Since the development of MUFASA, the launch of advanced generative AI tools like ChatGPT and similar AI technologies have showcased the potential for the seamlessly integrated of such tools into the daily workflow. Recently, a randomized trial of generative AI marked a significant milestone where it was found that these AI tools not only accelerated task completion and led to increased productivity, but also boosted user satisfaction [10]. This underscores the evolving role of AI across multiple sectors, including healthcare.
MUFASA has been demonstrated to be an invaluable asset in the daily workflow for MI/MSL teams, where its semantic search capability enhances the accuracy of case identification, even in scenarios where the initial inquiry from physicians were vague, as demonstrated in Table 1. MUFASA's ability to provide rapid access to responses from previous similar cases has been instrumental in saving time for each MI team member. In fact, with its implementation at LEO Pharma, MUFASA has been shown to save each MI team member approximately five hours per week, assuming an average consultation time of twenty minutes per case. This, along with promoting response consistency, reducing redundancy, and mitigating compliance risks, underscores MUFASA as a compelling solution for managing routine inquiries.
Applications of MUFASA for insight generation: the value of cluster analysisIn contrast to chatbot and generative AI solutions like ChatGPT, MUFASA is not designed for general public usage. Instead, its unique value proposition lies uniquely within the pharmaceutical industry. Though both tools stem from the broader field of AI, they each provide distinctly different capabilities. ChatGPT, for example, excels at generating responses based on existing data, while MUFASA offers a unique feature enabling users to visually map the relationships of inquiries and discern thematic clusters—a functionality not offered by ChatGPT or similar tools. This key difference allows MUFASA to encourage user-driven interpretation of themes and detection of business trends, which might necessitate intuitive skills not yet fully encapsulated when using tools such as ChatGPT. Thus, MUFASA accentuates the crucial role of medical affairs personnel instead of replacing it.
MUFASA's analytical system, which clusters inquiries, offers a profound understanding of the interests and questions of HCPs. In conjunction with insights from practicing pharmacists and PharmD residents, MUFASA paves the way for strategies that can be readily implemented, as exemplified in Tables 1, 2, and 3. Utilizing the comprehensive knowledge and patient counseling skills of these professionals, the trends identified by MUFASA can be translated into practical solutions, such as the development of pharmacist-oriented educational materials and bespoke content for MI websites.
MUFASA also serves as an agent for continuous improvement within pharmaceutical companies. Its ability to track trends allows for immediate adjustments to packaging or educational materials in response to changing needs. Moreover, its geographic visualization feature enables professionals to adapt their strategies to align with regional healthcare practices and patient requirements.
Furthermore, MUFASA's cluster analysis function is advantageous to the broader Medical Affairs department. The MI department can leverage unsolicited inquiries to obtain a broader perspective, eliminating potential sampling and selection bias that can be observed from advisory boards. The insights acquired from the MI database can aid in addressing concerns raised by healthcare professionals regarding new drugs, additional data needs, or recent studies. This tool thus can also contribute to the planning of Other Learning Activities or substantiate the need for such programs.
Ultimately, MUFASA's cluster analysis capability offers an efficient way of extracting valuable insights from unsolicited medical information. In the past, text data quantification and theme identification was a challenging task. Keyword and category searches often fell short due to language variations and potential inconsistencies in categorizing inquiries by different personnel. With MUFASA, the process of identifying trends becomes significantly less labor-intensive. This shift enables the MI team to concentrate more on specialized inquiries, thereby promoting a more comprehensive understanding of data and enabling the generation of actionable insights. As a result, this supports proactive departmental training initiatives, leading to increased value delivery by pharmaceutical companies.
Limitation and future of MUFASAAs a machine learning tool, MUFASA’s ability is fundamentally limited by the volume and quality of data available. In addition, semantic search results can be affected by variations in wording and phrasing, and those effects on the results are not clear at this moment. Similarly, what is considered clinically relevant may not be able to be processed by the Sentence Transformer in identifying relevant “similar cases”. Medical terminology and dictions are complex and nuanced, but Sentence Transformer models are trained based on usual language setting and not trained for medical situations. Therefore, while there may be an interest in training them from MI case handling to better understand how cases are categorized, there is an argument to not train the machine learning model for specific identification of clusters. This is because there is benefit in the lack of training as a lack of training removes the bias for which an individual thinks what a cluster should be.
It is also important to mention that semantic search does not factor in sentiment analysis. Continual development of MUFASA should explore the integration of AI’s ability for sentiment analysis as it is also important for a pharmaceutical company to analyze their social share of voice by sentiment and topic at the same time. For example, although a product may have a high social share of voice (SOV), it is important to address any negative comments.
For the cluster analysis, it was observed that one cluster may contain multiple themes. However, the purpose of clustering is to have the AI help identify themes that are not difficult to tease out manually. Although clusters are not fully isolated through the clustering process, separating the information from thousands of inquiries to smaller organized clusters of hundred inquiries allows for an easier identification of topics. Therefore, the utility of MUFASA ultimately made the process more efficient and effective despite such limitations.
MUFASA is still in its infancy in development and lacks integration with other data ecosystems at LEO Pharma. For example, the addition of geographical data can add tremendous potential to MUFASA. Geographical analysis will provide LEO Pharma a way to measure its breadth of reach of scientific messages in comparison to its competitors. When geographical locations are added to other demographic filters, such as gender, age, or occupation to identify important target audiences, this could lead to unique advantages against LEO Pharma’s competitors. The future of MUFASA should also explore the addition of other ecosystems that will allow the tool to evaluate large volumes of publications, clinical trials, and text insights from advisory boards which can help quickly identify, discern any new key topics of interest for HCPs.
Comments (0)