As the calendar turns to February, the beleaguered US healthcare system continues to battle an influenza season considered by Centers for Disease Control and Surveillance officials to be among the worst of this decade.
According to the public health agency’s most recent numbers, which include all cases of the disease reported to the CDC by January 20, this year’s flu season has seen a total of 11,965 lab-confirmed influenza-associated hospitalizations, representing a cumulative rate of 41.9 influenza hospitalizations per 100,000 US citizens.
Although these numbers are generally on par with 2014’s admittedly severe flu season, Dr. Daniel B. Jernigan, director of the CDC’s Influenza Division, noted during a January 26th press conference that it was the season’s escalating mortality rates that could set this year apart. According to the National Center for Health Statistics’ January 25th mortality surveillance data, 9.1 percent of that week’s deaths across the nation were attributable to influenza and pneumonia. At their height, rates in the 2012 to 2013 and 2014 to 2015 seasons peaked at 11 percent and 10.8 percent, respectively.
“We haven’t reached those peaks yet, but it’s still early in the season, and we may be getting to that level or maybe even surpass those numbers as the season progresses,” Jernigan said during the conference.
But while the situation may be grim, there is evidence of some relief coming out of budding digital and remote healthcare strategies. From alternative surveillance tools to triaging chatbots and virtual physician visits, many of these developing technologies are uniquely suited to addressing the challenges of influenza, John Brownstein, chief innovation officer at Boston Children’s Hospital, told MobiHealthNews.
“There’s many parts of the patient’s journey where digital tools can play a big role in terms of the flu epidemic,” Brownstein said. “Flu actually ties a number of exciting aspects of digital health together.”
More patients turning to digital tools
A number of consumer-facing digital health companies told MobiHealthNews that they’ve seen an increase in flu-concerned patients reaching for their services.
“Not surprisingly, there’s been a big increase in video doctor visits for the flu this year,” Dr. Sylvia Romm, medical director at telemedicine provider American Well, wrote in an email. “Just over the past month we have seen an increase of 300 percent for flu-related calls on American Well’s platform.”
Dr. Jason Tibbels, vice president of health services at Teladoc, described a similar increase.
“Our trends internally are tracking very closely with the numbers that the CDC is reporting. Our flu cases have way more than doubled [and are] significantly higher than anything we’ve ever seen in the past,” Tibbels told MobiHealthNews. “… The CDC reported that roughly 6 percent or so of brick and mortar visits are influenza-related, or influenza-like illnesses. I guess in part because of the type of illnesses that telemedicine attracts, our number’s closer to 20 percent of visits.”
Both telehealth providers said that the influx of patients resulting from the flu outstrips the standard growth rate they’ve seen over the past few years. This spike could be the result of increasing awareness of telehealth as a whole, as a representative from American Well noted that its hospital partners are, for the first time, telling patients to use telemedicine rather than visit the ER. For Tibbels, his company’s jump to roughly 8,000 total visits on peak days also signals a growing acceptance among consumers of telemedicine as an alternative for influenza care.
“Year over year in general, we find that [telemedicine] is becoming part of their normal flow, normal process in terms of getting healthcare … But there’s a lot of energy around this flu season. There’s a lot more communication in the press, people are way more aware of this option, and the numbers have shown that,” Tibbels said. “Largely, people are engaging more, and seem to be more comfortable engaging a virtual health encounter to address influenza.”
The increases do not seem to be limited to telemedicine alone. For instance, aggregated user data collected through Kinsa’s smart, connected thermometers indicated illness spikes across the country over the past few months. Meanwhile, Buoy Health — which makes a machine learning-powered chatbot that launched last March and had 2 million users as of late December — has roughly tripled its number of flu-concerned users since October.
“In October-November this year we were getting about 1,600 visits where their symptoms most closely matched the flu,” Dr. Andrew Le, CEO and cofounder of Buoy Health, told MobiHealthNews. “In December we’ve [increased] that to 2,300 visits, and in January we’ve gotten to 4,500 visits where it looks like they most likely have the flu. It made up 3.2 percent of all of our visits this month, and we had close to 139,000 [total] this month … so it’s definitely trending upward based on what we’re seeing.”
Brownstein noted that although these and other technology-driven care options appear to be growing in popularity among patients and providers, their total impact on care still pares when considering the huge number of people affected by influenza.
“To be perfectly honest, I still think these tools are in the hands of smaller numbers of people, whether it’s digital thermometers or crowdsourcing tools or symptom checkers,” Brownstein said. “I still think it’s early.”
