Source: MD Magazine | December 2018
Author: Chase Spurlock, Ph.D., and Michael Fleming, MD
When making a diagnosis, doctors have traditionally (and logically) relied on personal data directly from patients—their lab tests, examinations, and medical histories. But what if insights from population data were able to help doctors predict a potential diagnosis months or even years earlier and be used to monitor these patients after a diagnosis is made?
A growing body of research in the exciting field of predictive and prescriptive analytics suggests that if you input large datasets—drawn from millions of healthcare claims or electronic medical records, for example—sophisticated algorithms can identify patterns that deliver meaningful diagnostic information for patients with a wide range of conditions. These technologies can be used to uncover hidden risks in a population by detecting disease, correcting misdiagnosis, and monitoring disease progression. When patients are diagnosed earlier—and correctly—they will be able to start the right treatment plan sooner.
The power of this approach was demonstrated in a series of pilot studies by IQuity, a data analytics company, using proprietary algorithms to analyze healthcare claims for 20 million people in New York. Focusing on multiple sclerosis (MS), the trial included analysis of 4 billion data points, identifying patients who’d been correctly diagnosed with MS as well as those who’d been misdiagnosed. With high levels of accuracy, the approach was able to predict MS at least 8 months before it would typically be diagnosed using traditional methods.
While these pilot studies to date have centered on analyzing claims and social determinants data, predictive analytics can be enriched using a variety of resources. Real-time patient information can be integrated to fuel predictions, including data from Fitbits and mobile apps that monitor weight, blood pressure, and sleep patterns. When this level of personal data is layered on top of healthcare claims, electronic medical records, or other data representative of geographic, socioeconomic, and lifestyle factors, machine learning tools can offer doctors a more comprehensive view of a patient’s health outlook than has ever been seen before.