Published in PME Magazine
In the face of the coronavirus, innovative technologies such as artificial intelligence (AI) and machine learning (ML) have proven vital for the protection and promotion of public health.
According to a report by the British Medical Journal (BMJ), soon after COVID-19 was declared a global pandemic, the World Health Organization (WHO) indicated both AI and ML could be important tools for managing the crisis, triggering intense research and development in these fields. While a vaccine was being developed, AI played a significant role in helping to mitigate the impact of the pandemic and curb the spread of the disease through various means, including early detection and diagnosis, automated monitoring of affected individuals, contact tracing and the identification of clusters or ‘hot spots’, as well as the modelling of future cases and mortality. AI has also been helpful in facilitating research into the virus itself, improving both treatment regimens and prevention strategies, as well as supporting in the development of various drugs and vaccines. For example, it contributed significantly to the speed and success of the COVID-19 vaccine effort, particularly via the screening of molecules as potential drugs, both novel and repurposed.
The AI revolution powering healthcare
Beyond the pandemic battlefields, AI is revolutionising processes right across the life sciences industry, from the automation of medical literature reviews to drug re-innovation and re-purposing. Many leading manufacturers are using AI technology to advance their work and value proposition. For example, we are seeing AI used in drug development to predict which drugs are more likely to be effective against a specific target by rapidly screening drug compounds and predicting success rate based on biological factors. This will undoubtedly accelerate the drug development process, allowing for the faster discovery of drugs. The ability of AI to filter through data quickly also widens the opportunity for molecule optimisation and drug repurposing. For example, Novartis is using AI technology in the drug development process by employing ML algorithms to classify digital images of cells, each treated with different experimental compounds. The company then uses this data to predict which untested compounds are worth investigating in more detail, saving considerable time and costs.
"While a vaccine was being developed, AI played a significant role in helping to mitigate the impact of the pandemic and curb the spread of the disease"
Pharma companies are also beginning to use AI to revolutionise their clinical trial design and development processes. For example, Roche recently signed an agreement with British data analytics firm Sensyne Health, which will see them use AI technology to analyse anonymised patient data and electronic patient record information. The ability of AI to analyse these vast data sets at speed will support Roche in its clinical trial designs and further strengthen its approach in data and technology-driven personalised healthcare. Similarly, Pfizer and Concerto HealthAI, a Boston-based company that provides AI-driven solutions to life sciences researchers and healthcare providers, began a collaboration in 2019 to develop new study designs using real-world data and AI for various cancers. By using AI and machine learning technology to analyse medical records, insurance claims and economic measures, they aim to design more effective trials, identify and target potential patient populations, and ultimately help people living with some of the most devastating cancers.
Next level patient support
Away from the laboratories, life sciences companies are taking patient-centricity to the next level by using AI technology to assist with patient support programmes and improve compliance. In a study conducted for AbbVie, AiCure demonstrated that by using its AI platform, it was able to increase treatment adherence in patients with schizophrenia from 50% to 90% over a six-month period. The AiCure platform (via smartphone app) visually confirms when patients have ingested their medication correctly. This AI-driven method was found to be more effective than standard self-reported data and pill counts, as there was no margin for human error or tampering.
AI-powered mobile apps are also being used to support patients via teleconsultations, e-prescriptions and dosage reminders. The Roche app, mySugr, allows diabetes patients to access all relevant data regarding their condition in one user-friendly app. Patients can more easily log their blood glucose as the app automatically syncs their levels from their Accu-Chek meters. In addition, patients can manually add food diaries and photo updates. The AI technology also allows daily, weekly and monthly data analysis, helping patients to monitor and manage their condition and provide motivating feedback and personalised challenges to help patients achieve their treatment goals.
Physician perceptions and willingness to adopt AI
With seemingly limitless possibilities, AI has the potential to truly revolutionise patient healthcare. However, understanding perceptions of this technology is essential to successful development and adoption. In partnership with Medefield, we recently ran a poll with 70 physicians including cardiologists, rheumatologists and gastroenterologists across Europe (France, Italy, Spain, Germany), the UK and the US, to gauge their level of openness, concern and perceived benefit associated with AI-driven healthcare solutions.
