AI’s Healing Powers

Source details

Canonical URL

AI’s Healing Powers

Other formats

• Markdown version
• Structured JSON version
• Bundle manifest

Bibliographic details

• Authors: KERRY DOOLEY YOUNG
• Published: December 1, 2023

Summary and context

• Article title: AI’s Healing Powers By KERRY DOOLEY YOUNG
• Publication: F&D Magazine
• Date: December 2023
• Piece type: Case Study / short overview
• Central claim: Artificial intelligence shows promise in medicine, but there are recognized drawbacks and risks.

Demonstrated applications and empirical findings

• LumineticsCore system (Digital Diagnostics) used in a study in Rangpur, Bangladesh:
  • The system uses a special kind of camera designed to capture images of the eyes and evaluates them with AI.
  • Historical regulatory and payer milestones for the product:
  • 2018: the first AI-driven device to gain US Food and Drug Administration (FDA) clearance to check for diabetic retinopathy.
  • 2020: US Medicare agreed to pay for use of the device in primary care offices.
  • Trial productivity result (reported by Michael Abramoff et al. in Nature’s npj Digital Medicine, October):
  • When the AI tool was used, an estimated 1.59 patients an hour received what was deemed a high-quality visit, versus 1.14 in the control group.
  • Conclusion drawn by study proponents: LumineticsCore could help more people get screened for eye damage from diabetes, even in developing economies.

• AI in pharmaceutical research:
  • AI helps understand in early stages how compounds will work in the body, reducing the chance of side effects in later testing.
  • AI enables rapid synthesis of information about experimental drugs that would previously have taken years.
  • Quoted implication: “The efficiencies that have been built now on the discovery side are exponential,” and “We’re going to see a lot coming to the market soon.”

Risks, biases, and equity concerns

• Documented harm from biased algorithms:
  • 2019 Science paper found an algorithm widely used by large health systems and insurers underestimated the severity of Black patients’ illness, thereby risking denial of care.
• Source-of-data bias concerns:
  • Worries about developing AI tools based on data drawn with a tilt toward relatively wealthy people, who are often White and have good access to health care.
• Diversity needs:
  • Greater diversity is essential among patients whose data train AI tools and among the people who build these products.
  • Quotation: “You’re going to need to have multiracial, multicultural coders,” said Jerome Singh.
• Global South deployment challenges:
  • Comparative doctor density (WHO data cited in the article):
  • United States: about 36 doctors for every 10,000 people
  • United Kingdom: about 32 doctors for every 10,000 people
  • India: about 7 per 10,000 people
  • Infrastructure and institutional barriers cited: lack of electricity, lack of computer servers, shortage of workers able to translate AI-assisted diagnoses into effective treatment.
  • Framing of adoption difficulty: “In some settings, it’s going to be more of a sprint. In other settings, it’s going to be a marathon.”

Governance, regulation, and recommended safeguards

• Regulatory activity and guidance referenced:
  • 2021 WHO guidance report on ethics and governance of AI for health (advisors include Jerome Singh and Partha Majumder).
  • October (year not restated in text for WHO’s new report): WHO issued a new report outlining challenges of regulating AI in medicine, citing concerns about rapid deployment of tools derived from large language models without full understanding of benefits/harms.
  • A European Parliament report (issued last year in the article’s timeline) noted concerns about lack of transparency and privacy and security issues.
  • The FDA is working to refine its approach to regulating AI in medical products through formal guidance, specifying what evidence companies will need for FDA clearance.
• Policy and professional recommendations quoted or paraphrased:
  • Acceptance of AI adoption in medicine as inevitable: “We have to accept that this is reality,” said Partha Majumder.
  • Need for checks and balances to prevent inappropriate predictions and diagnoses.
  • Emphasis on building regulatory frameworks that define evidentiary standards for safety and effectiveness.

Framing and distinctions

• Terminology distinction offered by Michael Abramoff:
  • “Impact AI” — AI shown to help improve people’s health.
  • “Glamor AI” — products that garner tantalizing headlines but do not yet show hard evidence of benefit to patients.
  • Policy implication: Favor and invest in “impact AI,” and avoid paying too much for technology that does not improve outcomes.

• Ethical and practical imperative:
  • Vigilance required because AI can both help and harm patients.
  • Calls for more diverse training data, diverse developer teams, and careful regulatory oversight.

AI’s Healing Powers — KERRY DOOLEY YOUNG, F&D Magazine, December 2023