Rx Decision: A Decision Support Tool for Choosing Prescription Drug Plans for Patients Chris Anderson Joey Fadul Anupam Menon Harold Terceros Agenda • • • • • • Context Stakeholder Analysis Insurance Terminology Problem Statement Needs Statement Decision Support Tool • • • • • • • • • 5/28/2016 Task Analysis Video Description PUF Dataset Output Method of Analysis Monte Carlo Analysis Usability Test Recommendations/Future Work Project Management 2 Context • In 2009, the US spent $2.5 trillion dollars towards healthcare, almost 17% of gross domestic product (GDP) [1] • One step to reduce the total expenditure was the implementation of the Affordable Care Act, or “Obama-care” • Passed by Congress but currently being litigated in the Supreme Court • Constitutionality of un-insured fines • Medicaid expansions, etc. [1] Centers for Medicare & Medicaid Services-www.cms.gov 5/28/2016 3 Context: Affordable Care Act • Affordable Care Act offered the following proposed changes: [2] • Incentives of $40,000 for physicians to use an Electronic Medical Record (EMR) System • Gradual discounts to covered drugs (brand name and generic) to lessen the burden on patients in the “doughnut hole” • Creation of Heath IT (HIT) Extension Programs that would facilitate regional adoption efforts • Provision of funds to states to coordinate and promote interoperable EMRs • Accelerated construction of the National Health Information Network (NHIN) [2] www.healthcare.gov/law 5/28/2016 4 Context: Rx Costs • R & D phase has large attrition rate during development [5] • 5000 drugs screened • ONLY 5 enter testing • No income from this stage • Approximately 70% of cost during first 12 years (completion of clinical evaluations) Proportion of Chronically Ill Who Have Taken Fewer Rx Drugs Due to Cost Percent 25 20 15 10 5 0 Private Insurance Percent 5 Medicare/ Medicaid 20 [5] Rowberg, Richard. CRS Report for Congress, 2001 [6] Adapted from Center on an Aging Society, Georgetown University, 2004 5/28/2016 Uninsured 22 5 Context: Chronic Conditions in America • In 2005, 7 out of 10 deaths were from chronic conditions • Heart disease, cancer, strokes accounts for 50% of deaths each year [7] • Population using more healthcare services, including: • Physician visits, hospital care, etc. [6] [6] Center on an Aging Society, Georgetown University, 2004 [7] Center for Disease Control (CDC) 5/28/2016 6 Context: Patient Experience • The use of prescription drugs is a necessary component in our healthcare industry in order to treat/control America’s rising chronically ill population • Patients are expected to make health related decisions during encounters with their physician (mean encounter duration is ≈ 20 minutes) [8] • Doctor assesses patient’s concern and gives recommended course of action • Patient must make decision in the face of: • Uncertainty in outcome of prescribed course of action • Complexity of insurance decisions (coverage) [8] Medscape. Laurie Barclay, MD, 1 Nov. 2009. Web 5/28/2016 7 Agenda • • • • • • Context Stakeholder Analysis Insurance Terminology Problem Statement Needs Statement Decision Support Tool • • • • • • • • • 5/28/2016 Task Analysis Video Description PUF Dataset Output Method of Analysis Monte Carlo Analysis Usability Test Recommendations/Future Work Project Management 8 Stakeholders: Cause & Effect Physicians Awareness of Rx Interactions Or Availability Prescriptions, Procedures, Behavioral Changes Awareness of Insurance Coverage Government Regulated Research & Development Finding Information Patients Don’t Pick Optimal Uncertainty of Prescribed Outcomes Assess and Prescribe Course of Action Lawsuits Patients Will Ins. Cover Next Action Complexity in Decision-Making Trouble Comparing Qualitative Data Runaway Healthcare Cost Negotiate Pricing w/ Insurance Co’s PBM Pharmaceutical Sales To Physicians PBM Negotiation Districting Plan Pricing Coverage Offering Availability Physician Accepted Large Attrition Rate Pharmaceutical Companies 5/28/2016 Insurance Companies 9 Agenda • • • • • • Context Stakeholder Analysis Insurance Terminology