Soapbox

Disruptive Digital Health Models: Improving Access, Eliminating Bias and Delivering Higher Quality Care

By Jiayan Chen
March 3, 2022


• Categories: 01-Digital Health, 02-MedTech Economics, AI/ML, Care Delivery, Health Data, Soapbox

Digital health is experiencing a period of rapid, intentional disruption, simultaneously driven by and in search of developments in artificial intelligence (AI) and machine learning, improved data collection and curation methodologies, and rising consumer demand for tech-enabled health services. Fueled by the need to provide quality care during a global pandemic, healthcare stakeholders are acting quickly to identify new opportunities and overcome challenges in several key areas:

• Enabling improved scaling of digital health products and solutions, including unlocking more effective channels for their deployment
• Addressing systemic challenges in data quality
• Avoiding unintentional biases, overcoming digital gaps and solving other healthcare-delivery inequities
• Developing compelling healthcare-technology investment structures

While there may be myriad solutions to these issues, they are unlikely to succeed if pursued in isolation. To achieve positive, sustainable changes, developers of healthcare technology must identify ways to align incentives across healthcare industry stakeholders, such as payers, providers, social and other community service providers, government entities and patients.

Improving Patient Care and Population Health through AI-focused Payer-Provider Partnerships

The healthcare ecosystem is undeniably complex, but the ultimate goals are straightforward: Better care and improved outcomes for individual patients and populations. That said, despite notable strides, today’s healthcare system remains primarily designed around individual care moments rather than population health. And when quality comes into conflict with cost-cutting—particularly in healthcare services that are becoming increasingly commodified—the profitability imperative too often takes priority.

New digital health models can help prevent this type of unsustainable, zero-sum thinking. For example, some forward-looking companies are teaming with payers to implement center-of-excellence strategies powered by data analytics. These initiatives identify high-performing providers and align their incentives with payers, who are encouraged to construct their networks around these highly effective and productive providers. Another component of these programs is the deployment of AI and machine learning to facilitate learning and quality improvement across participants.

The results of such initiatives can be compelling. Not only are self-insured companies and healthcare plans seeing significant reductions in costs for plan beneficiaries who use such centers of excellence, they are also experiencing lower total costs of care and seeing improved patient outcomes.

Such real-world scenarios demonstrate a simple concept: By encouraging payers to deploy advanced technology at scale, the system as a whole can improve quality, enhance health outcomes, decrease costs, and increase activity for high-performance providers. Collaboration of this type can also replace what has often been a somewhat antagonistic relationship between payers and providers, in which incentives are aligned badly and one party’s profit is another party’s loss.

Improving AI Application Scaling and Data Quality

Healthcare is a global problem and stakeholders at every level must focus on global solutions. AI has the capacity to improve healthcare outcomes and address delivery challenges in virtually any setting. In developed countries such as the United States or Australia, this can mean helping overloaded healthcare providers make critical decisions in time-poor environments. In developing countries, AI can help overcome shortages of skilled expertise coupled with poor-quality health services and generally high healthcare needs.

But to achieve these benefits, AI needs to be scalable and the datasets upon which it is based must truly represent the populations it is meant to serve. Today, too much AI technology does not scale effectively, and it also fails many underrepresented groups such as ethnic minorities and women. This is in large part due to poor, incomplete, or nonexistent data for such populations.

From a flowchart perspective, before AI can be scaled, the datasets upon which it is based must be improved and expanded. This effort necessitates a focus on two key areas: gaining access to a greater pool of healthcare data, and ensuring consistent, quality data that is readily analyzable.

Success on the first point—gaining access to data—requires AI technology companies to gain the trust and cooperation of providers and patients in order to encourage data sharing, and to work with regulators and other governmental entities to allow access to and use of patient information while respecting individual privacy.

Today, there are a number of highly creative AI-based initiatives that have been developed that provide two-way benefits: Delivering effective care to underserved patients today, while collecting clean, relevant data to power the solutions of tomorrow. Transformational programs in women’s health—and in-vitro fertilization (IVF) treatments, in particular—provide examples of such innovation. AI-driven fertility applications can analyze images of patients’ embryos during the IVF treatment process to improve pregnancy outcomes, while at the same time gathering important medical and demographic information that strengthens existing datasets analyzed by machine-learning algorithms and that ultimately improves the next iteration of the technology.

These and other collaborative partnerships can benefit stakeholders from across the spectrum. Clinics that are engaged to contribute, prepare, verify and annotate data (which can account for the bulk of the costs associated with developing AI) can be assigned royalties for their contributions. Rather than trying to train the AI solution to manage poor-quality data, ensuring the arrival of high-quality data from the outset will itself promote improvements in AI technology, which can in turn enable providers to deliver better care that ultimately leads to better outcomes—and the cycle continues.

Depending on the jurisdiction, it may be more or less difficult for providers to share and technology companies to access patient data, particularly insofar as the data includes identifiers. Stakeholders across the spectrum should work with relevant legislative and regulatory bodies to design and update privacy laws and regulations that ensure the security and safe management of highly personal data while enabling its use to develop and deliver higher-quality care.

