Key opinion leaders (KOLs) play an important role in ensuring the success of commercialization plans. Identifying KOLs is vital for patient and provider engagement, clinical trial recruitment, clinical trial site selection, provider education, and more. There are several ways to identify KOLs, most of which are either tedious, based on luck or both. They range from tracking conference presentations to performing literature searches on research papers. They can even include searching social media postings and mentions, as well as analyzing clinical trial data. While claims data can provide very focused insights, the results can be misleading, costly, and possibly delay your plans, unless the raw claims are transformed into usable insights.
Claims data is notorious for having missing and inaccurate referral information, which can reduce the value and reliability of claims in your search for KOLs and even result in erroneous conclusions.
Referral information in the claims data can help you identify physicians with a colleague network large enough to reach clinical trial enrollment goals. Claims data coupled with Healthcare Providers
(HCP) referral data can help you identify influencers and who in a referral network should be educated and informed to increase clinical trial enrollment, a make-or-break for a successful clinical trial. According to the Tufts Center for the Study of Drug Development, 11% of sites in a given trial fail to enroll a single patient; 37% of sites under-enroll; 39% meet the enrollment targets; and 13% exceed their targets. This is where referral analytics comes into play. Through the use of statistical models and machine learning, both the fill rate and accuracy of referrals on claims can be improved, which may give you a competitive advantage in the search for KOLs.
When Kythera Labs analyzed 16,445,477,502 open claims, we found that the fill rate for referring physicians was less than half—such missingness results in a very incomplete picture of referrals. To add insult to injury, much of the provided referral information was inaccurate. Of the claims with a referring physician listed, 24% were physicians referring to themselves, with some in a manner that did not make logical sense. No wonder users are confused about the output of their referral data analysis. Kythera Labs leverages our healthcare data expertise to fix incorrect claims data with data science and machine learning resulting in double the number of accurate referrals in our claim set.
Those that believe that closed claims are a better source of referral data than open claims may be greatly disappointed by these findings. Individuals go in and out of networks routinely, which commonly results in missingness. While closed claims may help users come up with more accurate answers for some use cases, they do not have the answer to referral missingness in and of themselves.
At a high level, referral models typically consider the following information:
Despite the fact that claims data is known for having incorrect and missing information, healthcare claims are the source of Kythera Lab’s referral models and are, in fact, the source for other referral analytics available in the market. The impediments are diverse and vary from claim set to claim set. Some claims data sets are missing Type 1 NPI information (provider name), and others are missing Type 2 NPI information (facility name); both make referral insights and identifying the site of care unreliable. Claims may also be missing in entirety, which creates a problem with accurately understanding a patient’s care pathway. Analysis of claims data may also underestimate an HCP’s influence in the market, especially in cases of rare diseases.
Referral models cannot be built on data alone. Any model that connects claims logically in a more nuanced and accurate manner (than simply using common patient logic) requires incorporating the business logic that underpins the basis of patients’ care pathways.
Common patient logic is vastly overvalued in regard to solving for missing referrals. Just because two providers had an encounter with a common patient does not mean a referral took place. Simple business logic that relies primarily on the timing of the care pathway and does not include an understanding of provider specialties, the conditions they manage, and the services they provide will ultimately fail the user and produce referral relationships that do not make sense.
Kythera Labs creates mathematical representations of the actual, more complex business logic using statistical models that link together provider taxonomies, the conditions they manage, and the relationship between specialties across all the complex conditions that require multiple providers. With this information, we are able to answer which provider taxonomies are most likely managing patients and attribute the true referrals more precisely, enhancing our ability to identify KOLs more accurately.
When vetting healthcare claims vendors for KOL referral insights, you should be sure to ask questions such as:
Want to learn more about Kythera Labs’ referral model and how you can have more confidence in your KOL identification? Get in touch (firstname.lastname@example.org) or connect with me on LinkedIn: Russ Sacks, Co-Founder and EVP of Data Science and Special Projects at Kythera Labs.
Kythera Labs launches Patient Event Assets on the Databricks Marketplace.
Clinical Trial Enrollment, Site Selection, and Physician Engagement are more targeted when using Claims Data and Referral Analytics