Transcript Decision Analysis - Northwestern University

Decision Analysis: Aspects of
Medical Decision Making
Gordon Hazen
Northwestern University
Contributions

Decision analysis has contributed to
decision-making in business,
medicine,
medicine, engineering, and law
Stanford University October 2011
2
My Input to Today’s Discussion



The field of medical decision making
Relationships between medical
decision analysis and the broader
DA field
A nonscientific report based on my
impressions only …
Stanford University October 2011
3
The Field of Medical Decision Making



Modeling to aid diagnostic and therapeutic
choice by physicians and treatment choice
by patients
Cost-effectiveness modeling to inform
public policy
Decision psychology: Understanding/
predicting patient and physician choices
Stanford University October 2011
4
Relationships
Statistics
Health Economics
Cost
-Effectiveness
Analysis
Medical Decision
Analysis
Decision
Psychology
Decision
Analysis
Stanford University October 2011
5
Is methodological preparation enough?

No: Medical DA researchers are almost always attached
to institutions serving real-world stakeholders














Bloomberg School of Public Health
Center for Disease Control and Prevention
Centre for Evaluation of Medicines (CEM),
Cleveland Clinic
Dept. of Social Medicine
Erasmus University Medical School
Center for Health Policy, Palo Alto, CA
Institute for Health, Health Care Policy and Aging
Research
Mayo Clinic
MD Anderson Cancer Center
Merck
Minneapolis VA Ctr
Portland VA Medical Ctr
VA San Diego Healthcare System
Stanford University October 2011
6
Journals

Publication of clinical decision analyses in
medical journals is widespread



Frequently involve decision trees dovetailing into
Markov models
PubMed search for 2010 with abstracts containing
“Markov”: 1426 papers!
PubMed search for 2010 with abstracts containing
“Markov” and titles containing “Effectiveness”:
200 papers.
Stanford University October 2011
7
Journals

Journals with methodological focus






Medical Decision Making
Value in Health
International J. of Technology Assessment in
Health Care
Health Economics
J. of Health Economics
Medical Care
Stanford University October 2011
8
Examples of research (last three issues
of Medical Decision Making)

Decision Psychology







Dynamics of Trust in Medical Decision Making: An Experimental
Investigation into Underlying Processes
The 1-in-X Effect on the Subjective Assessment of Medical Probabilities
The Decision Making Control Instrument to Assess Voluntary Consent
The Influence of Narrative v. Statistical Information on Perceiving
Vaccination Risks
Information for Decision Making by Patients With Early-Stage Prostate
Cancer: A Comparison Across 9 Countries
Impact on Decisions to Start or Continue Medicines of Providing
Information to Patients about Possible Benefits and/or Harms: A
Systematic Review and Meta-Analysis
Preferences and Utilities



Estimation of a Preference-Based Carer Experience Scale
Eliciting Benefit–Risk Preferences and Probability-Weighted Utility
Using Choice-Format Conjoint Analysis
Predicting EQ-5D Utility Scores from the Seattle Angina Questionnaire
in Coronary Artery Disease: A Mapping Algorithm Using a Bayesian
Framework
Stanford University October 2011
9
Examples of research (last three issues
of Medical Decision Making)

Provider Decision Making





Shared Decision Making



How Long and How Well: Oncologists’ Attitudes Toward the Relative Value of LifeProlonging v. Quality of Life-Enhancing Treatments
Deceiving Numbers: Survival Rates and Their Impact on Doctors’ Risk
Communication
Electronic Notifications about Drug Substitutes Can Change Physician Prescription
Habits: A Cross-Sectional Observational Study
Overestimation Error and Unnecessary Antibiotic Prescriptions for Acute Cystitis in
Adult Women
Longitudinal Changes in Patient Distress following Interactive Decision Aid Use
among BRCA1/2 Carriers: A Randomized Trial
Are There Racial Differences in Patients’ Shared Decision-Making Preferences and
Behaviors among Patients with Diabetes?
Risk Communication



Informing Patients: The Influence of Numeracy, Framing, and Format of Side
Effect Information on Risk Perceptions
Influence of Graphic Format on Comprehension of Risk Information among
American Indians
Graph Literacy: A Cross-Cultural Comparison
Stanford University October 2011
10
Examples of research (last three issues
of Medical Decision Making)

