Medical Informatics Modeling Healthcare Outcomes

Document Sample
Medical Informatics Modeling Healthcare Outcomes Powered By Docstoc
					Medical Informatics
Modeling Healthcare Outcomes

     TRICARE ’03
       29 January 2003
  Digital Heretics

Maj Gen Lee P. Rodgers
            Just a Thought


―It is better to know some of the questions
  than all of the answers.‖

                            James Thurber
                            (1894-1961)
      Goal of the Enterprise

Organize
    Train
            Equip

Deployable Critical Care Resources
         Constraints
Manpower, dollars, tools

Hours - residents

Policies, NAS, appt templates

Systems - capabilities
      Response to Constraints
• Recognize paradigms - modeling
• Need to develop tools which
  – Identify impediments
  – Enhance training
• Jump the curve
• ―Separate‖ information from the tool
  Enhancing throughput

• Quickly assessing patients
  – (singly and in populations)
• Access to management data
• Access to decision support systems
    Envision alternate futures

•   Where are you?
•   What are your assumptions?
•   Where do you want to be?
•   What are the paths?
•   What are the future constraints?
          Model Concept…understanding our
                   assumptions
     Real World                  Model World                         MODELING



                                                Quantitative Model
                                                                                                   Qualitative Model
                                           • Elements
                 Abstraction                               •   Variables     Rationalism
                                           • Processes                                              • Assumptions
                                                           •   Constants
                                           • Activities                                             • Relationships
                                                           •   Randomness
                                                                                                    • Interactions
                                                           •   Constraints   Development




  Current System            Measurement
                                                                                Data       Input       • Deduction
                                                                                                       • Simulation




                                                                                                                          MODEL SOLUTION
  Judgment                                                                                               (Indirect Exp)
          Trial and Error                                                                              • Optimization
         Experimentation
                                                                              Validation

        Future
        System


                                                                              Conclusions
Resource Allocation                       Interpretation                     (hypotheses)
    Decisions
                                    ANALYSES
                                                                                       Bonder, Masterson & Miller 2000
            What do we need
• Databases which support the needs of the field e.g.
  ICDB
• Design provided compiles information which
  allows fact-based decision making
• Scenario-based models to facilitate design
  contingency plans
• Operations research (OR) focus
• Systems designed at the point-of-service by
  operators rather than in the lab by techies
Common Standards
Ask the Right Question

Col Stephen A. Jennings
     The Fundamental Relationship
Databases & Data ―Warehouses‖
     Data Sets/           Presentations     Queries
      Tables




     What we             How we classify   How we ask
      know                   things         questions

Ontologic Epistemology
     The Fundamental Problem
• ―Data‖
   – Many (all?) datasets can contain data for more than one
     domain
• Classification & relationships
   – Many (all?) datums pertain to more than one
     relationship
• Asking questions
   – There is no universally understood convention for
     asking questions efficiently
   – The asking of questions is not a common discipline
     The Fundamental Problem
• Programming
  – Doing ―stuff‖ with data & presenting it
  – Implicitly answers questions (pertinence/relevance)
  – Is done in a language
     • Which is the ―best‖ one? Has it been invented?
     • Most ambitious language efforts have died—only expedient
       ones have survived
     • Objects & encapsulation have been poorly understood
         – Have not made it into ―normal‖ education (cf., relativity)
         – Point out a deep semantic conflict between verbs & nouns
         – .NET; AppleScript == proprietary, not universal
  – No set of tools for easy exploration of heuristics
     • Back to the problem of how to formulate questions
           Observed Absurdities
• Serious users employ a command-line interface.
• There is no market for a GUI.
• The 5.25-inch disc is standard
    – There is no market for a 3.5-inch hard-shell floppy disc.
•   No one needs more than 768K of memory.
•   Apple® is dead. The Mac OS is dead.
•   UNIX is hard & ugly
•   MS-Windows® is the standard
                        Clearly So
• Software was more innovative in 1989 than it was in 1999
  (albeit more expensive)
   – Microsoft literally obliterated most niche markets/killed
     experimentation
• Microsoft developed the first portable document standard
  (―rtf‖). It is now the system format in Mac OS X
• The W3 Consortium developed WWW standards—not
  Microsoft!
   – There is a standard HTML, XML, etc. It is not Microsoft’s,
     although MS is a player in this open field.
• Proprietary standards are uniformly destructive—and
  seductive
• Whenever a commercial vendor pledges to adhere to open
  standards—he is generally lying.
• Vbasic is not portable.
               Possibly So
• Microsoft’s OS is dead
  – It is a dybbuk, hence deathless
  – Its code is impossibly large and secret
  – No one knows how secure Windows 2000 is
• In future, only 100% public code will be
  acceptable for operating systems
  – Obvious advantage to ―hackers‖
  – But a huge advantage to systems engineers
  Look @ the ICDB Home Page
• What’s there:
   –   One global variable
   –   One function
   –   One frameset
   –   A bunch of asp’s
• What there is not. Should be in every DHTML
  page
   –   A check for the DOM of the browser
   –   A check for the ID & version of the browser
   –   A check for required plug-ins
   –   Intelligent work-arounds
   –   JS routines have been available for nearly 10 years
Document Object Model
HTML Validation Service
                       So, What?
• Look forward—always
• Remember that the problem is not a simple
  systems one—it is an epistemological one
   – Use standard code, objects
      • Test all object actions in various systems
   – Avoid plug-ins
   – MS ―solutions‖ are noun-poor, verb-stealthy
   – Always follow up easy prototypes with good code
                     So, What?
• IE is not IE is not IE…
• MS’ concept of Javascript is non-standard
• MS’ form of Java is … different
   – but, it understands real Java
• Release your browser-based solution only if
   – It works on IE and Mozilla
      • on Windows
      • on Macintosh
      • on a Unix (X-windows) box
   – Some bells &whistles will have to be left out
Ya Gotta Dance with the One Ya Brung
   Getting the Right Data and Easy-CHCS