Problem Statement Needs Statement Decision Support Tool • • • • • • • • • 5/28/2016 Task Analysis Video Description PUF Dataset Output Method of Analysis Monte Carlo Analysis Usability Test Recommendations/Future Work Project Management 10 How Insurance Works • Insurance is analogous to a bet between you and the insurance company • They are betting that they will take in more money, in the form of premiums, than they will pay out in benefits • Policy is a contract that spells out what will be covered and how much the insurance will cover of your incurred medical costs 5/28/2016 11 Key Terminology • Prescription Drug Plan (PDP)-Insurance plan used to cover prescription drugs as part of Medicare Part D • Doughnut Hole-Point at which PDP stops offering coverage of prescriptions; then offering coverage again to patients in “catastrophic coverage” phase of plan • Premium-Amount paid each term (i.e. per month) to purchase insurance • Deductible-Amount initially incurred by patient before insurance policy begins paying for covered expenses • Cost Sharing-Co-payment amount-Set dollar amount for services rendered. Co-insurance %-covered split between patient and insurance, after deductible is met • Formulary-Listing of covered drugs [10] Centers for Medicare & Medicaid Services, 2012 5/28/2016 12 Medicare Terminology • Medicare-A national health program for those under age 65 with certain disabilities, above 65, and anyone with end stage renal disease (kidney disease requiring dialysis or a transplant) • Four parts to Medicare • Part A-Hospital insurance-covers inpatient care in hospitals, skilled nursing facilities, hospice, and home health care • Part B-Medical insurance-covers doctor services, outpatient care, durable medical equipment used in home healthcare • Part C-Advantage plans incorporating Parts A, B, and D (if elected); run by Medicare approved private insurance companies, and may include extra benefits for an extra cost • Part D-Prescription drug coverage-helps cover prescription costs and protect against higher costs associated with certain prescriptions, and run by Medicare approved private insurance companies [10] Centers for Medicare & Medicaid Services, 2012 5/28/2016 13 Medicaid Terminology • Medicaid-A state administered health insurance program available to certain low-income individuals and families who fit into a recognized eligibility group. • Specific requirements must be met (vary from state to state): -Age -Pregnancy Status -Disability -Blindness -Income -Resources -U.S. Citizen / Lawfully admitted immigrant • Dual Status-Those under age 65 who are eligible for Medicare because they receive Social Security or disability from the Railroad Retirement Board [11] Disabled Medicare Beneficiaries by Dual Eligible Status: California, 1996-2001 June F. O’Leary, Ph.D., Elizabeth M. Sloss, Ph.D., and Glenn Melnick, Ph.D. 5/28/2016 14 Medicare Part D Pros • Commonly Used Drugs • Generic or Brand Name • Mail Order (larger supply) • Out-of-pocket costs 5/28/2016 Cons • # of Insurance Providers • Formulary Listing • Drug Rates • Plan Changes • Out-of-pocket costs 15 Medicare Part D Out-of-Pocket Breakdown Medicare Part D Cost Responsibility Cost Responsibility Split (%) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Deductible Medicare Patient 0% 100% Cost Sharing Doughnut Hole 75% 0% 25% 100% Prescription Drug Plan Cost Stage Catastrophic Coverage 95% 5% [10] Derived using Centers for Medicare & Medicaid Services, 2012 5/28/2016 16 Medicare Part D Out-of-Pocket Breakdown Medicare Part D Cost Breakdown (Example Using $20,000 in Drug Cost) $25,000.00 Dollars ($) $20,000.00 $15,000.00 $10,000.00 $5,000.00 $Patient Total Drug Cost Cumulative Patient Outlay Deductible Cost Sharing $320.00 $320.00 $320.00 $652.50 $2,930.00 $972.50 Doughnut Hole $3,727.50 $6,657.50 $4,700.00 Catastrophic Coverage $667.13 $20,000.00 $5,367.13 Prescription Drug Plan Cost Stage [10] Derived using Centers for Medicare & Medicaid Services, 2012 5/28/2016 17 Agenda • • • • • • Context Stakeholder Analysis Insurance Terminology