Exploring model structure







A Systematic Comparison of Microsimulation Models of Colorectal
Cancer: The Role of Assumptions about Adenoma Progression
Clarifying Differences in Natural History between Models of Screening:
The Case of Colorectal Cancer
How Does Early Detection by Screening Affect Disease Progression?:
Modeling Estimated Benefits in Prostate Cancer Screening
Simulation of Quality-Adjusted Survival in Chronic Diseases: An
Application in Type 2 Diabetes
Bayesian Inference for Comorbid Disease Risks Using Marginal Disease
Risks and Correlation Information From a Separate Source
Can Life Expectancy and QALYs Be Improved by a Framework for
Deciding Whether to Apply Clinical Guidelines to Patients With Severe
Comorbid Disease?
Integrating Health Economics Into the Product Development Cycle: A
Case Study of Absorbable Pins for Treating Hallux Valgus
Stanford University October 2011
11
Examples of research (last three issues
of Medical Decision Making)

Calibrating models



Estimating the Unknown Parameters of the Natural History of
Metachronous Colorectal Cancer Using Discrete-Event Simulation
Bayesian Calibration of a Natural History Model with Application to a
Population Model for Colorectal Cancer
Representing Uncertainty in Models




A Concise Equation That Captures the Essential Elements of One-Way
Sensitivity Analyses in Health Economic Models
The Combined Analysis of Uncertainty and Patient Heterogeneity in
Medical Decision Models
A Framework for Addressing Structural Uncertainty in Decision Models
Accounting for Methodological, Structural, and Parameter Uncertainty
in Decision-Analytic Models: A Practical Guide
Stanford University October 2011
12
Activity: SMDM versus INFORMS DA
Cluster

SMDM 2006




314 Abstracts
67 DA or C/E
applications
13 DA or C/E
methodology
127 utility/
preference/ dec’n
psychology

INFORMS DA cluster
2006

93 Abstracts
26 DA applications

72 DA methodology


8 utility/ preference/
dec’n psychology
SMDM = Society for Medical Decision Making
C/E = Cost-effectiveness
Stanford University October 2011
13
Activity (Based on SMDM 2006
participation)

US











Harvard School of
Public Health
Stanford University
Tufts-New England
Medical Center
Boston University
Centers for Disease
Control and Prevention
Case Western Reserve
University School of
Medicine
University of Pittsburgh
University of Chicago
Dartmouth Medical
School
Duke University
UK/Europe





University of
Birmingham
University of York
University of Sheffield
University Medical
Centre Utrecht
Canada






Stanford University October 2011
University of Toronto
McMaster University
University of British
Columbia
Univ of Western
Ontario
University of Ottawa
Dalhousie University
14
Academic Preparation
SMDM Associate Editors: PhD areas



Psychology (4)
Epi-Biostat (2)
Economics (6)




Health Policy (2)
Health Econ (2)
Health Technology Assessment (2)
MD (4)
Stanford University October 2011
15
Academic Preparation: What do these
areas have in common?
Decision
Psychology
Applied
Math/Stat/OR
Values and decisions
under uncertainty
Economics
Stanford University October 2011
Health
16
Impacts of the DA field on medical
decision making

Decision psychology/ judgment and
decision making


Thriving. Active research contributions.
Prescriptive decision analysis



Basics understood and accepted
Nuances not broadly understood
Sophisticated tools not used in practice
Stanford University October 2011
17
Impacts of prescriptive DA on the
medical field

Basics understood, accepted and used


Probability, utility
Decision trees






TreeAge software quite popular
Markov chains
Sensitivity analysis
Information value
Probabilistic sensitivity analysis
Bayesian probability and statistics
Stanford University October 2011
18
Impacts of prescriptive DA on the
medical field

Nuances are not broadly appreciated


The distinction between a utility function and a
value function
Tacit belief that utility applies only to health
states as opposed to any outcome of interest



QALYs = 𝑥(𝑡𝑖𝑚𝑒 𝑖𝑛 ℎ𝑒𝑎𝑙𝑡ℎ 𝑠𝑡𝑎𝑡𝑒 𝑥) ∙ 𝑢(𝑥)
“Cost-utility analysis”
Models of joint health and consumption not used



Howard 1984
Smith & Keeney 2005
Lichtendahl & Bodily 2009
Stanford University October 2011
19
Impacts of prescriptive DA on the
medical field

Nuances are not broadly appreciated

The connection between preference assumptions
on the one hand, and expected utility/ expected
utility decomposition on the other.



