informs rocky mtn presentation 03-17-11
DESCRIPTION
Quantitative Approaches to Improve Healthcare Access and Quality, Rocky Mountain INFORMS Chapter Meeting, A panel presentation, featuring the work of: Linda LaGanga, Ph.D.,Steve Lawrence, Ph.D., C.J. McKinney, Ph.D. Candidate1, Antonio Olmos, Ph.D., Michele Samorani, Ph.D. Candidate (Mental Health Center of Denver, University of Colorado-Boulder, University of Colorado-Denver, University of Northern Colorado)Thursday, March 17, 2011TRANSCRIPT
1Rocky Mountain INFORMS, March 17, 2011
Quantitative Approaches to Improve Healthcare Access and Quality
Rocky Mountain INFORMS Chapter Meeting
A panel presentation, featuring the work of:Linda LaGanga, Ph.D.1,3
Steve Lawrence, Ph.D.2C.J. McKinney, Ph.D. Candidate1,4
Antonio Olmos, Ph.D.1Michele Samorani, Ph.D. Candidate2
1. Mental Health Center of Denver2. University of Colorado-Boulder3. University of Colorado-Denver
4. University of Northern Colorado
2Rocky Mountain INFORMS, March 17, 2011
Healthcare Issues we address
To overbook or not? If we schedule them, will they come? What would Deming do to improve
healthcare? To achieve efficiency and effectiveness
of healthcare
3Rocky Mountain INFORMS, March 17, 2011
Where is our work developed and documented? Experience and data from
the Mental Health Center of Denver Community mental health center serving over 14,000
people per year Surveys and interviews of other healthcare
providers/systems Presented at INFORMS annual conferences Other conferences:
Production & Operations Management Society Decision Sciences Institute Mayo Clinic Conference on OR/Systems Engineering in
Healthcare American Evaluation Association
4Rocky Mountain INFORMS, March 17, 2011
Read more about it…
Decision Science Journal (May, 2007) Journal of Operations Management
(2010, in press) Conference presentations and proceedings at
http://www.outcomesmhcd.com/Pubs.htm Research posters on the wall
opposite this room
5Rocky Mountain INFORMS, March 17, 2011
Appointment Scheduling and Overbooking Clinic Overbooking to Improve Patient Access
and Increase Provider Productivity LaGanga, L. R., & Lawrence, S. R. (2007).
Clinic overbooking to improve patient access and provider productivity.Decision Sciences, 38(2), 251 – 276.
6Rocky Mountain INFORMS, March 17, 2011
Simple Overbooking Example
7Rocky Mountain INFORMS, March 17, 2011
Model Assumptions Number of patients booked, K:
E(K) = SK = N S = Show rate, N = target n of patients K = N/S
Patients scheduled at even intervals throughout the day T = N/K = S Inter-appointment times compressed by the show rate
Patients arrive on time with probability S Patient service times deterministic
Added variability in final version
8Rocky Mountain INFORMS, March 17, 2011
Overbooking: Best Case
No patients waitWaiting
No overtimeD9D7D5D3D1Service
X10A9X8A7X6A5X4A3X2A1Arrivals
Expected number of patients (5) arrive, evenly spacedBest Case
65.554.543.532.521.510.50Start Time
13121110987654321Time Slot
OvertimeRegular Time
10 appointment slots / session; 50% show rate
5 patients seen; no provider idle time; no patients wait; no clinic overtime
9Rocky Mountain INFORMS, March 17, 2011
Overbooking: Bunched Early
W7W5W3W2Waiting
No overtimeD7D5D3D2D1Service
X10X9X8A7X6A5X4A3A2A1Arrivals
Expected number of patients (5) arrive, bunched earlyCase 1
65.554.543.532.521.510.50Start Time
13121110987654321Time Slot
OvertimeRegular Time
10 appointment slots / session; 50% show rate
5 patients seen; no provider idle time; 4 patients wait; no clinic overtime
10Rocky Mountain INFORMS, March 17, 2011
Overbooking: Late Arrival
No patients waitWaiting
OT D10ID7D5D3D1Service
A10X9X8A7X6A5X4A3X2A1Arrivals
Expected number of patients (5) arrive, one late arrivalCase 2
65.554.543.532.521.510.50Start Time
13121110987654321Time Slot
OvertimeRegular Time
10 appointment slots / session; 50% show rate
5 patients seen; 10% provider idle time; no patients wait; 10% clinic overtime
11Rocky Mountain INFORMS, March 17, 2011
Overbooking: Bunched Late
W9W8Waiting
OTD9D8D7IID3D1Service
X10A9A8A7X6X5X4A3X2A1Arrivals
Expected number of patients (5) arrive, bunched lateCase 3
65.554.543.532.521.510.50Start Time
13121110987654321Time Slot
OvertimeRegular Time
10 appointment slots / session; 50% show rate
5 patients seen; 20% provider idle time; 2 patients waiting; 20% clinic overtime
12Rocky Mountain INFORMS, March 17, 2011
Overbooking: Extra Arrival
W9W7W5W3W2Waiting
OTD9D7D5D3D2D1Service
X10A9X8A7X6A5X4A3A2A1Arrivals
More patients arrive (6) than expected (5)Case 4
65.554.543.532.521.510.50Start Time
13121110987654321Time Slot
OvertimeRegular Time
10 appointment slots / session; 50% show rate
6 patients seen; no provider idle time; 5 patients waiting; 20% clinic overtime
13Rocky Mountain INFORMS, March 17, 2011
Overbooking Utility Model
14Rocky Mountain INFORMS, March 17, 2011
Overbooking Utility Model
Maximize clinic “utility” Trade-off Patient access (number of patients seen) Average patient waiting times Expected clinic overtime
Note that provider productivity is implicit in this model
