Cost-Effectiveness Analysis in Health Care More on Value

Cost-Effectiveness Analysis in Health Care More on Value of Life and CBA Use of CBA in Health Problems with Willing-to-Pay for lives VSL vs. VSLY Is Modern medicine worth its cost? Slides on Other VSL topics RAND 1 © 2006 Emmett Keeler

Key Points • • We use WTP to get the Value of a statistical life (VSL), but there are problems in doing so. å Contingent valuation can help. “Lives saved” and QALYs can lead to different decisions How should VSL vary with age, wealth? Cutler: the health gains from technical progress makes higher medical costs worthwhile. RAND 2 © 2006 Emmett Keeler

Summary Table of Costs and Effects Measures of Effects Options Direct Cost, millions Indirect Cost Years of Restricted Satis. Life Exp. Activity faction Days A 4 100 B 1 50 C 10 200 1000 RAND 3 © 2006 Emmett Keeler

Cost Effectiveness and Cost Benefit • Cost-Effectiveness Analysis (CEA) å å • å Focuses on a single effect • e. g. number of kids vaccinated or aggregates all health effects into QALYs gives efficient way to allocate fixed budget Cost-Benefit Analysis (CBA) å å Aggregates all effects into dollars • even living longer or better answers “ Is a program worth it” RAND 4 © 2006 Emmett Keeler

Willingness to Pay for a statistical life (VSL) • • So CBA needs a value for health or life-saving gains. å “lost wages” was the historic measure, but WTP is now the economist’s favorite approach å comes from observed behavior, or surveys VSL is an estimate of an individual’s willingness to pay to avoid a small risk of death, scaled up to 1 death. VSL = WTP for risk reduction/(size of risk reduction) RAND 5 © 2006 Emmett Keeler

Example of WTP for safety • • • So suppose airbags reduce chances of dying in an car accident over the life of a car from 1/5, 000 to 1/10, 000 =. 0002 -. 0001 =. 0001 Air bags save the life of 1 driver per 10, 000 cars 10, 000 car buyers each paying $300 for an air bag is like paying $300 x 10, 000 = $3 million for each life saved å å if laws reflect values, then those buyers value life >= $3 million RAND 6 © 2006 Emmett Keeler

More Sophisticated Estimates • • • Use change in income needed to tolerate a change in risk as a measure of WTP Use data on job risks and wages to estimate the compensation wage necessary to induce someone to take a risky job. Regress wages on risk, other characteristics å å but personal risk tolerance is unobserved and effects job choice search for papers, books by Kip Viscusi on this. RAND 7 © 2006 Emmett Keeler

Summary: WTP Values for a statistical life • $/life comes from scaling up $/fraction of life. å • • å appropriate because most life-saving programs change risks only slightly. allows us to compare values from different risks Surveys, wage premia, prices of safer products Range VSL = $3 -15 million /life 5 -10 times larger than lost wages å EPA now uses $6 million. Reference. Hirth et al. Med Dec Making, 2000 å RAND 8 © 2006 Emmett Keeler

Problems with Willingness-to-Pay • People find it hard to guess å • å List paradox: What are you WTP for cutting your chance of getting breast cancer in half? å • what to do with bad answers? caused by confusion, gaming, distaste å of US age-specific incidence of breast cancer in half? of all US cancers in half? Each charity wants to be first in line. RAND 9 © 2006 Emmett Keeler

Contingent Valuation • • A method to get better WTP answers Forced choice: are you WTP $X for better school? å • • • å vary X randomly from 5 -100 in large survey. fit logistic regression line, find 50% value Say they must pay in taxes Mention substitutes, the full list Use telephone or face to reduce bad answers by help from interviewer References: see Hammitt handout RAND 10 © 2006 Emmett Keeler

Logistic regression of NOs ---> WTP Would not pay 1. 0 50% X $ for program RAND 11 © 2006 Emmett Keeler

Problems with Willingness-to-Pay -2 • • Willingness-to-Pay (WTP) higher for rich å Could value outcome by % of income Also high for old due to å “dead anyway” effect ( Pratt & Zeckhauser, 1996) • Money has fewer other good uses RAND 12 © 2006 Emmett Keeler

WTP and utility: Bill Gates’ headache utility = f(health, w) wtp is not aligned with utility, because MV $ lower for rich Good health ∆U A ∆U B b headache paraplegia a wtp wealth RAND 13 © 2006 Emmett Keeler

Measuring the value of life year (VSLY) • • • To base decisions on LE gains using BCA, we need to estimate VSLY. Harder to think of hedonic measures or realistic scenarios for VSLY than for VSL. In practice, VSLY = VSL / E(discounted QALE) å for example, all 42 studies reviewed in Hirth et al. “Willingness to pay for a QALY” Med. Dec. Making 2000 had estimated a VSL from which Hirth computed a VSLY RAND 14 © 2006 Emmett Keeler

Using Lives vs QALYs in CBA • • Historically EPA, NHTSA and other government agency have used CBA and “lives” valued at $6 million in fine-tuning regulations. Compared to using QALYs gained, “lives” gives more weight to people who: å • å won’t live long are already sick So reducing air pollution, which primarily kills people with COPD, is weighted heavily using lives compared to reducing environmental lead which affects kids. RAND 15 © 2006 Emmett Keeler