Informing patients, reducing burdens
What makes the apparent growth of these platforms exciting is the benefit they could bring to flu patients and the healthcare system as a whole.
Automated triage systems, for example, can help inform patients about their condition or their best course of action in the time prior to a healthcare encounter, and are often more user-friendly than the online resources many currently reach for.
“Symptom checker tools — these are about being patient-accessible conversational tools that allow patients to work in their own natural language,” Brownstein said. “This is a big difference from the world of WebMD where there’s not a lot of customization or personalization of answers. This is consumer tools that are giving very precise information, and more importantly giving advice around triage — is it an urgent event? If it isn’t urgent can you stay home altogether? Getting that kind of insight is huge for a patient.”
Alternatives to an in-person visit are especially effective when dealing with infectious diseases, Tibbels noted. For flu in particular, he said that there are many clinical scenarios in which antivirals can be given based on symptoms alone. Staying at home and receiving the prescription over a video call would allow patients to receive their treatment faster, while avoiding the risk of exposure to others.
“The CDC has actually for years put out a telehealth triage for influenza where they strongly encourage communities where influenza is circulating — which is pretty much every community right now — to [not] run to an inpatient setting, [but] call your doctor,” Tibbels said. “Giving people the ability to, when appropriate, come in and be treated empirically, remotely, and eliminate exposure absolutely benefits the population overall. It can decrease spread.”
A separate issue that comes with any epidemic is the staggering number of patients that can overwhelm brick and mortar health care facilities, Le explained. While there is already a need for new tools to increase healthcare capacity, an epidemic substantially raises the stakes.
“At the end of the day, there are not enough doctors for the number of people who are sick,” he said. “There are different levels of seriousness that you may have, and it’s not appropriate for you to just go to the emergency room whenever you want. It’s very costly to the system, it’s costly to you; that’s not the most efficient way, and that’s not the safest way, honestly, to take care of yourself. … There have to be more and more digital solutions that can help us at baseline, and then in turn prepare us better in moments of epidemics and crisis.”
Both Le and Tibbels highlighted the negative impact those who merely believe they are infected can have on urgent care centers, hospitals, public health programs, and other patients. While both of their respective products can reduce the stress on traditional sources of care, Le noted that software-based solutions, especially, are less likely to feel pressures of long-term user strain.
“During times of epidemic there is no fear on our end of being overwhelmed by traffic — our biggest fear is actually being on television shows, weirdly enough, because that kind of concerted spike is difficult for servers to handle,” Buoy said. “But when it comes to something like the flu, it’s actually a much easier problem, and one that we don’t have to make adjustments for on our end.”
Data collection drives influenza care
The benefits of tools that engage patients and direct treatment are easy to see, but digital technology also promises a wealth of timely data on these epidemics that could be used to guide relief strategies.
“There’s some delays that occur in the collection and reporting,” Brownstein explained. “It takes time for an individual who is sick to see a physician before that data can be collected, and then for that data to end up in the hands of public health, that can analyze it.”
Outbreak prediction strategies based on online data have been around for some time. Among the earlier and better known attempts was Google Flu Trends, a service launched in 2008 that used Google search trends as an early indicator of potential illness. While this effort ultimately proved inconsistent and was shut down in 2015, similar projects including one from Epidemico continue to collect data from social media and other public online conversations to model public health trends from influenza to food safety.
“That data is great, but the challenge is that you never know the reason for the search, and you’re also missing a lot of key demographic information about individuals that are reporting these things,” Brownstein, who cofounded Epidemico prior to its acquisition by Booz Allen Hamilton in 2014, said. “So it’s not highly specific to understanding what’s actually going on in a population.”
Brownstein is instead turning to active patient engagement for his team’s latest disease surveillance project, Flu Near You. Through a web or mobile app, users are pinged each week to take a quick survey disclosing their symptoms, which is then translated into an active disease map and other resources that Brownstein says his team shares with the CDC.
“Why aren’t we just engaging people more directly to be part of flu surveillance more broadly? Why not put the public back in public health, make them an active participant?” Brownstein asked. “That has multiple areas of value. If you make someone an active participant in surveillance you can get better data, you can get geography and age and vaccination status and you can understand the value of a vaccine. You get real-time data.”
By making individuals active participants (as opposed to “just a data point”), Brownstein said that his team’s project has also drives public flu education and vaccine awareness. Further, crowdsourced data can be just as effective as that collected by a professional or through passive surveillance, he argued, even if the sample sizes are smaller.
