There’s a catch-22 in the great AI debate. You’re unlikely to get a competitive, early-mover advantage unless you invest confidently in AI, but you’re reticent to invest until its success is proven. By which point your horse has bolted.
Healthcare is an industry that deals daily with big data sets of immense magnitude and it’s here that this conundrum feels particularly prescient. Waiting to formulate an AI strategy until more examples of its benefits are made apparent by AI-leaders is not advisable - by then it may be too late to catch up.
Take the threat that Medica faces from Google Health, for example. Medica provides tele-radiology reporting services to the NHS trusts that do not have sufficient capacity to carry out all reporting in-house. If Medica is not able to quickly implement AI in its service offering, and if Google was to provide its AI model for reading mammograms to the NHS, Medica’s business model would be at risk.
The healthcare data market
The core focus of the healthcare industry for the last few decades has been on productivity improvements through the use of technology. High throughput screening, genomics, in silico drug discovery, these are all techniques designed to reduce failure rates and drive better returns on investment. However, nothing has worked on a large and sustainable scale. Enter: AI and machine learning.
In 2020, the amount of health data is expected to double every 73 days[i]. The global market was worth $2.1bn in 2018, with exponential growth to $36.1bn predicted by 2025, at a CAGR of 50.2%[ii]. Investment is growing, too. VC investment in AI for healthcare totalled $1.3bn across 103 deals in 2017[iii]. VC backed deals and financing to healthcare AI startups grew to $2.7bn across 264 deals in 2018[iv]. In the first three quarters of 2019 alone, they reached $3.11bn across 261 deals, demonstrating strong year-on-year growth in investments in this field3.
Indeed, it already feels like a crowded space. There are nearly 200 start-ups using AI in drug discovery, and the competitive nature of this space will make it difficult to predict which companies and technology platforms will be successful.
Nevertheless, the opportunities are numerous and the reasons large pharma is pushing into this space are clear. AI confers obvious cost-saving advantages: faster drug discovery, including quicker identification of unmet medical needs; quicker lead identification and candidate optimisation; quicker ways to identify areas for drug repurposing; faster regulatory approval and speed to market. Add to that the advantages conferred by personalised patient benefits like earlier diagnoses, better treatment outcomes, better patient engagement, optimised operations and personalised medicines. Clinical trial can be managed better, medical imaging and diagnoses are easier to create, and better decision-making in the clinic is altogether faster and more efficient.
A short history of AI in healthcare
There are of course plenty of examples of healthcare companies that have already taken the AI bull by the horns. We outline a couple below – contact email@example.com to purchase the full report where you will find further examples and also a look at the AI-savvy companies for whom we see big things happening this year.
AI in drug discovery
Insilico Medicine is a US-based bioformatics company which announced in September 2019 that it used AI to design, synthesise and validate a novel drug candidate in 46 days – 15 times faster than the best pharma companies. The company developed GENTRL (Generative Tensorial Reinforcement Learning), a new AI system for drug discovery that was used to design a novel DDR1 kinase inhibitor from scratch in 21 days, then to synthesise and pre-clinically validate it in 25 days, a combined process that typically takes the pharmaceutical industry c.2 years.
In the initial 21 days, the AI program generated 30,000 novel small molecules that may work against fibrosis. In the subsequent 25 days, Insilico screened out and synthesised the six most promising compounds and performed in vitro tests for selectivity and metabolic stability. The lead candidate was then tested in live mouse models, where it displayed favourable activity.
AI in clinical trails
An analysis of clinical trial data from January 2006 up to December 2015 found that only 9.6%% of drug development programmes successfully progressed from Phase I trials to FDA-approval[v]. Most trials fail because the intervention does not demonstrate efficacy or safety, but other factors include flawed study design, participant drop-outs or incompliance, shortage of money, or a failure to recruit enough participants. AI has the potential to save billions of dollars through increasing the efficiency of clinical trials.
Take Deep 6 AI, which develops solutions to find patients for clinical trials. Researchers at the Cedars-Sinai Medical Center recruited two patients in six months for a cardio study. Using Deep 6 AI’s system, they identified 16 patients in 30 minutes, with eight patients recruited onto the study three weeks later. Such productivity gains allow small departments to increase the number of studies they can perform. The team of three recruiters at Cedars-Sinai is now able to find patients for 30 clinical trials in a year. Before utilisation of the Deep 6 system, the centre had one recruiter, who did one study a year.
AI in clinical decision making
Through utilising the vast amounts of genomic, biomarker, phenotype, behavioural, biographical and clinical data that is generated across the health system, AI has the potential to transform clinical decision-making processes.
For example, a Bayer and Merck & Co partnership is developing an AI-driven software to support clinical decision-making of chronic thromboembolic pulmonary hypertension (CTEPH). CTEPH is rare form of pulmonary hypertension, and its symptoms are very similar to those of other conditions like chronic obstructive pulmonary disease (COPD) and asthma, which can hinder the diagnosis of CTEPH. The CTEPH Pattern Recognition Artificial Intelligence received FDA Breakthrough Device designation in December 2018.
The tool uses machine learning to comb through image findings from pulmonary vessels, lung perfusion, and cardiac check-ups, as well as the clinical history of the patient. Bayer and Merck hopes the system will enable radiologists to analyse these diagnostic images faster and identify patients with CTEPH earlier, more efficiently and more reliably. The resulting earlier diagnosis of CTEPH would allow for the earlier use of therapies, benefiting patient care.
Overall, the software could support physicians with the intricate diagnostic decision-making process of CTEPH, which would benefit the physicians, patients, and also healthcare systems via earlier diagnosis.
We have looked at the universe of companies that we follow, outlining those that we feel are employing AI and where its application should have a meaningful benefit. These companies can be seen detailed in our full report. Please contact firstname.lastname@example.org for more on how to purchase the full report.
Talent and infrastructure
Usage of AI is likely to become a greater differentiator in the next 5-10 years, which means that companies should now begin building their AI infrastructure, or risk falling behind. Indeed, AI-experts are not an unlimited resource, especially given that most of the talent in this sector naturally gravitates towards traditional IT and technology companies. Even in companies dedicated to AI in healthcare, AI-experts typically only comprise c. 15% of staff[vi]. Thus, AI talent is a valuable and limited resource.
Meanwhile, building AI capability is not as simple as buying some new computer equipment. AI-infrastructure needs to be integrated across the entire company and have collaboration and buy-in from multiple departments. It may also require process changes in how data is procured and stored.
Another important motivator for building this infrastructure now is the inevitable entry of Big Tech into the healthcare field. Google, Microsoft, Amazon, Apple and Facebook have already announced different initiatives to enter the healthcare space - companies which will have fundamental advantages in terms of AI capabilities.
As such, waiting for more examples of AI’s efficacy is not advisable, companies should build their AI-capabilities now. Companies where Big Data is prevalent and which are not developing AI solutions and articulating the way in which they are being deployed in their businesses will, in our opinion, be competitively disadvantaged in the mid to longer term.