RICHARD ZECKHAUSER

PRESENTATION TO TSINGHUA UNIVERSITY

NOVEMBER 11 , 2010

Behavioral Decision and Movements Across Treatments

and Health Plans

Sources of Material

"Custom-Made Versus Ready-to-Wear Treatments: Behavioral Propensities in Physicians' Choices," Richard G. Frank and Richard J. Zeckhauser, Journal of Health Economics 26(6), 2007, 1101-1127.

  "Health Insurance Exchanges: Making the Markets Work,"

Richard G. Frank and Richard J. Zeckhauser, New England Journal of Medicine, 361(12), 2009, 1136-1137.

  "Selection Stories: Understanding Movement Across Health

Plans," David Culter, Bryan Lincoln, and Richard Zeckhauser, Journal of Health Economics, forthcoming.

“Blindness to the Benefits of Ambiguity: The Neglect of Learning Opportunities,” Stefan T. Trautmann and Richard J. Zeckhauser, October 2010.

Central Points

Individuals fall prey to behavioral decision Also true for highly important decisions True for professionals and individuals Health insurance

Also true for highly important decisions

Choice of medical treatment Investment allocation Health plan Whom to marry

True for professionals and individuals

Behavioral finance – 2008 financial meltdown as evidence

Doctors and choice of treatments – will review today

Health insurance

Adverse selection Inertia Adverse retention

Anchoring

RBRRBRBRRBR 7R 4B

Reddish or Blackish Bag Problem1/2 1/2

Winner’s Curse

 

Everyone bids for the company. High bid wins. Your best estimate of value is 200 million yuan. How much should

you bid? If others further shade down their bids, how does that influence

your optimal bid?

BIDDING FOR A COMPANY

Group Decision Processes

Group processes exacerbate behavioral propensities. o Agreement on why your candidate will win.

o Reinforcing beliefs on why the business deal makes sense.

o Herding suppresses information

RED, GREEN, GREEN, GREEN ….  Implication: Encourage alternative models and

contrary evidence.

Decision Biases

Loss Aversiono People reset their reference points (Gulf after BP spill; Chilean

mine rescue)o Small losses count much more than small gainso Equity premium relative to bonds?o Prescriptions: Dribble good news; Bad news all at once; Reset

your total portfolio value reference point Status Quo Bias

o Inheritance from Uncle Joe• Moderate-risk company A• High-risk company B• Treasury bills• Municipal bonds

2% OF Harvard faculty change health plan or retirement plan in a year

Overconfidence

Individuals systematically are more confident of their estimates than they should be.Population of Hainan ProvinceArea of Fujian ProvinceTotal fertility rate of Beijing(number per woman if has current age-specific fertility over her lifetime)

Total fertility rate of China (2007)

* China Statistical Yearbook, 2009** China 2000 Census

8,549,000*

120,000 sq. km.*

0.67**

1.75*

Option Blindness: Two Draws

Which jar would you choose?

Guess a color. If correct win $10.Replace the marble. Guess a color. If correct win $10.

Cholesterol Treatment

Your doctor has discovered that you have a high cholesterol level, namely 240. She prescribes one of many available statins. She says this will generally drop your cholesterol about 25%. There may be side effects.

Two months later you return to your doctor. Your cholesterol is now at 195. Your only negative side effect is sweaty palms, which you experience once or twice a week for half an hour. Your doctor asks whether you can live with this side effect. You say yes. She tells you to continue on the medicine.

What do you say?

Custom-Made Versus Ready-to-Wear Treatments

Prostate cancer – 7 or 8 primary treatments o "Custom-Made Versus Ready-to-Wear Treatments: Behavioral

Propensities in Physicians' Choices," Richard G. Frank and Richard J. Zeckhauser, Journal of Health Economics 26(6), 2007, 1101-1127.

 o "Health Insurance Exchanges: Making the Markets Work," Richard G.

