HCM 540

1
HCM 540 Healthcare Operations Management

• Process Flow Basics
• (Chapter 3 in MBPF)

2
General 4-stage framework for managing healthcare
resources (staff and physical capacity)

1. Demand/workload characterization and forecasting
2. Translation from demand to capacity
3. Scheduling
4. Short-term allocation

The details of these 4 stages all vary depending
on the specific healthcare context.
3
1. Demand/workload characterization

• Basic process flow physics
• How the work flows
• Occupancy/census/inventory/work in process
  analysis
• TOD/DOW nature of workload
• Healthcare operational data
• Getting data about workload
• Patient/work classification systems
• Different types of work require different levels
  of resources
• Forecasting
• Predicting future workload from past and other
  causal factors
• Work measurement and productivity monitoring
• Understanding the inputs and outputs relationship
• Important component of staffing analysis

4
2. Demand ? Capacity

• Labor and physical capacity costs dominate in
  healthcare
• Queueing and simulation models might be useful
  for helping to set capacity levels
• when tradeoffs between capacity cost and patient
  delay and/or access is important
• hospital bed allocation, ancillary staffing
• surgical block allocation, clinic capacity
• Staffing analysis
• standards, nurse-patient ratios, variable vs.
  constant tasks, benefit allowances, benchmarking

5
Good Resources for healthcare operations info and
ideas

• Institute for Healthcare Improvement -
  http//www.ihi.org/
• Family practice web site - http//www.aafp.org/
• Journal has nice toolbox - http//www.aafp.org/x75
  02.xml
• Healthcare management engineering mailing list
  HME group in Yahoo groups
• Very active practitioner forum about process
  improvement, operations management, industrial
  engineering, etc. in the healthcare industry
• Knoxville ED Study
• See course website for PPT, report and xls file
  for this nice study which was done by a professor
  at Univ. of Tennessee and a management
  engineering group

6
I. Business Process Perspective on Healthcare
Delivery
Process Management
Network of Activities
Inputs
Outputs

• patients, test results
• bill, resolved complaint

• patients
• specimens
• phone calls, charts
• complaints

• Uses resources (capital labor)
• Visit multiple locations
• nursing care, test processing, chart coding
• Value add and non-value (delays)

Information
7
Flow Units Attributes

• Flow units things that flow through business
  processes
• Ex patient, information, cash, people, supplies,
  test results, exams, paper
• Attributes characteristics of flow units
• Ex patient type, acuity, length of stay,
  admission origin, discharge status

Each attribute like index card in a pocket
HW1 examples of Processes, Flow Units, Attributes?
8
As Entities Flow

• Generated (enter system)
• ED, walk-in, call for appointment, specimen
  arrives at lab, charts to medical records and
  billing, patient admitted
• Attributes checked and/or set
• time of arrival, preliminary diagnosis, urgency
  status noted, surgical case type, IP or OP, DRG
• Resources gotten and released
• registration clerks, nurse, physician, bed,
  imaging equipment, transporters, biller, customer
  service rep
• Locations visited
• inpatient units, ED cubicle, waiting room,
  radiology, lab, waiting areas
• Get processed and/or transformed
• care delivered, procedure done, bill generated,
  chart filed, diagnosis made
• May be delayed, combined, split, rejoined, and
  eventually exit the system

9
An Urgent Care Clinic
Start/Enter
Wait
Register
Complete HHQ
Wait
Start/ntr
Vitals/ Assessment
Wait
Provider Contact Exam
Wait
Diagnostic/ Intervention
Wait
Provider Contact/ Results
Wait
Discharge
Collections
MCHC Pharmacy
Wait
Outside Pharmacy
Wait
Leave
Finish
Patients visit a series of queueing systems in
series
10
iGrafx Process
11
Basic Operational Flow MeasuresCh 3 of MBPF
Inputs
Outputs
Processing System
Flow Rate or throughput average number of
flow units (entities) that flow through a certain
point in a process per unit time
R
Flow time processing time wait time (total
time in the box)
T
Occupancy or Inventory number of flow units
within the boundaries of some process
I
I units of inventory T avg flow time
R units/time
R units/time
12
Throughput (Flow Rate) Concepts