15Rocky Mountain INFORMS, March 17, 2011
Relative Benefits and Penalties
= Benefit of seeing additional patient = Penalty for patient waiting = Penalty for clinic overtime
The values of , , and don’t matter Just their ratios or relative importance
16Rocky Mountain INFORMS, March 17, 2011
Utility Function
U SN
( )N OU U U A SN W O
Expected utility without overbooking
Expected utility with overbookingOU A W O
Expected net utility with overbooking
17Rocky Mountain INFORMS, March 17, 2011
Utility Function Described
( )NU A SN W O
Utility Benefit ofPatients that
“Show”
Less Utility Benefitw/o Overbooking
Less PatientWaiting Penalty
Less ClinicOvertime Penalty
Net Utility Benefit from Overbooking (could be negative)
18Rocky Mountain INFORMS, March 17, 2011
Simulation Experiments
Five clinic size levels N N = {10, 20, 30, 40, 50}
Ten show rates S S = {100%, 90%, … , 10%}
Full factorial experiment SN = 5 × 100 = 500 factor levels 10,000 replications per factor 500,000 observations
19Rocky Mountain INFORMS, March 17, 2011
Regression Analysis
Results from simulation analyzed using regression analysis
Regression equations obtained Expected patient wait times Expected clinic overtime Expected provider productivity
All coefficients significant R2 = 98%+
20Rocky Mountain INFORMS, March 17, 2011
Sensitivity to Service Uncertainty
-10
0
10
20
30
40
0.0 0.2 0.4 0.6 0.8 1.0
Service Time Variability
Aver
age
Net U
tility
N50R90N30R90N50R50N30R50N10R90N10R50N10R10N30R10N50R10
Average of net utility UN with overbooking as a function of service time variability cs , with and (=1, =0.5, τ =1.2)
21Rocky Mountain INFORMS, March 17, 2011
Conclusions
Overbooking is one solution for appointment no-shows
Can significantly improve performance Patient access (more patients seen) Clinic utility
But with a cost Increased patient waiting & clinic overtime
Good for some clinics, not for others
22Rocky Mountain INFORMS, March 17, 2011
Directions for Future Work
Scheduling policies Double booking Wave scheduling
Optimal overbooking policies Current overbooking policy is not “optimal” Dynamic programming
Nonlinear waiting & overtime functions Long waits much worse than short waits
23Rocky Mountain INFORMS, March 17, 2011
Lean Options for Walk-In, Open Access, and Traditional
Appointment Scheduling in Outpatient Health Care Clinics
© 2008 – Linda LaGanga and Stephen Lawrence
Linda R. LaGanga, Ph.D.Director of Quality SystemsMental Health Center of DenverDenver, CO USA
Stephen R. Lawrence, Ph.D.Leeds School of BusinessUniversity of ColoradoBoulder, CO USA
Mayo Clinic Conference on Systems Engineering & Operations
Research in Health CareRochester, Minnesota – August 17, 2009
Additional information available at: http://Leeds.colorado.edu/ApptSched
24Rocky Mountain INFORMS, March 17, 2011
Data Mining in Appointment Scheduling
Michele SamoraniPhD Candidate
Leeds School of Business, University of Colorado at Boulder
25Rocky Mountain INFORMS, March 17, 2011
Finding Patterns with Data Mining
26Rocky Mountain INFORMS, March 17, 2011
Young clients are more likely to keep appointments with no reminder call
DECISION TREE
27Rocky Mountain INFORMS, March 17, 2011
If clients are under the age of 26.3 years old and have low average CRM (<.5), then they are more likely to keep their appointments
CLUSTERING
28Rocky Mountain INFORMS, March 17, 2011
Using Data Mining to Schedule Appointments
29Rocky Mountain INFORMS, March 17, 2011
Overbooking – Shortcomings
Suppose service time = 30 minutes
Little waiting time and no overtime
Some waiting time and a high overtime
If we could predict which patients show up and which don’t, we could obtain a more controllable schedule
1 0 1 10
9:00
9:20
9:40
10:00
10:20
10:40
11:00
11:20
11:40
1 1 1 0
12:00
0 1 1 10
9:00
9:20
9:40
10:00
10:20
10:40
11:00
11:20
11:400 1 1 1
12:00
30Rocky Mountain INFORMS, March 17, 2011
The methodEvery time a visit request arrives:1)A classifier is used to predict if it shows or not (for each day)2)The visit request is scheduled by solving a stochastic program through column generation
Non‐controllable parameters•Service time•Revenue from seeing a patient•Clinic overtime cost•Waiting time cost
Controllable parameters•Number of slots K•Scheduling horizon h•Classification performance:
– Sensitivity (sn)– Specificity (sp)
How good we are at retrieving showing patients
How good we are at retrieving non‐showing patients
31Rocky Mountain INFORMS, March 17, 2011
Productivity vs Punctuality Productivity: number of patients seen. It is increased by:
Punctuality: 1/(overtime + waiting time). It is increased by:
32Rocky Mountain INFORMS, March 17, 2011
Real world case: MHCD
After playing for a few hours with the MHCD data set, I can achieve any of the following classification performances: sn = 0.9, sp = 0.5 sn = 0.7, sp = 0.7 sn = 0.6, sp = 0.8
Show rate Same day 1 day 2 days 3 days 4 days R
Low .74 .64 .65 .62 .61 .65
MHCD .87 .74 .75 .72 .71 .76
• Goal: Find the best policy for MHCD in terms of:– Overbooking– Open Access– Data Mining
33Rocky Mountain INFORMS, March 17, 2011
Policy DM OB OA .
