Tim Pletcher Presentation
- 1. The Analytics Opportunity in Healthcare
- 2. Agenda • Brief IHBI overview • Value of Analytics • Share best/worst project • Brief MiHIN introduction • ADT example • Adding intelligence to the network 2
- 3. IHBI@CMU: Snapshot Founded in 2001 as a center of excellence specializing in data mining and predictive analytics. Purpose • Promote the application of predictive analytics to solve social, economic and business challenges • Build bridges between the public and private sector • Train the next generation of data scientists 2010 • Moved under the auspices of the Herbert H. and Grace A. Dow College of Health Professions • Specifically charged to develop a focus on healthcare Team • PhD and MS in physics, statistics, economics, computer science, geography, etc. 3
- 4. Customers and Partners Health and Healthcare • Central Michigan District Health Dept. • College of Health Professions, CMU • College of Medicine, CMU • Eli Lilly • Henry Ford Health System • Michigan Health Information Alliance • Michigan Health Information Network • Partners Healthcare (Boston) • Spectrum Health System • Synergy Medical Manufacturing • The Dow Chemical Company • The Dow Corning Corporation • Ford Motor Company • General Motors • Harley-Davidson • Monsanto • Steelcase • Whirlpool Corporation Technology • IBM • SAS Institute • Hewlett-Packard (EDS) • Greenplum Pivotal Banking, Finance, Insurance • Auto-Owners Insurance • Comerica Bank Other • Proctor and Gamble • DTE Energy • Domino's Pizza • Gordon Food Service • State of Michigan 4
- 5. Data Science & Big Data Reporting Queries/drill down Alerts Statistical Analysis Forecasting Predictive Modeling Optimization What happened? Where is the problem? What action are needed? Why is this happening? What if these trends continue? What will happen next? What is the best that can happen? Analyticallyimpaired Adapted from Competing on Analytics: The New Science of Winning (Davenport, 2007). 5
- 6. IHBI Contextual Dataset One of the unique assets that IHBI has developed over the last few years is a substantial dataset called the Contextual Dataset • Based on a collection of population, socioeconomic, environmental, geographic, and health care variables • Designed to be integrated with private data to enhance modeling and knowledge • Data granularity is at the zip code level for all variables nationwide • We are gradually shifting the granularity of all variables to census tract by downloading the data at that level where available or otherwise using area and/or population weighted distribution methods to convert from a different granularity 6
- 7. 7
- 8. Research ICE3 Potential Reusable Data Sources 8
- 9. 9 A Time and Place • Guess/Intuition – Outcome does not matter-low value or (low accountability) – No other choice • Data-Driven Results – Reports Sales Are Down or performance metrics • Fact-based Assessments – Correlation vs. causation • Model- and Scenario- Assisted Decision Making – Predicting customer response – Predicting outliers – Simulation • Experimental Design “Test & Learn”
- 10. 10 Project Example Scenario: Hospital had a very low census. Relatively closed system where they owned the providers and only a small percentage of referrals came from outside the system. Goal: Provide a predictive model that forecasts total weekly admission rates by admitting specialty and weekly census on each nursing unit.
- 11. 11 Roadmap Patient Encounters PEMS Medical Records Classify records: a=I wi. E or O b= no I c=I w/o E or O a b c External referrals Map weekly admits to day-of-week Seq Rules Recent E/R & Outpt Internal referrals Admits by week by DR_SPEC Map DR_SPEC to LOS & NUR_UNIT Forecast integration DEMAND beds per NUR_UNIT Major Deliverable To Capacity & resource planning List of unusual events
- 12. 12 Total Forecasted vs. Actual Admits May04-April05 500 550 600 650 700 750 800 850 5/2/04 5/16/04 5/30/04 6/13/04 6/27/04 7/11/04 7/25/04 8/8/04 8/22/04 9/5/04 9/19/04 10/3/04 10/17/04 10/31/04 11/14/04 11/28/04 12/12/04 12/26/04 1/9/05 1/23/05 2/6/05 2/20/05 3/6/05 3/20/05 4/3/05 4/17/05 weekly_fcst actual admits
