/G/∞ system to model the queue of patients needing alternative levels of care in acute
care facilities whose treatment is completed and are waiting to be transferred to an extended care facility. Adeleke R. A et al [1]
considered application of queuing theory to the waiting time of out- patients in a hospital. The average number of patients and the
time each patient waits in the hospital were determined. Likewise in his paper Worthington [11] used queuing theory to model
hospital within list. He used an M/G/C queue with state dependent arrival rate to address the long- wait list problem. Through some
appropriate means he experimented with different kinds of management actions including increasing the number of beds decreasing
the mean service time etc. DeBruin et al [3] investigated the emergency in- patient flow of cardiac patient in an hospital in order to
determine the optional bed allocation so size of a hospital unit, occupancy rate and target admission rates. After analytically
estimating the required number of beds in first cardiac Acid (FCA) unit, they also used numerical method to estimate the number of
beds in the Critical Care Unit (CCU) and Normal care clinical ward (NC). Jonathan E. H et al [7] characterize an optimal admission
threshold policy using control on the scheduled and expedited gate way for a new Markov Decision process model. In their work,
they presented a practical policy base on insight from the analytical model that yield reduced emergency blockages, cancelations

©IJRASET (UGC Approved Journal): All Rights are Reserved 2163

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor:6.887
Volume 5 Issue VIII, August 2017- Available at www.ijraset.com

and off- units through simulation based on historical hospital data. Recently, Adeleke R. A et al [9] considered queueing Analysis of
patient flow in Hospital.
Application of queueing theory to model health care is growing more popular as hospital management teams are becoming aware of
the advantages of these techniques. In this research we will use both analytical techniques and simulation to study a simple queuing
network composed of only two service stations placed in tandem. In this paper, we studied all admissions into the Emergency and
accident Department (EAD) of a tertiary hospital. We will show that admissions into this system has a Poisson distribution, hence it
has exponential inter-arrival rate. We also examine the average length of stay, the occupancy rate and we determine the optimal bed
count in the Intensive Care Wards (ICW) and the Medical and Surgical Wards (MSW). Since the ICW and MSW have multiple
beds we will consider the M/M/c queue.

III. RELATED WORK

A. Finite Capacity
In an M/M/c/N queue only N customers can queue at any one time (including those in service). Any further arrivals to the queue are
considered "lost". We assume that N≥ c. The model has transition rate matrix
   
 
  (  )  
 2  (2  )  
 
 3  (3  )  
  
Q  
 C  (C  )  
 
  
 C  (C  )  
 
  
 c  c 

On the state space {0, 1, 2, ...,c, ..., N}. In the case where c =N, the M/M/c/N queue is also known as the Erlang–B model. This
model is same as M/M/C: (  /FIFO) except that the maximum number in the system is limited to N. where N  c. therefore
utilizing the steady state Probabilities of
n   if 0 n  N and
0 otherwise
 n  n if 0nc
c if cn N
We get,
 ( )n 
  p 0 nc 
0
 n! 
pn   
 (  )
n

 p c  n  N
 c! c nc 0 

Now for   1, the normalization condition  pn  1 gives
c n 0
1
 c1 (  ) n (  ) c  N c 1 
       c   
p0     1     
 n0 n! c!   c   c     
 
By Noting that the probability an arrival unit has to wait on arrival is given by the probability,
N ( )c  c pc
p( N  c)   pn  (1  ( ) N C 1 )( ) p0  (1   N C 1 )
n c c! c c   (1  )
We now proceed to compute some performance measures.

©IJRASET (UGC Approved Journal): All Rights are Reserved 2164

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor:6.887
Volume 5 Issue VIII, August 2017- Available at www.ijraset.com

B. Performance Measures
The expected queue length L can be computed as,
N
L=
 (n  c ) p
n c
n

(  ) c 1 p0 1   N C 1  ( N  c  1)(1   )  N c 

  
c c!  (1   ) 2 
Where     1 is referred to as the server utilization.
c
C. Expected number of busy and idle servers:
The expected number of busy servers E(B) is given by
c 1 N
E(B)=
 np   cp
n 0
n
nc
n


( )  n 1
(  ) n1 
   c1   N
p 
 p 
  n0 ( n  1)! 0 n0 (c  1)! c nc 0 
 
  m 
   c 2 (  ) (1     )  c1   N C 1  1 

  m0 m!

