Unlocking the Power of Safety Data with Cloud Technology

Brian Longo
Veeva Vault Safety

oday, the increasing number of data sources (clinical data, electronic patient health records, call center notes, medical literature articles, emails, and social media) means that there are more adverse event reports (AERs) to intake, process, and analyze for safety signal detection. Combination therapeutics, targeted medicines, and global, multi-market product launches with frequent regulatory changes keep adding layers to this information. These challenges are driving the need for drug safety organizations to more efficiently process and analyze a growing volume of complex information while continuing to meet new or evolving health authority requirements.

Many companies outsource the labor-intensive task of case intake and processing, creating a barrier between pharmacovigilance (PV) organizations and their data. Keeping safety operations in-house requires significant resources for repetitive manual data entry tasks in case intake and processing, which limits time for data analysis. Systems that store AERs are often not well-integrated with other safety applications and require a lot of overhead to keep up with new regulatory requirements such as data standards, or application versions. These challenges prevent companies from better leveraging their resources and maximizing the value of their safety information.

“We need to efficiently process the growing volume of data on a timely basis not only to meet reporting requirements but for better data analysis and detection of new signals,” suggests Ed Tucker, COO, Acerta Pharma B.V. (a member of the AstraZeneca Group).

Cloud technology is bringing these data out from behind their firewalls and making them more easily accessible to innovative technologies like artificial intelligence. By providing a foundation to aggregate learnings across industry, safety organizations will realize the greatest value from their data. By minimizing time and resources needed to manage and maintain safety applications, life sciences companies can instead focus on analyzing this information and turning its treasure troves of data into valuable insights.

Turning Unstructured Content to Structured Data for Analysis

One of the biggest challenges to analyzing safety data is the multiplicity of sources of unstructured information. AI-powered cloud solutions significantly reduce the overhead and manual data entry during case intake and processing. Using natural language processing (NLP), AI solutions automatically identify, extract, and convert text from both structured and unstructured data sources into the required fields for a drug safety case. Leveraging global libraries such as MedDRA and WHODrug, medical information is also coded and associated with a confidence score. With many required fields automatically populated, medically trained professionals become more efficient at verifying information instead of spending significant time manually entering data. Accuracy of extracted information will improve over time with machine learning and more data. Automatically recognizing identifiable patient and reporter, adverse event/reaction, suspect or interacting drug, and even detecting seriousness, AI solutions help accelerate data entry and increase efficiency – enabling processing of more information.

Better Prioritization and Risk-Based Approaches to Managing Adverse Events

With the number of AERs increasing, it is important to quickly prioritize them. Frequency of duplicate reports is growing due to multiple reporting sources, making it more difficult to accurately detect and evaluate patient safety risks. Almost 20 percent of the data in FDA’s adverse event reporting systems (FAERS) are duplicates. Machine learning (ML) can help determine if a report is a duplicate versus a follow up, linked to an existing case, or is a new report, enabling safety teams to be more efficient in case processing.

Risk-based approaches such as failure mode and effects analysis (FMEA) can also be applied to these data for better prioritization. Adverse event risk is calculated using a combination of severity, likelihood of occurrence, and ability to detect events. Less serious AERs can be automatically processed with minimal human oversight, while ones with greater risk are prioritized first for review and can even trigger a notification to a case processor.

“Accurately assessing risk allows us to better prioritize adverse events and allocate resources,” explains Betsy Reid, COO, Paidion Research, Inc. “More time can be spent analyzing highest risk events, while continuing to support timely processing of all cases.”

Continuously Improving AI with Industry Learnings

AI requires data, and the more data it gets, the more it learns. With ML, the AI engine gets smarter from learnings across an entire industry instead of from just one company. Cloud-based safety applications with comprehensive application programming interfaces (APIs) are easily integrated with other information sources to feed that AI engine with new data. More data enables refinement of existing, or development of new, data models to identify trends and relationships, and enable predictive analytics.

Safety also interacts with different functional areas throughout the PV lifecycle, which makes it even more crucial to enable seamless workflows. For example, integrating safety and EDC (Electronic Data Capture) applications enables real-time adverse event (AE) reporting. Additionally, as an adverse event is processed, continually feeding information to the AI engine at different decision points in the PV lifecycle will significantly improve the accuracy and quality of future upstream events. Eventually, more advanced decisions such as ranking adverse events, triaging, and suggesting determined expectedness, are possible.

Focusing on Strategic Data Analysis

Many companies spend a disproportionate amount of time and budget maintaining their safety systems and integrations instead of on critical data analysis. Upgrading systems require significant resources to install, deploy, test, and validate. With frequent regulatory changes, it is not uncommon for pharmacovigilance IT budgets to be entirely spent on keeping their applications up-to-date.

A cloud platform provides complete end-to-end safety solutions that are easy to maintain and upgrade. Integrating the submission process into the safety solution, for example, eliminates installation and setup of a submission application and gateway, or work with third parties to do so. PV teams can manage AERs from case intake to processing and submissions in a single system that includes full tracking and audit history. By removing silos and manual, disconnected steps, safety organizations can reduce errors and cycle times, and improve the quality of safety cases.

“With modern, cloud-based safety applications, we don’t have to manage gateways, databases, servers, or external reporting tools. I can spend more time analyzing instead of managing the data,” says Marc Morris, Vice President of Safety and Pharmacovigilance, Apellis.

More Efficient, Scalable, and Complete Safety Reporting and Analysis

PV organizations play a critical role in ensuring patient safety. With the number of reporting sources and the complexity of adverse event management continuing to expand, it is challenging for safety teams to efficiently process cases and scale. AI with NLP can quickly and automatically convert large volumes of unstructured information sources to structured data, improving the ability to accurately track and analyze adverse events to identify trends and detect safety signals.

The cloud makes data more accessible to innovative technologies and applications. The combination of AI and ML delivers platforms for accelerated, continuous learning from throughout the industry. As AI matures and risk-based approaches are more widely accepted in adverse event management, less serious AERs can be processed automatically with minimal human oversight, enabling PV teams to focus on AERs of greatest risk. With easy-to-maintain, cloud-based safety systems, companies can allocate more resources to data analysis and focus on what’s most important: patient safety.