Once comprehensive, inclusive and accurate data has been collected, the next challenge for technology companies is to find ways to scale delivery of their healthcare AI solutions. Scalability in this context means that data and solutions are accessible and affordable for thousands of clinics and potentially millions of patients around the world.

One way to solve the scalability problem is to take a social-network approach to healthcare. Just as social media platforms have transformed how we interact and access content, social networks can be used to transform the global healthcare industry. Decentralized networks can be established that connect clinics and patients globally and allow safe data sharing, giving access to the diverse datasets needed to create and deliver unbiased and equitable AI-based care.

Preventing Unintentional Bias while Overcoming Digital Healthcare Gaps

Improved data and greater scalability can help stakeholders take significant steps toward preventing the emergence of biases and shrinking the digital healthcare divide. In fact, bias prevention and increased access can work hand in hand.

As the often overused but indisputably concise saying goes, “Garbage in, garbage out.” Simply gathering more data is not the solution. The cleaner and more comprehensive the datasets upon which AI solutions are based, the more healthcare technology companies are able to deliver equitable, fair, reliable and accurate solutions to diverse populations. By expanding access to healthcare technology by developing nations and rural and remote areas in developed nations, more effective data can be collected that will further refine the delivery of services to individuals and communities.

Unevenness of care delivery is the result of numerous factors. Often, the first issue that comes to mind is patient income level—it is typically assumed that less-affluent individuals and families are less likely to have reduced access to broadband internet services. But in many rural areas, income is no discriminator when it comes to the strength and speed of internet connections; in some communities, rich and poor alike are unable to obtain broadband services. This also means that providers in such areas, regardless of their experience or training, may not have the technology they need to take advantage of digital health solutions.

Healthcare training systems can also impact access. In some communities, students in healthcare programs at every level have less technology training (or opportunities for such training) than their counterparts in other communities.

Even when technology is available, healthcare delivery is only as good as the people and the processes behind it. If one portion of a given healthcare ecosystem lacks efficient policies and procedures to implement and manage digital health solutions, overall results will be less than optimal. And as has been noted during the current pandemic, while technology is an enabler of healthcare services, it has placed a spotlight on the many, overlapping issues that prevent technology solutions from being exploited to the maximum degree.

No single organization—no teaching hospital, no clinic, no physician practice, no community center, no philanthropist and no government entity—can tackle the digital divide on its own. A collaborative, holistic, ecosystem-based approach is required in which all stakeholders embed accessibility in their mission statements, strategies and programs.

Patient engagement is another key factor in bridging digital gaps. There should be multiple doors into the various systems, and patients should have a voice in how they access providers and information. Patients may prefer engagement via physical (postal-service) mail, or they may prefer to receive and send information electronically. Some get their information via social media, others respond to text messages. By developing multichannel communication strategies, we can reduce the friction that sometimes arises during the patient-provider interaction.

Developing Compelling Healthcare AI Investment Structures

Many of the healthcare technology solutions under examination rely heavily on early infusions of cash from healthcare-industry investors. Such investors take a hard look at the underlying business models of the targets under consideration and attempt to determine which of them are going to provide the greatest positive effects, both in terms of healthcare outcomes and return on investment.

Typically, models that have focused on payers—self-insured companies and commercial health plans—have had the scale, resources and incentives to address healthcare inequities and roll out technology solutions more efficiently and with more durable impact.

However, when even smaller stakeholders work together, they too can use their collective scale and resources to test innovative healthcare technologies, including AI and machine learning solutions, against large, diverse data sets, uncover potential biases that reduce positive returns (again, in outcomes and in profits), and improve both technologies and the healthcare delivery system. Potential investors can seek out these types of creative, disruptive payer-provider partnerships when making their capital allocations.

The AI investment sector is also becoming increasingly savvy and turning away from technologies that require the AI do all the work to overcome dataset biases. Where developers can demonstrate the capability to augment the AI with new data collection models and workflows that access the right, high-quality type of data at the front end, prevent biases from arising in the first place, and support a smoother rollout of the solution, they are likely to make an impression as a more attractive investment target.

The Road Ahead

With all due respect to its complexity, the science behind disruptive healthcare technology solutions such as AI is not always the limiting factor—fundamental research and technology development are occurring at an ever-increasing rate.

Often, the real limiting factors stem from challenges in obtaining reliable, meaningful datasets that reflect diverse populations and eliminate bias; achieving acceptance of AI and other healthcare technologies by the universe of healthcare stakeholders; integration of these technologies into workflows and reimbursement models; developing innovative partnerships that align disparate healthcare stakeholders; and creating attractive investment opportunities.

Today, there is a strong community of innovators within the healthcare ecosystem. These innovators often want to work only with each other, as it only takes one less-effective layer of a network to slow down the entire system. Unfortunately, the unintended effect of this preference is to slow down overall, global acceptance and adoption of disruptive healthcare technology.

Where barriers and resistance to innovation can be broken down, however, opportunities abound. One of the greatest challenges ahead will be to provide more-traditional players with the evidence and assurances they need to dip their toes—even their entire feet—into the digital waters.