As pharmaceutical manufacturers bring more specialty drugs to market, the need to track adverse events across more geographies and reporting channels increases manual compliance costs. This diverts investment from higher value functions that drive growth, such as safety surveillance and risk minimization.
Between 2014 and 2020, the number of adverse event reports received by the US Food & Drug Administration grew by 84%, with industry analysts projecting global pharmacovigilance spending will grow at a compound growth rate of 11.5% between 2021 and 2028.
For these reasons, pharmaceutical companies are seeking to automate pharmacovigilance (PV) to speed the handling of adverse events, reduce costs, improve patient safety and unlock growth opportunities alongside compliance at scale. They want to move from reactive risk management to proactively predicting and preventing adverse events, while automating mundane tasks such as data entry.
Using AI to automate the pharmacovigilance process
Advanced artificial intelligence (AI) in the form of machine learning (ML) can not only help automate the extraction, classification and entry of adverse event data from source documents; it can also unlock new capabilities such as inferring safety outcomes across geographic and reporting domains and across multiple drugs.
Based on our work with pharmaceutical providers, we’ve found automated PV can reduce drug safety costs by 40% to 60%, allow aggregate analysis of multiple adverse events, reduce delays in reporting events and enable proactive risk management.
We’ve also identified ways in which PV data could be correlated with human genomic data to enable personalized treatment, and help regulators and manufacturers understand the true risks of treatment and reduce time to market for new medications.
However, many automation efforts stall because PV managers are reluctant to turn the tracking of potentially life-or-death adverse events over to automated processes that lack human control, or to algorithms that do not explain how they arrived at decisions such as whether to classify an adverse event as serious.
Many regulators have been reluctant to accept the results of systems whose behavior cannot be validated through deterministic testing that proves they generate consistent, correct outcomes that match expectations. As a result, regulatory groups are recommending and beginning to develop AI regulatory frameworks.
Boosting success of pharmacovigilance automation
As both AI technology and the regulatory landscape evolve, our work with clients has identified five steps pharmaceutical companies can take now to automate the pharmacovigilance process.
- Create an overall enterprise strategy. Allowing technical staff to experiment with the latest AI/ML tools without a defined vision for the desired business outcomes is a recipe for wasted time and effort. Instead, agree on measurable goals and build your automated PV plan around them.
Identify the right stakeholders and a sponsor who can help not only formulate strategy but also get the funding needed to implement it. While these steps are common to any business initiative, they are particularly important when implementing significant change such as the use of AI to automate drug safety monitoring.
Set goals that go beyond speeding current processes or reducing their cost — instead, consider AI and ML’s potential to achieve outcomes that would not otherwise be possible, such as enabling proactive risk management and even personalized treatments. Leverage your existing processes and relationships with regulators to ensure that AI/ML-enabled drug safety will protect patient health.
- Start small with robotic process automation (RPA). RPA is a relatively fast, low-cost and low-risk way to begin automating PV. Start with easily repeatable tasks — especially in the early, data collection steps that require extensive manual work and account for a large share of PV costs. This can allow pharmaceutical companies to monitor more drugs and adverse events with less resources, freeing staff for more value-added tasks.
For example, we helped a leading European pharmaceutical developer create a global operating model to increase the efficiency of processing more than one million adverse events per year. Among its features was an industry-first bot for case safety processing that saved 72,000 person hours of work per year and enabled 45% efficiency gains.
Our solution reduced end-to-end cycle time for processing adverse events by 30%, substantially reducing costs, increased first-time accuracy between 85% to 90%, and achieved 99.8% compliance for Individual Case Safety Reports regulatory submissions.
- Keep humans in the loop, at first. While pharmaceutical companies and regulators work to increase confidence in fully automated PV, pharma organizations can assign confidence scores to the results provided by automated PV systems. Begin by conducting a business process analysis of PV processes that don’t require additional, extensive human judgment, such as a doctor’s review to determine the importance of an event.
As confidence levels rise, humans may need to be involved less often, especially as pharmaceutical companies and regulators agree on more formal guidelines for the use of AI in PV.
- Consider tapping wider datasets to predict future events. Historically, companies have analyzed only adverse event data for their own products. However, for the last four to five years they have also been able to use quarterly data from the FDA to analyze adverse events associated with products from multiple companies, allowing them to conduct comparisons for more drugs to better predict future adverse events.
- Create a framework to evaluate and explain the business case. With the help of an advisory group of PV stakeholders, including the chief medical officer, head of safety operations and head of safety risk management, define “early win” use cases for PV automation with estimated costs and benefits.
Work closely with the advisory group to prioritize cases where, for example, staff can be freed from routine repetitive work for higher value activities. Explain in detail potential risks and delays. Where appropriate, go beyond cost savings to more strategic benefits such as faster processing of regulatory reports, improved compliance with reporting deadlines, and higher quality information about adverse events.
Pharma companies that automate PV process will not only speed the handling of adverse events and reduce costs; they will also move from a reactive to a proactive stance when it comes to predicting and preventing adverse events. By doing so, they can position themselves for greater insights into patient safety and growth opportunities for the foreseeable future.
Visit our pharmacovigilance webpage for more information about automating PV processes.