Conducting Cost-Effectiveness Analyses of Behavioral Interventions

1
Conducting Cost-Effectiveness Analyses of
Behavioral Interventions

• Todd H. Wagner, Ph.D.
• Mary K. Goldstein, M.D. 

2
Acknowledgements

• Partial funding was through a grant from the
  National Cancer Institute (2PO1 CA 55112-05A1).
• Dr. Goldsteins contribution was informed in part
  by work conducted with a grant from the National
  Institute on Aging (R01 AG15110).
• David Meltzer, Jodi Prochaska, Lisa Faulkner, and
  Stanford University seminar participants provided
  helpful comments.

3
Outline of the Talk

• Background on cost-effectiveness analysis (CEA)
• A bias in CEAs for behavioral interventions
• How to fix the problem
• Example
• Study design consideration

4
Economic Evaluations

• Most trials dont include economic analysis
• Economic analysis in only 0.2 of 50,000 trials
  conducted between 1966-1988
• Adams et al. Medical Care 30(3)231-43
• Economic analysis in randomized clinical trials
  is increasing

5
Dollars and Sense

• Economic studies inform decisions
• formulary
• adoption of new technology
• scope of benefits
• strategies for management of care
• organization of health care

6
Cost Effectiveness Ratio
Where C1 is the average cost of the
intervention group C0 is the average cost of the
control group E1 is the average effectiveness of
the intervention group E0 is the average
effectiveness of the control group
7
Standardization

• USPTF guidelines (Gold et al., 1996 summarized
  in three JAMA articles)
• Drummond et al. (1997)
• Created, along with journal editors, standards
  for both
• Methods
• Reporting

8
Whose Costs?

• Guidelines recommend a societal perspective for
  costs
• Include
• Provider
• Payer
• Patient

9
Effectiveness

• Guidelines recommend QALYs
• Values both quality and quantity of life
• Each year of life is multiplied by a weighting
  factor (utility)
• Utilities measure the preference of different
  health states on a 0-1 scale

10
Behavioral Interventions

• CEA framework holds, but caveats
• Behavior change is a slow process
• Treat many to prevent a few
• Use of intermediate outcomes (proxies)

11
Intermediate Outcomes

• Outcome is clinically relevant and predicts
  mortality or morbidity
• Receipt of a mammogram
• Substance use abstinence
• Change in dietary fiber
• QALYs would require huge and/or very long studies

12
CEA with an Intermediate Outcome

• Sufficient for publication
• Hard to interpret ICER
• Cant easily compare two CEAs with different
  intermediate outcomes
• Cant compare CEA to other CEA from another
  clinical area
• Sometimes only feasible approach

13
CEA with QALYs

• Measure QALYs or
• Translate intermediate outcome to QALYs
• Either build a model de novo or use an existing
  model
• Requires a lot of resources
• Most useful, but most challenging

14
CEA and Behavior Change

• CEA analysts treat behavior change as a
  dichotomous outcome
• Partial behavior change is not the same as no
  behavior change
• Getting the person to recognize that they have a
  problem is half the battle.

15
Whats Missing?

• Partial behavior change is missing from current
  models
• People who progressed in their process of
  changing their behavior but did not successfully
  change their behavior at the end of the study
  (Stages of Change)

16
Partial Behavior Change

• All behavioral interventions yield some partial
  behavior change
• Amount depends on
• the duration of the study
• baseline stage of change

17
Partial Behavior Change

• Should we and can we value partial behavior
  change in a CEA?

18
Should We?

• NO
• Intentions are not the same as observed behavior
  change
• Stage of change is flawed
• Too difficult

• YES
• Behavior change takes time
• Unobservable differences exist
• Future benefits are important
• Otherwise favors med/surg tx

19
Can We?

• Behaviors are like value chains
• A series of linked processes
• Interventions may be designed to improve a link
  (stage-matched design)
• Matching chemotherapy protocol to cancer stage
• Interventions may have differential effects on
  different links
• Stages of Change (TTM) Model by Prochaska and
  DiClimente

20
Stages of Change

• People progress through successive stages until
  they change their behavior
• Precontemplation no self-recognition of a
  problematic behavior.
• Contemplation self-recognition of a problem
  without action.
• Preparation Planning to change behavior soon.
• Action In the process of change.
• Maintenance adherence to the new behavior over
  time.
• Art of behavioral interventions is to achieve
  action, and hopefully maintenance

21
Stages of Change and CEA

• Stages of change is critical for interpreting CEA
  results
• Stage of change may be associated with
  receptivity or motivation
• Intervention effects can vary by stage of change
• Incremental cost-effectiveness ratio can vary by
  stage of change

22
Integrating Stages of Change in a CEA
23
Whats Needed?

• Data on stages of change
• Probability of moving from partial to successful
  behavior change
• Note these probabilities are not observed in the
  intervention group

24
Two Step Process

• Step 1
• Calculate ICERs by baseline stage of change
• This alone can provide much more useful and
  interpretable information for stage-matched
  interventions
• Step 2
• Estimate probability of people moving into
  successive stages of change
• For example, moved from precontemplation to
  contemplation and preparation

25
Source of the Probabilities

• Probabilities from the literature or
• The control group (if possible)
• Possible when control group gets usual care
• If 10 of precontemplators in the control group
  changed behavior, this is the probability for the
  model

26
Hypothetical Example
27
Randomized Controlled Trial

• 2700 participants enrolled in 3 arms
• Post card reminder (n900)
• Reminder phone call (n900)
• Personal motivational phone call (n900)
• Sample from managed care plan
• Managed care organization wants to know which
  reminder to use

28
Cost and Effectiveness
29
Report to MCO

• MCO should choose between postcard or reminder
  call

30
Transition Probabilities
31
Including Partial Behavior Change

• Note motivational call is now most effective
  overall

32
ICERs

• No single strategy is always preferred.
• Motivational call is dominated by reminder call
  for contemplators and those in action

33
Origin of These Ideas
34
Patient Reminders

• Mailed reminder vs no reminder for mammography
• A meta analysis of 16 studies
• US studies Pooled OR1.48 (participants come from
  provider files)
• Aust / NZ studies Pooled OR5.57 (participants
  from voter lists)

Wagner TH. The effectiveness of mailed patient
reminders on mammography screening a
meta-analysis. Am J Prev Med. 199814(1)64-70.
35
CEA and Stage of Change

• Fishman P. et al. Cost-effectiveness of
  strategies to enhance mammography use. Eff Clin
  Pract. 2000
• Compared three alternative methods for increasing
  mammography screening
• Reminder postcard
• Reminder call
• Motivational call

36
Fishmans (2000) results

• ICER varied by prior mammography status (i.e.,
  maintenance)

37
Feeling Lucky?

• The implications of a CEA should not vary by who
  enters the trial
• Two randomized trials with same intervention
• RCT 1 all participants are in preparation
• RCT 2 all participants are in precontemplation

38
Clinical Trial Design

• If the effect of the intervention might vary by
  stage of change
• Qx Enroll people from all stages?
• Enroll sufficient numbers in each stage
• Protect randomization

39
Conclusions

• Subgroup analysis by stage of change may be
  critical for interpreting the CEA
• Partial behavior change is important and can
  affect interpretation
• These methods are appropriate for stage-matched
  studies
• Need to consider study design implications