Category: Commercialization
Patients who could benefit go unidentified, limiting reach for rare and specialty therapies.
For rare and specialty therapies, the patients who could benefit are often undiagnosed, misdiagnosed, or simply invisible in the data a team can see. AI applied to claims and EHR data builds longitudinal patient journeys to identify and segment likely-eligible patients, sharpening commercial and medical targeting.
For rare and specialty conditions, the patients who could benefit are often undiagnosed, misdiagnosed, or scattered across records no single team can see. AI applied to large, de-identified claims and EHR datasets helps by learning what the journeys of confirmed patients look like — the tests, referrals, symptoms, and treatment patterns that tend to precede a diagnosis — and then finding other patients whose records resemble those patterns. The result is a prioritized list of people who may be worth a closer clinical look, not a verdict about any individual. These signals are probabilistic and meant to support awareness, education, and outreach planning rather than to label anyone. A model can suggest that a profile looks similar to diagnosed patients; only a qualified clinician, with the full picture in front of them, can diagnose. Treating the output as a starting point for human review — and handling the underlying data under the applicable privacy rules — keeps the tool useful without overstating what it knows.
Start with the data: ask how representative and complete the underlying claims or EHR sources are for your therapeutic area and population, because a model can only find patterns in the patients it can see. Probe how the model was built and validated, how it handles false positives, and whether it can explain why a given profile was flagged rather than returning an opaque score. Because this work touches sensitive health data, examine de-identification, privacy safeguards, and compliance with applicable regulations, and be clear about how outputs will and will not be used. Confirm the tool fits your existing CRM and analytics workflows, and weigh whether its findings are specific enough to act on responsibly rather than generating a long list a team cannot follow up.
A low-risk entry point is to run the model against a therapeutic area where you already understand the patient population, then compare who it flags against a cohort of confirmed patients. That shows how many known patients it would have surfaced, how much noise it adds, and whether the patterns it keys on make clinical sense — all before any output informs outreach or education. Keeping an early trial scoped to one indication keeps the evaluation clean and the privacy review manageable.
Many patients who could benefit are undiagnosed or misdiagnosed, and the clues to their condition are spread across labs, visits, and claims that no single system brings together. For rare and specialty conditions especially, symptoms can be nonspecific and the relevant signals are easy to miss, so eligible patients stay invisible in the data a team can actually see.
AI applied to de-identified claims and EHR data learns the patterns that tend to characterize confirmed patients — sequences of symptoms, tests, referrals, and treatments — and then finds other patients whose records look similar. It builds longitudinal patient journeys and segments likely-eligible patients so teams can prioritize where awareness and outreach efforts may matter most.
No. The output is a probabilistic signal that a profile resembles diagnosed patients, not a diagnosis. Deciding whether someone actually has a condition and how they should be treated is a clinical judgment a qualified clinician owns, with the full patient picture. The tool supports awareness and evaluation; it does not replace that judgment, and the underlying data must be handled under applicable privacy rules.
Ask how representative and complete their data is for your population, how the model was built and validated, how it handles false positives, and whether it can explain why a patient was flagged. Then examine de-identification and privacy safeguards, compliance with applicable regulations, how findings are meant to be used, and how the tool integrates with your CRM and analytics workflows.
Intercept Pharmaceuticals worked with Komodo Health to apply its Healthcare Map real-world data to primary biliary cholangitis, supporting patient-journey and treatment-pattern analyses for its Ocaliva franchise.
Vendor: Komodo Health · press release (Komodo Health (Business Wire)) · 2022-11-10 — Partner: Intercept Pharmaceuticals