Can AI Cancer Materials Match Clinician-Generated Contents?
A new study by Nigerian researchers has found that patient education materials written by ChatGPT match those written by clinicians on understandability, clarity, and usefulness. That matters for anyone building AI cancer materials for African patients. On LinkedIn, Runcie Chidebe, who serves as Education Administrative College Intern Interprofessional Learning at Cleveland Clinic, posted the following:
Earlier this week our paper, 'Comparing AI-ChatGPT and clinician-authored cancer patient education materials: A cross-sectional assessment,' appeared in Cambridge's Journal of Clinical and Translational Science. As far as we are aware, it is the first piece of research to examine AI-generated PEMs on cancer prevention in Nigeria and across Africa.
Why AI cancer materials matter where services are stretched
Tailoring cancer information to what patients need has become easier through digital tools, particularly where services are stretched. Large language models such as ChatGPT form a newer generation of these tools, able to produce responses that adapt to patients' questions.
Earlier research indicates that ChatGPT and comparable tools can produce cancer-related PEMs, meaning patient education materials, that readers consider simple to follow and acceptable. However, little evidence has so far emerged from Africa or other LMICs (nations with lower and middle national incomes) about cancer PEMs produced by AI, and what exists is only now beginning to surface.
In an effort to close that gap, we set ChatGPT-generated PEMs beside those authored by clinicians in Nigeria, judging each on how clear, how useful and how easy to understand it was. Our hypothesis was that, on all three of these measures, the ChatGPT content would perform just as well as the clinician-written content.
Also Read: Ordinary People Episode 4: Kofi Didn’t Deserve This!
What the study found about AI cancer materials
- PEMs generated by ChatGPT with GPT-4o matched those developed by clinicians on understandability, clarity and usefulness, and no significant difference was found overall.
- Even so, the clinicians' statements on 'healthy diet' and 'limit alcohol' scored higher for clarity than the AI versions, and both gaps were statistically significant.
- Only 14% of respondents said they used AI to look up medical guidance, yet 58% said they had faith in it, which suggests adoption could rise fast.
- When seeking medical guidance, 73% of respondents went to doctors first, then Google (52%), other social media (36%), 'Aproko Doctor' (22%), a Nigerian physician who is also an online influencer, and AI tools such as ChatGPT (14%).
- Most respondents (95%) got online using their phones, and 49% used laptops as well.
- On technology use, fitness wearables were used by 30%,
- Apps for health in general were used by 36%, and apps focused on cancer by a mere 4%.
- Nine in ten respondents said they would be keen to use cancer apps for both prevention and care
What happens next for AI cancer materials in Africa
The authors noted that the use of AI is growing across Africa, and said this makes it pressing to build more culturally relevant, evidence-based African information into what AI systems produce, so that misinformation does not spread among African people.
His gratitude went to the co-authors of the work, to Miami University and to Project Pink Blue.
He invited readers to share their views in the comments, and called for collaborators to repeat the study in other African countries, asking interested people to get in touch.
Also Read: Ordinary People Episode 8 ends with Tornado still on the run
According to the study, AI tools may allow health workers in Nigeria to turn out understandable information for patients more quickly, especially in settings where clinicians can scarcely spare the time to write it. Even so, the researchers emphasise that a clinician ought to check whatever an AI tool generates before patients get to see it.
The two areas where clinicians scored higher, diet and alcohol, show why that review matters. Advice on everyday habits must be precise to be useful, and small differences in wording can change how a patient understands it.
