Cost-Effectiveness of a Three Classification Prevalence Testing Design

1
Cost-Effectiveness of a Three Classification
Prevalence Testing Design

• Ben De Winkle
• and Eric Lunderberg

2
Setup

• Estimating prevalence
• Example AIDS in Africa
• Why?

3
Tests

• Cheap test
• Field Test
• Fallible

• Gold Standard
• Lab-based test
• Expensive
• Infallible

4
Sample Designs

• One-Phase
• Only gold standard
• Two-Phase
• Cheap test, then gold standard
• Reclassification
• Cheap test, then cheap test again

5
Comparing Sample Designs

• Example
• 6000 budget
• 1 cheap test cost
• 100 gold standard cost
• One-Phase 60 people
• Two-Phase 285 people
• Reclassification 3000 people
• More people vs. more reliable information

6
Which design is best?

• The one with the least variance
• In this example, reclassification is best.

7
Last Summer

• Two classifications was an improvement.
• Is three even better? 4? 5?
• The groundwork was laid.
• This summer exploration has begun.

8
Is it worth pursuing?

• Simulations
• Instances when three classifications are better
  than two
• Diminishing returns

9
Methods of Getting Estimates

• Explicit
• Plug test results into an equation
• Iterative
• Use the test results to make a guess.
• Then

10
The Results

• As prevalence increases, three classifications
  becomes better than two classifications.

11
Future Work

• More work with iterative method
• Algebraic solving
• Checking with simulations

12
Thank You!