7 Top Use Cases of AI & Machine Learning for Automated Case Processing in Pharmacovigilance
Life Sciences | 6 min READ
    
An elderly couple was getting their vitals assessed through a telehealth consultation in the aftermath of COVID-19, robots fulfilling warehouse orders at the instructions of a supervisor working remotely, airline check-in at kiosks with minimized human interaction¬¬—the pandemic has enforced an era of revolutionary transformation through the implementation of automotive technologies. Therefore, it is not a matter of surprise that the impact of these technologies, including robotics, big data and analytics, smart workflows, and natural language processing (NLP), among several others, has been predicted as revolutionary by business and technology leaders for years.
Sreenivas Reddy
Sreenivas Reddy

Global Program Director

Life Sciences

Birlasoft

 
However, up until early last year, it could be said that the pharmaceutical industry was relatively slow in adopting and embracing these technologies. Despite having healthcare databases with medical services providers and patient data of more than 300 million in units. At the same time, this delay may be accrued to aspects of regulatory requirements, other industries with similar or even heavier regulations to adhere to have been quick to implement advanced technology at a comparatively faster rate. The banking and financial services industry (BFSI) is a perfect example of this case in point.
AI & Machine Learning in Pharmacovigilance
Why is it Gaining Ground?
With the pandemic affecting every aspect of our lives, stakeholders in the biopharma industry and pharmaceutical companies were quick to realize that the time to unlock the capabilities of technology within their sphere of work had come. It was time to embrace automation and take massive strides in expanding operational potential.
While there had been some automation experiments in the pharmaceutical industry, most of them were limited to robotic process automation (RPA) in ops maintenance and patient servicing chatbots. Burdened by legacy processes of clinical trials in pharmacovigilance and ever-increasing volumes of case processing year-on-year, pharma professionals soon realized the need and scope to leverage artificial intelligence (AI) and machine learning (ML) at scale.
Stay Ahead
Visit our Pharmacovigilance page
Applying AI & Machine Learning to Case Processing
  1. Optimized Rapid Delivery and Cross-Industry Collaboration
  2. Connect Therapeutic Areas to All Strategic Capabilities
  3. Support Clinical Trial Efficiency, Interoperability, Effectiveness & Feasibility Goals
  4. Improve Automation and Observability of Regulatory and Safety Systems
  5. Measure Core Pharma Systems on Consumption and Experience
  6. Develop Community of IP Practice for Data, Open Source, and Patent Prosecution
  7. Social Listening for Accurate Health and Drug-Related Information
1. Optimized Rapid Delivery and Cross-Industry Collaboration
A European study on successful drug launches within the EU market closely looks at the benefits of cross-industrial collaboration in pharmacovigilance. New-age technologies enable a collaborative atmosphere enabling optimization of clinical trials and drug testing. To this end, AI and ML help in tearing down silos and automating 50-60% of simple post-marketing cases by using lessons learned in other biopharma studies upon identification of an appropriate opportunity to do so. Collaborative tendencies help standardize data and processes while optimizing signal detection techniques by the effective use of predictive analytics.
2. Connect Therapeutic Areas to All Strategic Capabilities
The pharma industry experiences growing volumes of adverse events (AE) data. The United States Food and Drug Administration (FDA) alone reported a 2.6-fold increase in the number of AE says they received in the eight years between 1998 and 2005. While volume increase is a global phenomenon, there is massive underreporting of AEs. This presents an opportunity to connect therapeutic areas across the world; mining new data sources can potentially unlock strategic capabilities within the pharmacovigilance domain. Intelligent workflow automation and cognitive agents help us reimagine a pharmacovigilance setup primarily focused on research and drug development. Such an end-to-end process overhaul can reduce trial time, eliminate operational bottlenecks, and minimize workflow laggards.
3. Support Clinical Trial Efficiency, Interoperability, Effectiveness & Feasibility Goals
A report by the US National Institutes of Health reveals the impact of emerging tools of AI on the case processing component of pharmacovigilance, which in effect has the most economic impact on its budget. Pharmacovigilance has a massive potential to automate patient safety data and optimize clinical trial efficiency. Technologies such as Signal detection, analyses of large volumes of data, and risk contextualization and tracking, ML-algorithms can be leveraged to achieve this objective. Pharmacovigilance professionals can leverage the combined accuracy measures to adjudicate different vendor algorithms upon successful training of adverse event database content. Automation proves to be an economically feasible and effective strategy since manual dependence on data entry, source document annotation, an otherwise strenuous activity, is eliminated.
Applying AI & Machine Learning to Case Processing
Applying AI & Machine Learning to Case Processing

Click to zoom in

4. Improve Automation and Observability of Regulatory and Safety Systems
The USNIH report also takes cognizance of a case study where two vendors showcased case-level accuracy higher than Pfizer’s internal benchmark within two database training cycles alone. 1/3rd of the cases that came under the purview of the clinical trials were processed up to 80% of their completion. The study demonstrated that case validity could be determined at the early stage of case intake in pharmacovigilance by leveraging machine learning. It also confirmed that through precision and accuracy in ML, case contents could be extracted from source documents, a pleasant welcome in place of outdated annotation methods.
5 Benefits of AI & Automation in Pharmacovigilance
5. Measure Core Pharma Systems on Consumption and Experience
Case processing often entails patients and trial subjects sharing their experience with medication. This can often trigger discussion of similar or opposite reactions by users in a separate community. A unified monitoring and evaluation system that tracks consumption patterns can act as an early warning system for adverse drug reactions. Social media also provides an opportunity to monitor these drug reactions outside the purview of traditional clinical trials.
For diseases with a social stigma associated with it, several drug users often share their consumption experiences on trusted circles on social media. More often than not, these experiences are underreported in individual case safety information reports (ICSR). Data mining under such scenarios can genuinely improve how traditional pharma systems measure case processing.
6. Develop Community of IP Practice for Data, Open Source, and Patent Prosecution
Cross-industrial collaboration in pharmacovigilance facilitates knowledge transfer among diversely placed stakeholders. It also gives the scope to develop set standards of intellectual property practices for data and open-source information and benchmark product safety. Therefore, the possibility of patent prosecution arises. To avoid such untoward incidents, the idea is to manage collaboration like a seamless process.
A Walden University research states that under such circumstances, right at the very outset, all parties involved should carefully plan and agree on their ways of working. The guidelines must be defined for managing intellectual property, resources, all within approved regulations of oversight and governance.
7. Social Listening for Accurate Health and Drug-Related Information
Several studies have clearly shown the potential of social listening on new media platforms to detect drug safety combinations and adverse events. A study concluded in 2015 analyzed five million-plus posts by approximately 7000 users on Instagram focused on associated symptoms of antidepressants. Upon text mining on medical dictionaries, the analysts identified co-mentions of drugs and adverse events for daily, weekly, and monthly periods. Additionally, they formulated proximity graphs to form associations between terminologies identified in the medical dictionaries and the probability of adverse events.
Considering the largely untapped potential of AI and ML in pharmacovigilance, Life Sciences companies stand at a juncture today where they can still benefit from a first-mover advantage. The benefits for implementation of ML are there for the taking in the long term. It is time for the pharmaceutical industry to adapt quickly, identify ML use cases rapidly, and leverage those at scale to move on an evolutionary road map on the way forward.
 
 
Was this article helpful?