In this year of the large language model, how is generative AI impacting the medical world? With this question in mind, CSOFT Health Sciences last month joined innovators from across the life sciences at the Drug Information Association’s 2023 Annual Meeting to explore the changing realm of medical communications, where representatives from an industry prone to the worst potential downsides of automation sought to illuminate its newfound role at the heart of diverse, patient-centric drug development. Strikingly, it is not a question of if or even when the technology behind ChatGPT, Bard, or custom models for medical settings will make its way into practice. Just half a year on from its release, generative AI is already assisting medical professionals with everything from plain language summaries and other types of medical writing to pharmacovigilance and the processing of unstructured data, all while improving experts’ efficacy in overseeing clinical outcomes.
Below, we look more closely at key themes from DIA surrounding the use of cutting-edge language AI and the implications for multilingual medical communications.
From Synthetic Data Generation to Significance Identification: Highlights of the Shift from Predictive AI to Generative AI
Perennial attendees of DIA, and those present in Chicago for the 2022 Annual Meeting, are likely familiar with the incredible impact of NLP algorithms in clinical research even preceding the arrival of models described as generative. At last year’s conference, predictive AI featured in areas such as patient recruitment and site selection, as well as the simulation of candidate molecules for development. While built from similar fundamentals, predictive and generative AI are astonishingly different in what they enable, however. On top of what the former offers for planning effective trials, the latter has added support for the actual conduct of those trials and the subsequent aggregation of clinical data in presentable form.
Achievements commonly mentioned in DIA talks and workshops included:
- Enhancing Pharmacovigilance (PV)
Signal detection and single case processing – processes that identify and evaluate potential issues indicated in clinical data – require the ability to infer real world patient experiences from troves of unstructured data collected during clinical trials. Algorithms that understand language have demonstrated a greater ability to do so than typical human professionals.
- Automating the Authoring of Plain Language Summaries
Medical writing presents a burdensome challenge, in terms of summarizing a vast breadth of information to clinical standards while clearly stating the conclusions to be taken from it. With generative AI, human medical writers can verify the accuracy of such summaries, rather than compose them, or even task the same type of AI with doing so through effective prompt engineering.
- Performing Automatic Document Reviews and Data Queries
When querying trial data for decisive information, generative AI can apply natural language processing capabilities to not only scan for information but also communicate that information concisely and in a readily usable format, saving researchers manual labor.
How far along is this technology, and what lies in store for the immediate future? While several DIA speakers referred to the outstanding question of AI regulation as a next hurdle, the degree of investment companies already exhibit toward it underscores a confident outlook. Audiences of DIA sessions frequently inquired about the practicalities of obtaining and running licensable models tailored to the unique subject matter of one or another medical field – an area that domain specific plug-ins can assist. Speakers meanwhile reinforced that far from replacing clinical professionals or even language services, linguistic AI is helping close gaps in labor intensive areas where talent is historically strained. For those seeking to maximize its potential, the ins and outs of prompt engineering may present an enduring challenge that will define the scope of its application.
As well as AI, DIA 2023 saw a renewed shift of focus to patient diversity and the need for decentralized clinical trials requiring coordination across borders. Wherever the industry leads, we are proud to support life sciences companies’ ever-growing needs for multilingual communications in an industry racing to better serve patients of all backgrounds. To learn more about our medical translations, visit lifesciences.csoftintl.com.[dqr_code size="120" bgcolor="#fff"]