The status of health tech after the pandemic was a key topic of discussion last week at CES, one of the largest technology shows in the world. It usually pulls thousands of attendees to Las Vegas but went completely virtual this year due to COVID-19.

Amid new product announcements from smart home tech to autonomous robots, one major theme was sustainability of the unprecedented acceleration in the telehealth industry. Another topic of debate was how virtual care is shifting from a one-to-one modality to a more comprehensive care model and what artificial intelligence designers can do to increase clinician trust in their diagnostic models.

Here are some of the top takeaways from the conference.

Digital health boom will continue post-COVID-19

Industry execs predicted digital health activity could even further accelerate after the coronavirus pandemic passes, despite looming debates over how it will integrate with existing systems and cost, among other question marks.

In 2020, digital health companies raised almost $14.7 billion in capital across more than 600 deals, according to Wainwright Fishburn, global health of digital health at law firm Cooley. That's a significant acceleration for the industry compared to previous years, one that's resulted in a flurry of deals, with the big one of the year being Teladoc's $18.5 billion acquisition of Livongo that closed in October.

Along with being the biggest digital health deal ever, it was the third largest deal in the U.S. overall in 2020, Wainwright said. Teladoc's stock grew almost 140% over the year, with peers like Amwell, Buoy and Ro also raising significant funding through investors and public offerings in 2020.

Investors are pinpointing trends like consumerism, the need to cut costs and expand health access to traditionally underserved populations like Medicaid beneficiaries or LGBTQ people as key drivers of continued growth in the industry.

"We're really long on direct-to-consumer models," Lynne O'Keefe, managing partner of early stage venture capital firm Define Ventures, said, noting COVID-19 has emphasized the need for a better consumer experience in healthcare. "We're really long on direct-to-consumer models."

But some investors also stressed the future of digital health is not DTC virtual care alone. Telehealth should integrate with other models and technology as part of a broader use case. Linking with wearables and sensors running in the background continuously will result in a more in-depth picture of a patient's health, driving more actionable insights — and making companies more attractive to potential investors.

"If it's straight telehealth, we're not interested," O'Keefe said. "To me, telehealth is a pipe ... we believe in hybrid models."

Digital health actors should also be looking for ways to drive value, instead of churning visits into other types of contact. In a healthcare system focused on end-stage diseases, digital therapeutics could be a way to facilitate self-service and prevention, Deneen Vojta, EVP of research and development for UnitedHealth Group, said Tuesday.

The software-as-medicine industry is growing quickly, with the global therapeutics market expected to reach $11.8 billion by 2027, bringing new treatment modalities to previously underlooked areas, like mental health, addition and chronic conditions.

"We feel the terms digital therapeutics and medical wearables will be interchangeable within the next few years," Steven LeBoeuf, CEO of wearables company Valencell, said. "Medical wearables need the digital therapeutic work cases in order to make them relevant."

Telemedicine: moving to virtual first

Experts forecast virtual care will continue becoming more of a front door for all healthcare needs, not just for select use cases. That shift is already evident, especially among millennials and Gen Xers who are more comfortable with the technology, moving from telehealth as a second or third option to a first choice for care access.

"It's never grown faster than it is growing right now," Teladoc CEO Jason Gorevic. "The technology enables us to break the models and rebuild them in ways that are much better."

The industry is moving to integrate more devices to collect data at home, and is increasingly specializing in once-niche markets like mental and behavioral health, experts said. 

In the next few years, "I think you’ll see many areas of specialization within telehealth that people will really gravitate towards," Varsha Rao, CEO of telehealth startup Nurx. In mental healthcare, for example, virtual care has the potential to expand access while avoiding stigma, by matching a patient with the right platform and specialist for their needs at a lower cost, with text, chat, video and online modules all being used interchangeably depending on a patient's preference.

Paired with remote patient monitoring and other devices, hospital-at-home will become more of a reality — besides the biggest machines that obviously can't be moved, Iris Berman, VP of telehealth services at Northwell Health, said.

And it's likely virtual care will continue to link with other clinical-grade technological tools to try and get patient data remotely that physicians normally would need to collect in office.

"Bottom line is, doctors need data," Drew Schiller, CEO of digital health platform Validic, said Tuesday.

But as startups zero in on pain points within the industry and look for solutions, it's important for developers to work with clinicians instead of striking out on their own, to build trust and ensure any new tech integrates with existing software, experts said at CES.

That's especially important as non-traditional companies, including tech giants like Amazon, Apple and Google, become increasingly involved in the healthcare industry.

"​One of the things tech companies get wrong is they think they can create something and then go look for the problem it solves for," Samir Qamar, CEO of diagnostic devicemaker Medwand, said.

CES speakers also worried about shifting reimbursement for telehealth, and how building this new tech-driven modality on top of America's fee-for-service payment infrastructure could drive up costs. New virtual services should try to integrate more AI and streamline operations to lower costs, and try to work within value-based models instead of volume.

"For them to become accepted widely, they have to show they're delivering value. They can't just be ways to ratchet up our healthcare costs," Vivian Lee, president of health platforms at Alphabet life sciences arm Verily, said.

Expanding trust in artificial intelligence

Artificial intelligence has the potential to transform healthcare, but many in the industry are wary of the technology's use as a diagnostic agent. Research on the efficacy of AI in the exam room is mixed, and many clinicians say it's difficult for them to trust suggestions from the inner workings of a machine they can't see or fully understand.

But taken along with the cloud, which enables researchers and developers to consolidate vast amounts of data, AI could help drive major insights in care, proponents say. Yet trust remains a huge problem, even as AI becomes more of a common tool.

Though machine learning often isn't explainable, transparency is key to increasing trust, Pat Baird, head of global software standards at Philips, said. Developers need to make sure they take steps to minimize bad data, ensure their algorithms are trained on a nonbiased data sample that's representative of the target population for the device and ensure user-centered design for target users.

Clinician input is important during this process. Also key is a clear regulatory framework so manufacturers and doctors know the rules of the road when it comes to AI, especially as total transparency may not be possible. The software is trained on comparable datasets to find its own pattern, and uses that to make a recommendation.

"We know how to do quality controls — period — regardless of the product or the type and I think we can reuse a lot of that," Baird said. "The details are different but I think overall we have a good head start."

But "communication really is key. You can't just set a piece of software in front of somebody and say trust me," Christina Silcox, digital health fellow at the Duke-Margolis Center for Health Policy, said.

And regulation of the products is also complicated by the fact that not all AI tools are considered medical devices, so aren't regulated by the Food and Drug Administration.​ Additionally, a handful of Supreme Court decisions mean AI software can't always get patented, so companies don't have to publicly release what patented software does that could be critical for understanding what its weaknesses are, Silcox said.

But there is information that's good to have to try to combat the black box nature of some of these algorithms, including performance data independent from training data, along with detailed information about intended use, input requirements, and how and whether the software will be evaluated over time to ensure efficacy.

"The key to trustworthy AI is for manufacturers to build AI that deserves trust," Silcox said.