“You don’t have to have millions and millions of people on crowdsourcing tools to get insight about what’s happening in the flu season, you know? We have tens of thousands right now and that’s done extraordinarily well,” Brownstein said. “Obviously you want a huge impact because you want adoption across the board, because that means education, that means better interventions and prevention in the home. But at the end of the day, from a pure surveillance perspective, existing data sets have done very well in capturing what’s happening.”
But while projects like Brownstein’s are focused solely on data collection and epidemic surveillance, a growing number of these digital health products and services have found additional epidemic-related uses for their own internal analytics. Kinsa, for example, recently announced the public launch of its disease surveillance platform, which uses temperature and symptom data collected passively through its digital thermometer and associate app to map the spread of illness throughout the country.
“The core hypothesis of the company is ‘Could we find a way to communicate with someone who’s just fallen ill to help them and, as a byproduct of that, understand where and when illness is spreading more broadly?” Kinsa CEO Inder Singh told MobiHealthNews a few weeks ago. “It works. It’s super cool and the fidelity of the data is so much even higher than what I expected.”
Singh said that the company waited through two and a half flu seasons to see how the company’s data matched up with the CDC’s flu statistics. Now, his team is able to boast an R-Squared of 0.96, which represents an extremely close match to CDC’s data.
“And we’re getting ours in real time, while their data streams in over the course of several weeks,” he said. “So we’re getting ours in realtime much earlier than any other data set, except perhaps social media, and certainly it’s more accurate than social media because it’s real illness data. It’s not a proxy from searches or conversations, it’s direct data.”
Singh said that his company’s data is ready to share with users, schools, and — for a price — health systems or manufacturers. Brownstein, when asked about this specific implementation, said that he would want to see more “rigorous scientific evaluation” of temperature and symptom-based surveillance, but that it can serve as useful point of entry.
“[Temperature is] a helpful indicator, but that’s not the only part of defining flu,” Brownstein said. “But it’s a great front door because with connecting a device and sensing a temperature, you can start to get additional information and put people through decision support tools.”
Both Le and Tibbels said that their services have also generated a substantial amount of location- and demographic-specific data. Like Kinsa, neither of these is a source representing lab-confirmed cases, although Le noted that his company is looking into ways for patients to confirm correct AI triage sessions after they’ve visited a professional.
Regardless, Tibbels said that data collected through their system could help public health target their efforts in the days before lab-confirmed data is available.
“We’re certainly open to that, and have had some discussions around that,” he said. “… What we’re seeing is helpful and worth sharing, we just haven’t traditionally had an avenue to do that. I actually have a relationship with a couple of docs at CDC, and we’re having those discussions.”
Whether or not the data is viable for public health, they can also be used internally to help Teladoc prepare its own network, Tibbels continued. This data, alongside influenza severity predictions from CDC and other experts, helps the company decide how many additional providers to recruit in each state or region, and to ensure that patients connect with a physician quickly.
“It’ll take efforts, sometimes, to do what I call ‘waking up’ the network,” he said. “We have over 2,000 certified physicians ready to talk to people within a median wait time of 10 minutes, but sometimes we’ve got to wake them up. That’s what we do to prepare our assets for these kinds of things.”
Being an end-to-end service, Teladoc also looks to use this data to better guide patients’ care after their session with a caregiver.
“As a physician I tend to focus on the diagnosis and treatment, but there’s a lot of support for patients on the front end and then the back end,” Tibbels said. “We’re seeing some local and regional shortages of Tamiflu, … our nurse team who handles helping fulfill prescriptions for our patients [is] hearing a lot of that, especially among the large chains. So, they’re working with patients. We kind of know where the shortages are and aren’t, and can redirect patients to pharmacies in their area that have supply.”
The consensus, it seems, is that the information being collected by healthcare devices and services is capable of delivering a tangible boost to the way patients receive and are guided to care. Still, Le wondered whether the abundance of logistical data generated by his product and other sources could be used to reduce influenza’s impact on a microbiological level.
“There’s so much work being done now to better model how flu virus is adapting year to year, and as technological solutions are helping patients on the front line understand symptoms, risk factors, and trends, it’d be really interesting to match that data alongside what’s really happening at a molecular level with the flu,” Le said. “[I wonder] whether that kind of holistic data set [can] help us to combat a virus that we haven’t largely understood in the long run. … Can we get richer and richer data sets to better inform our scientists, and ultimately defeat something that kills people every year?”