(min) (min) ∗ ∗
1 No No No 6.39 0.00 0.00 5.99 8 4
2 No No Yes 6.39 0.00 0.00 5.99 8 1
3 No Yes No 7.10 36.22 20.61 8.37 12 4
4 No Yes Yes 7.22 35.33 21.37 8.40 12 1
5
.6, .8 No No 6.82 0.00 0.00 6.44 8 5
.7, .7 No No 6.99 0.00 0.00 6.62 8 4
.9, .5 No No 7.36 0.00 0.00 7.00 8 5
6
.6, .8 No Yes 6.84 0.00 0.00 6.44 8 1
.7, .7 No Yes 6.83 0.00 0.00 6.43 8 1
.9, .5 No Yes 6.66 0.00 0.00 6.27 8 1
7
.6, .8 Yes No 7.24 21.11 14.96 7.78 12 3
.7, .7 Yes No 7.42 29.33 17.88 8.33 12 5
.9, .5 Yes No 7.58 40.78 23.56 9.03 12 2
8
.6, .8 Yes Yes 7.35 25.00 15.92 8.03 12 1
.7, .7 Yes Yes 7.44 28.44 18.51 8.28 12 1
.9, .5 Yes Yes 7.32 35.22 19.83 8.47 12 1
Data Mining
Overbooking
Open Access
34Rocky Mountain INFORMS, March 17, 2011
Policy DM OB OA .
(min) (min) ∗ ∗
1 No No No 7.28 0.00 0.00 6.88 8 4
2 No No Yes 7.27 0.00 0.00 6.87 8 1
3 No Yes No 7.47 29.07 15.32 8.39 10 5
4 No Yes Yes 7.52 28.00 15.62 8.39 10 1
5
.6, .8 No No 7.49 0.00 0.00 7.11 8 5
.7, .7 No No 7.56 0.00 0.00 7.18 8 2
.9, .5 No No 7.85 0.00 0.00 7.47 8 2
6
.6, .8 No Yes 7.56 0.00 0.00 7.17 8 1
.7, .7 No Yes 7.59 0.00 0.00 7.19 8 1
.9, .5 No Yes 7.52 0.00 0.00 7.12 8 1
7
.6, .8 Yes No 7.60 20.73 13.26 8.14 10 2
.7, .7 Yes No 7.65 12.11 8.69 7.83 9 5
.9, .5 Yes No 7.86 15.22 9.81 8.18 9 2
8
.6, .8 Yes Yes 7.62 21.87 13.83 8.20 10 1
.7, .7 Yes Yes 7.64 24.87 14.53 8.36 10 1
.9, .5 Yes Yes 7.57 28.13 15.82 8.44 10 1
35Rocky Mountain INFORMS, March 17, 2011
Conclusions Data mining can improve appointment scheduling in the
presence of no-shows If adopted in conjunction with overbooking, data mining can
either increase punctuality or productivity, depending on sensitivity and specificity
In case of low show rate, the advantage obtained by using overbooking is similar to the one obtained with data mining.
On the other hand, in case of high show rate, data mining is a superior technique
Interestingly, if we can achieve a decent classification performance, using open access is the worst choice
Thank you for your attention. Questions?
36Rocky Mountain INFORMS, March 17, 2011
What about the scheduling horizon h? h does not have any significant impact by itself:
But its interaction with sn and sp is significant:
37Rocky Mountain INFORMS, March 17, 2011
High sensitivity classifier
Classifier
38Rocky Mountain INFORMS, March 17, 2011
Driving Clinical Quality Driving Clinical Quality Improvement through Mental Health Improvement through Mental Health Recovery Control ChartsRecovery Control Charts
CJ McKinney, MA*Antonio Olmos, PhDLinda Laganga, PhD
Mental Health Center of DenverDenver, CO, USA
* - Corresponding Author
INFORMS Annual Meeting 2009San Diego, CA
October, 11th, 2009
39Rocky Mountain INFORMS, March 17, 2011
Literature Olmos-Gallo, P.A. DeRoche, K.K. (2010, August). Monitoring Outcomes
in Mental Health Recovery: The Effect on Programs and Policies.Advances in Mental Health (9)1, 8-16. http://amh.e-contentmanagement.com/archives/vol/9/issue/1/ contact P. Antonio Olmos for a copy of the publication
McKinney, C.J., Olmos-Gallo, P.A. McLean, C., LaGanga, L.R. (August 2010). Driving Clinical Quality Improvement through Mental Health Recovery Control Charts. Presented at the 3rd Annual Mayo ClinicConference on Systems Engineering & Operations Research in Health Care, Rochester, MN. Awarded First Place for Best Poster Presentation.
Clark, C.R., Olmos-Gallo, P.A. (2007). Performance Measurement: A signature approach to outcomes measurement improves recovery. National Council Magazine, 3, 26-28.