Problem Statement Needs Statement Decision Support Tool • • • • • • • • • 5/28/2016 Task Analysis Video Description PUF Dataset Output Method of Analysis Monte Carlo Analysis Usability Test Recommendations/Future Work Project Management 18 Problem Statement Patients are not able to make optimal health decisions regarding their physical AND financial well-being due to: • High level of complexity in the options with which they are faced • Insurance options and available coverage • High level of uncertainty in the prescribed course of action • Prescriptions/Procedures 5/28/2016 19 Need Statement A decision support tool (DST) is needed to consolidate PDP information • Draw consolidated information across insurance carriers • Minimize patient time spent on initial research • Answer basic questions about Medicare plans, products, and the companies entrusted to serve the general public 5/28/2016 20 Agenda • • • • • • Context Stakeholder Analysis Insurance Terminology Problem Statement Needs Statement Decision Support Tool • • • • • • • • • 5/28/2016 Task Analysis Video Description PUF Dataset Output Method of Analysis Monte Carlo Analysis Usability Test Recommendations/Future Work Project Management 21 Task Analysis: Preliminary 5/28/2016 22 Task Analysis: Decision Support Tool Physician Prescribed Course of Action Patient Condition Personal Attributes Decision Support Tool Insurance Plan Preferences Public Use Files 5/28/2016 Ranked List of PDPs Graphical Interpretation of Rx Cost vs. PDP cost Prescription Drug Plans 23 Decision Support Tool: Walkthrough 5/28/2016 24 Agenda • • • • • • Context Stakeholder Analysis Insurance Terminology Problem Statement Needs Statement Decision Support Tool • • • • • • • • • 5/28/2016 Task Analysis Video Description PUF Dataset Output Method of Analysis Monte Carlo Analysis Usability Test Recommendations/Future Work Project Management 25 Primary Data Source • Center for Medicare/Medicaid Services • Public Use Files (PUF) containing de-identified information for chronic diseases from year 2008 • Oracle 10g XE • Converted PUF data from .csv format into Oracle database • Manipulate and sift through data with SQL queries in order to obtain useful data • Lost roughly 25% of data due to lack of Medicare Part D data • End result: Medicare Part D data from 23.7 million beneficiaries 5/28/2016 26 PUF Data Breakdown • PUF data provides information on patient profiles consisting of • Age ranges (5 year cohorts) • Whether or not the patient has a given chronic illness • Diabetes, Heart Failure, Cancer, etc. • Whether or not a patient is dual eligible • Information on each profile includes • Prescription drug costs per year • Number of prescriptions per year 5/28/2016 27 Decision Support Tool: Sample Results 5/28/2016 28 Decision Support Tool: Sample Results 5/28/2016 29 Decision Support Tool: Sample Results 5/28/2016 30 Decision Support Tool: Sample Results 5/28/2016 31 Decision Support Tool: Sample Results 5/28/2016 32 Decision Support Tool: Sample Results 5/28/2016 33 “Doughnut Hole” and “Catastrophic” Coverage Doughnut Hole Catastrophic Range 5/28/2016 34 Agenda • • • • • • Context Stakeholder Analysis Insurance Terminology Problem Statement Needs Statement Decision Support Tool • • • • • • • • • 5/28/2016 Task Analysis Video Description PUF Dataset Output Method of Analysis Monte Carlo Analysis Usability Test Recommendations/Future Work Project Management 35 Weight Calculation • Mean Reciprocal Rank (MRR) is a statistical method to measure performance of the predicted results n 1 ∑Rank i MRR = i =1 # ofRanks • MRR used in tool for weight calculation where users rank their personal preferences of insurance attributes • Tier I (coverage attributes) • Plan Cost, Previous Experience, Formulary Choices • Tier II (plan cost breakdown) • Deductible, GAP Coverage, Cost Sharing, Premium Payment 5/28/2016 36 Utility Function • Insurance plans ranked in terms of Plan Cost, Formulary Choices, and Previous Experience 3 CoverageUtility = Σ [WCoverageAttribute •U CoverageAttribute] n =1 • Plan Cost is further