AI is the new paradigm in forecasting infectious disease risk – Healthcare Finance News

Optum Director of Research Danita Kiser, speaking at HIMSS21.

LAS VEGAS Optum is using artificial intelligenceand machine learning to create infectious disease forecasting that is as accurate as weather forecasting.

Optum researchers first began studying the COVID-19 pandemic while tracking an early flu outbreak across the United States in 2019, according to Director of Research Danita Kiser, speaking at HIMSS21.

Flu season began early that year, around Oct.1, according to the Centers for Disease Control and Prevention.

For 2019, "we had an early onset of influenza B," Kiser said. "It was scary for those of us in healthcare [research]. It usually doesn't start to take off until December."

The first COVID-19 cases were reported in the United States in January 2020.During the early stages, surveillance networks couldn't tell the difference between COVID-19 and the flu, Kiser said.

Forecasting saves lives, as public measures can be taken to keep people safe, according to Kiser. The CDC estimates that between 12,000 and 61,000 people die each year from the flu.

Optum's big data and AI-based computational epidemiology system can make a global impact, she said, sending text messaging to caregivers giving early warning so patients can be isolated.

COVID-19 has increased the urgency for accurate forecasting. In the United States, 618,000 lives have been lost to the virus.

A pandemic is much harder to forecast than an epidemic, Kiser said. COVID-19 came with no historical data to use as a future predictor, and medical coding for the coronavirus wasn't widely used until March 2020.

For flu predictions, researchers useddata sets, Google searches, doctor's visits, academic research and UnitedHealth Group intelligence to produce a network of precise indicators of seasonal outbreaks.

In 2020, they used machine learning to find hidden patterns within masses of disease indicator data. The information then was used for predictive models that forecast when and where flu activity increases in states and cities around the country.

Optum uses a modeling approach called nowcasts. The model has 243 unique forecasts for each state. Real-time signals come into the network. Before nowcasts, forecasting the flu was retrospective, Kiser said.

Flu can now be forecasted out weeks, while the average accuracy rate is 80% precision forecasting for COVID-19two to three weeks in the future.

"In 2021, through a network of signals and indicators, combined with machine learning models, we can predict the outbreak of flu and COVID throughout the country," Kiser said. "If we can forecast a week or two in the future, we can put prevention measures in place to save people's lives."

Optum has been sharing the information internally with OptumCare leaders and is developing plans to share forecast information with external providers.

The worry now is around "long COVID," as the Delta variant continues to push up the number of cases.

Whether the next pandemic can be predicted, Kiser said, "that's the Holy Grail of forecasting."

Twitter: @SusanJMorseEmail the writer: susan.morse@himssmedia.com

Read the original post:

AI is the new paradigm in forecasting infectious disease risk - Healthcare Finance News