Introduction to Decision Analysis
Outline
- Introductions
- Motivation
- Examples of Decision Analysis
- Workshop objectives
Course Website
https://graveja0.github.io/v-brch-nigeria-2023/
- All course materials (slides, case studies) are posted here.
- Our (likely evolving) schedule will also be posted here, and updated regularly.
Introductions
- Please introduce yourself!
Motivation
The Past Two Decades …
- Cured Hepatitis C
- Significantly reduced incidence of HIV
- Potential cure for relapsed/refractory leukemia & lymphoma
- Perfected vaccines (e.g. HPV vaccine) to prevent diseases such as cervical & other cancers
- Strides in preventing cardiovascular disease
Despite these advances …
- Burden of non-communicable diseases (NCDs) like cancer, cardiovascular disease, and diabetes is growing.
- NCDs are the cause of 74 percent of deaths globally, with most (86%) of those deaths in low- and middle-income countries (LMICs)
- Source: WHO
- Rates also remain high in many parts of the world for malnutrition, unmet need for sexual and reproductive health services, and maternal mortality.
Examples of Decision Analyses
Ex 1. HIV
You have been appointed as Director of a funding allocation committee responsible for prevention & treatment initiatives for HIV.
How will the committee decide on the proportion of funds for prevention efforts versus treatment?
Should any of the funds be used for research?
How do you respond to a member who argues that the funds are better spent on childhood vaccinations?
Ex 2. Birth Defects
A hypothetical birth defect is present in every 1 in 1,000 children born
Unless treated, this condition has a 50% fatality rate
Ex 2. Birth Defects
Should we test for this hypothetical birth defect?
A hypothetical birth defect is present in every 1 in 1,000 children born.
Unless treated, this condition has a 50% fatality rate.
Ex 2. Birth Defects
Should we test for this hypothetical birth defect?
Diagnostic test: Perfectly accurate
All newborns in whom the defect is identified can be successfully cured
BUT the test itself can be lethal:
- 4 in every 10,000 infants tested will die as a direct and observable result of the testing procedure
Ex 2. Birth Defects
Objective: Minimize total expected deaths
Ex 2. Birth Defects
Objective: Minimize total expected deaths
Consider a population of 100,000 newborns
Testing produces: (0.0004 x 100,000) = 40 expected deaths
No testing produces: (0.001 x 0.5 x 100,000) = 50 expected deaths
Ex 2. Birth Defects
Objective: Minimize total expected deaths
Consider a population of 100,000 newborns
Testing produces: (0.0004 x 100,000) = 40 expected deaths
No testing produces: (0.001 x 0.5 x 100,000) = 50 expected deaths
Looks like TESTING WINS!
Ex 2. Birth Defects
Objective: Minimize total expected deaths
Consider a population of 100,000 newborns
Testing produces: (0.0004 x 100,000) = 40 expected deaths
No testing produces: (0.001 x 0.5 x 100,000) = 50 expected deaths
Looks like TESTING WINS!
Anyone got a problem with this??
Different lives are lost
With testing, virtually all 40 deaths occur in infants born without the fatal condition.
With no testing, all 50 expected deaths occur from “natural causes” (i.e. unpreventable birth defect)
Different lives are lost
“Innocent deaths” inflicted on children who had “nothing to gain” from testing program
We may treat one child’s death as more tolerable than some other’s – even when we have no way, before the fact, of distinguishing one infant from the other.
Ex 3. Colon Cancer
- 3rd leading cause of cancer death.
- Men, women, all races.
- Early detection aids in prevention & improves health outcomes .
- \sim80\% preventable.
Ex 3. Colon Cancer – Prevention
- When to screen?
- Which test & how often?
- If a polyp is found/removed, when to repeat?
Ex 3. Colon Cancer – Prevention
- Cost-effectiveness of alternative screening tests?
- Alternative frequencies of screening?
- How should screening vary by risk groups?
Ex 3. Colon Cancer - Estimating Probabilities
- Predict colon cancer risk/incidence
- Family history, past polyps, other factors
- Probability of getting colon cancer
- Predict impact from colon cancer
- Probability of cure
- Probability of death
- Probability of morbidity
Ex 3. Colon Cancer - Estimating Probabilities
- Predict “benefit” from screening
- All previous probabilities
- Probability of early detection
- Probability of complication from testing
- Probability of false positive
- Probability of false negative
- Probability of positive tests = cancer
- Probability of negative test = no cancer
Estimating probabilities is fundamental to decision making
- Cannot readily obtain needed probabilities
- Varying time periods / lengths
- Methods to estimate probabilities
Commonality of cases
- Unavoidable tradeoffs
- Different perspectives may lead to different conclusions
- Multiple competing objectives
- Complexity
- Uncertainty
Decision Analysis
Aims to inform choice under uncertainty using an explicit, quantitative approach
Aims to identify, measure, & value the consequences of decisions under uncertainty when a decision needs to be made, most appropriately over time.
Workshop Design
- We’re flexible – if there is a topic that is unclear to you, or that you would like expanded upon, please let us know!
- Mixed content
- Lectures
- Small group case studies
- Large group case studies and “hands-on” Excel exercises
Workshop Content
- Basics of decision analysis (Days 1-2)
- Decision trees
- Bayes theorem and treatment thresholds
- Uncertainty and prioritizing future research
Workshop Content
- Basics of Cost-Effectiveness Analysis (Days 3-4)
- Valuing cost and health outcomes
- Incremental cost-effectiveness analysis
- Introduction to Markov Modeling
Workshop Content
- Advanced Topic Preview (Day 5)
- Sensitivity Analysis
- Advanced CEA modeling frameworks.
Questions?
