We’ve covered a lot of ground with examples about how pharmas are using data to their advantage, mostly structured to this point, so now it’s time to discuss the challenges of unstructured data in the Outcomes and Adherence category. If you missed the last piece in this series, Part V – Sales & Marketing, be sure to check it out.
Tracking how drugs are used in the real world after they go to market can provide great insights into how a pharmaceutical company can optimize the way the drug is manufactured, distributed, marketed and/or sold. However, one of the primary challenges with “after market” data is it is often trapped as unstructured information that is difficult to harvest.
Take the example of clinical notes in an EMR. Doctors and their support teams fill out all the structured data fields about you, and then they put the real “juicy” information in plain text within a free-form notes or comments field. The question becomes, what could be learned by looking within the unstructured clinical notes?
Below are some fields from a Japanese EMR, with the information translated into English. When you read the comments, you can see that there are descriptions about the drug’s dosage, efficacy, etc.
By using Natural Language Processing (NLP) on the clinical notes, we can gather relationships between the terms and then even map them visually as shown in the following diagram. What are the symptoms that are associated when drug X is prescribed? What drugs were prescribed when symptom Y was found?
There are many other observations around outcomes and adherence that can be made, but integrating disparate data sources—and unstructured data formats—is something that becomes important. There are many ways to overcome this.
If you’d like to learn more, check out my recent webinar titled “Overcoming Healthcare & Pharma Data Challenges for Competitive Advantage: Lessons from other industries.”
Have you run into data challenges around outcomes and adherence? I’d love to hear more about them, if so.
Until next time,