Photo by Tima Miroshnichenko from Pexels
The volume of audio files that potentially contain valuable data on adverse events (AEs) has exploded, outpacing the pace pharmacovigilance (PV) operations originally anticipated in the COVID era. Healthcare company call centers, which in some instances saw up to 75% of adverse drug reaction (ADR) notifications reported through this channel, are challenged to handle this drastic increase in source data that requires tedious review for potential safety events or risks contained in audio interactions and transcriptions. As the increase in source volumes surpasses normal hiring capabilities, pharmaceutical companies face a growing challenge to ensure compliance with AE reporting regulations.
The need for a technology solution
Anyone contacting a healthcare provider (HCP), healthcare regulator or a medical commercial call center could disclose potential adverse events (PAEs) during their conversation. Regulatory standards require marketing authorization holders (MAHs) to screen call recordings to identify adverse events or safety risks.
Human workers, typically trained call center agents, medical information (MI) agents or MAH employees, screen call recording files and live interactions. Manually screening audio files requires considerable time to listen, identify and transcribe PAEs to a text format that can be processed onwards to the MAH safety inbox. In addition, manual screening is inconsistent in quality and speed due to reliance on humans to perform the activity and also introduces layers of duplication and opportunities for human error.
The regulated timeline for reporting AEs is very tight, and the clock starts ticking as soon as critical information surrounding that AE is made available to the MAH. Typically, a fatal or life-threatening event must be triaged, processed in a safety database, and submitted to a regulator within seven days of receipt. Any other AEs must be reported within 15 days. Once an MAH receives an audio file, that information must be analyzed and processed as soon as possible to ensure safety compliance is maintained. In times of increased virtual adverse drug reaction (ADR) reporting, as seen with the COVID pandemic when telehealth and virtual doctor visits became standard, it’s impossible to ramp up manual analysis of audio files to meet that rise in demand.
Adverse drug reactions (ADRs) are currently in an explosive phase. COVID vaccines and treatments have ushered in a new age of ADR reporting with rates between five to eight times the anticipated volume. The global adverse event (AE) volume forecast for 2021 could be close to 4.5 million ADRs, which (if this trajectory is maintained) could increase to 12 million ADRs in 2022, even as the pandemic begins to wane. To compound this issue, the formats (audio versus visual versus text), channels (social media versus apps versus telehealth) and types (structured versus unstructured) of reports now vary dramatically from industry norms. With accelerating call volumes outpacing hiring capabilities, there must be a backlog risk-mitigation strategy to assess risk and ensure regulatory compliance.
Audio transcription and record processing automation
Automating audio transcription and adverse event identification provides multiple advantages to patient support programs, commercial and MI call centers, virtual agent collection programs and telehealth. When scaling processing volumes to meet increased demand, technology paired with human personnel can provide a much faster process than human personnel alone. This pairing takes the burden off of MAHs to significantly expand by competing for the limited talent available in today’s highly competitive job market and de-risks the scaling process.
Automating audio file transcription and adverse event identification also optimizes audio file analysis, making regulatory compliance maintenance more achievable. This is because AI technology and natural language processing (NLP) can automatically detect PAEs in transcribed text, enabling an expedited path to human validation, the safety database and onwards to regulators. Automating this process also helps prevent wasted time on voice messages that are blank, beeps or those communications that do not reference a PAE. In addition, it reduces the risk of human error by ensuring no PAEs are missed during analysis. If a PAE is detected, a qualified human PV analyst is notified of the point in a conversation where it is discussed to rapidly validate that the event meets AE-processing criteria.
Auto prioritization, the process of automating the prioritization of files that may contain a PAE or risk according to the system, reduces regulatory compliance risk. As an automated system analyzes files to detect PAEs, they are ranked by importance for qualified human reviewers. This tactic reduces the risk of late reporting of safety events, as no time is wasted on files that don’t contain useful information.
The future of audio transcription and NLP
Technology solutions that automate the processing of audio files and identification of AEs decrease risk and optimize call center overload and backlogs by taking the burden off organizations. In addition, automated audio processing and identification of AEs can minimize noise by using auto- prioritization of audio files to help analysts promptly identify and report valid AEs to ensure regulatory compliance.
Transcription as a service is limitless and broadly applicable to multiple industries, including insurance, legal, marketing, sales and more. Multiple sectors have a huge opportunity to take unstructured data such as audio files and turn it into usable structured data sets. For the pharmaceutical industry, automating the transcription and identification of adverse events also opens the door to more efficient record-keeping. Also, notes from doctor visits can be dictated by voice rather than handwritten— both eliminating the need to analyze illegible handwritten notes and speeding up the HCP’s ability to document and submit patient interactions. The companies that enable automated audio transcription and identification of adverse events today will be more equipped to facilitate HCP demand for enhanced options for note-taking and submission such as voice in the future.
Alison Sloane is general manager, Vigilance Detect at IQVIA. As general manager of Detect powered by AETracker, Alison’s focus is on driving the vision to provide customers with a tech-enabled approach to adverse event and risk detection in structured and unstructured data. Alison joined Quintiles Drug Safety more than 20 years ago. Shortly thereafter, she assumed a customer-managed secondment to a pharmaceutical company for 15 months in the U.K. On return to Quintiles, she expanded her roles in clinical trials, endpoint management, regulatory reporting and operations management.