How deep learning helps match the right patient with the right healthcare Posted on : Nov 17 - 2018

Most healthcare systems and insurance providers know who the high-risk patients are: Usually people living with chronic conditions—diabetes, asthma, heart disease—who aren't managing their health well enough. These individuals are at a high risk of ending up in the ER. There are several problems with this high-risk list approach to caring for patients.

First, these risk scores are often compiled using insurance claims. This is a limited data source, and it is all about the past, not the present. This is not the kind of decision support that prevents readmissions or gets the right care to the right person at the right time.

Second, insurers and physician practices often use phone calls to check in with these individuals. This method is as inefficient as it sounds.

Finally, this approach doesn't provide any clue as to what is preventing an individual from being healthy. The healthcare providers know who the sickest people are, but they don't always know what the person needs to stay healthy. The missing element could be transportation to doctor visits, medication reconciliation, financial support to buy food or prescriptions, or even more frequent doctor visits.

The promise of artificial intelligence in healthcare is to help doctors and nurses—and even insurance companies—match the right care at the right time to the right person. That change—mass personalization in healthcare—is the promise of the specialized version of AI called deep learning.

Deep learning is branch of machine learning. Deep learning systems are modeled on the human brain. These artificial neural networks learn by passing data through layers of algorithms. Training data is fed into the bottom layer. Each node in the layer assigns a value to a data point. If the value passes a certain threshold, the data moves on to the next layer, until it arrives at the output layer. During training, these thresholds are adjusted until analysis of similar data sets yield similar outputs.

MIT is using this technology to power to simulate a clinical trial to determine the lowest possible dose of chemo for people with brain cancer. They are also working on a model that could suggest treatments for sepsis.

Health tech companies are using deep learning to, for instance, predict which person will develop pressure sores during a hospital stay or which heart attack patient will be back in the hospital within a week. Doctors need deep learning tools to compile data from multiple sources, look for patterns, and rate risk at the patient—not the population—level. View More