Rapid innovation in the life sciences industry has led to faster product launches and an increase in the volume of product-related content. In 2021, the life sciences industry, witnessed an average of 3x increase in content production, thereby putting pressure on organizational budgets. To add to this, information consumption patterns of healthcare professionals (HCPs) and patients have moved toward on-demand, personalized, and real-time digital content. There is an increasing trend of HCPs and patients to decide on the “What, When, and How” of brand experiences. As per the Digitally-Savvy HCP survey report by Indegene, 62% of HCPs stated a preference for relevant personalized content for more insightful interactions.
Source: Veeva Connect Summit 2021
Source: Digitally-Savvy HCP survey report by Indegene
67% of companies think that their organization has ambitious digital plans but only 27% are satisfied with their company‘s progress.
The Medical Affairs Digital Strategy Council and DT Consulting, an Indegene consulting business, recently conducted a survey on the digital preparedness of Medical Affairs businesses. The survey results revealed that with respect to digital strategy and execution, 67% of companies think that their organization has ambitious digital plans, but only 27% are satisfied with their company‘s progress.
Organizations are increasingly realizing the importance of relooking at their MLR review processes in order to meet the ever-increasing volumes of content. They acknowledge the value that artificial intelligence (AI) can bring into the processes in taking away the mundane tasks from the reviewers and helping them focus more on new content that comes in. Within the content supply chain, the review and approval processes lend themselves relatively more towards automation when compared to the other processes, given the proportion of repetitive and rule-based work in review and approval.
Typically, the industry takes up to 40 days to complete the review/approval process right from the draft content (promotional materials) to the distribution stage. Today, given the amount of time that review occupies in the supply chain, the opportunity to shorten the time to market is significant.
Most companies understand that there is a consistent need for new ways of incorporating data science in the ways of working. Currently there are new ways that are specifically addressing the quality and the quantity of data that comes through. AI will be the best bridge to be able to convert that data into business value.
Head of Commercial Promotional Review and Marketing Operations (US/Global and Japan) Alexion Pharmaceuticals
With automation and AI progressively taking over marketing functions such as orchestrating campaigns and learning from market feedback, the future is moving toward “creation to distribution” of content in <24 hours. As customers interact with the campaign, AI will be able to pick up the right content modules to assemble tactics from a bank of approved modules, respond dynamically, and tweak campaigns on-the-go based on content effectiveness scores.
AI, if implemented right, can take over 80% of review and approval responsibilities, while reviewers focus their efforts on high-risk content and cocreation of de-novo content.
AI, trained to mimic human reviewers, will evolve over the next few years and can be broadly charted across 2 phases (Figure 3).
Figure 3: The roadmap for MLR of the future can be segmented into 2 phases
In phase 1, AI can be used to address specific point problems or to automate repetitive error-prone tasks, like in the following cases:
Figure 4: Claims validation against approved repository
Figure 5: Content comparator highlights changes compared to previous versions/original content
Figure 6: Phase 1 frees up substantial capacity for reviewers
In a few years, AI will mature to “general intelligence” applications and cross the barriers of having to decipher the medical meaning and contextualize business rules into review checks. In Phase 2, AI can be deployed in the background to carry out review checks on assets being created much like how Google today predicts what we might be typing in the search bar in real-time. What it means is that the author will receive real-time alerts to correct any errors while the reviewer is working on the asset, as AI continuously scans the document acting very much like a human reviewer. Following are a few examples:
Figure 7: Verification of data for factual and scientific accuracy
Figure 8: Suggestions on sentence structure to bring the right medical meaning
While in Phase 2, AI will be ready to take over a majority of review decisions, a concept of “Risk Score” can be used to determine if the asset needs to be reviewed and approved by a human reviewer. A risk score is an indicator of the compliance and quality risk an asset is likely to carry after AI has conducted checks, as well as the subsequent corrective actions made by the author. A low-risk asset is one that is reviewed and approved by AI to meet all the required compliance and quality standards. A high-risk one is where there are deviations to the approved content and is perceived by AI as to requiring human judgment for approval. The process of risk scoring is one that would involve ML to evaluate risk as it reviews multiple assets and learns from the differences between decisions made by itself and that by a human reviewer (Figure 9).
