Veteran Health Network CDR Bard, LCDR Campbell, MAJ Ford 5 June 2012 Outline • Backstory / Problem • Abstract Network • Network Operation • Measure of Effectiveness • Network Analysis • Summary and Conclusion 2 U.S. Dept. of Veteran Affairs (VA) VA Mission: Fulfill Lincoln’s promise Serve and honor America’s Veterans Lincoln’s Promise: “to care for him who shall have borne the battle, and for his widow, and his orphan” 3 VA 2011-2015 Strategic Plan Strategic Goals Access to Care Optimal Value Mindful of President Lincoln’s promise 4 Veteran Health Administration (VHA) Largest integrated health care system in United States 152 medical centers 1400 community-based outpatient clinics 21 Veteran Integrated Service Networks (VISNs) Meets needs of 8.3 million Veterans each year 5 VA Sierra Pacific Network • VISN-21 serves 1.2 million Veterans in northern and central California, northern Nevada, Hawaii, the Philippines, and Guam • Consists of 40 sites across six Systems • Each Health Care System is sub-network of larger VISN-21 network 6 Abstract of VISN-21 Examine three independent Health Care Systems VA Palo Alto VA Sierra Nevada VA Northern California Representative of urban, rural and hybrid areas 7 Community Terms • Urban (Palo Alto): Consists of major population centers • Rural (Sierra Nevada): Sparsely populated with a few small urban areas • Hybrid (Northern CA): Consists of major population centers surrounded by rural areas 8 Nodes and Edges • • Nodes: • SUPPLY - Veteran populations by county • DEMAND - Treatment facilities Edges: • Connect each county with network facility • Cost is distance in miles 9 Simplified Graph Veterans Treatment Treatment Demand COUNTIES CLINICS 10 Abstract of VISN-21 Siskiyou County 1 9 2 8 VA Northern 7 California (Hybrid) Yreka 3 4 5 6 11 VA Northern California (Hybrid) COUNTIES S U P P L Y CLINICS D E M A N D Network Analysis • Purpose: Provide outpatient care to Veterans • Data Tracked: • Cumulative Distance Traveled • Per Capita Distance Traveled (outputs) • Patients Assigned to Clinics 13 Measures of Effectiveness • • Model allocates Veterans to treatment facilities • Minimum-Cost Flow Modeling • Minimize Veteran travel distance to treatment Objective Function: min S cij yij cij: cost (distance) per unit flow yij: number of veterans (flow) on arc 14 Assumptions • All eligible Vets receive care from VA System • One City per County for distance calculations • • No population distribution for veterans in county • Community near geographic or population center Health Care Systems (HCS) operate independently • Ability for interchange among HCSs for specialty care • Not modeled for simplicity and tractability 15 Modeling • Begin with an unconstrained model • Add network design constraints and evaluate responses • Patient limits • Patient limits with buffers • Facility closure or patient capacity reductions • Open a new clinic • Year 2030 veteran populations 16 Unconstrained Results • All patients go to nearest clinic Distance to Care 35 30 Per Capita Distance: • • Urban: 13.69 miles Hybrid: 17.37 25 Distance (miles) • 20 15 10 5 • Rural: 30.02 0 Urban 17 Hybrid Network Rural Patient Limits • Capacities chosen to ensure no unmet demand • • • Urban / Suburban Outpatient Clinics • Capacity: 30,000 (urban / hybrid) • Capacity: 25,000 (rural) Rural Outpatient Clinics • • Modeled as upper bound on Clinic – Treatment Arcs Capacity: 10,000 Hospitals • Capacity: 75,000 (hybrid) • Capacity: 50,000 (urban / rural) 18 Patient Limits Results • Per Capita Distance: Distance to Care 45 • Urban: 16.26 miles 40 35 • Hybrid: 22.36 Rural: 40.93 30 Distance • 25 Urban Hybrid 20 Rural • 20 - 35% increase 15 10 5 0 Unconstrained Patient Limits Model 19 Patient Limits with Buffers • 1% buffer below capacity for all facilities • Allow for patient transfers • New sign-ups • Recently moved • Flexibility 20 Patient Limits with Buffers Results • Distance to Care Per Capita Distance: • 45 40 Urban: 16.32 miles 35 • Hybrid: 23.0 Rural: 41.25 30 Distance • 25 Urban Hybrid 20 Rural 15 • Baseline • Closest to Reality 10 5 0 21 Unconstrained Patient Limits Model With Buffers Budget Cuts • • All HCSs must close a clinic or reduce staffing to save costs • Force a clinic closure in each region • Reduce patient limits across the board to simulate staffing cuts Model chooses optimal clinic to close and redistributes patients 22 Closures Hybrid: Yreka, CA Rural: Winnemucca, NV Urban: Capitola, CA 23 Budget Cuts Results • Per Capita Distance (optimal): Distance to Care 50 • • Urban: 16.99 miles 45 40 Hybrid: 23.60 35 • Rural: 41.25 Optimal Decision • Urban: Staff Cuts (+0.67) 30 Distance • 25 Urban Hybrid Rural 20 15 10 • • Hybrid: Closure (+0.60) Rural: Closure (+1.50) 5 0 Unconstrained 24 Patient Limits With Buffers Model Closure Staff Reductions Budget Cut Takeaways • • Can safely close one clinic in each network without disruption • Two or more closures trigger unmet demand (untreated patients) • Network is efficient but vulnerable • Redundancy is expensive and not an efficient use of limited resources Maximum reductions in patient capacities (staff cuts) without disruption • Urban: 10 percent – unmet demand • Rural: 5 percent – unacceptable patient assignments • Hybrid: 2.5 percent – unmet demand 25 Open New Clinic • Political Pressures • Can’t close a clinic and displace vets • Must open a new clinic in each network • Modeled after VA’s Rural Outreach Program • • Opening new small clinics in rural, underserved areas • Yreka (CA) and Winnemucca (NV) are examples Optimal choice from among three communities in each region 26 Urban Three possible locations Rancho Calaveras, CA Tracy, CA Hollister, CA 27 Hybrid Three possible locations Weaverville, CA Orland, CA Colusa, CA 28 Rural Three possible locations Austin, NV Fernley, NV Mammoth Lakes, CA 29 Open Clinic Results • Per Capita Distance: Distance to Care 50 • Urban: 16.06 miles 45 40 • • Hybrid: 21.92 Rural: 49.72 Reduction from Baseline 35 30 Distance • 25 Urban Hybrid 20 Rural 15 • • Urban: 0.26 (1.5%) Hybrid: 1.08 (4.7%) 10 5 0 • Rural: 1.53 (3.7%) Unconstrained Patient Limits With Buffers Model 30 Closure Staff Reductions Open Clinic Open Clinic Takeaways • • Not worthwhile in urban network • Slight decrease in objective function • No patient load decreases on full capacity clinics Effective in hybrid and rural networks • Larger decreases in objective function • Decreased patient loads at full clinics 31 2030 • What does the future hold? • • Veteran population projections by county from the VA for 2030 • 40% reduction from current level • Fewer WWII, Korea, and Vietnam era vets • Drafts vs. Volunteer Force Assumed all current clinics remain • • Not likely to be true Will VA system be folded into National Health Care System? 32 2030 Results • Distance to Care Per Capita Distance: • 50 45 Urban: 14.88 miles 40 • Hybrid: 17.69 35 • • Rural: 32.2 Similar to Unconstrained model results Distance 30 25 Urban 20 Hybrid Rural 15 10 • • Clinic capacities become inconsequential 5 0 Future closures? Model 33 Conclusion • Network satisfies strategic objective • Network resilient to limited disruption • Offers insight to VA network of networks • Project results in alignment with VA practices • Flexibility for future Veteran population 34 Future Work • Add competing objective function(s) • Minimize Veteran traveling distance • Minimize cost per patient • Minimize overhead costs • Increase granularity • Determine Optimum Staffing Levels • Model to help VA meet strategic goals considering 35-40% decrease in Veteran population by 2030 35 QUESTIONS? 36 BACKUP 37 Patient Limits and Buffers Results Urban Patient Loads 45 80000 40 70000 35 60000 Patients Assigned Distance (miles) Distance to Care 30 25 Unconstrained 20 Patient Limits 15 With Buffers 50000 40000 Unconstrained 30000 Patient Limits 20000 10 10000 5 0 With Buffers 0 Urban Hybrid Network Rural Facilites Rural Patient Loads Hybrid Patient Loads 60000 120000 50000 80000 60000 Unconstrained Patient Limits 40000 Patients Assigned Patients Assigned 100000 40000 30000 Unconstrained Patient Limits 20000 With Buffers With Buffers 10000 20000 0 0 Facilities 38 Facilities Budget Cuts Results Distance to Care Urban Patient Loads 80000 45 70000 40 60000 Patients Assigned 50 Distance (miles) 35 30 Unconstrained 25 Patient Limits With Buffers 20 15 50000 Unconstrained 40000 Patient Limits 30000 With Buffers Closure 20000 Closure Staff Reductions 10000 Staff Reductions 10 0 5 0 Urban Hybrid Network Rural Facilites Rural Patient Loads 120000 60000 100000 50000 80000 40000 Unconstrained 60000 Patient Limits With Buffers 40000 Patients Assigned Patients Assigned Hybrid Patient Loads 30000 Unconstrained Patient Limits With Buffers 20000 Closure Closure Staff Reductions 20000 Staff Reductions 10000 0 0 Facilities 39 Facilities Open Clinic Results Urban Patient Loads 45 80000 40 70000 35 60000 30 25 Unconstrained 20 Patient Limits With Buffers 15 Patients Assigned Distance (miles) Distance to Care 50000 40000 Unconstrained 30000 Patient Limits With Buffers 20000 Open Clinic Open Clinic 10000 10 0 5 0 Urban Hybrid Network Rural Facilites Rural Patient Loads Hybrid Patient Loads 60000 120000 50000 80000 60000 Unconstrained Patient Limits 40000 With Buffers Patients Assigned Patients Assigned 100000 40000 30000 Unconstrained Patient Limits 20000 With Buffers Open Clinic Open Clinic 10000 20000 0 0 Facilities 40 Facilities 2030 Results Urban Patient Loads 45 80000 40 70000 35 60000 Patients Assigned Distance (miles) Distance to Care 30 25 Unconstrained 20 Patient Limits With Buffers 15 50000 40000 Unconstrained 30000 Patient Limits With Buffers 20000 2030 2030 10000 10 0 5 0 Urban Hybrid Network Rural Facilites Rural Patient Loads 120000 60000 100000 50000 80000 40000 60000 Unconstrained Patient Limits 40000 With Buffers Patients Assigned Patients Assigned Hybrid Patient Loads 30000 Unconstrained Patient Limits 20000 With Buffers 2030 2030 10000 20000 0 0 Facilities 41 Facilities