AskX: ChatGPT Integration with Amplity’s Insights Database

An industry-first application of ChatGPT on medical transcription records

By: Joe Turner
Head of Insights, Amplity Health

This is the third article in a series exploring Amplity Insights, the industry’s only database containing unstructured full-text transcripts of dictated physician notes from physician–patient encounters. To learn the benefits of conducting research using unstructured data from medical transcription records, see Closing the Insight Gaps Left by Structured EMR Datasets, and to see a case study of Amplity Insights at work, see Amplity Insights Database to Determine the True Prevalence of Disease.

In this installment, we introduce AskX, a new search channel that merges the deep resources of the Amplity Insights database with ChatGPT, a natural language processing tool driven by AI technology. AskX allows our clients to use ChatGPT to query the Insights database, receiving real-time responses to questions (simple or complex) about the treatment of patients and the rationale of providers.

Introduction
What happens when you combine the pioneering AI technology of ChatGPT with the >70 million medical transcription records in the Amplity Insights database? The answer is AskX, Amplity’s proprietary AI-powered search channel where you can—using your own words—explore healthcare analytics in the way that makes the most sense to you. Gain access to a rich index of insights through conversational questions that yield easy-to-understand results. AskX sums up complex data simply and naturally.

AskX leverages the power of the ChatGPT transformer model but applies it to a closed data set that is safely firewalled from public-domain sources. Our quality control standards and processes ensure the legitimacy of AskX responses.

Brief Overview of AI and ChatGPT
What is AI? In basic terms, it’s a computer system that can learn on its own. One of the advances in AI is called deep learning, which uses neural networks to mimic how the human brain learns new things. It can recognize complex patterns and produce accurate predictions.

In 2017 there was an AI breakthrough moment. Google invented something called a transformer model. This was a new AI system that you could train just by throwing billions of examples of text at it. It could figure out relationships between different words and phrases, and use that information to predict the next set of text in a sequence.

Then in 2020, an artificial intelligence startup named OpenAI released GPT-3 (generative pretrained transformer), which was the third iteration of its transformer model. It was the most powerful transformer model ever built. It used the biggest supercomputers and was trained on the most data, including blogs, news sites, Wikipedia, social media, and even public-domain medical records—everything OpenAI could get its hands on. It could do all kinds of language-based tasks much better than any AI system before.

In December 2022, OpenAI released a new tool called ChatGPT, a chatbot based on OpenAI’s GPT model. It became an overnight sensation, reaching 100 million active users within 2 months; by comparison, Instagram took a year to reach just 10 million users. Microsoft invested $10 billion in OpenAI, and has already integrated it into their Bing search engine and the latest version of Windows.

And just recently ChatGPT 4 was released, reportedly trained on 500 times more data than the previous model and with stronger reasoning skills. It can do your tax return, turn a simple sketch into a working website, and create video games in under a minute. Unlike typical search engines, it doesn’t respond to your query by giving you a list of website links. Instead, it responds with a single textual answer using fully formed sentences and paragraphs. You’ve probably experimented with ChatGPT by now. Ask it a question and it gives you an actual answer. You can use it for free at https://openai.com/blog/chatgpt.

Brief Overview of Amplity’s Insights Database
The Amplity Insights database is the only commercial database containing full-text transcripts of dictated physician notes from physician–patient encounters. It provides data, analytics, and insights not available from any other single source, offering a direct view into the treatment of patients and the rationale of providers, while remaining fully HIPAA compliant.

The Insights data set covers the United States and includes >150,000 individual multispecialty healthcare providers and >70 million patient records. What’s more, we’re adding 2 million new records every month.

AskX: The Integration of ChatGPT with Amplity’s Insights Database
Recognizing the opportunity to use ChatGPT’s immense reasoning skills to help answer specific research questions, we were excited to experiment with it to see what meaningful data it could extract when we fed it Insights data exclusively. So, using OpenAI’s secure APIs that enable the ChatGPT transformer model to communicate directly and exclusively with our own proprietary data set (using robust definitions and protocols), we fed our Amplity Insights records to ChatGPT. We then set up a secure chat portal to enable users to type in any question related to the point-of-care records within the Insights database and receive accurate, well-formulated, easy-to-read responses within a matter of seconds.