Figure 9: Risk scores can be used to determine if content needs to be reviewed and approved by a human reviewer
Figure 10: Phase 2 enables reviewers to play a key role in content strategy
The role of technology in automating the review function will not stop at creating content right the first time but will expand in scope to solve for
Digitization of the knowledge base, IT infrastructure, and processes to maintain the knowledge base and development of AI models will pave the way to full-scale AI adoption in review functions.
For AI to automate review, it needs a knowledge repository of approved claims, data facts, references, clinical and scientific evidence bank, and regulatory and business rules, among others, to conduct checks on the assets. It will also require models that can separate the necessary data for extraction from noise, data pipelines to store the extracted data in a query-able repository, and a model that can use the stored data to conduct asset reviews accurately.
For example, a case for automation of data fact-checking in assets can be divided into
The collection of such use cases, AI models, and repositories will help propel organizations toward Phase 2.
In the current landscape, the majority of the burden of mundane tasks rests on the shoulders of review teams, but by shifting the review function and the responsibility onto AI, the role of reviewers will transition to be more consultative in nature.
As organizations adopt technology in review, reviewers will play the role of subject matter experts and help design the AI models, train the models to perform better, improve accuracy through vital supervised learning inputs, conduct quality audits, and provide the crucial human judgment required for high-risk material flagged by AI. A few examples could be
It is a journey and a long-term aspiration plan to make sure that we build our confidence in AI. It is not something that will happen overnight, though, we need to consider how can we facilitate it. We need to be partners and engage in proof of concepts, experiments, and try to see how can we help. We need to be in a place where we trust AI and machine learning and natural language processing to truly not require second tech afterward. But that will take time, not only from the AI capabilities point of view, but also from our team‘s engagement and building the trust with this platform. Experimenting with technology will to some level help us move forward towards big steps in the next few years.
Head Global Oncology Medical Information and Review, Takeda
Think about where do you want to be in the long term, and start with small baby steps. Each time you move forward, you bring incremental value. And on the change management if you can find a value proposition for medical review teams, then that makes it significantly easier to move forward.
Associate Director Content Factory, Eli Lilly and Company
When you bring something new, some people are going to embrace it right away, while with others it will take a little bit of time. So change management and putting in some really well thought out and processes around it is a very important aspect.
Director, Marketing Operations, Gilead Sciences
To manage the change as organizations progress through Phase 1 and Phase 2, they need to start drawing a roadmap for AI adoption in reviews; identifying use cases for automation; prioritizing them; and mapping capabilities, gaps, and timelines for executing the plan. In parallel, processes and review flows need to be redrafted to weave in automation use cases and reviewers trained to work side by side with technology. That said, change management from a human-acceptance perspective is also going to be a key aspect that organizations will need to consider. Teams will need to be involved in the process right from the strategy phase so that they are able to see the value that AI can bring into their day-to-day review tasks, and how their roles are going to transition towards high-impact and strategic ones, while AI takes care of the mundane tasks for them. By involving teams in the change process from the beginning, acceptance of change can be accelerated; moreover, the machine can be trained faster to tackle more real-world challenges through on-the-ground perspectives that the teams can bring into the picture.
Content consumption patterns in the life sciences industry are changing at a very fast pace. Organizations are already seeing and acknowledging the value that AI can bring to their MLR operations, though they are at a juncture where they need a guiding path to take the next step forward. Although adopting AI into the MLR processes might not happen overnight, a phase-wise approach where organizations can transition from manual reviews to AI-assisted reviews, and finally to AI-conducted reviews, which would allow teams with enough time to prepare their organizations to trust the technology as well as to adapt to new ways of working where AI takes on the repetitive tasks, and review teams transition to more strategic review roles.
Note: All quotes in this whitepaper have been sourced from the "Webinar- MLR of the future: Accomplish review, approval, and delivery in <24 hours"