Instructions: You will work together on this exam. Each part contains shared questions and individually tailored components (marked in green for Styvers, orange for Fatma). Your responses should demonstrate command of global health evidence, climate science, and the ability to use AI tools critically. All data sources, assumptions, and Claude interactions must be documented.
Submission Format: Your submission must be a Cloudflare Pages site (yourname.pages.dev) built with Claude’s assistance during the exam. The site should contain all your analyses, figures, tables, DAGs, the policy brief, and your reflection — presented as a professional, navigable web document. Use Claude Code to build and deploy the site. Submit the URL at the end of the 2-hour period.
Burden Estimation
Using the Global Burden of Disease (GBD) 2021 Results Tool, extract DALYs, mortality, and prevalence data for the risk-outcome pairs most relevant to your dissertation research.
Extract data for unsafe water, sanitation, and handwashing (WASH) as a risk factor for diarrheal diseases, lower respiratory infections, and nutritional deficiencies among children under 5. Disaggregate by county-equivalent subnational units if available. Also extract the broader "Environmental/occupational risks" category for Kenya.
Extract data for dietary risks, air pollution (household and ambient PM2.5), and occupational risks attributable to agricultural practices. Extract both national and subnational estimates where available.
Required Analysis
For your country:
- Quantify the current disease burden (DALYs per 100,000) attributable to the environmental risk factors most relevant to your dissertation
- Analyze the 2000–2021 trend — is the burden increasing or decreasing, and why?
- Compare your country to 3 peer countries in the same region and 3 countries where similar interventions have been implemented
Climate-Health Nexus
Using climate data from at least two of the following sources, analyze how climate change is expected to modify the disease burden you identified in Part A within your study regions.
IPCC AR6 NASA POWER ERA5 / Copernicus ND-GAIN Index WHO Climate & Health Profiles
Northern Kenya’s arid and semi-arid lands are among the most climate-vulnerable regions globally. Using climate projections (temperature, precipitation variability, drought frequency) for your five study counties (Marsabit, Turkana, Isiolo, Wajir, Garissa):
- Model how projected changes in drought frequency and groundwater recharge affect borehole functionality and the WASH-attributable disease burden
- Estimate the additional DALYs that could result from a 1.5°C vs. 2.0°C warming scenario, using your DRIP FUNDI sensor data on borehole downtime as the exposure pathway
- Analyze how climate-driven displacement and livestock migration (relevant to your pastoral study populations) compound water insecurity and health outcomes
Türkiye’s agricultural sector faces both heat stress and water scarcity under climate projections. Using data for your priority agricultural basins:
- Analyze how projected temperature increases and precipitation changes affect agricultural water demand, crop yields, and the economic viability of your carbon-financed drip irrigation model
- Estimate the health co-benefits of transitioning from flood to drip irrigation — reduced pesticide runoff exposure, reduced farmworker heat stress through changed labor patterns, and improved household nutrition through sustained crop productivity
- Quantify these co-benefits in DALYs averted using GBD risk factor attributable fractions
Intervention Impact Modeling with Claude
Using Claude as an analytical tool, construct a health impact model that connects your dissertation intervention to measurable global health outcomes. You must complete all four components below.
C.1 — Causal Pathway Diagram
Use Claude to help you construct and critique a Directed Acyclic Graph (DAG) linking your intervention to health outcomes through environmental, behavioral, and economic mediators. The DAG must include at least:
- Your primary intervention (borehole O&M programs / drip irrigation adoption)
- Environmental exposure pathway (water quality or agricultural/air quality)
- Climate modifier (how climate change amplifies or attenuates the pathway)
- Carbon finance mechanism (how it sustains the intervention)
- Health outcome (specific GBD cause or risk factor)
- At least two confounders you must address
C.2 — Health Impact Estimate
Using your GBD baseline data, climate projections, and intervention effect sizes from published literature, estimate the DALYs averted per year by your intervention at current scale and at projected 2030 scale. Use Claude to:
- Identify the strongest published effect sizes for your intervention type (cite the specific studies)
- Propagate uncertainty through your model (provide 95% uncertainty intervals)
- Identify the parameters your estimate is most sensitive to
C.3 — Cost-Effectiveness Analysis
Using your dissertation’s cost data (DRIP FUNDI program costs / drip irrigation lifecycle costs), calculate the cost-effectiveness of your intervention in $/DALY averted. Compare to:
- WHO-CHOICE thresholds for your country
- The cost-effectiveness of alternative interventions addressing the same disease burden
- The implicit cost-effectiveness of the carbon finance mechanism — i.e., does the carbon revenue make the intervention cross the cost-effectiveness threshold it wouldn’t otherwise meet?
C.4 — Policy Brief
Write a 1-page policy brief addressed to a specific decision-maker, arguing for your intervention using the health + climate + carbon finance evidence you’ve assembled. Claude may assist with drafting, but you must demonstrate that you directed the analysis and can defend every number.
Address your brief to a Kenya County Governor in one of your five study counties.
Address your brief to the Turkish Ministry of Agriculture and Forestry.
Reflection on AI Use
500 words. Describe specifically how you used Claude in Parts B and C. Address:
- What did you ask it to do? Provide specific examples of prompts and outputs.
- Where did it help most?
- Where did it make errors or produce outputs you had to correct?
- What does this tell you about the role of AI tools in global health research — specifically, where do they complement versus substitute for domain expertise?
Evaluation Criteria
| Component | Weight | Assessed On |
|---|---|---|
| A. GBD Data Extraction & Interpretation | 20% | Correct use of GBD tools, appropriate disaggregation, trend analysis quality, peer country comparison logic |
| B. Climate-Health Analysis | 25% | Quality of climate data integration, plausibility of projections, strength of connection to study context, scenario analysis rigor |
| C. Impact Modeling with Claude | 30% | Rigor of causal model (DAG), defensibility of DALY estimates, appropriate uncertainty quantification, cost-effectiveness methodology, policy brief clarity |
| D. Reflection on AI Use | 10% | Honesty, specificity of examples, critical evaluation of AI limitations, insight into human-AI complementarity |
| Overall Integration & Collaboration | 15% | Evidence of genuine collaboration between students, cross-fertilization of methods/findings, coherent shared framing, quality of writing |