Case Study: ED Congestion at

Saintemarie University Hospital

Prof. Sean Zhou
Decisions, Operations and Technology
CUHK Business School
Learning Objectives

• Learn the impact of variability in process performance

• Be able to identify the sources of variability
• Be able to propose concrete measures to reduce
waiting in the process

2
Hospital Background
• Only ED department in the metropolitan area, with an alternative
50 miles away
• More than 2,000 beds, ED employs more than 250 staff including
60 doctors (30 interns, 25 junior, 5 senior), 150 nurses, 40 admin
staff
• Numbers of patients are stable over the years, but patients time
in ED has increased from 4 hours in 2006 to 5 hours in 2009
• Congestion/long waiting in ED is a common problem in many
countries/regions, e.g., US, Hong Kong
• ED delay causes increased mortality and length of hospital stay
for critically ill patients

3
What is the admission process of a patient visiting ED?

4
Process Flow Diagram

Initial Wait Patient Management

Triage and
registration and Discharge

3 mins plus 10 mins of 1 hour 10 min 3 hours and 50 mins (40 mins of
registration discharge time)

5
Patient Admission

• Once arrive, first seen by a first-line nurse to conduct

the triage, 2-3 mins; then officially registered (10 mins)
• Classified into four groups depending on acuteness
• Degree 1 (most acute): 8/day
• Degree 2: 33/day
• Degree 3: 119/day
• Degree 4: 5/day

6
ED Initial Wait

• Initial wait: after triage and registration, patients wait

for a room in a dedicated area under supervision of a
nurse, average 1 hour and 10 mins, but could be as
long as 10 hours; approximately 5 patients left without
being seen by a doctor per day
• 40 exam rooms, 25 for the whole day, 15 closed for
11pm-8am

7
Patient Management

• Patient management: average 3 hours and 10 mins,

while acute cases may take up to 10 hours
• 40% require lab tests, two hours between prescription for the
test and the test results come in
• 30% require X-ray, 15% require CT scan; CT takes 30 mins,
getting results takes 3 hours. CT scanner empties for 10
mins between patients
• 25% need advice from another specialist: need additional 2
hours, one hour for the specialist to come down to ED and
another to examine the patient
• Once all results are in, it takes 45 mins to make decision
(diagnosis)

8
Patient Discharge
• Patient discharge
• On average, it takes 40 mins between diagnosis and leaving exam room
• Three possible destinations:
• home (60%)
• observation unit (20%), 32 bed, one hour for the transfer, with many are waiting
to be admitted as inpatient
• another department of hospital (20%), the beds of departments have very high
occupation rate (90%), the transfer took slightly more than 1 hour

• Patients spend 3 hours and 50 mins on average in patient management

and discharge (occupying a room), with the standard deviation of 3
hours.
• Can Emma DuPont reduce the waiting time in ED without increasing
costs?

9
Detail Flow of Patient Management and Discharge Phase

Vital First
Laborat Radiology Final
signs evaluati Specialist discharge
ory test exam decision
on

40% of 30% X-ray 25%, 2 hours 45 mins 40 mins

patients, 2 15%t CT
hours
average

10
Question 1: What operational problems is the
Saintemarie ED facing? What is your assessment of the
current performance? What do you think is driving these
problems?

11
Problems of ED Operations
• Long waiting time
• Service Quality: degree 1 patients waiting time target is met; but only
two-thirds of degree 2 patients can be seen by a doctor within 20
mins. Another concern is that 5 patients left the hospital per day.
• HR: low morale, stress, patient dissatisfaction
• Economic: due to long wait time, some patients chose private clinics
and some left without being seen by a doctor, resulting lost of
revenues; additional staff costs of monitoring waiting room; risk of
medical complications
• Reputation

12
Question 2: What are the main sources of variability?

13
Patient Arrival Process

• Around 164 patients visit ED per day, stable, no

seasonal/weekly trend, except Monday is busier and
Sunday lighter
• Arrival times (inter-arrival time) of patients are
variable/uncertain
• Peak hours: 9am-8pm, average 111 arrivals
• Seasonality vs. variability

14
Patient Management and Discharge

• Different priority classes/diseases

• Some patients need lab tests (40%), others no
• Some need radiation (30%, 15% of CT scan), others
no, CT scan has unnecessary idle time
• Some need to see specialists (about two hours), other
no
• Some need to wait for bed to be admitted into the
hospital (becoming inpatient), others no
• 15 rooms closed from 11pm to 8am

15
Question 3: Evaluate the average time that patients wait
before entering the care process. Assume patient arrival
follows a Poisson process.
• If not distinguishing peak and off-peak time, what is the
utilization in the ED and the average waiting time?
• The utilization in the ED during the peak time (9am-8pm) and
off-peak time
• The average time a patient would have to wait before being
seen during the peak time and the offpeak time . How does
your result compare to the actual average wait time provided
in the case (1h.10mins)?