Glover, H. (2005). Recovery based service delivery: Are we ready to transform the words into a paradigm shift? Australian e-Journal for the Advancement of Mental Health, 4(3), www.auseinet.com/journal/vol4iss3/glovereditorial.pdf (accessed 15 May 2009)
Montgomery, D. C. (2005) Introduction to Statistical Quality Control, Fifth Edition. Hoboken, NJ: John Wiley and Sons, Inc.
Olmos-Gallo, P. A., DeRoche, K. K., McKinney, C. J., Starks, R., & Huff, S. (2009). The Recovery Markers Inventory: Validation of an instrument to measure factors associated with recovery from mental illness. Working paper
40Rocky Mountain INFORMS, March 17, 2011
The Heart of Recovery Measurement
41Rocky Mountain INFORMS, March 17, 2011
CheckCheck DoDo
PlanPlanActAct
Continuous Improvement
42Rocky Mountain INFORMS, March 17, 2011
Quality Components in Mental Health Services
Performance
Reliability
Durability
Perceived Quality
Conformance to Standards
How well are MH services working? Are consumers improving in their recovery?
How often do we see improvements in recovery? How consistent are the outcomes across consumers?
How long does the consumer retain the recovery-supportive skills and tools taught through MH services?
How does the consumer perceive our ability to support MH recovery? Community?
Are we meeting program fidelity standards?
Quality Components Relationship to MH Services
43Rocky Mountain INFORMS, March 17, 2011
Quality Control in Mental HealthQuality Control in Mental Health
Allocate and reallocate clinical resources more efficiently
Improve and maintain clinical program fidelity Reduce length of treatment, while sustaining same
level of recovery and recovery supportive factors Increase the number of consumers served, while
decreasing burden on case managers/therapists Identify most effective programs based upon
consumer needs
44Rocky Mountain INFORMS, March 17, 2011
Mental Health Recovery Concept of Recovery has taken root
around the world Working Definition (MHCD): “A non-linear process of growth by which people move
from lower to higher levels of fulfillment in the areas of hope, safety, level of symptom interference, social networks, and activity.”
Federal Grant (SAMHSA) for Transformation to Recovery-Oriented Mental Health Systems
For information on the Recovery Transformation Summit, see http://www.gmhcn.org/files/RRecovery_Newsletter_Fall2010.pdf
45Rocky Mountain INFORMS, March 17, 2011
Mental Health Recovery Outcomes
MHCD has developed 3 consumer specific recovery outcomes Consumer Recovery Measure – (Consumer Perspective)
Hope, Safety, Activity, Level of Symptom Management, Social Networks
Recovery Marker Inventory – (Clinician Perspective) Housing, Employment, Education, Active Growth, Participation, and Symptom Management
Recovery Needs Level – (Clinical Algorithm) Provides for one of 5 levels of treatment based upon clinical criteria
The examples in this presentation will utilize the Consumer Recovery Measure.
46Rocky Mountain INFORMS, March 17, 2011
47Rocky Mountain INFORMS, March 17, 2011
48Rocky Mountain INFORMS, March 17, 2011
Relationship among Recovery Outcomes(1) Recovery Marker
Inventory (RMI)(Longitudinal data to support
clinical decision making)
(3) Consumer Recovery Measure (CRM)
(Consumer’s perception of their own recovery)
(2) Promoting Recovery in Organizations (PRO)
(Consumer’s perceptions of how well specific programs and staff are
promoting recovery)
To what degree is RECOVERY
happening for consumers at MHCD(Formative and summative
evaluation of recovery)
(4) Recovery Needs Level
(RNL)(Appropriate level of services)
49Rocky Mountain INFORMS, March 17, 2011
Consumer Recovery Measure v3.0
The CRM V3.0 includes the 15 items listed below:2. Lately I feel like I’ve been making important contributions (active-growth)4. I have hope for the future (hope)5. I am reaching my goals (active growth)7. I have this feeling things are going to be just fine (hope)8. Recently my life has felt meaningful (hope)9. Recently, I have been motivated to try new things (active-growth)11. There are some people who cause me a lot of fear (safety)12. I get a lot of support during the hard times (social network)14. In most situations, I feel totally safe (safety)15. My life is often disrupted by my symptoms (symptom interference)16. Sometimes I’m afraid someone might hurt me (safety)17. I have people in my life I can really count on (social network)18. Life’s pressures lead me to lose control (symptom interference)19. I have friends or family I really like (social network)20. My symptoms interfere less and less with my life (symptom interference)21. When my symptoms occur, I am able to manage them without falling
apart (symptom interference)
50Rocky Mountain INFORMS, March 17, 2011
Quality Control Issues in Recovery
Multiple sources of variability Measurement Consumer System
Changing environmental, treatment, and consumer specific factors affect outcome measurements.
Difficulty in detection of small changes due to large variability within and among consumers
51Rocky Mountain INFORMS, March 17, 2011
Multilevel Modeling and Recovery
Multilevel modeling allows for the partitioning of variance among multiple levels of nesting, i.e. measures within consumers within therapists
Allows for regression based correction of expected outcomes for any unit at any level, i.e. conditional estimates based upon consumer characteristics in environment or treatment.
Can be used to simultaneously monitor multiple aspects of the system from measurements to clinical sites.