According to our study, over 70% of physicians consider AI-powered patient support tools, including mobile apps and wearable devices, to be valuable for healthcare and likely to satisfy patient needs. One US cardiologist commented that AI “could provide objective metrics on patients” while another talked about a “potential role in using AI to identify gaps in care and to help with patient compliance”. Furthermore, 64% of physicians believe their patients would be accepting of and likely to use AI-powered support tools should they become more widely available. However, physicians in the US felt much more strongly about this, with 28% anticipating patients would be ‘extremely likely to use’ AI tools versus just 8% in Europe.
In the practice setting, 72% of physicians consider AI tools aimed directly for use by clinicians (including programmes to manage administrative tasks and diagnose patients) as valuable and likely to benefit them. Both the logistical benefits, in particular billing (“Billing, billing, billing, billing, billing!”, said one US cardiologist) and therapeutic advantages of AI systems were identified by physicians, with the latter mainly focused on the diagnostics and treatment solutions likely to help patients. For example, one rheumatologist noted the key benefits of AI were the “ability to look for and identify rare and unusual diseases, help evaluate early subtle radiographic changes to predict the future course and evolution of the disease, and to help select the optimal therapy”.
Despite this generally positive perception, physicians’ readiness to adopt these tools into their practice was very mixed. Only one in three physicians said they would adopt these tools as soon as they were offered. The remaining two-thirds of physicians were far more cautious, saying they would wait until at least a few months after launch, and another one in three physicians stated their need for peer recommendation or concrete real-world evidence prior to adopting the AI tools. There may be many reasons for this, not least the generational differences in trust and comfort of modern technologies. For example, one gastroenterologist based in Europe noted that AI is “for the young, not for the old” while another said, “for young patients, AI will be useful, but not for older patients”. Likewise, one physician commented: “I consider it like a gadget which is not able to help me. It needs more and more adaptations. A little bit like our computers, which are so different [now compared] to what they were like 35 years ago.”
Looking to the future
Despite recent innovation in healthcare-based AI, most solutions remain in their infancy, far from being standardised and widely used. In the coming months and years, we predict AI will continue to become integrated into healthcare systems in innumerable ways, becoming increasingly patient-facing. Like the COVID-19-induced paradigm shift in vaccine development timelines, AI will revolutionise healthcare by optimising both efficiency and productivity alongside quality of care. Arguably, the utopian vision is that AI technology and data sharing are so sophisticated, integrated, widespread and regulated that successful preventative medicine is normalised, both on an individual and population level. Although this is a long way off, each shorter-term AI solution adopted across the healthcare industry moves us closer to that goal.
On an administrative and operational level, AI tools automating appointment scheduling and billing can reduce costs, delays and errors, liberating resources for direct patient care and improving the patient experience. On a clinician-centric level, extended reality (XR) can improve how specialists are trained, for example allowing surgeons to gain confidence before treating patients. XR can also play a direct role in patient treatment, for example enhancing cognitive behavioural therapy (CBT) in the treatment of anxiety disorders. Patient treatment can also benefit from a range of other AI solutions, such as data from wearable devices enabling disease identification, monitoring and treatment personalisation. This is especially relevant in chronic conditions like diabetes or cancer, which are hugely burdensome globally. More generally, AI solutions can increase the analysis speed and accuracy of many standard diagnostic tools such as MRI or histopathology, allowing clinicians to spend more time interacting directly with patients.
There are, however, important obstacles to the widespread implementation of such AI tools in healthcare systems. Acceptance and trust, by both clinicians and patients, is a central challenge. Closely linked, the success of AI requires extensive and diverse data pools. Standardising and regulating both how this data is collected and shared is a critical barrier. Overall, only strong evidence of the effectiveness of these tools, as well as the cooperation of healthcare professionals and patients in collaborating with creators and the systems themselves, will enable the AI revolution. Cooperation will be achieved when the exact needs of healthcare systems, clinicians and patients are listened to, understood and catered for with user-friendly and practical solutions.