Updesh Dosanjh, IQVIA04.05.23
As the regulatory framework for devices continues to evolve, we are seeing a greater emphasis on Post Market Surveillance and the appropriate management and reporting of adverse events. Whether we call it PMS, Safety or Pharmacovigilance, a common theme is that there is significant work to transfer information from one format to another with all the inefficiency that entails.
When it comes to medical device safety, the time and effort necessary to manually search for adverse events and enter information into management systems is a distraction from focusing on data that would enable organizations to deliver safer, more effective products to market. As individuals increasingly report adverse events the need to sift through large data sets is only growing in volume and complexity, driving costs ever higher and increasing the risk that something important will be missed.
As margins tighten throughout healthcare, and in increasingly competitive markets for therapeutics, organizations can ill afford to spend many hours each month manually managing reports and searching for trends of potential adverse events. Fortunately, advances in artificial intelligence (AI) and machine learning (ML) have made it possible to automate as much as 80% of this critical safety work, freeing teams to focus more effort on analytics and product improvements.
The challenge is that reports of adverse events can be found in a wide range of datasets. There are enterprise applications such as customer relationship management (CRM), quality management, and safety systems and peer-reviewed literature and on top of that, data comes in a range of formats, including audio and video files, emails, PDFs and Microsoft Office documents. In some cases, this data is structured, but increasingly it is not – especially if it is coming from outside the organization.
This complex mix of data sets and formats leaves teams with the unenviable task of managing potentially thousands of data points for adverse events. Even if this data has been compiled in a spreadsheet or database, significant manual effort is necessary to look through it all, flag adverse events and follow up on them.
In many organizations, this takes hundreds of hours per month. It forces teams to take a reactive approach to their work, and it leaves little time to dive deeper into data sets to identify patterns, generate insights and take more proactive steps to mitigate risk or improve safety or devices.
Given the narrow window in which adverse events must be reported to regulatory agencies, these volumes must be processed quickly. Not only does automation expedite the process; it also does a more thorough job that manual reviewers facing deadline pressure. Web based, collection tools with smart querying allows organizations to receive high quality and more complete information by asking smart questions of the reporter. This allows cleaner data to be easily sent in an ingestible format to safety teams.
Here, several forms of automation can play a pivotal role. Optical character recognition (OCR) can convert handwritten or typed text into machine-readable text, while natural language processing (NLP) can look for patterns or context that may refer to an adverse event. Beyond simply extracting data, this combination of technology tools can also convert this largely unstructured data into cases within a case collection system. All told, leveraging OCR and NLP for intake can cut the manual workload by 70%.
Further use of automation has two key benefits. First, it helps Safety automate up to 80% of its current manual workload. As discussed, this is an important step in enabling teams to transition away from completing repetitive administrative tasks and reactively responding to adverse events that have already been reported.
The second, related benefit is the ability to transition to proactive safety and investigation. By spending significantly less time on detection, collection and intake, teams are better positioned to leverage the structured adverse event data that AI tools generate to gain real-time insight into the safety process. This lets organizations get ahead of potential compliance concerns, identify and mitigate risks before they become large-scale problems and even detect adverse event signals as they are occurring.
Medical device companies are continually searching for ways to empower their employees to devote more time to the higher-level work they are passionate about doing. Automating routine manual tasks can play an important role in making this happen. The benefit is acutely felt in pharmacovigilance, where reducing manual workflows by 80% makes it possible to focus on the end-to-end safety process, not repetitive data entry – and to ensure that the devices a company brings to market are safe and effective for as many patients as possible.
Updesh Dosanjh is Practice Leader, Pharmacovigilance Technology Solutions at IQVIA.
When it comes to medical device safety, the time and effort necessary to manually search for adverse events and enter information into management systems is a distraction from focusing on data that would enable organizations to deliver safer, more effective products to market. As individuals increasingly report adverse events the need to sift through large data sets is only growing in volume and complexity, driving costs ever higher and increasing the risk that something important will be missed.
As margins tighten throughout healthcare, and in increasingly competitive markets for therapeutics, organizations can ill afford to spend many hours each month manually managing reports and searching for trends of potential adverse events. Fortunately, advances in artificial intelligence (AI) and machine learning (ML) have made it possible to automate as much as 80% of this critical safety work, freeing teams to focus more effort on analytics and product improvements.
