Der Markt für Software für Arzneimittelsicherheit erfährt mitten in der COVID-19-Krise ein enormes Wachstum. Im Jahr 2020 wurde der globale Markt auf 143 Millionen Dollar geschätzt und soll bis 2027 auf 201,8 Millionen Dollar anwachsen. Traditionell waren viele pharmazeutische Unternehmen bei der Integration von maschinellen Analysetechnologien in einem Workflow zur Arzneimittelsicherheit zurückhaltend, vor allem aufgrund von Compliance-Bedenken. Da jedoch die Datenmengen und die Anzahl der Quellen exponentiell wachsen, haben viele Unternehmen maschinelle Analysetools getestet und in ihre Workflows zur Überwachung der Arzneimittelsicherheit zu integrieren.
Weiter geht es hier mit dem englischen Originalbeitrag, der zuerst im Blog unseres Mutterunternehmens Copyright Clearance Center veröffentlicht wurde.
The pharmacovigilance and drug safety software market is experiencing tremendous growth amid the COVID-19 crisis. In 2020, the global market was estimated at $143 million and expected to grow to $201.8 million by 2027. Traditionally, many pharmaceutical companies have been conservative in incorporating machine analysis technology into a pharmacovigilance workflow, largely due to compliance concerns. However, with the amount of data and number of sources growing exponentially, many companies have started to test and integrate machine analysis tools into their pharmacovigilance workflows.
A pharmacovigilance workflow involves many different data sources, including patient cases, healthcare reports, scientific literature, and even social media. As pharmacovigilance teams know best, the process of monitoring, identifying, and analyzing adverse events across all these sources can be very manually intensive. The workflow and process of applying machine analysis tools to enhance a pharmacovigilance workflow can be very different depending on the data source.
Automating Adverse Event Detection in Scientific Literature:
Many companies today are leveraging machine analysis tools to automate the process of identifying adverse events across scientific literature. Scientific literature, especially in XML format, can be more easily consumed and interpreted by text analytics tools than other data sources. For many companies, this has been the starting point for automating parts of their pharmacovigilance workflow. GlaxoSmithKline K.K.(GSK), a leading company in the pharma industry, has been vocal with the success of leveraging SciBite’s Semantic Platform to enhance their pharmacovigilance capabilities. By leveraging SciBite’s TERMite Expressions (TExpress) module, GSK Japan was able to automate the process of searching for phrases that suggest adverse events within the text. The result of this automated workflow was ultra-fast processing of text and accurate identification of adverse events.
Automating Adverse Event Detection in Patient Cases:
Out of all the different data sources involved in pharmacovigilance, processing patient cases can be the most costly and resource intensive. Pfizer and three other vendors set out to create a pilot to test the feasibility of automating patient case processing. The pilot tested three machine analysis tools and measured their accuracy with extracting and evaluating specific entities within patient cases. The result of the pilot established a feasible and effective use case for leveraging machine analysis tools for patient case processing. All three tools were successful with accurately identifying and evaluating adverse events within patient cases.
Automating Adverse Event Detection in Social Media:
Social media is another potential data source to be monitored as part of a pharmacovigilance program. Quality and adequate turnaround time are special concerns in this scenario, due to the non-standard way the data is generated. A study by Comfort et al in 2018 looked at the benefits of utilizing machine learning models to identify adverse events in social media postings. The results of the study showed 83% accuracy with identifying adverse events in social media with tremendous efficiency. For context, this task took 48 hours for the machine learning model to complete while it would’ve taken human experts an estimated 44,000 hours.
The use cases detailed above provide practical and scalable solutions for automating a pharmacovigilance workflow. Leveraging machine analysis tools is becoming increasingly more popular across the industry. Regardless of the data source, machine analysis tools have proven their effectiveness in reducing cost and improving turnaround time for pharmacovigilance teams. These enhancements are not only beneficial for a company, but also for the patient. Which, in hindsight, is the most important factor of all.