Artificial intelligence can be a boon for some jobs in healthcare that involve a lot of administration or rote activities, but AI cannot yet replace the nuance or flexibility of human decision-making, a new study asserts.
“The AI revolution has started in earnest,” Harvard economics professor David Cutler, PhD, observes in a new study report.
But as long-term care and the rest of the healthcare industry adopts or develops tools at a record pace, it can be important to specify when and where the technology is most beneficial.
Cutler outlined five categories in which to review AI’s efficacy in healthcare in a new article for the JAMA journal network.
AI can lead to cost savings and staff reduction in routine office work roles as remote monitoring, but it cannot replace physicians or researchers, and probably shouldn’t, report authors assert.
The five categories, briefly, are as follows:
- For simple, repetitive tasks like office billing and appointment scheduling, AI can replace people outright.
- AI cannot replace physicians’ more subtle patient reads, particularly of patients they have been seeing long-term.
- AI can enhance efficiency in healthcare with patient monitoring and specific diagnoses like blood biomarkers.
- AI must be designed not to mimic or exacerbate human biases across racial, ethnic, economic or age divisions.
- AI’s conclusions should help generate hypotheses and spur clinical trials, but AI-generated findings shouldn’t be taken at face value.
Overall, Cutler’s tone is both more optimistic and supportive of AI’s overall adoption, noting that at times it “can display more compassion than physicians”, but that “building software code and algorithms that replicate human mistakes and bias is not good enough.”
The latter warning echoes other recent studies which show that racial and ethnic bias, as well as current gaps in internet access, could exacerbate existing healthcare inequities.
Cutler has written extensively in the recent past about AI, concluding with a research team that AI could cut healthcare costs in the US by as much as 10% — up to $360 billion — annually, while noting that widespread adoption of AI within healthcare has been slower than in other industries.
