Misdiagnosis costs lives, increases healthcare expenditures, and reduces reimbursements for providers in a value-based care system:
Can AI diagnostics fix this problem?
Written by Spring DigitalDx Ventures Fellow Sam Lambert
Introduction
Over the past several years, artificial intelligence has been increasingly utilized. AI has the potential to improve patient care and reduce costs. This is particularly true in the field of diagnostics. Currently, misdiagnosis is a significant problem in the U.S. healthcare industry. In fact, current diagnostic errors cost the U.S. economy nearly $750 billion each year and leads to an estimated 40,000 to 80,000 deaths.
By utilizing AI for diagnosing diseases, errors can be significantly reduced. This is particularly useful at a time when the U.S. healthcare system is transitioning from a volume-based reimbursement system to a value-based care system, where insurance reimbursements are based on factors such as patient outcomes and experience. Using AI in diagnostics can improve efficiency and accuracy, leading to better patient outcomes and thus improved reimbursements and bottom-lines for hospitals.
AI Diagnostic Market and Applications
The market for AI diagnostics is growing rapidly. In 2019, the global market for AI diagnostics was valued at $378.4 million. This is expected to grow to $1.98 billion by 2024 at a CAGR of 39.3%.
The rapid growth of the AI diagnostics market can also be seen by the rapid increase in investments and patent publications. Between 2016 and 2019, more than $1.1 billion was invested in companies developing AI diagnostic technologies. Furthermore, in 2018 alone, there were more than 1,100 new patent families related to AI diagnostics. The applications for these new technologies span multiple medical specialities and disease states.
Applications and Impact
Given the importance of diagnostics in medical care, applications of artificial intelligence can have many different applications across disease states. DigitalDx Ventures, a niche strategy impact fund that focuses solely on AI diagnostics startups, has shared examples of their portfolio companies that demonstrate the potential of this technology. These examples can be found in table 1 below:
Value-Based Care Overview
Value-based care refers to the payment for healthcare services based on the quality of services provided, as determined by factors such as patient outcomes and experiences. It can be contrasted with fee-for-service models, where providers are paid for each service given to a patient.
The model of value-based care started to emerge in 2008 with the passage of the Medicare Improvements for Patients and Providers Act. It was further advanced in 2010 with the passage of the Affordable Care Act. Payments based on value-based care have been steadily increasing, with an estimated 34% of healthcare payments in 2017 tied to this model. Various forms of value-based care exist. Some examples of these models can be found in table 2 below.
Though these models differ in the way they approach value-based care, they share the central tenet of basing reimbursements on quality rather than quantity, shifting risk to healthcare providers. There is some concern from providers that this can reduce reimbursements. For example, in the 2013–2014 fiscal year, an estimated 44% of hospitals that participated in the Hospital Value-Based Purchasing Program, which adjusts payments based on patient satisfaction scores and other compliance measures, saw a decrease in reimbursements due to value-based measurements.
Since two of the important factors comprising value-based reimbursements are patient satisfaction and outcomes, properly diagnosing and triaging patients the first time is critical. AI diagnostics can improve these factors, mitigating the decrease in reimbursements experienced by hospitals. By streamlining the diagnostic process and reducing the chance of misdiagnosis, AI diagnostics can drastically improve the patient experience. Furthermore, improved accuracy and earlier diagnosis can increase patients’ chances of survival, thus improving their outcomes. Because patient experiences and outcomes are critical factors in determining reimbursements for value-based care models, AI diagnostics can help increase reimbursements by improving both of these measures.
There are also concerns that value-based care programs distort incentives for providers, causing them to treat less risky patients while under-treating patients with a higher likelihood of bad outcomes to maintain profitability. These concerns can be embodied by patients with more advanced diseases or other risk factors. Providers may be incentivized to turn away from these patients in value-based care systems as their reimbursements could be reduced. Furthermore, because providers no longer receive payments for each service they provide, they may be incentivized to provide fewer services, even ones that may be medically necessary, in order to reduce costs. Therefore, these distorted incentives could cause value-based care systems to reduce the quality of care for certain populations.
AI diagnostics can also address these concerns. With widespread use, AI diagnostics can catch diseases earlier, reducing the risk of patients who would have otherwise been diagnosed with a disease in a higher-risk stage. Furthermore, greater diagnostic accuracy can reduce the number of unnecessary procedures, allowing providers to perform only medically necessary procedures and still remain profitable.
Conclusion
The healthcare diagnostic market needs disruption. Inaccurate diagnostics lead to unnecessary procedures, high costs, lawsuits, bad experiences, and poor outcomes. Utilizing artificial intelligence in diagnostics has shown great potential for improving the accuracy of diagnostics and the ability for diseases to be caught earlier. In a world of emerging value-based care, where providers are reimbursed based on patient outcomes and experiences, AI diagnostics can not only improve healthcare affordability and effectiveness for patients, they can also improve reimbursements for hospitals.