• Throughput rates are the number of flow units per
  unit time
• admits/day, tests/hour, phone calls/hour, /month
• Flow is conserved what flows in, must flow out
• Inflow and outflow fluctuate over short term
• In gt Out ? Occupancy, queue or inventory grows
• Out gt In ? Occupancy, queue or inventory shrinks
• Long term stable process
• Flow In Flow Out
• Can combine and split flows

Ri2 clinic walk-in patients per day
Ro total flow of patients out of clinic per day
Process (Tflow time in clinic)
Ro Ri1 Ri2
Ri1 scheduled clinic patients per day
13
Flow Time Concepts

• Flow time is amount of time spent in some process
• May include both waiting and processing
• Its a duration and measured in units of time
• length of stay, exam length, processing time for
  a test, procedure length, time to register,
  recovery time
• Service rate 1/avg flow time
• Example avg flow time 0.5 hours ? service rate
  of 2/hr
• Flow time varies for individuals and/or different
  types of flow units
• consider average flow time for now

What is overall average time in dotted box?
R1
20 pats/hr
R1 type 1 flow in
Type 1 Flow Time10 mins
R1R2
Type 125 mins
5 pats/hr
R2 type 2 flow in
Type 2 Flow Time20 mins
R2
14
Flow Time, Flow Rate, and Inventory Dynamics
Ri(t) instantaneous inflow rate at time t Ro(t)
instantaneous outflow rate at time t DR(t)
instantaneous inventory (occupancy) build up rate
at t
DR(t) Ri(t) - Ro(t)
If Ri(t) gt Ro(t) ? get buildup at rate DR(t) gt 0
If Ri(t) Ro(t) ?get no change in occupancy
If Ri(t) lt Ro(t) ? get depletion at rate DR(t) lt 0
15
Example Constant DR during (t1,t2)
In other words, during the time period (t1,t2),
occupancy is being depleted or is building up at
a constant rate DR.
Occupancy change Buildup Rate x Length of Time
Interval
O(t2)-O(t1) DR(t2-t1)
Example If system empty at t1, and DR3
people/minute, how many people are in the system
after 10 minutes?
16
(No Transcript)
17
Occupancy Inventory can be averaged over time
for stable processes
At 1010 the inventory will start to build again
for next flight.
Inventory 0 from 943-1010
(27 mins)
So, whats the average inventory in here (from
910-943)? Hint How can we interpret the AREA
of this triangle?
Avg inventory (33(30) 27(0))/60 minutes
16.5 people
18
Littles Law IRTAverage occupancy Throughput
x Avg. Flow Time
Stuff in system Rate stuff enters x How long it
stays
x
I
T
R

/
/
T

I
R
R

I
T

• If you know any two, you can calculate the third
• You choose what to manage and how
• Relationship between some important averages
• Can be applied to many different types of
  business processes
• Put Littles Law into Google and youll see the
  wide variety of applications of this basic law of
  systems

19
Simple Applications of Littles Law

• Avg Customers in Line Customer arrival rate
  Avg Time in line
• Length of billing cycle in Accounts Rcv / Avg
  Sales per Month
• Avg Hospital Daily Census Admission Rate Avg
  Length of Stay
• Avg customers at web site Hit Rate Avg Time
  Spent at Site
• Work in process work input rate Avg
  Processing Time

20
In class flow analysis (handout)

• Patient Flow Model 01
• one patient type, one unit, infinite capacity
• average arrival rate and length of stay given
• Patient Flow Model 02
• two patient types with different average length
  of stay
• Exercise 3.10 in MBPF
• A little Hotel Occupancy problem (we can always
  learn from other industries)

21
Littles Law in action

• Typical daily census report
• Monthly summary similar may include comparison
  to previous month or same month last year
• What does this show?
• How created?
• What doesnt this show?