Frank and Richard J. Zeckhauser, New England Journal of Medicine, 361(12), 2009, 1136-1137.

Antidepressants – o multiple classes, different biochemical processeso each class might have half dozen different drugs

Individuals use norms in making decisions. What about professionals?

Costs of Customization

Communication costs Prostate cancer “…there is potentially significant benefit to be gained through using an impartial, evidence-based decision model that explicitly accounts for the preferences of each individual patient.”

Cognition costsSatisficing and use of heuristics

Coordination costsOur green-red example. Worse if there is cost to reveal.

Adam Smith’s pin factory counterfactual – sequential customers Stronger pins for pants

Longer ones for flannelSmoother ones for silkStill might standardize, particularly if also produce staples and paper clips

Medicine has three key features: specialization and division of labor; varied product mix; customers with different needs served in succession, not batches

Capability costs – know some therapies better than othersHeart attack treatments vary strongly across areas

Chemotherapy Treatment

Standard protocol o Fixed interval, say monthlyo Dosage tradeoff between benefit and side

effectso White blood count closely monitoredo If low, delay treatmento Dosage not changed VERY ODD

Three Levels of Rationality

Hyper rationality – Doctor optimizes for each patient.

Decision cost rationality – Doctor recognizes costs of customization. Take reasonable decisions in light of the costs they face.

Heuristic behavior – Doctor simply employs ready-to-wear treatments.

Use of Norms Hypotheses

Sensible Use of Norms (SUN) Hypothesis – Doctors use norms when they make the most sense. Thus, would customize for chronic conditions more than acute conditions.

Therapeutic Norms Hypothesis – Doctors select treatments for a representative patient in each category.

My Way Hypothesis – For many important conditions, doctors will regularly prescribe a treatment quite different than the choice of other physicians. Thus, the choice might depend on past luck, which drug encountered first (detail men).

Multi-Armed Bandit Problem

Benefits depend on:o Drug, patient, time

Optimize given prior distributions. Complex decision under uncertainty, with

substantial learning. Well studied problem in applied mathematics.

Lottery with highest Gittins index played at each point.

Implications of Use of Norms

1. Prescribing concentration by physician greater for acute conditions than chronic conditions. (Supports SUN Hypothesis)

2. Prescribing concentration depends on physician. Is high relative to share for top-selling drug. (Rejects SUN Hypothesis; Supports Therapeutic Norms and My Way Hypotheses)

3. Length of visit little affected by complexity of decisions. Strongly affected by physician-specific factors. (Rejects SUN Hypothesis; Supports Therapeutic Norms and or My Way Hypotheses)

4. Multiple drug choice. Many drugs available for same condition, and dosages vary. Alternative findings:

A. Drug switching and dosage driven by patient response to treatment. (Supports SUN Hypothesis)

B. Drug switching and dosage driven by demographic, immediate clinical, or physician-specific factors. Patient response plays little role. (Supports Therapeutic Norms and/or My Way Hypotheses)

DATA

National Ambulatory Care Survey (2004) – 25,000 visitso Reason for visit, diagnosis, medication prescribed, tests,

referrals, duration, demographics of patient and physician, insurance, type of practice, specialty, location.

USE TO TEST FOR TREATMENT OF CHRONIC VERSUS ACUTE CONDITIONS, AND VISIT TIMES

Quality Improvement for Depression Study – four randomized effectiveness trialso Individuals diagnosed with major depression. Detailed

clinical, treatment and demographic data. Collected four times over two years. Scores on depression scale 0-100.

Full responders had a reduction in score over 50%; partial responders reduction 25-50%; non-responders less than 25%.

Concentration of Prescriptions

1. Mean concentration for most used drug by physician for condition is greater than 60%.

2. Greater than market share of any drug for any of the conditions.

3. Concentration for chronic conditions is 13% lower.

First two findings support Therapeutic Norms and My Way Hypotheses.

Third finding gives some support to SUN Hypothesis.