The results, frankly, are remarkable. Just AskX a question in plain English and receive a real-time answer. For example, what is the prescriber sentiment surrounding your drug or a competitor’s drug? Or, why are patients prematurely discontinuing a particular therapy? Questions can be related to factors impacting the care of patients, lines of therapy, complications, provider preferences, etc. These inquiries can focus on marketing, commercial efforts, medical, compliance, access, and so on. Importantly, there’s no need to hunker down with your IT team to formulate “the exact right question,” then iterate repeatedly until you get a meaningful answer.

Specific Uses for AskX
Based on our clients’ experience so far, we see 4 immediate and important uses for the AskX search function. Let’s take each of these in turn.

1. Market Research at Scale
If you're running market research prescriber surveys or advisory boards or talking to key opinion leaders, you’ll typically gather responses from a relative handful of individuals over several weeks or months. And those respondents may just tell you what they think you want to hear, or what they think they do in a particular situation, but some level of inherent bias is likely. On the other hand, with AskX you can drill into the actual experiences of thousands of doctors in real time. Get legitimate, unbiased point-of-care data, reported in an appealing format that you can download and use however you wish. And because the answers are coming from historical data, you’re learning what they really think and how they really act in practice.

2. Enhanced HCP Target Lists
The Amplity Insights database tracks prescribers at the National Provider Identifier (NPI) level, which means that AskX can query records from the Insights database and tie them back to an HCP target list. By merging ChatGPT’s functionality with clients’ HCP target lists, we empower our clients on both the medical and commercial sides of their organizations to sharpen and hone their messaging. As a result, both medical and field teams gain access to a wealth of qualitative information to help drive and inform their outreach efforts.

3. Raw Records Licensure
In some cases, it makes sense for a client to license an entire set of Insights raw records so their in-house analytics teams can use their own proprietary algorithms and queries to help answer key business questions. By giving our clients’ data and analytics teams the ability to leverage this AskX component, they can filter the subset of the raw records that contains the most pertinent information relative to their business questions. This cuts down on the iterative nature of leveraging standard data science techniques to filter subsets. AskX enhances the value we provide our clients when they choose to engage with the data in this manner.

4. “Bring Your Own Data” Capability
We believe that our Insights database—with >70 million patient records from >150,000 multispecialty HCPs—is the best-in-market qualitative healthcare data set available. However, we also recognize that manufacturers have their own extensive data sets that hold immense business value. Amplity has established a secure environment for ingestion of such client data, giving internal manufacturer stakeholders the opportunity to query their own internal data by leveraging AskX and the ChatGPT model. For example, a client’s MSL notes can be securely stored within our platform, enabling the client leadership team to query those notes and gain the benefit of the resulting intelligence.

It is important to note that in each of these 4 use cases, AskX leverages the power of the ChatGPT transformer model but applies it to a closed data set that is safely firewalled from public-domain sources. Our quality control standards and processes ensure the legitimacy of AskX responses.

Conclusion
In many respects, the integration of ChatGPT with Amplity’s Insights database is the AI breakthrough for qualitative unstructured healthcare data that the industry has been waiting for. With its first-to-market utilization of ChatGPT on medical transcription records within the Insights database, AskX gives you the power to engage directly, in your own words, with the data itself at scale and in real time. Regardless of your needs—market research, enhanced HCP target lists, raw records licensure, or bring-your-own-data capability—AskX gives you complete access to full-scale integration and instant results, adding competence and precision to your data analysis capabilities.

Try the public version of ChatGPT for free at https://openai.com/blog/chatgpt. And to learn more about how Amplity’s proprietary AskX can help in your situation, visit Amplity Insights. Or choose Contact Us to begin a conversation.