Who should control the physician's Medical AI Assistant (MAIA)?

Now that MAIAs are beginning to gain traction, it's worth considering how control of these emerging tools will impact physicians and patient care.

For clinicians, who believe AI could bring benefit across clinical activities such as diagnoses and clinical imaging, 94% are likely to use a reliable and secure AI assistant to assess symptoms and identify a disease or condition.

The key words here are "reliable" and "secure." 

The utility, credibility and reliability of a MAIA depends on the quality of the training of the generative AI the and patient data the MAIA has access to.  Clinicians overwhelmingly expect the results from generative AI-dependent tools to be based on high-quality, trusted sources.

However, physicians recognize that institutional EHRs are often incomplete and frequently inaccurate. Patient-reported information, including outcomes and social determinants of health, can be lacking.  Data collected outside of the institutional facilities, such as certain lab results, wearables that track glucose, sleep, activity, diet, mood, and other lifestyle monitors, and second opinions from independent providers may be missing. Data breaches are commonplace. When institutions or private equity control the MAIA, how confident will the physician be that the available data is reliable and secure?

Institutional control of MAIAs can turn physicians into indentured servants

Like most of us, physicians are on the lookout to improve their situation and burnish their reputation. On-the-job experience and education are sometimes kept by the physician in so-called “teaching files”. There’s no reason to treat digital teaching files differently from paper ones. When physicians switch employers, they lose access to employer-provided tools previously at their disposal. The new employer may have a different EHR system and require the use of their own AI which likely does not have access to machine learning on a competitor’s digital system, including MAIA. When a MAIA cannot travel with the physician, valuable history and training unique to that physician's experience is lost.  The possibility of losing access to a robust and valued MAIA can be one more tie binding the physician to a suboptimal situation. It also commoditizes physicians, as institutional MAIA obscures differences and subspecialization. 

Control of the MAIA belongs with the physician

For the majority of us, our smartphones and laptops travel with us when we change jobs. These tools contain programs that we have tailored to our particular needs.  Physicians take their smartphones with them….why not their MAIAs? Whether the MAIA is paid for by the physician or the new employer, it needs to be accessible to the physician. Patient information belongs to the patient, not the institution, and cannot be withheld from patients or their designated physicians. 

According to the Hippocratic Oath, medical licensure, and the law, a physician’s first responsibility is to the patient. When physician controls their MAIAs they take responsibility for the technology. A hospital or EHR vendor’s first responsibility is to their board or stockholders, not the patient, and control of a clinical tool like MAIA could be seen as a conflict of interest.

Optimal physician use of a MAIA requires an accurate, transparent, and complete health record

HIE of One is developing Trustee®, an open-source, patient-controlled health record designed to provide accurate, complete, and verifiable medical information. A patient-controlled record—one involving only the patient and his or her authorized clinicians—ensures accurate and complete information, which can also include patient-provided input and information around social determinants of health. The combination of a physician-controlled MAIA and patient-controlled Trustee® puts the practice of medicine where it belongs—between the patient and physician—wherever the care is provided, with no third party interference.

The reasons for putting physicians and patients in control of health records and machine learning AI are clear. The technical and business issues are discussed in other posts on this channel. Join us here as we discuss the role of patient communities and modern security protocols to leverage the new patient rights regulations. Please invite others and stay tuned for upcoming demonstration access that allows you to combine your health records with state-of-the-art AI from OpenAI and Google.