Based upon Mixed-Effects ANOVA design
52Rocky Mountain INFORMS, March 17, 2011
Example of Multilevel modeling concepts
Consumer Level Effect
CRM Scores
=
Intake
+
Time in Tx =
=
Intercept
Intercept
Mood Disorder
Mood Disorder
+
+
=Intercept
ACT Tx+
=Intercept
ACT Tx+
=Intercept
ACT Tx+
=Intercept
ACT Tx+
Typical SLR Model System Level Effect
Higher Level Effects
53Rocky Mountain INFORMS, March 17, 2011
Multilevel Regression Corrected Control Charts
CUSUM for Consumers (between consumer comparisons)
Allows for determination of a consumer’s progress as compared to peers in same treatment with environmental and demographic similarities
54Rocky Mountain INFORMS, March 17, 2011
Example MRC-CUSUM Self Comparison
55Rocky Mountain INFORMS, March 17, 2011
Example MRC-CUSUM Peer Comparison
56Rocky Mountain INFORMS, March 17, 2011
Utilization of MRC-CUSUM
Improved allocation of resources – by allowing consumer comparison to peers
Identification of factors that may promote/inhibit recovery
Provide feedback regarding progress and relapse more quickly to clinicians
57Rocky Mountain INFORMS, March 17, 2011
Multivariate Control Chart
Bivariate Control Chart for plotting of regression parameters (intercept and slopes)
Corrections may be made based upon environmental, treatment, and demographic characteristics
58Rocky Mountain INFORMS, March 17, 2011
58
I II
IVIII
59Rocky Mountain INFORMS, March 17, 2011
I II
IIIIV
BELOW AVG.
ABOVE AVG.
Decreasing
Increasing
Recovery InterceptRe
covery Slope
NOTE: ANY Outlier within a quadrant indicates it is farther away from the average than would be expected under typical circumstances.
60Rocky Mountain INFORMS, March 17, 2011
Utilization of Bivariate Control Chart
Identify “outlying” consumers to help determine aspects of a program that promote self-perceived recovery, and those aspects that may be a deterrent to improvement in self-perceived recovery.
Allow for identification of consumers who may need further resources or different treatment.
Allows for overview of consumer progress, where comparisons over time may allow for evaluation of process changes and overall consumer effect.
61Rocky Mountain INFORMS, March 17, 2011
Summary of Benefits
Allow for more efficient allocation of treatment and resources.
Identify program aspects that promote or deter improvement in self-perceived recovery.
Identify consumer in need of additional treatment or resources.
Allow for the identification of consumer and system factors that affect or interact with consumer outcomes and program effectiveness.
Being able to cater to differing needs of the wide variety of consumers served.
Identification of Episodes of Care
62Rocky Mountain INFORMS, March 17, 2011
Moving forward in recovery models to drive quality improvement
Statistical Models
Information Technology
Knowledge Building& Dissemination:Learning CollaborativesStaff Involvement,Training
63Rocky Mountain INFORMS, March 17, 2011
Future Directions to Drive Recovery System Improvement
Identify clinically significant patterns Expand to other recovery measures and aspects. Coordinate with data mining to identify
relationships between services and recovery outcomes
Automate quality control process Integrate fully into clinical quality review processes Develop accessible reporting and dashboard
systems for clinicians and managers
64Rocky Mountain INFORMS, March 17, 2011
More information
If you would like to see more information concerning MHCD’s research and work with Recovery please visit:
http://www.outcomesmhcd.com/http://www.reachingrecovery.org/
Or contact [email protected]
65Rocky Mountain INFORMS, March 17, 2011
Extra slides that were mentioned but not presented on 3/17/11 due to time limitations
From Mayo Clinic Conference on Operations Research & Systems Engineering in Healthcare
Lean Options for Walk-In, Open Access, and Traditional Appointment Scheduling in Outpatient Health Care Clinics (LaGanga & Lawrence, 2009)
Includes further development to appointment scheduling models toinclude metaheuristic optimization of overbooking levels
Comparison of traditional scheduling, open-access,and walk-in policies
Lean process improvement to reduce no-shows and expand intake capacity.
Condensed slide set. See http://www.outcomesmhcd.com/Pubs.htmfor complete, original presentation.
Driving Clinical Quality Improvement Through Mental Health Recovery Control Charts (McKinney, Olmos, McLean, LaGanga, 2010)
Poster presentation First Place Award
66Rocky Mountain INFORMS, March 17, 2011
Lean Options for Walk-In, Open Access, and Traditional
Appointment Scheduling in Outpatient Health Care Clinics
© 2008 – Linda LaGanga and Stephen Lawrence
Linda R. LaGanga, Ph.D.Director of Quality SystemsMental Health Center of DenverDenver, CO USA
Stephen R. Lawrence, Ph.D.Leeds School of BusinessUniversity of ColoradoBoulder, CO USA
Mayo Clinic Conference on Systems Engineering & Operations
Research in Health CareRochester, Minnesota – August 17, 2009
Additional information available at: http://www.outcomesmhcd.com/Pubs.htm
67Rocky Mountain INFORMS, March 17, 2011
1. Background on Appointment Scheduling
68Rocky Mountain INFORMS, March 17, 2011
Motivation Healthcare Capacity Funding restrictions Demand exceeds supply Serve more people with limited resources
Manufacturing Scheduling Resource utilization Maximize throughput
Healthcare Scheduling as the point of access
Maximize appointment yield
69Rocky Mountain INFORMS, March 17, 2011
2. Lean Approaches
Rapid Improvement Capacity Expansion (RICE) Team January, 2008
Article in press, Journal of Operations Management (2010). Available at http://dx.doi.org/10.1016/j.jom.2010.12.005
70Rocky Mountain INFORMS, March 17, 2011
Lean Approaches Reducing Waste Underutilization Overtime No-shows Patient Wait time
Customer Service Choice Service Quality Outcomes
71Rocky Mountain INFORMS, March 17, 2011
Lean Process Improvement in Healthcare Documented success in hospitals
ThedaCare, Wisconsin Prairie Lakes, South Dakota Virginia Mason, Seattle University of Pittsburgh Medical Center Denver Health Medical Center
Influences Toyota Production System Ritz Carleton Disney
Hospitals to Outpatient Clinics run by hospitals Collaborating outpatient systems
72Rocky Mountain INFORMS, March 17, 2011
Lean Process Improvement: One Year AfterRapid Improvement Capacity ExpansionRICE Results
Analysis of the1,726 intake appointments for the one year before and the full year after the lean project
27% increase in service capacity from 703 to 890 kept appointments) to intake new consumers
12% reduction in the no-show rate from 14% to 2% no-show
Capacity increase of 187 additional people who were able to access needed services, without increasing staff or other expenses for these services
93 fewer no-shows for intake appointments during the first full year of RICE improved operations.
