Hongyu Zhou

How can we... Use LLMs to understand whether disease research happens where it's most needed? Read the blog post by Hongyu Zhou on the Accelerate Programme website. Image shows a photo of Hongyu with his quote.

📢 New blog! @zhou-hy.bsky.social used LLMs to investigate whether disease research happens where it's most needed.

Analysing 300k papers revealed stark disparities in global health research distribution - 43% of clinical trials occur in low/middle-income countries.

Find out more: bit.ly/47RyL2L

The Changing Geography of Medical Research Medical research remains concentrated in high-income settings, risking misalignment with global health needs. We build a geography-aware knowledge graph linking articles in the 524 leading medical jou...

🔗 Read the full article now on medRxiv: medrxiv.org/content/10.1...

Thanks for following along! Comments & feedback welcome!
🙏🧵

10/🧵
📈 The result: health shocks trigger rapid, durable rises in research attention — both domestic & global.
Responses are strongest for high-lethality threats, showing how the system mobilizes when risks are greatest.

9/🧵
🦠 What about sudden shocks — like Ebola?
Do countries ramp up research when health emergencies hit?

To test this, we use 3,134 WHO Disease Outbreak News alerts as quasi-random shocks to disease burden. 📢

8/🧵
📊 Responsiveness in low-income countries depends heavily on these actors.
Without philanthropy, responsiveness growth would shrink by 38%.
Without government support, by 32%.

7/🧵
💰 Funders fund differently.
🔹 Philanthropies → target neglected burdens
🔹 Corporations → profitable chronic diseases
🔹 Governments → somewhere in between

6/🧵
🔹 Responsiveness has grown since 1990
🔹 Strongest growth in upper-middle-income countries
🔹 Rises modestly with GDP per capita — but varies widely among the wealthiest

5/🧵
📊 Is research becoming more responsive to health needs?
We modeled publications vs. disease burden (DALYs) to test how output shifts with a 1% rise in burden.

Answer: yes — but unevenly. 👇

4/🧵
📈 Promisingly, we see diversification of global research output across almost all disease areas.
Over time, regions with the heaviest burdens have also emerged as research leaders, e.g. in HIV/AIDS & STIs, where African institutions gained visibility as key contributors.

3/🧵
🌍 Some regions are often studied but less often authored.

For neglected tropical diseases & malaria, Africa is the context in 32% of research — but just 14% of authorship.

2/🧵
Findings:
Most research centers on NCDs in wealthier settings (cardiovascular, neoplasms).
But neglected tropical diseases, maternal & neonatal health remain less-studied despite heavy burdens.

1/🧵
We analyzed 300k+ research articles from top medical journals (1990–2021).

Using LLMs + crosswalks, we linked:
🌍 Countries’ research output
🦠 Disease burden (via DALYs)
⚡ WHO health shocks

🚨Thrilled to share our paper🚨
Does science focus on the diseases that hurt people the most?
When deadly outbreaks hit, how does research respond?

Our work with:
@prashantgarg.bsky.social
@trfetzer.com

shows how medical research worldwide responds to both endemic burdens and emergencies.
🧵👇

7/🧵
💰 Funders fund differently.
🔹 Philanthropies → target neglected burdens
🔹 Corporations → profitable chronic diseases
🔹 Governments → somewhere in between

6/🧵
🔹 Responsiveness has grown since 1990
🔹 Strongest growth in upper-middle-income countries
🔹 Rises modestly with GDP per capita — but varies widely among the wealthiest

5/🧵
📊 Is research becoming more responsive to health needs?
We modeled publications vs. disease burden (DALYs) to test how output shifts with a 1% rise in burden.

Answer: yes — but unevenly. 👇

4/🧵
📈 Promisingly, we see diversification of global research output across almost all disease areas.
Over time, regions with the heaviest burdens have also emerged as research leaders, e.g. in HIV/AIDS & STIs, where African institutions gained visibility as key contributors.

3/🧵
🌍 Some regions are often studied but less often authored.

For neglected tropical diseases & malaria, Africa is the context in 32% of research — but just 14% of authorship.

2/🧵
Findings:
Most research centers on NCDs in wealthier settings (cardiovascular, neoplasms).
But neglected tropical , maternal & neonatal health remain less-studied despite heavy burdens.

1/🧵
We analyzed 300k+ research articles from top medical journals (1990–2021).

Using LLMs + crosswalks, we linked:
🌍 Countries’ research output
🦠 Disease burden (via DALYs)
⚡ WHO health shocks

🚨Thrilled to share our Paper🚨
Does science focus on the diseases that hurt people the most?
When deadly outbreaks hit, how does research respond?

Our work with
@prashantgarg.bsky.social
@trfetzer.com

shows how medical research worldwide responds to both endemic burdens and emergencies.
🧵👇