broken into Monthly Premium, Deductible, Cost Sharing, and GAP Coverage 4 CostUtility = Σ [WCost Attribute •U Cost Attribute] n =1 • Utility function determines ranking of 30 Virginia based Medicare Part D Prescription Drug Plans 5/28/2016 37 Sensitivity Analysis • According to the customer’s preferences, each attribute was ranked and used to solve the weights • The following formula was used to solve the sensitivity analysis for weights of the ranked PDPs: o w j w j 1 wi o w k k i , j i • The weights were calculated for the attributes important to the user in order to rank the suitability of insurance plans 5/28/2016 38 Sensitivity Analysis • There were three combinations under which the sensitivity analysis was conducted • When the insurance/cost attribute ranking was equally distributed • When the insurance/cost attribute ranking was incrementally distributed • When two of the insurance/cost attribute rankings were tied 5/28/2016 39 Agenda • • • • • • Context Stakeholder Analysis Insurance Terminology Problem Statement Needs Statement Decision Support Tool • • • • • • • • • 5/28/2016 Task Analysis Video Description PUF Dataset Output Method of Analysis Monte Carlo Analysis Usability Test Recommendations/Future Work Project Management 40 Simulation • Crystal Ball was used to perform Monte Carlo simulation on the suggested insurance plans • Enters random values as inputs based on assigned distributions • Discrete Uniform Distributions • Random # generation for health profiles from PUF dataset • 10,000 trials Resulted in multiple plans remaining in top 10 5/28/2016 41 Simulation: Outcome • Sensitivity # 30 Ranked PDP-Envision Rx Plus Silver • Never makes the top 10 by reducing the price • Ranked # 10 if formulary increases from 2,446 to 6,395 • (2.6 times its original formulary listing) 5/28/2016 42 Simulation: Outcome Mitigation • Additional criterion defined by user; maximum monthly premium for insurance plan • Resulted in wider range of plan rankings 5/28/2016 43 Simulation: Estimated User Savings • 95% Confidence Interval: $77.66 to $1877.99 • Range: -$332.40 to $1978.10 • Estimated Average Savings: $1,243.00 w/ 50% saving more than $1,389.74 per year 5/28/2016 44 Agenda • • • • • • Context Stakeholder Analysis Insurance Terminology Problem Statement Needs Statement Decision Support Tool • • • • • • • • • 5/28/2016 Task Analysis Video Description PUF Dataset Output Method of Analysis Monte Carlo Analysis Usability Test Recommendations/Future Work Project Management 45 Usability Test: Human Subjects Review Board • The goal of the Human Subjects Review Board (HSRB) is to protect the rights of human research subjects [11] • Of the six exemptions to perform human studies without prior approval, item four qualified the following study to be done without approval of the HSRB [12] Office of Research Subject Protections-http://research.gmu.edu/ORSP/HumanSubjects.html 5/28/2016 46 Usability Test: Design • Usability test designed to evaluate • Ease of use of tool (intuitiveness) • Understandability of tool • Was administered to subjects with minimum age of 55 years old • Subjects were asked to • Participate in timed run through decision tool • Complete timed checklist consisting of 8 questions regarding their results • Find costs, interpret graphical output, etc. 5/28/2016 47 Usability Test: Hypotheses • Total time to use DST • Ho: μ = 20 minutes • HA: μ > 20 minutes • Evaluation of checklist • Ho: μ = 2 incorrect answers • HA: μ > 2 incorrect answers • Time to complete checklist • Ho: μ = 5 minutes • HA: μ > 5 minutes 5/28/2016 48 Usability Test: Results 25 Time to Complete Decision Tool n = 11 20 15 12.39 Time to Complete Results Checklist Expected Actual 10 8.43 Number of Errors in Checklist 5 1.09 0 1 5/28/2016 2 3 49 Usability Test: Results Total time to use decision support tool Ho: μ = 20 minutes HA: μ > 20 minutes T-test Results: to = -10.504 < t.05, 10 =1.812 DO NOT REJECT Ho Evaluation of checklist Ho: μ = 2 incorrect answers HA: μ > 2 incorrect answers to = -3.178 < t.05, 10 = 1.812 DO NOT REJECT Ho Time to