73Rocky Mountain INFORMS, March 17, 2011
Lean Process Improvement:RICE Project System Transformation
Appointments Scheduled and No-Show Rates
050
100150200250300350400450
Mon Tue Wed Thu Fri Mon Tue Wed Thu Fri
App
oint
men
ts
0%
5%
10%
15%
20%
AppointmentsNo-Show Rate
Year Before Lean Improvement
Year After Lean Improvement
74Rocky Mountain INFORMS, March 17, 2011
How was this shift accomplished? Day of the week: shifted and added Tuesdays and Thursdays
Welcome call the day before Transportation and other information Time lag eliminated Orientation to Intake Assessment
Group intakes Overbooking Flexible capacity
75Rocky Mountain INFORMS, March 17, 2011
Lean Scheduling Challenge Choice versus Certainty Variability versus Predictability Sources of Uncertainty / Variability No-shows Service duration Customer (patients’) Demand
Time is a significant factor Airline booking models?
76Rocky Mountain INFORMS, March 17, 2011
3. Response to Overbooking
77Rocky Mountain INFORMS, March 17, 2011
Sample Responses Campus reporter’s visit to student health
center “Not now and never will” Patient waits 15 – 20 minutes New administration, new interests
Morning News Radio “Overbooking leading to increased patient
satisfaction? That just doesn’t make any sense!” Public Radio Interviewer Benefits of increased access at lower cost
78Rocky Mountain INFORMS, March 17, 2011
78
Other Responses Practitioners
Dentists General medicine Child advocacy
How should we overbook? Other options Lean Approaches Open Access (Advanced Access) Walk-ins
79Rocky Mountain INFORMS, March 17, 2011
4. Enhanced Appointment Scheduling Model
0%
5%
10%
15%
20%
0 1 2 3 4 5 6 7 8 9 10 11 12
Number Waiting (k)
Prob
abili
ty
80Rocky Mountain INFORMS, March 17, 2011
Objectives of Research Optimize patient flow in health-care clinics Traditionally scheduled (TS) clinic Some patients do not “show” for scheduled
appointments TS clinic wishes to increase scheduling flexibility Some capacity allocated to “open access” (OA)
appointments, OR Some capacity allocated to “walk-in” traffic
Balance needs of clinic, providers, and patients
81Rocky Mountain INFORMS, March 17, 2011
Objectives of Research
Study impact of open access and walk-in traffic When is open access or walk-in traffic
beneficial? What mix of TS, OA, and WI traffic is
best? What are trade-offs of TS, OA, and WI
on clinic performance?
82Rocky Mountain INFORMS, March 17, 2011
Relative Benefits and Penalties = Benefit of seeing additional client = Penalty for client waiting = Penalty for clinic overtime Numéraire of , , and doesn’t matter Ratios (relative importance) are important
Allow linear, quadratic, and mixed (linear + quadratic) costs
83Rocky Mountain INFORMS, March 17, 2011
Linear & Quadratic Objectives
1, 1,1 1
ˆˆ 1ˆN k
jk N k N kj k k i k
U A k i kA
S
Linear Utility Function
Quadratic Utility Function 2 2
1, 1,1 1
ˆˆ 2 1 1ˆN k
jk N k N kj k k i k
U A k i kA
S
Benefit from patients served
Patient waiting penalties during normal clinic ops
Patient waiting penalties during clinic overtime
Clinic overtime penalties
84Rocky Mountain INFORMS, March 17, 2011
Heuristic Solution Methodology 1. Gradient search
Increment/decrement appts scheduled in each slot Choose the single change with greatest utility Iterate until no further improvement found
2. Pairwise interchange Exchange appts scheduled in all slot pairs Choose the single swap with greatest utility Iterate until no further improvement found
3. Iterate (1) and (2) while utility improves4. Prior research: Optimality not guaranteed, but
almost always obtained
85Rocky Mountain INFORMS, March 17, 2011
How does Open Access contribute to leaner scheduling?1. It provides a more reliable method of
overbooking.2. It eliminates the uncertainty of demand for
same-day appointments.3. It guarantees that patients will be seen when
they want.4. It reduces uncertainty caused by no-shows.5. It eliminates waste caused by unfilled
appointments.
