Leveraging Answer Engines and Generative AI to Improve the Patient and HCP Experience

Life sciences companies are continuing to evolve from being brand-centric to a more customer and healthcare professional (HCP)-centric approach by supplementing traditional customer engagement strategies with timely, personalized, and consistent messaging delivered via convenient and customer-preferred channels. While the face-to-face approach will never become obsolete, it will end up becoming just one part of the new hybrid work model. This critical shift to hybrid needs to be addressed by pharma companies to ensure that they are equipped to succeed in these newer modes of engagement in order to gain a competitive edge in this evolving landscape.

For example, timely access to the right medical information is the number one priority as it has become essential to the omnichannel experience in the pharmaceutical and biotech industries. In some cases, it’s even the difference between life and death. HCPs and patients routinely need to locate critical medical information to save lives, reduce risks, and answer detailed questions from patients who have done their own personal online research. An NIH study¹ revealed that 88.5% of HCPs search for information on medical websites, apps, and general online search engines daily or several times per week. Not only do they need accurate answers, but they also need them fast. A different NIH research study² showed that, on average, healthcare practitioners have only 10 minutes or less to answer a patient’s question and 70% of them reported that they need to seek information for their patients while they are waiting.

Obstacles to providing relevant real-time information

However, providing relevant and reliable medical information located in many different systems and knowledge repositories is no small task. After engaging with many large pharma and biotech companies, it’s clear there are several major obstacles preventing individuals and providers from getting accurate medical information in a timely manner. These include:

• Unstructured data is hard to access: the underlying data characteristics are unique to the medical information space. Most health-related insights are hidden in text-based medical documents and journals, all of which are unstructured data or free-form text, including drug websites, research papers, and FDA sites. This makes finding relevant information in this type of data extremely challenging.
• Domain-specific terms and nomenclatures: pharma and biotech have no shortage of specific terminologies and phrases, such as drug brand and chemical names, technical terms like “contraindication”, and acronyms, and they can change over time. As a result, the tools given to users must be able to answer questions that contain domain-specific terms and understand medical vernacular as well.
• Limited visibility into specific user questions: with traditional search tools, content creators do not have clear insight into all the questions that HCPs and patients may have and as a result must “guess” at the specifics to include in their content strategy. Beyond the boilerplate information on drugs and treatments, life sciences organizations lack the understanding of what end users really want to know from their microsite. Chatbots and legacy FAQs end up being extremely limited and narrow in their question and answer pairs. In turn, content creators cannot provide all the relevant medical information users want on an ongoing basis.
• Meeting FDA regulations: because the FDA requires that only approved information be shown to patients and HCPs, pharma companies have extremely limited options to try and make their information easy to discover. Chatbots and legacy search have not improved the end users’ ability to find answers either.

How can pharma organizations respond?

To address these ongoing and emerging needs, pharma and biotech organizations are rethinking the way they share information and knowledge with HCPs and patients. Fortunately, advancements in natural language processing, which refers to the branch of computer science that is concerned with giving computers the ability to understand text and spoken words in much the same way human beings can, have paved the way for modern and life-changing solutions known as generative AI answer engines. The problem with existing tools such as search engines and chatbots is they are not always equipped to comprehend human language, nor can they effectively process the complexity of medical documents needed to deliver the right answers from them. Many existing search engines and chatbot solutions use a simple keyword or rule-based matching method that prevents them from understanding domain-specific terms, user intent, and other nuances inherent in human language. They also tend to be limited when used to locate the precise and comprehensive medical information needed to address specific medical issues. Although newer generative AI tools like ChatGPT or Bard are trained on large corpuses of information, they will still “hallucinate” and will not understand domain- and company-specific words and phrases without substantial labeling and training, which is time-consuming and costly.

In contrast, generative AI answer engines can interpret not only keywords, but also synonyms, industry- and company-specific terms, and semantics, such as user intent, in medical questions. With a deeper understanding of the question, those answer engines can intelligently process the documents containing the answers and quickly deliver the most accurate answers for the end users. Unlike traditional search engines that return a long list of potentially relevant results, or modern solutions such as ChatGPT and Bard which often produce inaccurate responses, an answer engine provides users with specific information from trusted sources that they can use right away, without having to sift through a long list of answers or validate the results. More specifically, they provide all the relevant answers to a medical question. For example, when a patient asks, “What are the side effects of Xanax?”, the answer engine will return complete results about all the side effects a patient may experience instead of a partial result, which is required to satisfy regulatory constraints. This ensures that the patient is informed on all potential outcomes, resulting in less patient harm and risk.ChatGPT and Bard cannot guarantee they will repeatably deliver all results.

These solutions also provide useful user insights to content and omnichannel teams so they can generate relevant content for review prior to publishing. Not only can they leverage this intelligence to better understand the content needs of users, but they can also use it to close content gaps, improve information relevance, and anticipate users’ future interests. More importantly, and because an answer engine only leverages existing and proven content from the organization, content teams can quickly and efficiently add information to the microsites to answer emerging questions as they become more important.

In today’s digitally transformed world, technology plays a critical role in how pharmaceutical companies do business. The advent of AI is helping them to support diverse use cases and meet the needs of both internal and external stakeholders such as HCPs, clinical researchers, and patients. However, care must be taken to ensure that only trusted information is being presented. Using generative AI as the backbone, answer engines help HCPs to quickly identify and publish potential drug interactions, adverse reactions, and other important medical information that can improve patient outcomes. They also allow researchers to accelerate the time to discover relevant clinical studies and data to support their research and enable patients to learn more about their medical conditions and treatment options on their own, without having to call their doctors every time or wait on the helpline.

Thanks to built-in natural language capabilities, these answer engines are reducing the time it takes to find the right medical information and provide not just one, but all the relevant answers to a medical question, so no critical information is missed. Providing answers at unmatched precision, speed, and transparency, they quickly becoming a trusted channel for pharma and biotech companies to share critical medical information. As the technology continues to evolve, it will play an increasingly important role in the medical industry, with the potential to improve patient outcomes and experience, enhance research, and ultimately save lives.

References

¹ Hermes-DeSantis, E. Hunter, Welch, J., Bhavsar, R., Boulos, D., Noue, M.. “Preferences for Accessing Medical Information in the Digital Age: Health Care Professional Survey”. PubMed, June 2021, https://pubmed.ncbi.nlm.nih.gov/36260374/

² Novak, K. “Know Your Audience: HCPs Savvy Information Seeking Behaviors”. 2040 Digital, March 2019, https://www.2040digital.com/matching-the-needs-of-hcps/know-your-audience-hcps-savvy-information-seeking-behaviors/

About the author

John Reuter is the Chief Strategy Officer at Kyndi, a global answer engine provider that empowers people to do their most meaningful work.