39 th NIC, Washington DC, 21 March Daniel A. Allwine, M.S. This work is supported by Contract No from the Centers for Disease Control and Prevention (CDC). CDC Project Officer: Bruce G. Weniger, M.D., M.P.H. Immunization Safety Branch, NIP, CDC The Vaccine Formulary Selection Algorithm: A Web-Based Tool for Determining Best- Economic Value Pediatric Immunization Formularies
39 th NIC, Washington DC, 21 March GOVERNMENT: Centers for Disease Control and Prevention Bruce G. Weniger, M.D., M.P.H. - Project Officer, SBIR Contract No Vaccine Domain knowledge. INDUSTRY: Austral Engineering and Software, Inc. Enrique Medina, M.S., Daniel Allwine, M.S., M. QadeerAhmed -Software and user interface design and web implementation. -User interaction. ACADEMIA: Southern Illinois University Edwardsville University of Illinois at Urbana-Champaign Edward C. Sewell, Ph.D. Sheldon H. Jacobson, Ph.D. - Modeling and algorithm development. A collaborative effort between government, industry, and academia. Participants and Contributions
39 th NIC, Washington DC, 21 March The economic factors involved in establishing a cost-effective pediatric immunization formulary are not straight forward. Some tangible and intangible costs: –Vaccines and related products –Administration, storage, and delivery –Pain and suffering (leads to avoidance) –Potential treatment for unvaccinated individuals Combination vaccines are making the procurement landscape more complicated.
39 th NIC, Washington DC, 21 March Assist health care professionals in making vaccine formulary choices for purchasing. Automate determination of best-value formularies to procure. Facilitate use of vaccine selection algorithm through a user-friendly website Provide access to large number of users throughout the health care system. Allow users to customize costs for their particular situation.Objectives
39 th NIC, Washington DC, 21 March Purchasers of Vaccine Products Public Sector Agencies Federal, state, and local agencies Physician Groups Health Maintenance Organizations Pediatricians and Family Practitioners Health Insurance Companies Vaccine Manufacturers Potential Users of Website
39 th NIC, Washington DC, 21 March CDC/Academia collaboration since Operations research modeling and economic vaccine selection algorithm within the childhood immunization schedule. Weniger, B.G., Chen, R.T., Jacobson, S.H. Sewell, E.C. Deuson, R., Livengood, J.R., Orenstein, W.A., 1998, "Addressing the Challenges to Immunization Practice with an Economic Algorithm for Vaccine Selection," Vaccine, 16(19), Jacobson, S.H., Sewell, E.C., Deuson, R., Weniger, B.G., 1999, “An Integer Programming Model for Vaccine Procurement and Delivery for Childhood Immunization: A Pilot Study,” Health Care Management Science, 2, 1-9. Sewell, E.C., Jacobson, S.H., Weniger, B.G., 2001, “ “Reverse Engineering” a Formulary Selection Algorithm to Determine the Economic Value of Pentavalent and Hexavalent Combination Vaccines,” Pediatric Infectious Disease Journal, 20(11), S45-S56. Jacobson, S.H., Sewell, E.C., 2002, "Using Monte Carlo Simulation to Determine Combination Vaccine Price Distributions for Childhood Diseases,” Health Care Management Science, 5(1). Jacobson, S.H., Sewell, E.C., Allwine, D.A., Medina, E.A., and Weniger, B.G., “Designing pediatric vaccine formularies and pricing combination vaccines using operations research models and algorithms,” Special Report. Expert Review of Vaccines 2(1), 15-19, Jacobson, S.H., Karnani, T., and Sewell, E.C., “Analyzing the economic value of the hepatitis-B - Haemophilus influenzae type B combination vaccine by reverse engineering a formulary selection algorithm,” Vaccine 2003;21: Technical Content
39 th NIC, Washington DC, 21 March Operations research model currently captures: ACIP recommended childhood immunization schedule requirements. Cost components (e.g., vaccine purchase prices, labor costs, office visit costs, injection costs). Vaccine constraints (e.g., number of injections, age limitations, brand matching). Vaccine wastage by expiration. Operations Research Model
39 th NIC, Washington DC, 21 March Problem Characteristics Goal: Find optimal formulary –Lowest-cost set of vaccines that satisfies the Recommended Childhood Immunization Schedule Puzzle Analogy: Put pieces together to form the picture –We know the desired picture –Each piece has a price –There are many extra pieces –A piece can be used in more than once –100,000 to 5,000,000 ways to assemble the puzzle –We must find the least expensive way
39 th NIC, Washington DC, 21 March Recommended Childhood and Adolescent Immunization Schedule — United States — 2005 Approved by the Advisory Committee on Immunization Practices, the American Academy of Pediatrics, and the American Academy of Family Physicians. Influenza (yearly)
39 th NIC, Washington DC, 21 March Vaccine Selection Algorithm Uses operations research principles to determine the best-value (lowest cost) vaccine formulary that satisfies all the constraints. Efficiently search across all possible sets of vaccine products –There are approximately 100,000 to 5,000,000 vaccine formularies to examine for each set of cost data. –As additional vaccines (particularly combination vaccines) are added to the market, this number will grow exponentially!
39 th NIC, Washington DC, 21 March Search Tree Organized procedure to examine all the possibilities At each level, must choose a vaccine for a disease/month. Preprocessing eliminates vaccines that cannot possibly be in best-value formulary. Pruning eliminates useless branches. MRKNoneGSKMRK HIB-HBV AVPGSKAVP GSK HIB Month 2 HIB Month 2 DTPa Month 2 HIB Month 2 HIB Month 2 DTPa Month 2 HIB Month 2 HIB Month 2 DTPa Month 2 HIB Month 2 HIB Month 2 DTPa Month 2 HBV Month 2
39 th NIC, Washington DC, 21 March Search Engine and Validation “Back End” Search Engine is written in C++ Typical problem –Solved in less than half a minute –500,000 to 1,000,000 branches (partial solutions). 50,000 to 100,000 full solutions are examined Problem was also modeled as an Integer Program (IP) Used AMPL to generate the IP from the data, CPLEX to solve the IP Randomly generated 1,000 test problems –Input parameters randomly chosen –Prices randomly generated –Existing vaccines randomly chosen for inclusion/exclusion –Hypothetical combination vaccines randomly created Test problems solved using both the search engine and the IP. Same solutions were found by both methods
39 th NIC, Washington DC, 21 March Developed in state-of-the-art Microsoft ®.NET platform Server Software: .NET components: solver in C++ and a managed wrapper in C# Default and user-customized vaccine data in SQL database Middleware Software: ASP.NET scripts generate HTML and Javascript code Served to web browser (client) by Internet Information Server Client software: user interface through common browsers Best when viewed in Internet Explorer Recently released a more friendly user interface for occasional users Similar to common tax preparation software packages Interview driven Website Architecture and User Interface
39 th NIC, Washington DC, 21 March Hypothetical AVP DTPa-HIB-IPV A hypothetical DTPa-HIB-IPV pentavalent product from AVP can be considered. Hypothetical AVP DTPa-HIB-IPV Product Product Price ($) Occurrences in Solution Total Cost ($) TotalInjections
39 th NIC, Washington DC, 21 March finds best-value vaccine formularies (all antigens as of today) Enables “what-if” scenarios Web interface supports Government, “Private”, and/or user-customized prices Website use is currently free Additional services available -Analysis of your particular vaccine procurement scenario -Automated what-if scenarios, price selection, model structure customization, price change notifications Closing Remarks