The AI Boom in Healthcare: How Manufacturers Can Prepare

December 12, 2019

Kellie Rademacher identifies some of the key questions to consider when choosing an AI partner.

A vast amount of data is generated daily in the US healthcare industry requiring mechanisms to aggregate, store, process, interpret, and evaluate it at a pace and with accuracy that goes beyond human capital. Healthcare organizations like manufacturers, physician practices, institutions, health plans, and pharmacy benefit managers have invested in their own internal data infrastructure based on independent needs and with varying degrees of capabilities.1 The result: a slew of healthcare data across stakeholder types, populations and diseases being siloed.  Our healthcare system needs to build the capabilities and efficiencies to improve outcomes and reduce costs of not only delivering healthcare but developing it.2 A streamlined approach is vital for access to robust data sets and mechanisms to aggregate, interpret, evaluate and act on limitless amounts of data more rapidly.   Artificial intelligence (AI) has torpedoed to the surface as a mechanism capable of addressing the data needs of the healthcare industry. AI startup companies jumping on the drug development bandwagon have grown 12-fold (60%) in just two years. Some estimates show healthcare AI reaching $6.6 billion in value by 2021 with projections to save the US healthcare industry $150 billion annually by 2026.3   It should be no surprise then to see pharmaceutical manufacturers and other healthcare industry stakeholders dabbling in partnerships with not only big data companies like IBM Watson4 but also startup organizations5 on matters ranging from understanding EHR usability and how that supports precision medicine, to accelerating the development and adoption of digital tools directed at improving patient care and driving innovation, to partnerships like that between Novartis and Microsoft to put AI on every researcher’s desktop computer-an alliance that the company expects will help them reimagine medicine to improve and extend patients’ lives.6     The 60% growth in AI startup organizations inserting themselves into healthcare will provide healthcare companies with many partners to choose from but all may not be equal. Drug manufacturers will need to determine where they believe the demand for AI is internally needed, assess if that demand would best be met for the company by building the in-house infrastructure or by choosing to partner with an AI organization. Choosing the right AI partner will be imperative. A miss here can end up costing substantial dollars and valuable time without delivering any return on investment. IBM Watson and MD Anderson Cancer Center didn’t deliver on a $62 million investment.   When choosing an AI partner a few key questions have been identified. What is the return on investment going to be? How experienced is the company? How much can the AI company tell you about their own system? What is the real cost to the healthcare stakeholder?7 In order to make the best choice manufacturers or any stakeholder deciding on an AI engagement will need knowledgeable persons to help guide them, whether that be internal employees or consulting organizations skilled in understanding machine learning algorithms and with proven industry experience. AI companies looking to be successful in the healthcare space should be prepared to understand the industry as well as the unique needs of manufacturers and other healthcare stakeholders and be able to demonstrate competencies in the space. As the AI boom progresses, and consistent with what we are seeing across the healthcare industry today, it’s likely we may see pharma manufacturers, institutions, consulting organizations, and payers engage in M&A activity to develop an internal infrastructure or level of AI expertise. The potential absorption of smaller AI companies will increase as positive results of existing partnerships are published in the public domain.    It’s common in healthcare for parts of the industry to “watch and wait” as the pioneers take the first steps into new endeavors, which is one reason why in the short term, manufacturers are likely to partner with technology organizations piloting projects to understand the utility, scalability, efficiency, and overall economic impact of AI-generated outputs versus investing in the necessary infrastructure out of the gate. According to a recent article, the investment community is seeing major increases in three areas as it pertains to AI utility: digitization (making operational processes less expensive); engagement (improving the patient interaction with healthcare); and diagnostics (new products and services developed using AI algorithms).8  If predictions by Accenture are correct and AI could address approximately 20% of unmet clinical demand-coupled with Forbes’ reporting AI startups had their best year ever, raising $9.33 billion9-it is unlikely we will see a drought in funding for these organizations in the near term.     An outstanding question is if AI will be used primarily to mine existing data or if the algorithms will be used to generate new data. Most likely there will be a mix of both machine learning and neural networks. Machine learning is an algorithm that allows computers to learn independently without additional or explicit programming, and as the machine encounters more data the algorithm performs better. A subset of machine learning is deep learning, which takes the algorithm one notch further: it allows the AI application to draw its own conclusion, using multiple algorithms in tandem, to mimic a human neural network.10 With these types of capabilities, it’s clear the industry is moving in parallel with AI applications generating new data and algorithms mining existing data. It will take time and more importantly education by companies using AI-generated data to the consumers of this data for the healthcare industry to understand AI algorithms. Experience will also be needed for the industry to trust evidence being generated from this new technology for inclusion in decision-making processes. This represents an opportunity for pharma manufacturers to engage with providers, payers, and patients to help these stakeholders understand what AI is, how the application was used to create relevant evidence, and highlight positive outcomes from AI-generated learnings.   Consistent with the development and deployment of applications, tools, medical devices, prescription digital therapeutics, and drug therapies in the US healthcare industry, regulation and guidance are needed for the assurance of safety and efficacy of products generated using AI applications and algorithms. The FDA is revamping policies, developing pathways, and proposing a regulatory framework to approve and monitor AI software and technologies as medical devices.11 It is yet unclear how the agency, providers, payers and patients will be able to keep up with products new to the market that have been developed using machine learning and AI applications; in addition, how will stakeholders manage the evaluation of the continuous influx of updates to products already agency approved or cleared? How will software or algorithm updates be tracked? Will the agency require developers to provide notifications to all stakeholders of updates? How will the agency provide guarantees of continued safety and efficacy?   Our healthcare system is rapidly changing-including the focus on value, the push to reduce the cost of care, the need to improve patient outcomes and increasing accessibility for patients putting ever-increasing demands on healthcare professionals and policy makers to work and think differently. AI applications may or may not be a solution. Time will tell.   Kellie Rademacher is Vice President, Access Experience Team, at Precision for Value.  