16
Waiting Time for Multiple, Parallel Servers (Approximation)

Utilization :
customers Waiting Time Formula for Multiple (m) Servers:
in service Ls

Queue length Lq

( )(
𝑠𝑒𝑟𝑣𝑖𝑐𝑒𝑡𝑖𝑚𝑒 𝑈𝑡𝑖𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛√ 2 ( 𝑚+1 ) −1 𝐶𝑉 𝑎 +𝐶 𝑉 𝑝 1
)
Inflow Outflow 2 2
𝑊 𝑠= × × +
𝑚 1−𝑈 𝑡𝑖𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛 2 𝜇

Entry to system Begin Service Departure

By Little’s law,
Time in queue Wq Service Time 1/
• The number of customers in
Flow Time Ws=Wq+1/ queue:
• The number of customers in
system:

17
Inputs
• Arrival rates (Poisson arrival)
• peak: /hour
• off-peak: /hour

• Number of servers
• peak hours (11 hours): =40 servers
• Off-peak hours (13 hours): servers
• Daily average:

• Service time (total in patient management)

• Mean: 3 hours and 50 mins (including 40 mins of discharging time) or /min=
• Standard deviation: 3 hours

18
Utilization of Servers
• Overall

• Utilization during Peak hours

• Utilization during off-peak hours

19
Variability

• Service time coefficient of variation CV p :

Standard deviation of service time 3  60

CV p  
Average service time 3  60  50

• Interarrival time variability CVa :

• Poisson arrival process exponential interarrival
time standard deviation = mean

20
• Question 4: Shall DuPont consider investing in
dividing the ED into two separate units: inpatient
(acute cases) and outpatient (non-acute cases)?
• Pros and cons? Pool vs. independent servers

21
• Pros
• Reduce the wait time of non-acute patients, then likely reduce
LWBS
• Allow process and staff specialization
• Avoid switching costs between outpatients and inpatients
when pooled

• Cons
• Separating resources reduces efficiency
• Average waiting time likely will increase
• Most acute patients’ waiting time will increase
• Classification errors may occur
• High infrastructure and equipment investment
22
Question 5: What additional (or alternative) measures
would you recommend for improving Saintemarie’s ED
performance?

23
Shorten Service Time/Reduce Variabilities/Adjust Capacity

• Optimize CT scan usage

• Optimize discharge process
• Increase or decrease the number of servers, e.g., rooms, doctors
• Reschedule the intern lecture, currently scheduled at 11am, the
peak of patient arrival
• Standardize processes, reduce the need of advice from
specialists
• Outsource the reading of lab results, economies of scale
• Pool different EDs in the area (think about the practice in HK)
• Manage demand so as to reduce the number of patients to ED

24
Optimizing CT

• Current process: technician calls the nurse to pick up

the patient and then calls the nurse of the next patient,
10 mins CT idle time
• Current processing time: 40 mins
• How about calling the nurse of next patient 10 mins
earlier?
• One server, Poisson arrival (general, , service time std:
assume 0.5 (optimistic), when average CT service time
changes from 40 mins to 30 mins, what is the reduction of
waiting time?

25
Optimize discharge process

• In a multi-stage process, performance of one stage

often relies on the others
• Centralized admissions of hospital, bed pooling of
different units
• Shorten admission delay
• Hard to implement in practice, each doctor wants to
control admission of his/her unit

26
Summary of What are Implemented

• Real-time allocation of nurses and doctors

• Improve of CT scan utilization, wait time of CT scan
reduced by 30 mins
• Improve collaboration with specialist consultants, wait
time reduced by 25 mins

27
Takeaways

• Variabilities introduce great challenges in managing

processes. Therefore, in the presence of variabilities,
a small improvement, e.g., reduce cycle time of some
step, can have a significant impact
• The performance of separate systems/processes can
be highly correlated, e.g., the CT scan process affects
the overall waiting at the ED; only when interacting
systems are improved, the full benefit of improvement
can be achieved

28