Substitution/Independence (von Neuman &
Morganstern)
Utility independence
The “preference assumptions  utility
decomposition” game is not understood or
played. Exceptions:



Pliskin, Shepard, Weinstein
Wakker
Miyamoto
Stanford University October 2011
20
Impacts of prescriptive DA on the
medical field

Sophisticated tools not employed


Influence diagrams
Utility decompositions beyond additive






One key exception: HUI (Feeney, Torrence, Furlong)
Copulas for constructing joint probability
distributions
Scoring rules
Risk tolerance
Stochastic dominance
Combining expert judgments
Stanford University October 2011
21
Impacts of prescriptive DA: Current
research on QALYs


QALYs are the “utility function” for medical decision
analyses
Foundations in preference theory




Pliskin, Shepard, Weinstein 1980
Miyamoto, Wakker, Bleichrodt, Peters 1998
Miyamoto 1999
More for the prescriptive DA field to do?


Equity/distributive issues (Lipscomb 2009, Drummond 2009)
How to aggregate health impacts over time? Is QALY =
𝑥 𝑡𝑥 𝑢(𝑥) too simplistic? (Lipscomb 2009)
Stanford University October 2011
22
QALYs: More for the prescriptive DA
field to do?

McHorney 2004




Fryback 2004


Close to two dozen generic QOL instruments
Hundreds of disease-specific instruments
In cancer, over 75 different QOL measures exist
 83 instruments for “General status and quality of life”
6 generic instruments widely adopted






36-Item Short Form Health Survey (SF-36)
WHOQOL-BREF Quality of Life Assessment
Quality of Well-Being scale (QWB-SA)
EuroQol EQ-5D
Health Utilities Index Mark 2 and Mark 3
SF-6D, a preference-based measure derived from the SF-36
Stanford University October 2011
23
QALYs: More for the prescriptive DA
field to do?



Instrument developers use psychometric techniques
 classical test theory (CTT)
 item response theory (IRT)
… to assess reliability, validity, difficulty, stability
These are not tasks for prescriptive DA.
Stanford University October 2011
24
QALYs: More for the prescriptive DA
field to do? Important issues:


Community preferences vs. preferences of those who have
experienced a health state? (see Drummond 2009 )
Different methods for valuing health yield different results






Consensus that a standardized “reference method” is
required for assessing QALYs
Reluctance to endorse “yet another summary measure”
(Fryback 2004)



Standard gamble
Time tradeoff
Rating scale
Person tradeoff
loss in ability to compare to previous research findings
Measure would not be universally adopted
These are not prescriptive DA issues …
Stanford University October 2011
25
Health as a multiattribute concept


This is universally agreed to be the case.
Existing instruments all have multiple
dimensions/attributes



but not the same ones …
HUI 2/3 both already based on multiattribute utility
theory
More for prescriptive DA to do? Maybe not?


Reluctance to endorse “yet another summary
measure”
Anything more complicated than additive or
multiplicative might be too much of a “black box” to be
accepted by practitioners
Stanford University October 2011
26
ISPOR Development Workshop 2007 on “Moving
the QALY Forward: Building a Pragmatic Road”

Eight-Item Consensus statement
1.
2.
3.
4.
QALYs are a health-based input, but there are
other inputs to health decisions.
QALYs can be used for population-wide and
individual health decisions
Little is known about the relationship between
health and general well-being
Who should provide health value inputs?
a)
b)
5.
Those who have experienced the health state?
Representative sample of community members?
Distributive/ equity issues need to be addressed.
Stanford University October 2011
27
ISPOR Development Workshop 2007 on “Moving
the QALY Forward: Building a Pragmatic Road”

Eight-Item Consensus statement
6.
7.
A need to better understand why different
methods of QALY assessment give different
answers
Need for better ways to aggregate health
impacts over time
HYEs theoretically superior but practically infeasible
8.

A “Reference Method” is needed for QALYs
Which of these issues can be addressed by
prescriptive DA?


Distributive/equity issues?
Health impacts over time?
Stanford University October 2011
28
Summary

Decision psychology continues to make contributions
to medical decision analysis





Prescriptive DA forms the basis for medical DA



Decision psychology
Preference/ utility
Shared decision making
Risk communication
Basic concepts and tools widely used
More sophisticated methods not understood and not
applied
Health quality (QALYs)


Some opportunities for prescriptive DA contributions
The biggest stumbling blocks to progress do not appear
solvable by prescriptive DA
Stanford University October 2011
29
Questions/ comments to follow …
Stanford University October 2011
30