Artificial Intelligence (AI) technology is showing promising growth in the life sciences and healthcare industries. Companies such as Bristol Myers-Squibb have recently announced AI partnerships, and the global AI healthcare market is set to hit $13 billion by 2025.1
To ensure that AI reaches its full potential for biopharma, it is essential that our use cases mirror the full scope of what AI technology can achieve. Currently, AI is sometimes applied in too restrictive a fashion, which limits its transformative potential. Within healthcare, for example, AI is mainly used to group patients into different categories (e.g. subtypes, risk profiles) which are then used to find the best treatment for that patient group. This is what we call "stratification medicine". By doing this, we mainly focus on urgent cases where patients are already ill, based on direct observation of their symptoms. However, this doesn’t use the full power of AI technology. AI can learn from vast quantities of data, detecting hidden patterns and predicting future occurrences based on that data. Essentially, AI can predict whether patients will be ill in the first place, and what treatment would be best suited to them. This is precision medicine, and the difference here is illustrative of the mindset change required across the healthcare ecosystem, to capture the potential of AI.
From policy to thought patterns, we must overcome a number of key, cross-industry challenges to enable greater impact for AI in the life sciences and healthcare industries.
Firstly, we must consider how AI could contribute to an improved pharmaceutical gold standard, and how that impacts our ideas of adequate reporting, for the benefit of patients and businesses. Traditionally, the biopharma industry relies on one-off clinical studies, which are validated with traditional statistics, based on averages in selected patient groups. These studies are not representative of the population at large; for example, only 2-4% of adult cancer patients participate in clinical trials, and they do not represent the complexities of the real-world environment. With AI, we can move away from the average to the individual, and include all cases instead of just the few that fits the model. This will ensure that every patient receives the best possible treatment. In this scenario, AI takes a supportive role to existing analytics teams, ensure that data is processed more efficiently, and is more readily available. The commercial side of pharma are the first movers here; our work at OKRA Technologies started on the commercial side, where the business case for increased efficiency is clear. We are now also working with R&D departments, and are moving to collaborate at the point of care.
Secondly, we face technical challenges, notably a lack of data integration and lack of access to data. AI technology can integrate data that ranges from genomic to clinical data, both structured and unstructured, and this requires great technological sophistication. As part of the High-Level Expert Group on AI, I am part of developing recommendations and frameworks that would support more effective use of and access to health data. Stakeholders must collaborate across the industry, which is challenging, but I was encouraged to see a range of biopharma technology champions come together, as they recently did for a multi-stakeholder AI healthcare workshop in Brussels, co-hosted by OKRA. With respect for different interests, and for data privacy, new frameworks and collaborations are in the making.
Thirdly, as a direct user of AI, you may feel that AI systems do not give enough "reason" or explanations behind their output, or worry that your job will be replaced by new technology. To the first point, many companies are responding to the need for transparency: at OKRA, we have built-in reasons behind every recommendation, explaining what data it draws on and why a certain decision would benefit the user or patient, so they can be confident in their decisions. To the second point, the key here is that AI will support professionals in their current work; speeding up, improving accuracy and precision, and eliminating routine tasks. This has come at the right time, as pharmaceutical executives face intense pressure to compete and greater efficiency is needed. However, it seems some life science companies are attempting to build their AI...
(All Rights Reserved)