Manual Workflows Slow Down Adverse Event Reporting
Adverse events are important to the life sciences industry for two important reasons. One is regulatory, as companies must report adverse events to the U.S. Food and Drug Administration, EU, PMDA and other agencies in the world. The other is safety, as companies want to ensure that their products are not harming patients.The challenge is that reports of adverse events can be found in a wide range of datasets. There are enterprise applications such as customer relationship management (CRM), quality management, and safety systems and peer-reviewed literature and on top of that, data comes in a range of formats, including audio and video files, emails, PDFs and Microsoft Office documents. In some cases, this data is structured, but increasingly it is not – especially if it is coming from outside the organization.
This complex mix of data sets and formats leaves teams with the unenviable task of managing potentially thousands of data points for adverse events. Even if this data has been compiled in a spreadsheet or database, significant manual effort is necessary to look through it all, flag adverse events and follow up on them.
In many organizations, this takes hundreds of hours per month. It forces teams to take a reactive approach to their work, and it leaves little time to dive deeper into data sets to identify patterns, generate insights and take more proactive steps to mitigate risk or improve safety or devices.
AI and Automation Detection, Collection and Intake Save Significant Time
This is where AI and ML come into the picture. In three key steps of the safety process – detection, collection and intake – process automation leads to the discovery of the adverse events, at higher quality, but with less effort and in less time.Step 1. Detection
AI is critical to detection given the significant volume of data. However, for every signal, there are huge volumes of noise, as fewer than 2% of reports are adverse events. Using artificial intelligence to review all sources, including audio and text files, streamlines this task significantly, reducing the manual workload by 50% to 60%. It is also more accurate: Three independent audits of the IQVIA process found that AI enabled review did not miss a single adverse event across thousands of searches.Step 2. Collection
The scenario is similar for collection. It’s not uncommon for organizations to review multi-language cases from a range of partner organizations, patients and HCPs all using email/paper/PDF forms, generating a huge volume of work in the simple act of receiving and transcribing information from sender format to the safety systems.Given the narrow window in which adverse events must be reported to regulatory agencies, these volumes must be processed quickly. Not only does automation expedite the process; it also does a more thorough job that manual reviewers facing deadline pressure. Web based, collection tools with smart querying allows organizations to receive high quality and more complete information by asking smart questions of the reporter. This allows cleaner data to be easily sent in an ingestible format to safety teams.
Step 3. Intake
The intake process is a bit different, as it involves looking at data from source documents submitted to an organization rather than seeking out signals of adverse events externally.Here, several forms of automation can play a pivotal role. Optical character recognition (OCR) can convert handwritten or typed text into machine-readable text, while natural language processing (NLP) can look for patterns or context that may refer to an adverse event. Beyond simply extracting data, this combination of technology tools can also convert this largely unstructured data into cases within a case collection system. All told, leveraging OCR and NLP for intake can cut the manual workload by 70%.
A Greater Focus on the End-to-End Safety Process
Transformative as this may be, it is only the beginning. Organizations can achieve further efficiency – and savings – through a range of business process engineering efforts. Possibilities include, but are not limited to, automated follow up once cases have been received and processed; generation of quality review scorecards, data visualizations, and other reports; and automated tracking of corrective actions and preventive actions (CAPA) for quality and compliance purposes.Further use of automation has two key benefits. First, it helps Safety automate up to 80% of its current manual workload. As discussed, this is an important step in enabling teams to transition away from completing repetitive administrative tasks and reactively responding to adverse events that have already been reported.
The second, related benefit is the ability to transition to proactive safety and investigation. By spending significantly less time on detection, collection and intake, teams are better positioned to leverage the structured adverse event data that AI tools generate to gain real-time insight into the safety process. This lets organizations get ahead of potential compliance concerns, identify and mitigate risks before they become large-scale problems and even detect adverse event signals as they are occurring.
Medical device companies are continually searching for ways to empower their employees to devote more time to the higher-level work they are passionate about doing. Automating routine manual tasks can play an important role in making this happen. The benefit is acutely felt in pharmacovigilance, where reducing manual workflows by 80% makes it possible to focus on the end-to-end safety process, not repetitive data entry – and to ensure that the devices a company brings to market are safe and effective for as many patients as possible.
Updesh Dosanjh is Practice Leader, Pharmacovigilance Technology Solutions at IQVIA.