FutureDocs Forecasting Tool An Open Source Physician Projection Model

1. FutureDocsForecasting ToolAn Open Source Physician Projection Model Erin P. Fraher, PhD, MPP with G. Mark Holmes, PhD and Andy Knapton, MSc Cecil G. Sheps Center for Health Services Research University of North Carolina at Chapel Hill July 29, 2014

2. In case your editor calls, here’s why our model is innovative in 2 slides: Slide 1 • Starts with different question-what services will patients need versus how many doctors will we need? • Uses different geography-state and sub-state data highlight differences between “haves” and “have nots” • Displays data in interactive format-model is web-based, open source and designed to be customized, challenged and improved • Seeks different outcome-designed to educate and engage stakeholders in redesigning system

3. In case your editor calls, here’s why our model is innovative in 2 slides: Slide 2 Mindset — it’s a tool, not an answer. Tool allows user to choose from: • 3 models — supply, utilization, relative capacity (a.k.a. “surplus/shortage”) • 3 types of visualizations — maps, line charts and population pyramids • 3 geographic views — national, state and sub-state level • 5+ alternate futures — “what if” scenarios regarding ACA implementation, physician FTEs, retirement, use of NPs and PAs, and redistributing graduate medical education (GME)

4. Presentation Overview • Highlight why our model is a disruptive innovation • Explain (at ~60,000 feet!) methodology • Demonstrate sample findings • Answer your questions

5. Our model is a disruptive innovation that challenges traditional workforce modeling • Silo-based projections by physician specialty • No “what if” scenarios • Proprietary (read: black box) & uncustomizable models • Not regularly updated • Lack friendly and interactive user interface

6. https://www2.shepscenter.unc.edu/workforce

7. Innovations in modeling physician supply Supply Side Innovations • Modeled patient carehours (patient care FTE) • First model to include detailed GME training pathways, including sub-specialization trends • Includes scenario that redistributes GME slots to states and specialties where greatest need

8. Innovations in modeling patient use of health care services • Created Clinical Service Areas (CSAs) to capture why and where people seek care • 19 Clinical Service Areas (e.g., respiratory conditions, circulatory conditions, endocrinology, mental health, preventative care, etc.) • Modeled use of health care in 3 settings: • outpatient (including physician offices and hospital outpatient settings) • inpatient settings • emergency departments

9. Developed sub-state unit of geography, Tertiary Service Areas (TSAs) To capture sub-state variation, created TSAs • Based on Dartmouth’s Hospital Referral Regions • TSAs are markets that encompass primary and specialty care services • Health system consolidation and ACOs and ACO-like structures create need for region-based data

10. Plasticity matrix brings supply and service use together by mapping physicians to services • Starting question: what health services will patients need? • Next question: which physician specialties can provide those services? • Innovation: plasticity matrix maps services provided by physicians in different specialties to patients’ visits

11. Plasticity—Providers and Services: A sample matrix for outpatient settings Number of outpatient visits, select specialties and CSAs

12. Plasticity—Providers and Services: A sample matrix for outpatient settings Number of outpatient visits, select specialties and CSAs Within a CSA, how are outpatient visits are distributed across specialties?

13. Plasticity—Providers and Services: A sample matrix for outpatient settings Number of outpatient visits provided per FTE per year, select specialties and CSAs

14. Plasticity—Providers and Services: A sample matrix for outpatient settings Number of outpatient visits provided per FTE per year, select specialties and CSAs Within a specialty, how are visits distributed across CSAs?

15. These Innovationsturn workforce modeling upside down • Model does not produce estimate of counts of physicians needed by specialty • Instead, it asks: what are patients’ needs for care and how can those needs be met by different workforce configurations in different geographies?

16. “Relative Capacity”: Indicator of how well physician supply matches use of services Model calculates “relative capacity”—a measure for each clinical service area in each geography = supply of visits physicians in that TSA/State can provide utilization of visits needed by population in TSA/State <.85=shortage .85-1.15=in balance >1.15=surplus

17. You end up with a picture that shows capacity of workforce to meet demand for different types of health services Shortage/Surplus for Outpatient Circulatory Visits by TSA, 2014 Bangor, ME Rochester, MN Boston, MA New York, NY Boulder, CO Washington, DC San Francisco, CA Huntington, WV Los Angeles, CA Raleigh-Durham, NC Atlanta, GA Slidell, LA Dallas-Fort Worth, TX Miami-Fort Lauderdale, FL Houston, TX

18. Just a few interesting findings to show model capacity FutureDocsForecasting Tool

19. Supply model: Pediatric surgical FTEs double between 2011 and 2030 1,402 699

20. Supply model: 12% decline in general internal medicine FTEs 58,849 51,553

21. Not much change in shortage/surplus for all visits at the national level In Balance Shortage

22. But looking at national data conceals broad variation between geographies Shortage/Surplus for All Visits, All Settings by TSA, 2014 Bangor, ME Rochester, MN Aurora, IL Melrose Park, IL Boston, MA New York, NY Boulder, CO Washington, DC San Francisco, CA Huntington, WV Los Angeles, CA Raleigh-Durham, NC Atlanta, GA Slidell, LA Dallas-Fort Worth, TX Miami-Fort Lauderdale, FL New Orleans, LA Houston, TX

23. Health system is rapidly changing.Need scenarios to model “what ifs” Model includes scenarios for: • Baseline model (2011) assumes ACA not implemented • Utilization side — implement exchanges and different Medicaid assumptions • On supply side — change retirement, FTE, use of NPs/PAs, and redistribute GME

24. What would be effect on shortage/surplus if all states expanded Medicaid? Shortage/surplus for all visits in all settings if all states expand Medicaid In Balance Shortage

25. The model highlightsdifferential effect between states: Medicaid expansion has larger effect in Texas…

26. …And a smaller effect in Massachusetts

27. We modestly redistributed GME slots to specialties/states where demand > capacity in 2030 Specialty that loses the most slots: Emergency medicine, 60 slots Specialty that gains the most slots: General internal medicine, 61 slots (Note that family medicine gains 55 slots.)

28. Example: GME scenario removes 40 emergency medicine slots from New York, reducing emergency medicine physician supply in that state 3,785 patient care FTEs under baseline, 2030 3,499 patient care FTEs under GME redistribution, 2030 -7.6% difference

29. But New York still has surplus capacity for ED visits, even when 40 residency slots are removed Surplus In Balance 2.58 supply/visits under baseline, 2030 2.41 supply/visits under GME redistribution, 2030 -6.6% difference Shortage

30. GME redistribution has differential effect by state: Nevada benefits but IM still declines 171 patient care FTEs under baseline, 2030 189 patient care FTEs under GME redistribution, 2030 +10.5% difference

31. Now let’s field some questions… • We’ll take the next 10 minutes for any questions • For interview requests please contact physfnd@cooperkatz.com