Factored stochastic tree modeling for medical decision making

About This Presentation
Title:

Factored stochastic tree modeling for medical decision making

Description:

A decision tree with stochastic nodes added ... Roach et al. (1988): Prostate cancer in a man with asymptomatic HIV. 6. Markov. cycle ... – PowerPoint PPT presentation

Number of Views:53
Avg rating:3.0/5.0

less

Transcript and Presenter's Notes

Title: Factored stochastic tree modeling for medical decision making


1
Factored stochastic tree modeling for medical
decision making
  • Gordon Hazen
  • Northwestern University
  • Rowland Chang
  • Northwestern University
  • James Pellissier
  • Loyola University/Merck Pharmaceuticals

2
Outline of talk
  • What is a stochastic tree?
  • Basic concepts..
  • Stochastic tree transformation and rollback.
  • Approximating human survival.
  • Factoring out mortality..
  • Factoring stochastic trees.
  • Discounting / Risk aversion.
  • Influence diagrams for stochastic models
  • Our THA model...
  • The StoTree modeling environment.
  • Cost-effectiveness for THA..

3
What is a stochastic tree?
  • A stochastic tree is
  • A decision tree with stochastic nodes added
  • A continuous-time Markov chain with chance and
    decision nodes added
  • A multi-state DEALE model
  • A continuous-time version of a Markov cycle tree

Stochastic Trees
Decision Trees
Continuous-time MCs
4
What is a stochastic tree?
Matchar Pauker (1986) Transient ischemic
attacks in a man with coronary artery disease
5
What is a stochastic tree?
Roach et al. (1988) Prostate cancer in a man
with asymptomatic HIV
6
Markovcycletree
7
Transforming stochastic trees
  • Superposition / Decomposition


8
Transforming stochastic trees
  • Eliminating self-transitions

9
Transforming stochastic trees
10
Stochastic tree rollback

v(x) Quality rate at x L(x) Mean
quality-adjusted duration beginning at
x Recursive formula
11
Stochastic tree rollback
12
Approximating human survival
13
Coxian approximation to human mortality
14
Coxian approximation to human mortality
15
Factoring out mortality
16
Factoring out mortality
Background mortality
Stroke morbidity
17
Factoring out mortality
Background mortality
Stroke morbidity
18
Equivalent product tree
19
Rollback with Coxian mortality
20
Factored stochastic trees
Cancer
AIDS
Background mortality
21
Factoring stochastic trees
Tsevat et al. (1986) Warfarin for dilated
cardiomyopathy
  • Systemic embolism

Pulmonary embolism
Systemic hemorrhage
22
Discounting / Risk aversion
  • Utility function yielding quality-adjusted
    duration
  • Utility function yielding discounted
    quality-adjusted duration
  • Rollback with discounting

G
23
Modeling risk attitude
Vaccine scenario What chance p of immediate
death would you take to reduce your ongoing
mortality rate by a percentage c?

Undiscounted quality-adjusted duration forces p
c
Discounted quality-adjusted duration allows p lt c
(risk aversion)
24
Continuous-risk utility assessment
25
Influence diagrams
Decision tree
Influence diagram
26
Influence diagrams with stochastic nodes
27
THA model
28
ACR functional status
29
THA vs. Conservative Management
30
ACR Functional Status / Initial THA Outcome
31
Prosthesis Status After THA
32
Last Surgery
33
Conservative Management
34
The StoTree modeling environment
35
StoTree modeling environment
36
...StoTree modeling environment
37
Rollback in the THA model
85-year-old white male
38
Cost-effectiveness for THA
85-year-old white male
39
THA Cost-Effectiveness Results
Write a Comment
User Comments (0)