The numbers reported in the Free Press a few
years ago.
22
Beyond Averages

• Littles Law is about averages
• Average may be meaningless
• Example bimodal distribution from pooling long
  and short procedure times, extreme DOW volume
  swings
• Upper percentiles
• 90 of calls answered in less than 1 minute
• 95 of the time we have lt 200 patients in house
• Time of day and/or day of week (TOD/DOW) effects
  may be significant
• Seasonal effects may be significant
• Range
• be careful with minimums and maximums
• Example from ED consulting report
• Hands on lets create some histograms of real
  healthcare data
• Well do this with some real length of stay data
  momentarily

23
Hospital Census Data

• Hard to tell if DOW effect present
• Impossible to see TOD effect since data is daily
• Seasonality?
• At time exceed capacity?
• data quality?
• is capacity correct?
• census reflects patient type

24
Enhanced Census Reporting Examples

• Bed Allocation Committee Monthly Report
• Used _at_ monthly meeting of stakeholders to assess
  occupancy issues
• Daily, weekly census, Overall M-Thu summaries,
  30-60-90 day trends, unit group summaries,
  validity checks
• Obstetrical Occupancy Reports
• Used as part of planning for OB expansion

Note Data values and sources have been modified
to preserve confidentiality.
25
Raw Data
Summary Data
26
Discharge timing by hour of week
TOD/DOW Avg. and 95ile
DOW
Occupancy frequency distribution
Discharge timing by hour of day summary
27
Analysis of Time of Day Dependant Data

• Many processes in healthcare have important
  TOD/DOW effects
• high variability and uncertainty in timing of
  arrivals and length of stay (or duration of
  process)
• overall averages simply not that useful
• timing of arrivals, occupancy and discharges
  drives staffing and capacity planning
• Examples recovery holding areas, emergency, IP
  OB, walk-in clinics, call centers, short-stay
  units
• Applies to any units of flow such as tests, phone
  calls, patients, nursing requirements

28
If Arrivals and LOS are Random Variables
29
Then, occupancy is certainly a random variable
that depends on TOD and DOW
Question See p34 in IHI Guide. What exactly is
Figure 3.1 showing?
30
Hillmaker A Tool for Empirical Occupancy
Analysis

• Data has in/out date-timestamp
• admit/discharge, start/stop, enter/exit, etc.
• Example entry and exit times from a surgical
  holding areas was available in surgical
  scheduling system
• Interested in arrival, discharge, occupancy
  statistics by time of day and day of week
• mean, min, max, and percentiles
• Time bins ½ hr, hr, 2hr, 4hr, 6hr, 8hr
• Example mean and 95ile of occupancy with ½ hr
  time bins
• Want statistics by some category or
  classification of interest as well as overall
• Example category created was combination of
  location (which holding area) and phase of care
  (preop, phase I, phase II)
• Freely available from http//hillmaker.sourceforge
  .net/

31
Why Hillmaker needed?

• Many processes in healthcare have important
  TOD/DOW effects
• high variability and uncertainty in timing of
  arrivals and length of stay
• overall averages simply not that useful
• timing of arrivals, occupancy and discharges
  drives capacity planning
• Examples recovery holding areas, emergency, IP
  OB, walk-in clinics, call centers, short-stay
  units
• Applies to any units of flow such as tests, phone
  calls, patients, nursing requirements, dollars,
  specimens, staff, etc.
• Provides important first step in applying
  stochastic patient flow models such as simulation
  or queueing
• Estimation of arrival rate parameters
• Standard hospital information systems usually are
  very weak in area of TOD/DOW metric reporting
• Consider the traditional inpatient census report
• Can you explain percentile again to me? said
  the manager.
• Obsession with averages and uncomfortable with
  distributions
• Yes, Im amazed that such tools arent standard
  fare in a healthcare managers arsenal

32
What Hillmaker Does
Scenario data (in/out/ category)
Hillmaker (Access)
Graphing Templates
Arrivals, discharges, occupancy by
DateTime-category
Arrivals, discharges, occupancy summaries by
TOD-DOW-category
33
In/Out Data
34
Hillmaker Interface
Data source inputs
Date/time related inputs
Algorithmic options
Output products
35
(No Transcript)
36
A portion of Excel graphing engine
37
Day of week graphs
38
Getting Hillmaker

• http//hillmaker.sourceforge.net/
• Isken, M. W., Hillmaker An open source
  occupancy analysis tool. Clinical and
  Investigative Medicine, 28, 6 (2005) 342-43.
• Ceglowski, R. (2006) Could a DSS do this?
  Analysis of coping with overcrowding in a
  hospital emergency department, Nosokinetics News
  (http//www2.wmin.ac.uk/coiec/Nosokinetics32.pdf),
  3(2) 3-4.

