BLOG: We need AI most in the room where it happens
Click Here to Manage Email Alerts
Artificial intelligence has become a routine part of our lives.
Siri texts my wife that I’m on the way home. My city’s water department automatically emails me if I have a break in a sprinkler line. Gmail offers to finish sentences for me. In medicine, AI identifies critical care patients at risk for decline based on vital sign changes. It scans terabytes of heart rhythm data from take-home monitors, finding patterns of arrhythmia. It detects melanoma better than board-certified dermatologists. It’s all wonderful, but I would argue that AI is still missing from the most important place we need it in health care. That would be, to borrow a phrase from the musical Hamilton, “the room where it happens.”
Without question, the room where it happens in health care is wherever a doctor and a patient meet for a traditional encounter, be it in the inpatient or outpatient setting. It’s there that a doctor hears a patient’s story, interprets data, explains his or her thoughts, and articulates a plan. The therapeutics we use today to treat patients are far better than a generation ago, but the ways we record that doctor-patient encounter have become worse, if we are honest about it. Electronic health records have made charting far more cumbersome, both to create and to interpret. Chart notes are filled with auto-complete templates punctuated with small, hard-to-find nuggets of actual patient-specific information.
A properly trained voice-driven AI “scribe” could mitigate this significantly, listening to the doctor-patient conversation and summarizing the history accurately, displaying relevant diagnostic tests on a screen without requiring 47 mouse clicks to find it. It could record verbalized exam findings and, taking all the available information, apply the right ICD-10 codes (yes, including all the specificity that no doctor has time for) as well as the treatment plan. The document would be completed before anybody left the room where it happened, and it would be more accurate than 99% of today’s work product.
What’s more, an AI assistant could give the doctor hints of all sorts: relevant history to collect, exam findings to seek, diagnoses to consider, drug alternatives to avoid interactions or to lower co-pays — all specific to the individual patient. It could collect and deliver to the patient relevant learning material and follow-up to be sure instructions were followed.
This type of virtual assistant would help doctors practice medicine that is both more individualized and malpractice-proof. It would facilitate far better review of quality, and maybe most importantly, it would allow doctors to look patients in the eye rather than stare at a monitor.
Which brings up this question: Do we even need a doctor in the room where it happens? Isn’t AI already smarter than us? Maybe, but that’s exactly the point: We doctors have spent 15 years letting EHRs degrade our patient interaction. It’s time that computers start to enhance it. Patients don’t care who records the chart note, but they sure want an unhurried doctor, a warm smile and an empathetic word from a person they trust. Humans make medicine humane. Computers can help us be more so.
Collapse