complete checklist Ho: μ = 5 minutes HA: μ > 5 minutes to = 3.713 > t.05, 10 = 1.812 REJECT Ho Since we reject estimated time to complete the checklist, but not the amount of errors in results interpretation and time to complete the tool, we can conclude that the tool is easy to use, results are easy to interpret, but the results tab is too complex to interpret given a 5 minute time limit. • If the time to complete the checklist was increased to 6 min, 30 sec we would not reject the null hypothesis 5/28/2016 50 Agenda • • • • • • Context Stakeholder Analysis Insurance Terminology Problem Statement Needs Statement Decision Support Tool • • • • • • • • • 5/28/2016 Task Analysis Video Description PUF Dataset Output Method of Analysis Monte Carlo Analysis Usability Test Recommendations/Future Work Project Management 51 Recommendation • We recommend the designed DST be implemented in the waiting rooms of physician offices; given the ease of use, and relatively quick time to complete the tool • Secondly, we recommend the DST be implemented on insurance broker websites in order to give user’s an unbiased, no obligation estimate…allowing the user to compare all plans in their selected region; thus breaking the barrier to the insurance industry 5/28/2016 52 Future Work • • • • • • 5/28/2016 Implement prescription lookup system Medicare Part C (Advantage Plans) Special Needs Plans (SNPs) Expand beyond Virginia Expand data files to include health outcome data Dynamic health behavior for patients 53 Agenda • • • • • • Context Stakeholder Analysis Insurance Terminology Problem Statement Needs Statement Decision Support Tool • • • • • • • • • 5/28/2016 Task Analysis Video Description PUF Dataset Output Method of Analysis Monte Carlo Analysis Usability Test Recommendations/Future Work Project Management 54 Work Breakdown Structure (WBS) HCAS Project 1.0 Research 2.0 Design 4.0 Analysis 5/28/2016 3.0 Management 5.0 Communicate Results 55 Budget Breakdown • Budget calculated with a rate of $60 per hour • Hours determined by task length from Gantt chart and which group members will working on each task Task Total Hours Total Cost 1.0 Research 65 $3,900 2.0 Development/Design 622 $37,320 3.0 Management 115 $6,900 4.0 Analysis 242 $14,520 5.0 Communicate Results 296 $17,760 Total Project 1340 $80,400 5/28/2016 56 Earned Value and Hours Management Earned Value Tracking 1.20 % Complete 1.00 0.80 Earned Value 0.60 Estimated Earned Value 0.40 0.20 0.00 0 5 10 15 20 25 30 Time (Weeks) • Today • SPI = 1.05 • CPI = 1.52 5/28/2016 1600 1400 Hours Worked • Beginning of Semester • SPI = 0.67 • CPI = 1.21 Hours Management 1200 Planned Hours 1000 800 Actual Hours 600 400 200 0 0 5 10 15 20 Time (Weeks) 25 30 57 Gantt Chart • Start of Fall semester through Briefing # 2 (October, 24) 5/28/2016 58 Gantt Chart • Briefing #2 (October, 24) through Faculty Presentation (December, 2) 5/28/2016 59 Gantt Chart • Spring Semester through SIEDS Competition (April, 27) 5/28/2016 60 Questions? 5/28/2016 61 References [1] Centers for Medicare & Medicaid Services-www.cms.gov [2] www.healthcare.gov/law [3] Congressional Budget Office-Technological Change and the Growth of Health Care Spending [4] www.globalsecurity.org-Russian Military Budget [5] Rowberg, Richard. CRS Report for Congress: Pharmaceutical Research and Development: A Description and Analysis of the Process, 2001 [6] Center on an Aging Society, Georgetown University, 2004 [7] Center for Disease Control (CDC) [8] Medscape. Laurie Barclay, MD, 1 Nov. 2009. Web [9] Testimony of David A. Balto To the Committee on the Judiciary, Subcommittee on Antitrust, Competition Policy and Consumer Rights United States Senate, Dec. 2011 [10] Centers for Medicare & Medicaid Services, 2012 [11] Disabled Medicare Beneficiaries by Dual Eligible Status: California, 1996-2001 June F. O’Leary, Ph.D., Elizabeth M. Sloss, Ph.D., and Glenn Melnick, Ph.D. [12] Office of Research Subject Protectionshttp://research.gmu.edu/ORSP/HumanSubjects.html 5/28/2016 62 BACKUP SLIDES 5/28/2016 63 Context: Collapse & Increased Spending The Soviet