Rare-disease patient finding: Identifying the “right” patients at the “right” time.

Getting rare disease patients the right treatments at the right time is vitally important to the pharmaceutical industry and the patients it serves. Each day of delay compounds the human costs of poor patient outcomes, from their avoidable medical costs to disease progression. From an economic viewpoint, it is estimated that mis-diagnosed rare-disease patients drive over $200,000 in avoidable medical costs per patient, and over $500,000 per patient when not treated promptly. 1 In an environment where bringing a new medicine to market still costs the pharmaceutical industry an average of $2.6 billion, 2 connecting therapies to eligible patients accelerates appropriate starts, reduces waste, and increases Return on Investment (“ROI”). The challenge to more efficiently and effectively identify patients for whom these drugs are developed is readily apparent.

The key elements of “patient finding” success

When launching or scaling a therapy, the goal is to identify the patients who are eligible–or soon-to-be–during the narrow time window before initial diagnosis, therapy change, or referral. Once these patients have been identified, the next critical step is to link each patient to the physician most responsible for initiating or influencing treatment for that patient so that time from diagnosis to treatment can be shortened.

Ideally, conditions are such that you have:

• Timely Signals: Recent and comprehensive patient medical and pharmacy claims activity that indicate proximity to diagnosis, escalation, or therapy starts.
• Precise Attribution: Confidence that the linked HCP is a key decision-maker for treatment.
• Explainable Logic: A clear “why” behind each relevant HCP so field teams can prioritize and act efficiently.
• Operational Integration: Outputs that flow to the field to trigger next-best actions, and field inputs that flow back to the solution for continuous improvement.

The solution to identifying patients quickly

Phase 1: Simple patient identification with advanced HCP attribution

The first step in finding patients is to start with confirmed patients suffering from the disease in question and then find their appropriate HCP decision-makers. Using simple identification by claims diagnoses, Phase 1 finds all known patients in the available data. For each patient, the most relevant HCPs are identified based on a custom algorithm ranking the recency and frequency of meaningful interactions (visits, procedures, prescriptions, lab testing) as well as disease-specific features such as relevant specialties, affiliations, or referral patterns.

However, relying solely on what we see in the claims data can lead to issues of:

• Incomplete Visibility: Claims coverage is partial on both a macro and a micro level – not all patients appear in claims, and those who do often have gaps in data captured such as unrecorded diagnoses, the specialty pharmacy not being visible, and/or sparse lab/genetic testing results.
• Delayed Action: Claims data typically has a latency of two to four weeks from time of service/diagnosis. As a result, waiting for the diagnosis in the claims could be too late for appropriate action.

To combat these issues, patient signatures and predictive models are deployed to find undiagnosed or yet-to-be-diagnosed patients.

Phase 2: Predictive expansion to high-likelihood patients

Phase 2 builds on the diagnosed cohort, extending reach to patients who present like confirmed cases but lack a recorded diagnosis in the claims. This approach is two-tiered:

Tier 1 – Signature-based rules: Patient signatures are derived from the diagnosed population capturing common patterns in diagnoses (comorbidities), frequent procedures, lab or genetic testing orders, and specialty-visit patterns. A weighted rules approach flags patients who look similar to our diagnosed cohort but are likely missing a coded diagnosis in claims or nearing diagnosis in clinic.
Tier 2 – Predictive modeling: Supervised machine learning models are trained on the same features built to create patient signatures. With the addition of machine learning, the model (rather than a human) now determines the importance and correlation of the features to produce a likelihood of patient relevance.

Phase 2 still makes use of the advanced HCP attribution logic but now expands the potential patient universe to scale the solution.

This process is particularly useful in identifying rare disease patients whose diagnosis can remain elusive. For instance, to identify children who are likely to be diagnosed with a rare dwarfism called achondroplasia, we established a set of business rules that often point to the disease without an official diagnosis. These signatures included claims for recurrent ear infections (diagnoses), pharyngeal surgeries (procedures), growth hormone treatments (prescriptions), or frequent specialist visits (specialties) – all uncommon for typical pediatric patients but common for achondroplasia patients. By weighting the importance of these and other features, we could expand our outreach from strictly diagnosed patients to likely diagnosed patients.

 “In patient identification, it’s not just about finding the right patients – it’s finding the right HCP to engage their patient exactly when it’s relevant to drive a diagnosis and treatment decision.”

– Isabelle Amick, Senior Director of Data Sciences, Propensity4

Overcoming the data challenge

The volume, latency, and complexity inherent in claims make rare-disease patient finding uniquely challenging. Anchoring on diagnosed patients, then extending with signature-based rules and predictive modeling, provides the speed, transparency, and scale required to successfully meet the challenges in finding the right patients at the right time served by the right HCPs.

Learn more about how we can help you throughout your product lifecycle: Propensity4 | Product Lifecycle Optimization
_______

1 Every Life Foundation: Groundbreaking Study Reveals Staggering Economic Toll of Delayed Rare Disease Diagnosis, accessed October 28, 2025, https://everylifefoundation.org/delayed-diagnosisstudy/.
2 The Alchemist’s Playbook: Transforming Drug Patent Data into …, accessed October 28, 2025, https://www.drugpatentwatch.com/blog/the-alchemists-playbook-transforming-drug-patent-data-into-financial-gold-with-advanced-ip-valuation-and-financing-models/