Visit Times

17 minutes is mean visit time.New patients get more time.Upper respiratory problems get 2.2 minutes

more.No other characteristics of diagnosis affects

visit length by as much as 2 minutes.Complexity does not matter. Strongly reject SUN Hypothesis.Support Therapeutic Norms Hypothesis.

Multiple Drug Choice and Dosage

Changes in medications explained by schooling, age and ethnicity.

Changes in medications not related to clinical indicators.   Strongly rejects SUN Hypothesis. Supports Therapeutic Norms and/or My Way Hypotheses.  Dosage increase related to schooling and age. Dosage increase unrelated to response to treatment or level

of symptoms.   Strongly rejects SUN Hypothesis. Supports Therapeutic Norms and/or My Way Hypotheses.

Conclusions

Physicians rely on norms foro Selecting prescriptionso Lengths of visito Switching medication

They tend to use ready-to-wear rather than customized treatments.

Significant evidence of My Way behavior.

Movement Across Health Plans

Economists focus incessantly on adverse selection. Implication is that individuals closely monitor costs and

benefits of alternative health plans, and move amongst them.

Initial study on Status Quo Bias* examined choice of health plans by Harvard employees. 3% switch plans in a year Compare old and new employees

Harvard had to cancel its high-cost fee-for-service plan. Premium differential was $1,300/year, and deductible difference was $400/person/year. In a family of four, all had to go to hospital to be better off.* William Samuelson and Richard Zeckhauser, "Status Quo Bias in Decision Making," Journal of Risk and Uncertainty 1, March 1988, 7-59.

Table 9. Health Plan Choices 1986 by Age Group and Enrollment Year

Plan 1985-1986

Ages 21-31Year of

Enrollment1983-1984

Before 1983

BCBS 6.4 8.8 27.4 (.001)HUGHP 50.2 50.4 29.4 (.001)HCHP 22.2 22.3 25.0MGHP 3.7 2.8 2.7Bay St 6.4 9.6 3.4 (.01)Tufts 3.4 1.4 3.0Lahey 2.1 .8 6.1 (.001)BC Low 4.1 4.0 3.0TotalNumber

100.01304

100.0649

100.0296

From: William Samuelson and Richard Zeckhauser, "Status Quo Bias in Decision Making," Journal of Risk and Uncertainty 1, March 1988, 7-59.

Premiums Across Plans

THE LAW OF ONE PRICE. In an efficient market, identical goods must have the same price.

Medigap insurance covers coinsurance payments for Medicare (plan for elderly). The coverage is identical, and only relates to financial payment.

The LAW OF ONE PRICE indicates prices across providers will be very close:1. Expensive item.2. Easy to search.3. Little or no quality differential.

Actual Premiums Across Plans

Table 5: Range of Annual Premium Prices for Selected Plans for a 65-Year-Old in Five States with Large Medicare Populations

Plan A Plan C Plan F Plan J

Illinois $467 to $1,202 $802 to $1,633 $854 to $1,861 $2,247 to $3,502

New York $864 to $1,560 $1,408 to $2,385 $1,617 to $2,800 n/aa

Ohio $612 to $1,284 $924 to $2,064 $996 to $1,944 $2,028 to $3,156

Pennsylvania $500 to $1,373 $761 to $1,964 $802 to $1,649 $2,312 to $2,976

Texas $300 to $1,683 $664 to $2,125 $880 to $2,171 $2,059 to $5,658

a Only one insurer reported offering plan J in New York’s consumer guide, with a premium of $3,552. Source: State consumer guides prepared by state insurance departments for premiums typically offered in 2000 or 2001.

Conclusions

Behavioral considerations strongly influence choice behavior.

Even well-trained Chinese economics professors suffer from behavioral decision.

Health-care involves critical decisions. Doctors show tremendous inertia in their choices. Contrary to economic models, patients do not respond

strongly to premium differences. QUESTION: How should we change our thinking

about economics given such findings?