(c  1)! 
( )  1  (
c
)  (1   )  p0
 
 
   
  m ( ) c1 ( )c
  c 2 (  )    N C 1     N C 1  1 
  1 ( )   c! 1  ( c )  (1   )  p0
 m 0 m! (c  1)!  c    
 

  
 m ( )c
  c 1 (  )    N C 1  1 
   1  ( c )  (1   )  p0
 m0 m! c!   
 
 sin ce   1 so  N c1  0
  c

Hence the expected number of idle servers E(I) is given by
E(I)=E(c-B)=E(c)-E(B) =c-c  = c(1   )
Applying little’s formula we also obtain expected waiting time in the queue.

w
Lq
 c 0

 p p  1   N C 1  ( N  c  1)(1   )  N c  

   (1   ) 2 

IV. LENGTH OF STAY DISTRIBUTION

The number of days in hospital for a patient is described by the term length of stay (LOS). LOS is defined as the time of discharge
minus time of admission. Following, the average length of stay is abbreviated as ALOS.
total disch arg e days
The average length of stay(in days)= or
total disch arg es
total inpatient days of care
The average length of stay(in days)=
total admission
Below are the definitions for each of the four data items included in the above calculations.
TOTAL DISCHARGE DAYS-the sum of the number of days spent in the hospital for each inpatient who was discharged during the
time period examined regardless of when the patient was admitted.

©IJRASET (UGC Approved Journal): All Rights are Reserved 2165

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor:6.887
Volume 5 Issue VIII, August 2017- Available at www.ijraset.com

TOTAL DISCHARGES- the number of inpatients released from the hospital during the time period examined. This figure includes
deaths. Births are excluded unless the infant was transferred to the hospital’s neonatal intensive care unit prior to discharge.
TOTAL INPATIENT DAYS OF CARE- sum of each daily inpatient census for the time period examined.
TOTAL ADMISSIONS- the total number of individuals formally accepted into inpatient units of the hospital during the time period
examined. Births are excluded from this figure unless the infant was admitted to the hospital’s neonatal intensive care unit.

V. BED OCCUPANCY
It is common practice in health services to estimate the required number of beds as the average number of daily admissions times
average length of stay in days and divided by average bed occupancy rate(average number of occupied beds during a day ) Huang X
(1995)
average no.of daily admissions
Bed requirement=  average length of stay
average bed occupancy rate
Hospital bed capacity decisions have been made based on Target occupancy rate (TOR)- the average percentage of occupied beds
and the most commonly used occupancy target has been 85% Linda V. Green [5]. Another metric often cited in the literature is the
target access rate (TAR), which measures the percentage of the time that a census count will show that the hospital contains at least
one empty bed, Kumar and john[2].

VI. NUMERICAL SOLUTION

The occupancy rate (  ) is related to the real demand (  ) and LOS (  ) and can be defined as follows,
average number of beds occupied (1  pc )
 
number of beds available c
Table .1 Server utilization for M/M/C:N/FIFO Model
 0.01 0.02 0.03 0.04 0.05

C=1 0.0019 0.0036 0.0052 0.0067 0.0081

C=2 0.0009 0.0019 0.0029 0.0039 0.0048

C=3 0.0006 0.0013 0.0019 0.0026 0.0033

C=4 0.0004 0.0009 0.0014 0.0019 0.0024

C=5 0.0003 0.0007 0.0011 0.0015 0.0019

0.015

0.01

0.005

0
1 2 3 4 5

Graph(1)  versus( , c)
From the table and graph, we conclude that, the server utilization factor(  ) increases when the arrival rate increases. Also, we find
that the server utilization factor(  ) decreases when the number of beds increases.

©IJRASET (UGC Approved Journal): All Rights are Reserved 2166

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor:6.887
Volume 5 Issue VIII, August 2017- Available at www.ijraset.com

The term (1  pc ) can be entitled as the effective demand as the refused admissions are subtracted from the real demand.
Furthermore, the product  which is the expected number of patients in the system is also known as the workload of the system.
Many hospitals use the same target occupancy rate for all hospital units, no matter the size of the unit. The target occupancy rate is
typically set at 85% and has developed into a golden standard (green [5]). The conclusion is clear and important. Larger hospital
units can operate at higher occupancy rates than smaller ones while attaining the same percentage of refused admissions. Therefore,
one target occupancy rate for all hospital units is not realistic.
Total admission into ICW=50, ALOS in ICW=4.44 days, percentage of patients reneged, k=3.4%. we also have the following set of
data for the MSW. Total admission into MSW=100, ALOS= 6 days.
From the parameter values specified, we estimate the arrival rate to each station as,
N ICW N
ICW   1.67days 1 MSW  MSW  3.33days 1
30days 30days
But the queue leading to the MSW is composed of new arrivals and blocked patients from the ICW. Also we have only a fraction 1-
k=96.6% of patients arrived into ICW without reneging during service. So that the effective arrival into the ICW is
eICW  ICW (1  k )  1.613days 1
eMSW  eICW   ICW  3.283days 1

