First up, we’re looking at how researchers are trying to fine-tune flu forecasting with the help of AI. Flu season reliably arrives around this time every year — but where the virus heads and how it spreads can seem wildly unpredictable.

One startup, Kinsa, is collecting data from smart thermometers to develop algorithms that might be used to derive insights about flu activity — which it could then sell to retailers, consumer-goods companies, and maybe even health systems. At the same time, academic researchers are trying to use machine learning and statistical methods to better predict flu’s path and peak. 

Interest in those high-tech approaches is growing. One sign of the boom: When the CDC first invited researchers in 2013 to use their data to forecast the flu season every week, researchers submitted 11 models. Last flu season, there were 38. 

But despite the excitement over AI’s promise, there are plenty of obstacles that could stop it from meaningfully improving flu forecasting. The spread of flu is notoriously difficult to predict, even with high-tech tools. The most notable example: the failure of Google Flu Trends.

The underlying data used to build the algorithms can be low-quality, messy, and unrepresentative of who’s actually sick with the flu. It’s not clear whether there’s money to be made in developing the models. And it also isn’t clear whether public health officials can be convinced to use them in their preparations for flu season. 

“The challenge is to see how they can incorporate what we’re sharing with them every week to change the way they perform their activities. In other words, we can do a lot of math, a lot of technical progress — but if it doesn't become useful to change behavior in any meaningful way in our societies, then the efforts lose their value or their relevance,” said Mauricio Santillana, a leading flu forecasting researcher at Boston Children’s Hospital and Harvard. 

STAT’s Rebecca Robbins and I have more here. 

The talk of health tech

BLOOD SUGAR MONITORING MEETS DIGITAL HEALTH COACHING — Care management company Omada Health has inked a deal with Abbott to pair that company's FreeStyle Libre continuous glucose monitor with Omada's online health coaching program. Patients with type 2 diabetes will be able to link their blood sugar readings to Omada's platform, where they can get personalized coaching about how to better manage these levels. The deal follows on the heels of Omada's $73 million funding round announced in June, as well as a recent strategic investment in the company from Intermountain Ventures, the terms of which weren't disclosed. 

COLOR SCORES ANOTHER HIGH-PROFILE PARTNER — Verily announced a partnership with genetic testing company Color to provide participants of Verily’s Project Baseline Healthy Study with “actionable” results from DNA tests. Participants in the project — which aims to build a high-tech map of population health and help people get more involved in clinical trials — will now be able to access Color’s testing services and genetic counselors. The partnership will give Project Baseline a trove of new data and could pull in more participants who want access to Color’s services. The collaboration is the second high-profile partnership for Color in as many months. Last month, the NIH’s massive All of Us research project gave Color $4.6 million to serve as the initiative’s genetic counseling service. 

IT’S EXPENSIVE TO STORE AND ANALYZE GENOMIC DATA — … like, really expensive. While it’s becoming more and more affordable to sequence the human genome and mine it for insights, storing data on the billions of letters in our genetic code can run up a steep bill. In a talk last week, Facebook’s Mark Zuckerberg highlighted that big expense as a challenge for the Chan Zuckerberg Initiative, which funds a slew of scientific projects. CNBC reports that Zuckerberg called out Amazon Web Services and fellow billionaire Jeff Bezos for the hefty price tag on the company’s cloud services. “Let’s call up Jeff and talk about this,” he said.

Image of the week

Researchers are testing AI to catch white eye, a sign of serious eye diseases. (kietzman et al. / science advances)

AN APP AIMS TO CATCH EARLY SIGNS OF EYE DISEASE — A prototype app is testing the power of AI to find something called white eye, a milky reflection on the eyeball in flash pictures that can signal serious eye conditions. The app, called CRADLE, was developed by researchers at Baylor University including Bryan Shaw, who came up with the idea for the software after his son Noah was diagnosed with retinoblastoma at 4 months old. The researchers tested the app using 50,000 pictures of 40 children, half of whom had eye disease and half of whom didn’t. CRADLE picked up on white eye in pictures that were taken, on average, 1.3 years before a child was diagnosed. The caveat: The app missed some cases and turned up false positives, which could lead to unnecessary medical visits if it were to be used more widely.

Study of the week

ANOTHER ROUND IN THE DEBATE OVER REAL-WORLD DATA — There’s a lot of talk about whether real-world data plucked from electronic health records, wearables, insurance claims, and apps could one day replace clinical trial data. But a new study suggests that isn’t likely to happen any time soon. Researchers looked at 220 U.S.-based clinical trials published in 2017 and analyzed whether they could be replicated with just observational data. Only 15% of the studies could have been replicated using only claims or EHR data. But the authors say real-world evidence can still bolster clinical trials in two ways: looking at whether findings from randomized trials and observational studies line up, and looking at whether a group of trial participants is representative of the real-world people who would use the intervention.

Coming up 

TALKING PATIENT ENGAGEMENT AND DIGITAL HEALTH — The Connected Health Conference kicks off today in Boston, bringing together everyone from chief technology officers and physicians to government researchers and regulatory experts. One topic on the agenda that caught my eye: why changing patient behavior is so hard, even with high-tech help. Follow updates from the conference with #Connect2Health on Twitter, and check back in next week’s newsletter for highlights and insights.