39
Sources of Internal Workload DataMeasuring Flow
Time Rate

• Departmental information systems
• lab, radiology, surgical scheduling, nursing, ED
  patient tracking, patient transport
• Hospital information systems
• Reg ADT, billing, appointment scheduling, finance
• Data warehouses and data marts
• Management engineering, finance, planning,
  marketing
• Clinical data repositories
• Log books, tally sheets, hard copy reports
  (yuck!)
• Will devote a session to business intelligence
  technology
• data warehousing, OLAP, data mining
• Getting data out of information systems
• Tips for data collection
• See p38 in IHI Guide
• Ill show you some techniques for Excel based
  data collection tools

40
Patient Classification

• What are our products and services?
• What types of workload drives demand?
• classifying workload into a manageable number of
  different classes facilitates forecasting and
  capacity planning models that are robust to
  changes in workload mix
• A myriad of classification schemes exist for both
  patient types, procedures, tests
• Well look in detail at productivity monitoring
  schemes and nursing classification schemes when
  we discuss staffing in a few weeks

41
Guiding Principles for Classification Schemes

• Similar bundle of goods and services in diagnosis
  and treatment of patients
• similar resource use intensity
• Based on readily available data
• administrative data, clinical data
• Manageable number of classes
• Similar clinical characteristics within a class
• medically meaningful

42
Sampling of Patient Classification Systems

• MDC, DRG the basic for PPS
• CCS Clinical Classification Software
• AHRQ developed for health service research
• CSI, Disease Staging, MedisGroups, RDRG, APR-DRG,
  SRDRG severity based systems
• APG, APC outpatient version of DRGs
• Service a simple proxy often used internally
  (e.g. based on attending physician, surgeon,
  etc.)
• Nursing Unit / Unit Type - another simple proxy
• ignores effect of overflows

43
Why is classification hard?

• Not all diseases well understood
• Treatments for same disease differ
• Coding illnesses is difficult
• some classes too narrow, some too broad
• Tradeoff between manageable number of classes and
  within class homogeneity
• Severity matters
• Administrative easily available but other data in
  chart more expensive to obtain
• Different classification schemes needed for
  different purposes
• resource allocation, financial reimbursement,
  outcomes analysis

44
DRGs

• Originally intended as production definition for
  hospitals (devd _at_ Yale by Fetter et al 70s
  early 80s)
• To serve as basis for budgeting, cost control and
  quality control
• Adopted by Medicare in 1983 for PPS
• Based on MDC (medical and surgical), ICD9-CM
  codes, age, some comorbidities complications
• Statistical clustering along with expert medical
  opinion
• See Fetter article in Interfaces for very nice
  description of DRG development

Diagnosis Related Groups Understanding Hospital
PerformanceFetter, Robert B.. Interfaces.
Linthicum Jan/Feb 1991. Vol. 21, Iss. 1 p. 6
(21 pages)
45
Refinements to DRGs

• DRGs questioned on ability to describe resource
  use
• Limited account of severity
• Numerous severity based refinements to DRGs
  proposed
• Computerized Severity Index
• Fetter et al developed Refined DRGs which better
  reflect severity and resource use
• will be phased in by HCFA (now CMS)
• Bottom line no one perfect classification
  system for resource management
• become familiar with many and use each as needed
• important to use SOMETHING as gross aggregate
  measures are not extremely useful for detailed
  resource management

46
IHI Reducing Delays and Waiting Times

1. IHIs process improvement framework
2. General guidance on delay reduction
3. 27 Change concepts for delay reduction
4. Redesign the system
5. Shaping the demand
6. Matching capacity to demand
7. Four key examples
8. Surgery
9. Emergency Department
10. Within clinics and physicians offices
11. Access to care