           LCDR Emory Fry
                   Lessons Learned
• Development cycles are 2-5 years
   – Far too long for our business environment
• Products difficult to integrate into current clinical business
  process
   – ―Designed‖ by engineers, not providers
• Central solutions astronomically expensive
   – All available R&D funding centrally controlled
• MTF solutions often stove-pipes
   – Initiatives often use multiple proprietary vendors, don’t scale well,
     nor integrate easily
    Yet We All Agree That…
….clinically relevant information should flow
seamlessly from one application to another, from
one user to another, without duplicate data entry,
and should be presented in whatever manner
enhances our ability to provide quality patient care
with measurable improvements in outcome as
economically as possible….
      Is There A Better Way?
• How can we limit development expense?
• How can we reduce costly hardware
  requirements?
• How can we improve data integrity?
• How can we ensure functional relevance for
  the field?
     Easy-CHCS: One Solution
• Language extensions for MUMPS
   – Transforms CHCS into a modern, object-oriented
     database
• Enables reuse of expensive, debugged, business
  rules and functionality
• Provides CHCS with all modern development
  tools and capabilities
• Open-source license reduces expense
           Additional Benefits
• CHCS preserved as database tier
   – Server hardware reused
   – No additional database licenses
   – Economical web-based technologies
• Data transformation errors avoided
• Provides open standard, vendor agnostic,
  bidirectional interface into CHCS
• Enables rapid, MTF driven development cycles
Model the System

Lt Col Brian Masterson
          Model Concept…understanding our
                   assumptions
     Real World                  Model World                         MODELING



                                                Quantitative Model
                                                                                                   Qualitative Model
                                           • Elements
                 Abstraction                               •   Variables     Rationalism
                                           • Processes                                              • Assumptions
                                                           •   Constants
                                           • Activities                                             • Relationships
                                                           •   Randomness
                                                                                                    • Interactions
                                                           •   Constraints   Development




  Current System            Measurement
                                                                                Data       Input       • Deduction
                                                                                                       • Simulation




                                                                                                                          MODEL SOLUTION
  Judgment                                                                                               (Indirect Exp)
          Trial and Error                                                                              • Optimization
         Experimentation
                                                                              Validation

        Future
        System


                                                                              Conclusions
Resource Allocation                       Interpretation                     (hypotheses)
    Decisions
                                    ANALYSES
                                                                                       Bonder, Masterson & Miller 2000
Simulation of the Wilford Hall
Medical Center Intensive Care
            Unit
Build a Virtual Critical Care
        Environment
Exercise Alternative Futures
  WHMC has a Memorandum of
Understanding with Texas to provide
  trauma services in San Antonio
95 %
Air Force’s Largest GME Center
                               WHMC ICU Demand Model

             Civilian Emergency         Beneficiary Trauma     Beneficiary Non-trauma less DM     COPD, CHF, Diabetes Patients

             3500

             3000
                     TC
Admissions




             2500
                          DM

             2000

             1500

             1000

              500

                0
                    FY 02      FY 03   FY 04   FY 05   FY 06   FY 07   FY 08   FY 09   FY 10    FY 11   FY 12   FY 13   FY 14   FY 15

                                                                          Year
                       Conceptual Model
                        Alternate ICU configuration
                        extends model applicability

