Opinion: Machine learning can improve health care

ajc.com

This story was originally published on Oct. 27, 2018.

To what extent can your doctor’s functions be automated — replaced or enhanced by intelligent machines? How might such automation improve care and reduce costs? These questions are central to understanding Clover Health — a California-based company providing Medicare Advantage insurance plans in seven states: New Jersey, Pennsylvania, Tennessee, Georgia, Arizona, South Carolina and Texas.

A while back, I hosted a dinner in New York for a dozen-plus health care innovators — entrepreneurs, medical school professors, futurists, etc. Someone in the room asked, “How much of today’s physician services can be reduced to algorithms?”

An algorithm is a set of instructions (like a computer program) leading to unambiguous results. Feed in a patient’s biological metrics, desires and other characteristics and receive a diagnosis of the patient’s illness, causal factors, and a prescribed treatment regimen.

At the New York dinner, attendees speculated that between 80 percent and 97.5 percent of what doctors do today could ultimately be distilled down to algorithms — theoretically allowing computers to substitute for physician expertise (and leaving doctors free to perform higher-level cognitive tasks).

Clover Health straddles medicine and technology. Its chief technology officer, Andrew Toy, came not from health care but rather from Google, Android, MTV and Morgan Stanley. His LinkedIn biography describes Clover as “a technology oriented Medicare Advantage company focused on changing healthcare by developing rule-and-ML based models and interventions that improve member outcomes.”

With ML (machine learning), algorithms rewrite themselves as the machine “learns” more and more about patient care. I’ve written about a recent case where IBM’s Watson computer “read” a Japanese leukemia patient’s medical records, genetic data, and 20 million journal articles on leukemia (all in around 10 minutes) and concluded that teams of doctors had misdiagnosed her illness and treated her with the wrong medications. Watson effectively, continually reprogrammed itself to analyze the patient’s illness in ways no human had done.

Watson’s counter-intuitive leukemia diagnosis is an extreme example of machine learning, though it may suggest what medicine will look like in, say, 25 years. Clover Health is attempting to use similar logical tools to manage routine care. Let me explain, based on recent conversations with Toy and Jason Alderman, Clover’s chief communications officer.

Clover aims to be ahead of the curve on applying machine learning to day-to-day patient management. They collect data from enrollees’ hospital and provider visits, examine the data for insights, and allow their algorithms to test hypotheses, learn, and find trends that in all likelihood no human would discover.

Goals include getting patients to seek care outside of high-cost emergency rooms, taking their medications, and managing chronic diseases (such as diabetes). Machine learning enables Clover to predict with 85 percent accuracy whether a patient is likely to face hospitalization within 28 days.

Clover, acting on the advice of its algorithms, has its call center contact enrollees when problems seem likely. If systems indicate more serious need of intervention, nurse practitioners and medical assistants visit enrollees’ homes. Every one of these actions is recorded and fed back into Clover’s data warehouse, which allows machine learning to constantly improve future interventions.

Medicare’s rules pose challenges to Clover’s model. Stringent network access rules make it difficult to expand the insurer’s territory. Mandated minimum wait-time and access requirements may be mechanical and differ substantially from realities of time and access.

A particular challenge is Medicare’s Star Ratings system. Medicare rates each plan on a scale of 1 (poor) to 5 (excellent). A 1-star plan can only enroll new members during select portions of the year; a 5-star plan can recruit anytime. High ratings also earn plans bonuses for patient compliance.

Hence, Clover argues, plans often play to the test rather than to their enrollee’s health and personal interests. As an example, African-American enrollees are less likely than white enrollees to get annual flu shots. Hence, the rating system provides an incentive to avoid African-American enrollees.

Alderman says Clover will always focus on the well-being of patients over its Star Rating. He said their goal is to do the right thing and will work with the Centers for Medicare and Medicaid Services to enhance the Star Ratings system and to link metrics far more directly to patient health and quality outcomes.