Young old lifesaving tradeoffs age prevent. deaths life years QALYs disc at 7% 5 1 73 65 13 35 1 44 37 12 75 1 12 9 6. 1 Ratio 5/75 1 6. 1 7. 1 2. 1 RAND 16 © 2006 Emmett Keeler

Cutler: Is Technology Worth it? • • Even if medical spending is rising, social welfare is enhanced if the benefits exceed the costs Cutler paper estimates the Ro. R from 4 clinical areas that have experienced rapid technological change Estimating medical spending is easy å See Levitt et. al. 2004 Benefits are harder to estimate Your money or your life, Oxford UP 2004 has more RAND 17 © 2006 Emmett Keeler

Outline of Cutler’s Approach For each disease area, • Measure Years of life added in QALYs • Assume value of a current QALY = VSL/LE = $4 million/ 40 years = $100, 000* • Compute all discounted costs and benefits å 3% discount rate * How does this VSLY compare to a value of a discounted QALY in theory? RAND 18 © 2006 Emmett Keeler

Advances in Treatment of CVD • • Roosevelt dies of hypertension å Eisenhower heart attack å • Largely untreatable in 1945 å Bed rest recommended No treatment except anticoagulant Mortality from CVD has fallen by ½ since 1950 RAND 19 © 2006 Emmett Keeler

Case 1: Heart Attacks • • • Thrombolytic drugs to dissolve clots Coronary Artery Bypass Graph (CABG) Angioplasty Use of stents to open blockages Drugs for CHD (e. g. ACE inhibitors) Statins to lower cholesterol (e. g. Lipitor) RAND 20 © 2006 Emmett Keeler

Calculations for CVD • • Average 45 year old can expect to live 4. 5 years longer relative to 1950 å 2/3 rds attributable to medical treatment (2. 8 years) å PV of these three years is $108, 000 Cost of keeping folks alive in non-working years is $10, 000 more than in 1950 PV of costs for CVD medical care is $30, 000 (zero costs in ’ 50) Net return = 260% å Or $3. 60 gained for every $1 invested RAND 21 © 2006 Emmett Keeler

Value of LBW Medical Care • Low birthweight å å • • å Since 1950, LBW babies live 15 more years About. 04 of all infants are LBW So 0. 6 years saved per infant (. 04*15) NICU, CVD Tx have increased lives by 3. 4 years Do the results generalize, i. e. å å Is it true (generally) that the benefits of medical spending have been worth it? Cutler et al. NEJM 2006 says yes for young people, maybe not for those over 65. RAND 22 © 2006 Emmett Keeler

Summary: Is technical progress worth its cost? • David Cutler looked at changes in life expectancy in the US since 1960. å å • • å multiplied by a value of QALY or life year Compared them to the estimated change in medical costs. On average, benefits are bigger than costs. Others have done similar analyses for new drugs. Still, US could do much better at the margin. å I. e. there is a lot of waste in the system RAND 23 © 2006 Emmett Keeler

Supplementary slides RAND 24 © 2006 Emmett Keeler

Setting Treatment cutoffs Let A(x) = E(b(c) | c > x) Discounted Years saved by treatment b(x) Let Strategy 1 be treat one more Strategy II be screen more, treat if above x Let (x) be cdf of X = P(c < x) contains (x) X contains 1 - (x) of kids

Nordhaus’: fun vs. health and VSLY • Suppose you’re offered a choice between å å • • All the consumption gains since 1950 (better cars, computers, TV’s, veggies etc…) OR All the health improvements (9 years of life and better health while alive). Difficult choice based on informal surveys Nordhaus estimates it’s $750, 000/9 years å Just under $100, 000 per year RAND 26 © 2006 Emmett Keeler

Reconciling lifetime wages with willingness to pay for safety. • • • Young adults will work 10 -20% of their future living hours So if they value leisure hours like work hours, Value of life = 5 -10 times future lifetime wages Keeler, The value of remaining lifetime is close to estimated values of life, J Health Econ, (20) 141 -3 (2001) RAND 27 © 2006 Emmett Keeler

Discounting as decreasing returns? 0 . 01 . 03 . 05 . 07 How many years for sure are equivalent to a SG with a 50% chance of living 50 years + a 50% chance of living 1 week?

WTP for a QALY studies • • Researchers have patients rate their current HRQL, and give their WTP for a lifetime cure This gives low values of WTP/QALY, e. g. å å å $12, 500 - $32, 000/QALY in VA patients facing possible neurosurgery King et al MDM 2005 average HRQL ~. 7, WTP ~ $100, 000 • use of average lifetime, income effects? 5 other such studies gave similarly low values. RAND 29 © 2006 Emmett Keeler

Estimating waste using CBA • Waste is defined as resources spent unnecessarily in US health care system å å å Inefficient production of given services In producing the wrong things • whose value < the cost of producing them. • Artificial heart Preventable asthma hospitalizations? We just finished a paper on this. RAND 30 © 2006 Emmett Keeler