- 13. 13 Predicted vs. actual admits in June 05 for major doctor specialty codes 6/5/05 6/12/05 6/19/05 6/26/05 Total Ratio CAR 89 ( 79 ) 89 ( 75 ) 90( 78 ) 90 ( 83 ) 358 ( 315 ) 1.14 GYN 54 ( 49 ) 54 ( 69 ) 53 ( 63 ) 53 ( 52 ) 214 ( 233 ) 0.92 IMG 192 ( 198 ) 193 ( 168 ) 193 (209 ) 194 (164 ) 772 ( 739 ) 1.04 NES 33 ( 43 ) 33 ( 33 ) 33 ( 30 ) 33 ( 35 ) 132 ( 141 ) 0.94 NEU 34 ( 22 ) 34 ( 35 ) 34 ( 39 ) 34 ( 42 ) 136 ( 138 ) 0.99 ONC 28 ( 30 ) 27 ( 25 ) 27 ( 39 ) 27 ( 33 ) 109 ( 127 ) 0.86 ORT 26 ( 31 ) 26 ( 47 ) 26 ( 39 ) 26 ( 38 ) 104 ( 155 ) 0.67 PUL 61 ( 55 ) 61 ( 71 ) 62 ( 46 ) 62 (50 ) 246 ( 222 ) 1.11 SUR 67 ( 55 ) 67 ( 58 ) 67 ( 52 ) 66 ( 55 ) 267 ( 220 ) 1.21
- 14. 14 Length of Stay by Doctor Specialty SPEC LOS Pts % pts %beds COL 1 96 0.0592 1.0000 COL 2 111 0.0685 0.9408 COL 3 161 0.0993 0.8723 COL 4 301 0.1857 0.7730 COL 5 269 0.1659 0.5873 COL 6 183 0.1129 0.4213 COL 7 145 0.0895 0.3085 COL 8 89 0.0549 0.2190 COL 9 76 0.0469 0.1641 COL 10 39 0.0241 0.1172 COL 11 21 0.0130 0.0932 COL 12 21 0.0130 0.0802 COL 13 19 0.0117 0.0672 COL 14 10 0.0062 0.0555 COL 15 10 0.0062 0.0494 Example: COL = Colon & rectal surgery; 1621 patients in first 5 months of 2004 COL- LOS Early 2004 0 50 100 150 200 250 300 350 0 2 4 6 8 10 12 14 16
- 15. 15 Predicted vs. actual of bed occupancy for the 2nd week of June 05 Sunday Monday Tuesday Wednesday Thursday Friday Saturday Unit 12-Jun-05 13-Jun-05 14-Jun-05 15-Jun-05 16-Jun-05 17-Jun-05 18-Jun-05 ICU's C5M 25 (28) 26 (29) 27 (29) 28 (28) 28 (28) 28 (29) 27 (31) C6N 10 (11) 10 (10) 11 (10) 11 (9) 11 (10) 11 (10) 11 (9) Cardiology H5 25 (26) 27 (30) 29 (30) 29 (27) 29 (23) 30 (26) 27 (23) I5 25 (30) 27 (30) 28 (30) 29 (28) 29 (24) 29 (23) 27 (22) Medical GPU's B1 26 (30) 27 (30) 29 (28) 29 (27) 30 (28) 30 (28) 28 (29) B2 25 (29) 27 (27) 28 (28) 28 (26) 29 (26) 29 (27) 27 (23) B6 20 (19) 21 (21) 22 (22) 23 (20) 23 (19) 23 (18) 22 (16) F6 11 (11) 12 (13) 12 (16) 13 (15) 13 (13) 13 (11) 12 (10) Surgical GPU's B4 21 (16) 22 (19) 24 (24) 24 (28) 24 (28) 25 (21) 23 (14) F4 24 (28) 25 (27) 26 (26) 27 (28) 27 (27) 27 (25) 26 (23) Total predicted 511 542 571 579 587 592 553 Total census 551 586 603 596 589 575 547
- 16. 16 Patient History DATE Type SITE DR SPEC CD 03/19/2001 O 20 GAS 03/30/2001 O 20 GAS 04/30/2001 O 20 GAS 06/18/2001 O 20 GAS 06/20/2001 O 20 GAS 06/26/2001 O 20 GAS 07/19/2001 O 20 GAS 08/21/2001 O 20 GAS 09/27/2001 O 20 GAS 11/27/2001 O 20 GAS 01/10/2002 O 20 GAS 01/15/2002 O 20 GAS 05/23/2002 O 20 GAS 08/19/2002 O 20 GAS 09/25/2002 I 20 SUR MRN=5213569 Visits: O.GAS.20 Leads to: I.SUR.20
- 17. 17 Examples of the rules • There are 406 rules in which ‘E’ or ‘O’ are followed by ‘I’ COUNT SUPPORT CONF RULE 1755 2.017798013 71.98523 O.NES.20 ==> I.NES.20 1131 1.30035872 32.48133 O.ONC.20 ==> I.ONC.20 930 1.069260486 21.57272 O.NEP.20 ==> I.IMG.20 1014 1.165838852 17.61946 O.GAS.20 ==> I.IMG.20 2478 2.84906181 13.18576 O.CAR.20 ==> I.IMG.20 475 0.546127667 10.86957 O.CAR.36 ==> I.CAR.20 642 0.738134658 9.610778 O.URO.20 ==> I.CAR.20 493 0.56682303 8.56646 O.GAS.20 ==> I.SUR.20 504 0.579470199 4.474432 O.IMG.30 ==> I.GYN.20 448 0.515084621 13.37713 O.IMG.81 ==> I.IMG.20
- 18. 18 Distribution of Time Lags for one rule
- 19. 19 Proj of DSC(i) by week Proj of DSC(i) by week Projection with Rules Various DR_SPEC_CD Proj of DSC(i) by week 1-52 Wk 52 Wk -50 Wk -51 Wk 0 Rules with Weibull distribution Rules Left-Hand-Side Wk 2 Wk 1
- 20. MiHIN Role • Manage statewide legal trust fabric for data sharing • Maintain statewide “master data” in Active Care Relationship Service, Health Provider Directory, Trusted Identities, Consumer Preferences • Connect HIEs, Payers, Pharmacies, DCH, Federal Government, others • Align incentives or regulations to fairly share data and promote data standardization (via Use Cases) • Convene groups to identify data sharing barriers, reduce provider burdens, engage consumers, & enable population health Copyright 2015 Michigan Health Information Network Shared 20
- 21. State-wide Shared Services MDCH Data Hub Medicaid MSSS State LABS Doctors & Community Providers HIE QOs (Qualified sub-state HIEs) Network of Networks: Data Warehouse 21 Health Plan QOs (more coming) Single point of entry/exit for state Virtual QOs Pharmacies (more coming) Immunizations Mi Syndromic Surveillance System Mi Disease Surveillance System Consumer QOs (more coming) Federal Copyright 2015 Michigan Health Information Network Shared Service PIHPs