                           Medical        Stepdown
 ICUs Partitioned By
      Specialty

                           Cardiac         Surgical
                                                             • Investigative Outcomes:
                       • Changes in:                               • Patients Deferred or
                            • Staffing Levels                       Turned Away
                            • Bed Availability        Comparative   • Bed Occupancy Rates
                            • Closure Policy             Outcomes   • MOU Compliance
                            • Internal Transfer Policies            • GME Compliance
   ICU                                                              • Costs
Admissions
                            Medical Stepdown
 Non-Partitioned,
Multi-Specialty ICU         Cardiac    Surgical
Model Demo
                                            Scenario Output
                                   Base     Option 1;    Option 2;                Option 3;   Option 3;
                           Base   Case yr   less PCU,   no PCU, all    Option 2   relocate    relocate     Option 4;   Option 4;
                           Case    2014     more ICU        ICU       year 2007   Peds ICU    Peds ICU    MilCon Proj MilCon Proj
       Patient Arrivals     186     214        186          186          196         186         214          186         214
       Available Beds
                     M       8       8         11           14           14          14          14           14          14
                     S      12      12         12           12           12          14          12           12          12
                     C      10      10         10           10           10          10          10           10          10
                     P       6       6          3            0            0           6           6            6           6
    Occupancy Rates
                     M    87.1%    89.5%     81.2%        74.6%        77.8%       70.1%       80.2%        74.3%       82.3%
                     S    84.8%    87.0%     84.7%        85.1%        86.2%       79.2%       83.5%        84.0%       86.6%
                     C    77.4%    85.1%     71.0%        65.7%        70.8%       63.5%       72.1%        64.9%       75.3%
                     P    41.8%    42.3%     59.6%        0.0%         0.0%        42.5%       43.6%        42.5%       43.3%
                Overall   76.1%    79.6%     77.7%        75.7%        78.7%       67.7%       74.4%        70.3%       76.3%
     Open to Trauma       88.6%    82.1%     93.3%        96.3%        95.1%       98.0%       94.9%        96.7%       92.3%
           Turn Aways
                     M      5.7    11.3        3.3         1.7          2.9          1.0         2.8          1.7         4.4
                     S      4.0    8.1         2.6         1.1          2.0          0.5         2.0          1.0         3.3
                     C      5.1    8.5         2.7         1.7          2.5          0.9         2.6          1.6         3.5
                      T     5.4    10.2        3.3         2.2          2.7          1.0         3.0          1.8         4.1
                     P      1.1    1.1         7.6         27.9         27.8         1.1         1.1          1.1         1.1
                 Costs
               Revenue    106330 110375     108169       101679       104512      115371      126949       113835      123689
          Cost to Treat   3062225 3200810   3106307      2913014      2996030     3326129     3673382      3281795     3582954
      Net Cost to Treat   2955896 3090435   2998139      2811336      2891519     3210758     3546433      3167960     3459265
Cost of MHS Turn-away      54460  119173     47818       106816       130152       41257       35906        21329       51154
             MHS Total    3010356 3209608   3045957      2918151      3021671     3252015     3582339      3189289     3510419
 Cost of Civ Turn-away     64201  117690     46263        49501        57093       7065        28199        18486       40122
                    Occupancy by ICU Type
                                                   Medicine         Surgical     Cardiac       PCU

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

 0%
       Base Case;   Base Case;    Option 1; less    Option 2; no    Option 2; yr 07;    Option 3;       Option 3;     Option 4;      Option 4;
          p186         p214      PCU, more ICU;     PCU, all ICU;        p196        relocate Peds   relocate Peds   MilCon Proj;   MilCon Proj;
                                       186             p186                            ICU; p186       ICU; p214         p186           p214



                                                           Scenario
                              Comparison Occupancy and
                                    Compliance
                                                            Overall Occupancy       Open to Trauma

                 85%                                                                                                                     100%
                                                                                          98.0%
                                                                  96.3%                                            96.7%
                                                                                95.1%                  94.9%                             95%
                 80%
                                                      93.3%
                                                                                                                                 92.3%
Occupancy Rate




                                                                                                                                                Open to Trauma
                                                                                                                                         90%
                 75%         88.6%

                                                                                                                                         85%

                 70%                      82.1%
                                                                                                                                         80%


                 65%
                                                                                                                                         75%


                 60%                                                                                                                     70%
                       Base Case; Base Case; Option 1; less Option 2; no Option 2; yr  Option 3;  Option 3; Option 4;      Option 4;
                          p186       p214     PCU, more PCU, all ICU; 07; p196 relocate Peds relocate Peds MilCon Proj;   MilCon Proj;
                                               ICU; 186        p186                   ICU; p186  ICU; p214     p186           p214




                                                                    Scenario
Leaders may obtain feedback in many ways
Modeling is a tool that can help
  us to plan for the future by
 questioning our assumptions,
checking our data, refining our
  metrics and designing our
         infrastructure.
      What are the impacts of
 alternative resource partitioning
policies on facility effectiveness?
    What are the impacts of
alternative closure policies on
    facility effectiveness?
What is the most effective size a
            facility?
What policies and procedures
best mitigate the impact of a
major medical deployment?
    Put it all together
                PLAN
                 PLAN
                 Model
                 System
           EVALUATE   IMPLEMENT




           Healthcare
            Delivery
EVALUATE    System                IMPLEMENT