References

1. Panch T, Mattie H, Celi LA, "The 'inconvenient truth' about AI in healthcare,"

npj Digit Med. 

August 16, 2019. https://www.nature.com/articles/s41746-019-0155-4. Accessed November 25, 2019.   2. Kent J., "FDA sets goals for big data, clinical trials, artificial intelligence," September 4, 2018. https://healthitanalytics.com/news/fda-sets-goals-for-big-data-clinical-trials-artificial-intelligence. Accessed November 25, 2019.   3. Uzialko AC, "Artificial intelligence will change healthcare as we know it," June 9, 2019.  https://www.businessnewsdaily.com/15096-artificial-intelligence-in-healthcare.html. Accessed November 25, 2019.   4. Bresnick J., "IBM Watson Health teams up with hospitals for AI, EHR research," February 20, 2019. https://healthitanalytics.com/news/ibm-watson-health-teams-up-with-hospitals-for-ai-ehr-research. Accessed November 25, 2019.   5. Eddy N., "Partners HealthCare creates funds for AI development, digital tools," October 25, 2019. https://www.healthcareitnews.com/news/partners-healthcare-creates-funds-ai-development-digital-tools. Accessed November 25, 2019.   6. Hale C., "Novartis to put AI on every employee's desk through Microsoft partnership," October 1, 2019. https://www.fiercebiotech.com/medtech/novartis-to-put-ai-every-employee-s-desk-through-microsoft-partnership. Accessed November 25, 2019.   7. Catlin J., "How to choose an AI vendor: 4 questions to answer, December 27, 2017. https://www.lexalytics.com/lexablog/choose-ai-vendor-data-analytics. Accessed November 25, 2019.   8.

Forbes

, "AI and healthcare: a giant opportunity," February 11, 2019. https://www.forbes.com/sites/insights-intelai/2019/02/11/ai-and-healthcare-a-giant-opportunity/#598ae3a54c68. Accessed November 25, 2019.   9. Su J., "Venture capital funding for artificial intelligence startups hit record high in 2018," February 12, 2019. https://www.forbes.com/sites/jeanbaptiste/2019/02/12/venture-capital-funding-for-artificial-intelligence-startups-hit-record-high-in-2018/#5945c60f41f7. Accessed November 25, 2019.   10. Uzialko AC, "Artificial intelligence will change healthcare as we know it,"  June 9, 2019. https://www.businessnewsdaily.com/15096-artificial-intelligence-in-healthcare.html. Accessed November 25, 2019.   11. US Food and Drug Administration, "Artificial intelligence and machine learning in software as a medical device." https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-software-medical-device. Current as of November 5, 2019. Accessed November 25, 2019.