86Rocky Mountain INFORMS, March 17, 2011
How does Open Access contribute to leaner scheduling?1. It provides a more reliable method of
overbooking.2. It eliminates the uncertainty of demand for
same-day appointments.3. It guarantees that patients will be seen when
they want.4. It reduces uncertainty caused by no-shows.5. It eliminates waste caused by unfilled
appointments.
87Rocky Mountain INFORMS, March 17, 2011
5. Computational Results
0
1
2
3
4
5
6
7
8
9
10
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Net
Util
ity p
er P
rovi
der
Open Access (OA) Traffic (% of capacity)
Walk-ins
Open Access
-6.190
1
2
3
4
5
6
7
8
9
10
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Net
Util
ity p
er P
rovi
der
Open Access (OA) Traffic (% of capacity)
Walk-ins
Open Access
-6.19
88Rocky Mountain INFORMS, March 17, 2011
Computational Trials 44 sample problems solved Session size N = 12 Appointment show rate = 70% Number of providers P = {1, 2, 4, 8} OA call-in rate = {0%, 10%, …100%} capacity With P = 4 and N = 12, then = 24 is 50% of capacity
Walk-in rate = {0%, 10%, …100%} of capacity With P = 4, then = 2 is 50% of capacity
Quadratic costs Parameters =1.0, =1.0, =1.5
89Rocky Mountain INFORMS, March 17, 2011
Patients Seen
10
11
12
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Patie
nts
Seen
per
Pro
vide
r
OA or WI Traffic (% of capacity)
Walk-ins
Open Access
2 Providers (P=2)10
11
12
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Patie
nts
Seen
per
Pro
vide
r
OA or WI Traffic (% of capacity)
Walk-ins
Open Access
2 Providers (P=2)
N=12, P=1, =0.7, =1.0, =1.0, =1.0, =1.5
90Rocky Mountain INFORMS, March 17, 2011
Patient Waiting Time
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Expe
cted
Wai
ting
Tim
e / P
atie
nt
OA or WI Traffic (% of capacity)
Walk-ins
Open Access
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Expe
cted
Wai
ting
Tim
e / P
atie
nt
OA or WI Traffic (% of capacity)
Walk-ins
Open Access
N=12, P=1, =0.7, =1.0, =1.0, =1.0, =1.5
91Rocky Mountain INFORMS, March 17, 2011
Clinic Overtime
0.0
0.5
1.0
1.5
2.0
2.5
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Exp
ecte
d P
rovi
der O
vert
ime
(dtim
e un
its)
OA or WI Traffic (% of capacity)
Walk-ins
Open Access
0.0
0.5
1.0
1.5
2.0
2.5
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Exp
ecte
d P
rovi
der O
vert
ime
(dtim
e un
its)
OA or WI Traffic (% of capacity)
Walk-ins
Open Access
N=12, P=1, =0.7, =1.0, =1.0, =1.0, =1.5
92Rocky Mountain INFORMS, March 17, 2011
Provider Utilization
60%
65%
70%
75%
80%
85%
90%
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Expe
cted
Pro
vide
r Util
izat
ion
OA or WI Traffic (% of capacity)
Walk-Ins
Open Acess
60%
65%
70%
75%
80%
85%
90%
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Expe
cted
Pro
vide
r Util
izat
ion
OA or WI Traffic (% of capacity)
Walk-Ins
Open Acess
N=12, P=1, =0.7, =1.0, =1.0, =1.0, =1.5
93Rocky Mountain INFORMS, March 17, 2011
Net Utility
N=12, P=1, =0.7, =1.0, =1.0, =1.0, =1.5
0
1
2
3
4
5
6
7
8
9
10
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Net
Util
ity p
er P
rovi
der
Open Access (OA) Traffic (% of capacity)
Walk-ins
Open Access
-6.190
1
2
3
4
5
6
7
8
9
10
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Net
Util
ity p
er P
rovi
der
Open Access (OA) Traffic (% of capacity)
Walk-ins
Open Access
-6.19
94Rocky Mountain INFORMS, March 17, 2011
6. Insights and Recommendations
95Rocky Mountain INFORMS, March 17, 2011
Managerial Implications TS appointments provide better clinic utility
than does WI traffic or OA call-ins Any WI or OA traffic causes some decline in utility
An all-WI or all-OA clinic performs worse than any clinic with some TS appointments Even a relatively small percentage of scheduled
appointments can significantly improve clinic utility Degree of improvement depends on number of
providers A mix of TS appointments with some OA or WI
traffic does not greatly reduce clinic performance (utility)
96Rocky Mountain INFORMS, March 17, 2011
Insights from the Model Loss of utility with WI traffic is due to the long
right-tail of Poisson distribution Excessive patient waiting & clinic overtime
Loss of utility with OA traffic is due to uncertainty about number of OA call-ins
TS appts reduce patient waiting and clinic overtime Binomial distribution has truncated right tail
Multiple providers improves clinic utility Portfolio effect – variance reduction
97Rocky Mountain INFORMS, March 17, 2011
Lean Options for Walk-In, Open Access, and Traditional
Appointment Scheduling in Outpatient Health Care Clinics
© 2008 – Linda LaGanga and Stephen Lawrence
Linda R. LaGanga, Ph.D.Director of Quality SystemsMental Health Center of DenverDenver, CO USA
Stephen R. Lawrence, Ph.D.Leeds School of BusinessUniversity of ColoradoBoulder, CO USA
Mayo Clinic Conference on Systems Engineering & Operations
Research in Health CareRochester, Minnesota – August 17, 2009
Questions and comments? [email protected]([email protected]), [email protected]. Further information at http://www.outcomesmhcd.com/Pubs.htm
98Rocky Mountain INFORMS, March 17, 2011
INTRODUCTION
Driving Clinical Quality Improvement Through Mental Health Recovery Control ChartsC.J. McKinney, Pablo A. Olmos, Cathie McLean, Linda R. LaGanga,
Division of Quality Systems, Mental Health Center of Denver, Denver, CO
Every community mental health center focuses on clinical quality. Benefits of effective service delivery support quality through:
• optimize resource allocation,• increase consistency in consumer outcomes, • increase service fidelity, • decrease administrative load on clinicians, and • increase access to consumer services.