Union’s collapse due to military spending occurred at a point where their military expenditure reached $33 billion/year [15-17% of its GDP], but western authorities believed this only reflected the operations and maintenance costs, and that the expenditure was actually double the amount reported [due to weapon/ship building being coded as “free goods to the Ministry of Defense”]. [4] [3] Congressional Budget Office-Technological Change and the Growth of Health Care Spending [4] www.globalsecurity.org-Russian Military Budget 5/28/2016 64 Context: Rx Costs • Correlating the two graphs show that almost 70% of cost is reached by year 12 in R&D process • ~34% cost due to clinical evaluations • Rx attrition rate during development • 5000 drugs screened • ONLY 5 enter testing • No income from this stage [5] Rowberg, Richard. CRS Report for Congress: Pharmaceutical Research and Development: A Description and Analysis of the Process, 2001 5/28/2016 65 Context-U.S. Percent Excess in Prescription Pricing [8] Socolar, D., Sager, A. Boston University School of Public Health. Calculations from Patented Medicine Prices Review Board annual reports (2000 & 2002). 5/28/2016 66 Insurance Terminology • • • • • • • • Individual Coverage-Most expensive; offered to people who don’t have employer coverage Group Coverage-More cost effective due to the larger number of people in the group. Typically seen in employer context Premium-Amount paid each term (i.e. per month) to purchase insurance Provider-Insurance company providing coverage in return for premium payment Co-Pay-Fee associated with services rendered Deductible-Amount initially incurred by patient before insurance policy begins paying for covered expenses Co-Insurance-Amount of covered split between patient and insurance, after deductible is met Reasonable & Customary Charges-Amount that insurance is willing to pay for a procedure performed [10] Centers for Medicare & Medicaid Services, 2012 5/28/2016 67 Medicare Part D • Stand Alone Plan: • Prescription Drug Plan (PDP)-additional plan to cover prescription drugs • Packaged Plans (Medicare Approved/Privately Insured Plans): • Advantage Plans (Medicare Part C)-Packaged with Part A/B, and offer further benefits for extra cost • Special Needs Plans (SNP)-Packaged with Part A/B, and cover chronic conditions specifically • Has emphasis on the needs of a chronic condition [10] Centers for Medicare & Medicaid Services, 2012 5/28/2016 68 Stakeholder Diagram General Public Voting Patients use Medicare, Made Available by Gov’t Patients Government Regulators FDA Patients Seek Information Alone Regulations on Drugs Medicare Doctors FDA Accepts or Rejects Drugs Contracts w/ Private Ins. Co’s Private Insurance Companies Special Needs Plan Pharmaceutical Companies R&D Approach Insurance & Doctor’s for Pricing Structures Prescription Drug Plan PBM [9] Adapted from Deloitte, Life Sciences Division 5/28/2016 69 Stakeholders: Pharmacy Benefit Managers (PBMs) • The PBM market has an expensive history of lawsuits brought on by unethical conduct • Between 2004-2008, a coalition of 30 state Attorneys General have brought cases against the “big three” PBMs securing over $370 million in damages and penalties 5/28/2016 • Fraud • Misrepresentation to plan sponsors, patients, and providers • Unjust enrichment through secret kickback schemes • Failures to meet ethical and safety standards Case Damages Reason(s) for United States v. Merck & Co., Inc. $184.1 million Government fraud, secret rebates, drug switching, and failure to meet state quality of care standards. United States v. AdvancePCS (now part of CVS/Caremark) $137.5 million Kickbacks, submission of false claims, and other rebate issues United States v. Caremark, Inc. Pending Submission of reverse false claims to governmentfunded programs. State Attorneys General v. Caremark, Inc. $41 million Deceptive trade practices, drug switching, and repacking. State Attorneys General v. Express Scripts $9.5 million Drug switching and illegally retaining rebates and spread profits and discounts from plans. 70