- 22. Transitional care management Medicare & BCBSM fees Jan 2013 Payer Code Non-Facility Facility Locality Medicare 99495 $120.39 $99.38 Detroit Medicare 99496 $169.65 $145.70 Detroit BCBSM 99496 $329.33 $281.38 All 22 99495 - • Communication (direct contact, telephone, electronic) with the patient and/or caregiver within 2 business days of discharge • Medical decision making of at least moderate complexity during the service period • Face-to-face visit, within 14 calendar days of discharge 99496 - • Communication (direct contact, telephone, electronic) with the patient and/or caregiver within 2 business days of discharge • Medical decision making of high complexity during the service period • Face-to-face visit, within 7 calendar days of discharge Copyright 2015 Michigan Health Information Network Shared Services
- 23. Statewide Health Provider Directory 23 • Contains Electronic Service Information (ESI) used to route information to providers • Flexibly maintains multiple distribution points for single provider or single distribution for organization • Manages organizations, providers and the multiple relationships between them Copyright 2015 Michigan Health Information Network Shared Services
- 24. Patient Provider Attribution Service (in Michigan we call this ACRS™) • Enables providers to declare active care relationships with patients – this attributes to a patient the active members of their care team • Accurately routes information (e.g. Admit-Discharge-Transfer messages, medication reconciliations) • Improves care coordination • Reduces readmissions • Allows better outcomes • Enables alerts to providers in active care relationships with patients • Coordinates entire care team with changes to patient status in real time • Allows searches by authorized persons or organizations: • Health systems and provider/physician organizations • Care coordinators • Health plans • Consumers (who can dispute asserted relationships) 24Copyright 2015 Michigan Health Information Network Shared Services
- 25. Active Care Relationship Service (ACRS™) 25 Patient Information Source Patient ID First Name Middle Initial Last Name Suffix Date of Birth Gender SSN – Last 4 digits Address 1 & Address 2 City, State, Zip Home & Mobile Phones Physician Information NPI First Name Last Name Practice Unit ID Practice Unit Name Physician Organization ID Physician Org Name Physician DIRECT Address DIRECT Preferences Copyright 2015 Michigan Health Information Network Shared Services
- 26. Data Sharing Organization (DSO) Data Sharing Organization (DSO) Supports Seamless Exchange Alerting Disconnected Entities Patient to Provider Attribution Health Provider Directory 1) Patient goes to hospital which sends message to DSO then to MiHIN 2) MiHIN checks patient-provider attribution and identifies providers 3) MiHIN retrieves contact and delivery preference for each provider from HPD 4) Notifications routed to providers based on electronic address and preferences Primary Care Specialist Care Coordinator 26 Patient Copyright 2015 Michigan Health Information Network Shared Services
- 27. Adding Intelligence to ADTs Copyright 2015 - Michigan Health Information Network Shared Services 27 Scoring activity Patient Risk Score/s Hospital Patient Risk Score/s
- 28. Adding Intelligence to ACRS Copyright 2015 - Michigan Health Information Network Shared Services 28
- 29. Next Generation Care Coordination Predictive Scores can help care coordinator s prioritize effort and eventually automate next steps ALERT! Readmission Avoided Schedule follow up visit within 7-14 days Primary Care Specialist Care Coordinator Existing Statewide Capability (90% of All Admissions Available) Intelligence Opportunity Patient to Provider Attribution (ACRS) Health Provider Directory Patient Copyright 2015 - Michigan Health Information Network Shared Services 29
- 30. Thank You & Contact Info Tim Pletcher, DHA Director Institute for Health & Business Insight 989.621.7221 pletc1ta@cmich.edu http://www.youtube.com/watch?v=elHjESwQ8_o&feature=youtu.be 30 Tim Pletcher, DHA Executive Director Michigan Health Information Network 989.621.7221 pletcher@mihin.org