This poster presents our development of a set of reliable and valid mental health recovery measures, which we combine for a multi‐perspective assessment of recovery progress, which anchors an objective clinical quality control system.
RECOVERY ASSESSMENTMHCD consistently collects, reviews, and analyzes data across all consumers on four different recovery‐oriented outcome measurement tools. The combined data from these assessments provide multi‐perspective viewpoints for a more comprehensive picture of the consumer’s recovery experience and what factors may be impacting their recovery. It also provides supporting information to ensure the consumer is placed at a level of care that appropriately reflects their needs.
Recovery Marker Inventory – Clinician AssessmentAssessments are recorded on seven factors associated with recovery: Employment, Learning/Education, Activity/Growth Orientation, Symptom Interference, Participation in Services, Housing, and Substance Use.
Documentation of this data provides the clinician with a longitudinal perspective – from both an overall standpoint, as well as more specific recovery dimensions. These observations can then be used to help guide clinical discussion with the consumer, and indicate treatment focus.
RECOVERY ASSESSMENT continuedRecovery Needs LevelThe Recovery Needs Level is a series of indicators that through an objective algorithm assigns the consumer to an appropriate clinical service level. The RNL is completed by the clinician every six months and as needed. The measure consist of 15 different dimensions such as the GAF, Residence, Case Management, Substance Abuse, and Service Engagement.
Promoting Recovery in OrganizationsThe PRO survey is completed by the consumer, and consists of 7 sections covering all major service positions at MHCD, i.e. front desk, nursing/medical, case management, and rehabilitation. This data is collected annually through a random sampling of consumers. The survey summaries are then utilized to determine how well the teams and system are promoting recovery ideals.
The Recovery Outcome Tools have enabled us to develop a quality review system to monitor individual consumer outcomes and recommend review in cases where the consumer may not be progressing as expected. We are able to do this in three ways:
1.The Consumer Recovery Profile provides a snapshot of a person’s current mental health recovery progress. It demonstrates through graphs and tables the current status of a consumer to aid in service planning.
QUALITY CONTROL CHARTS
Self‐Comparing Control Chart Peer‐Comparing Control Chart
Utilization Review Form
Consumer Recovery Measure – Consumer AssessmentWith the Consumer Recovery Measure, the consumer rates agreement or disagreement with statements regarding their current recovery experience. These responses gauge consumer perspective on five dimensions of recovery: Symptom Management, Sense of Safety, Sense of Growth, Sense of Hope, and Social Activity.
Graphic representation of this data is shared with the consumer to initiate clinical discussion about changes in these areas, what the consumer attributes the changes to, and possible relationships between categories. This promotes insight, and empowers the consumer to share their story in a new and different way.
2. The Recovery Change Chart automatically identifies consumers needing further review by flagging those with substantial change in their recovery outcomes. A flag occurs whenever a consumer deviates from their expected outcomes for an extended period of time or if the deviations are large.
3. The Utilization Review Process: When a consumer is “flagged” by the Change Chart they will be an automatic candidate for a utilization management review. This review is done by other clinicians reviewing a consumer’s medical record to determine if a gap in services has occurred and if other services should be considered. The recommendations from this review are forwarded to the program manager for further review and implementation.
CONCLUSION & FUTURE DIRECTIONS
For more information about research or mental health recovery at MHCD, please view conference
presentations on our website:www.outcomesmhcd.com
Consistent with continuous quality improvement, integration of these tools into the clinical workflow is a constantly evolving process. We feel the following are basic needs to meet, and opportunities for operational enhancement:
• Education of Clinical staff, Executive Management, Consumers, and other stakeholdersas to the value of outcomes data collection and analysis and integration into the clinical practice
• Technological ability to “communicate” with the Electronic Medical Record ‐ the Recovery Profile is connected to the Electronic Medical Record, so it can be easily accessed by clinicians by bringing the information to them, without having to log in or open other data storage sites
• Integration into the daily clinical work flow – clinicians can review outcomes information with consumers during individual sessions, so as to make the information more meaningful; it is employed as part of the Peer Review process; and can be used during six month case reviews
•Automation of Quality Review process – control charts “flag” concerning outcomes outliers and identify them for Utilization Management Review, so as to address and redirect treatment inefficiencies in a timely manner
• Exploration of “super performer” characteristics to identify benchmarks for teams/programs
• Consumer Recovery Portal – consumers will have access to their outcomes data for increased engagement in the recovery process
•Integrate physical and mental healthcare
••Maximize outcomes to improve human lives!Maximize outcomes to improve human lives!
Qualitative Identification of Service Outliers
QUALITY CONTROL CHARTS continuedPresented at the Mayo Clinic Conference on Systems Engineering & Operations Research in Health Care (August 2010), Rochester, MN. Awarded First Place for Best Poster Presentation.