Table .2 Performance measure for ICW

C1 % server utilization Mean waiting time in the Mean waiting time in
queue the system

10 83.5 0.0002 4.4402

11 75.9 0.0001 4.4401
12 69.58 0.0000 4.4400
13 64.23 0.0000 4.4400
14 59.64 0.0000 4.4400
15 55.66 0.0000 4.4400
16 52.18 0.0000 4.4400
17 49.11 0.0000 4.4400
18 46.38 0.0000 4.4400
19 43.94 0.0000 4.4400
20 41.75 0.0000 4.4400

100
% server utilization
80
60
Percenta

40 % server
utilization
ge

20
0
1 2 3 4 5 6 7 8 9 10 11
no of beds

©IJRASET (UGC Approved Journal): All Rights are Reserved 2167

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor:6.887
Volume 5 Issue VIII, August 2017- Available at www.ijraset.com

Table .3 Performance measure for MSW

C2 % server utilization Mean waiting time in the queue Mean waiting time in the
system

21 95.14 0.0235 0.1835

22 90.81 0.0225 0.1825
23 86.86 0.0215 0.1815
24 83.25 0.0209 0.1809
25 79.92 0.0202 0.1802
26 76.84 0.0196 0.1796
27 74 0.0191 0.1791
28 71.35 0.0185 0.1785
29 68.89 0.018 0.178
30 66.66 0.0175 0.1775

% server utilization
100
90
80
70
60
Percentage

50
40
30
% server utilization
20
10
0
1 2 3 4 5 6 7 8 9 10

no of beds

Table 2 and 3 shows the result for various values of C1 and C2 . From the tables we can see that C1=12 guarantees that there is no
waiting at the EAW, since urgent Patient needing urgent care are brought in through it. In the MSW, C2=28, will guarantee an
approximate of 71.35% server utilization and a minimum waiting time in queue.

VII. CONCLUSION
In this work, we analysed a queueing network model with reneging to study how waiting time in the Emergency and Accident
department (EAD) of an Hospital is influenced by the number of beds in the ICW and MSW. The system was decomposed into two
independent multi-server queues so as to obtain estimates for the required number of beds in the wards. We found that the required
number of beds to ensure that emergent patients are promptly attended and there is easy flow is approximately 12 in the ICW and
28 in the MSW for the test hospital under consideration.

©IJRASET (UGC Approved Journal): All Rights are Reserved 2168

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor:6.887
Volume 5 Issue VIII, August 2017- Available at www.ijraset.com

REFERENCES
[1] Adeleke. R. A, Ogunwale O.D, Halid O. Y(2009) Application of queueing theory to waiting time of out patients in hospitals, Pacific Journal of science and
technology Vol.10(2) 270-274.
[2] Arun Kumar and John Mo (2010) Models for bed occupancy Management of a hospital in Singapore. Proceedings of the 2010 International Conference on
Industrial Engineering and Operations Management Dhaka, Bangladesh.
[3] Debruin A.M, A.C Van Rossun MC Visser, GM Koole (2006) Modeling the Emergency Cardiac In-Patient flow : An application of queueing theory. Health
care Management Science Vol. 10, No 2, pp125-137.
[4] Derlet R W, John R Richard, Richard L.kravitz(2006) Frequent overcrowding in U.S Emergency Departments, Academic Emergency Medicine, vol 8,issue
2,pp 151-155.
[5] Green L.V (2002) How many beds Decision, Risk and Operation. Working Paper series
[6] Green L.V, soares J, Giglio JF, Green R A(2006) using queueing theory to increase the effectiveness of emergency department provider staffing. Academic
Emergency Medicine 13(1):61-68.
[7] Janathan. E H,Shervin Ahmed Beygi, Mark P. Van Oyen(2009), design and Analysis of Hospital of Hospital Admission Control for Operational Effectiveness
Technical Report 09-05. University of Michigan. Michigan
[8] Murray J. Cote (2000) understanding patient flow. Production/Operations Management Decision Line pp8-10
[9] Olorunsola S. A, Adeleke R.A and ogunlade T.O (2014) Queueing analysis of patient flow in Hospital, IOSR Journal of Mathematics Vol.10 ISSN:2278-5728.
[10] WeissE.N, J.O Mcclain (1986) Administrative days in acute care Facilities: a queueing analytic approach Operations ResearchVol.35 no 1 pp 35-44.
[11] Worthington D.J (1987) queueing models for Hospital waiting list. The journal of the Operational Research Society Vol.38, No 5,pp413-422.

©IJRASET (UGC Approved Journal): All Rights are Reserved 2169