Capital S Consulting
Patient-Level Targeting That Learns From Every New Patient
Most firms get their commercial teams to HCP or institution-level targeting and stop there. We go one level further, to the patient, using de-identified data your firm already licenses
Most Targeting Stops at the Prescriber
HCP tiering ranks who prescribes today. It cannot see the patient who is still years from the right specialist. So we start from the other end: the patients already on therapy, often from an expanded access program or a clinical trial, and the common signals in their diagnostic history. Those signals and the strength of their correlation with the condition go into what we call the knowledge workbench. Weekly claims and lab feeds run against it and produce a constantly refreshing set of leads for your team to reach out to. This is not prediction for its own sake. It is giving your reps a reason to have a conversation, based on data as it arrives. We built this exact system for a specialty pharma firm, and it now identifies more than half of their closed leads across their therapies
Five Steps, and the Last One Never Stops
The method is not exotic. Look backward at the patients you already treat, work out what their records showed before diagnosis, then let each weekly refresh flag the patients whose records match the profile, and the providers treating them
Start With Patients on Therapy
The model starts from the patients already on therapy or with a confirmed diagnosis, often participants in an expanded access program or a clinical trial. On a de-identified basis, their diagnostic history is the training set: the diagnosis codes, lab results, procedures, and prescriptions that showed up before anyone wrote the correct diagnosis down. The wrong specialist, the panel run twice, the procedure that ruled something else out
The Knowledge Workbench
We build the workbench with your medical team: the diagnostic codes, procedure codes, and prescription codes that indicate a potential patient's likelihood of having the condition and their level of candidacy for the therapy. Each code is weighted by how strongly it correlates with a confirmed diagnosis, and the weights are rebuilt as the treated population grows
Weekly Claims and Lab Feeds
Each refresh runs against the workbench and produces an updated set of patient leads, linked to the providers treating them, in your commercial data warehouse. Claims and lab records are joined on a tokenized patient identifier, so no patient PII enters your environment
Specific Leads for the Field
Reps and MSLs get patient leads tied to named providers pushed into Salesforce, so conversations start from data instead of broad education outreach. For a rare disease launch, those leads sit in the same rare disease CRM your field already works in
Smarter With Every Patient
Every new patient on therapy broadens the diagnostic history the model reads and re-weights the workbench. We stay on to refine it with your team as the program grows, so the leads your field works next quarter reflect what this quarter's patients taught the model
Where It Fits in the Launch
This is not a pre-launch deliverable you check off and file. It works differently depending on where you are in the launch, and it gets better after launch rather than worse
Before Launch
Before the first commercial patient, the initial profile comes from the diagnostic history of your clinical study participants, analyzed on a de-identified basis, with claims-based criteria filling in around it. The model starts with a real signal set on day one and sharpens as commercial patients start therapy
After the First Patients
The retrospective analysis begins as soon as a treated cohort exists. A handful of patients is enough to start reading diagnostic history. The signal set is rebuilt as the cohort grows, and the provider ranking moves with it
Alongside Tiering
Tiering tells you who prescribes now. The workbench adds the providers whose patients match the pre-diagnosis profile. The target list should use both, weighted the way your commercial team decides they should be weighted
As a Managed Service
After launch, every data refresh delivers updated patient leads as one workstream inside the rare disease launch platform your team already keeps running
Why Commercial Teams Bring Us In
Efficient Reps, Specific Leads
Firms bring us in so a lean field team spends its time on specific, data-backed leads instead of generic disease education. The list refreshes every week, and time in territory follows it
Modeled for Your Therapy
We work to understand your specific therapy and patient population and model for it directly, then stay on to improve the model as more patients start therapy
Built With Your Medical Team
The knowledge workbench is built working directly with your medical team. It holds the diagnostic, procedure, and prescription codes that indicate likelihood of the condition and candidacy for the therapy
Frequently Asked Questions
What data does patient-level targeting need?
De-identified claims and lab data, usually from IQVIA, Komodo Health, or Symphony Health, plus a starting population of patients already on therapy or with a confirmed diagnosis, often from an expanded access program or a clinical trial. Most firms are already licensing at least part of the claims and lab data for other purposes.
If you have not licensed a feed yet, we help define the specification before you sign: which diagnosis and procedure fields, which lab result formats, what refresh cadence, and what level of patient-level detail the vendor will release. The wrong cut of the data is expensive to unwind, and the vendor will not tell you which cut a predictive model needs.
How is this different from HCP tiering?
HCP tiering ranks providers on what they prescribe today. It is backward looking by design, and it works well when the treating universe is established and prescribing volume is a fair proxy for opportunity.
Patient-level targeting starts from the patient. It reads the de-identified diagnostic history of patients already on therapy, learns which pre-diagnosis signals predict the condition, and then points to the providers whose current patients show those same signals. The two answer different questions, and most commercial teams should run both.
Does patient PII enter our systems?
No. The data is de-identified before it reaches us, and patients are keyed to a tokenized patient identifier that lets us join claims and lab records for the same person without knowing who that person is.
Your compliance and privacy policy sets the rules for what is permitted, and we configure to it. We document the joins, the identifier handling, and the retention for your privacy team to review.
How long before the model is useful?
Every patient population is different, so the answer depends on yours. Before launch, your clinical study participants give the knowledge workbench its starting point: their de-identified diagnostic history sets the initial signals, with claims-based criteria filling in around them.
From there the model gets more useful as more patients are confirmed on therapy and the diagnostic history behind them broadens. Each refresh re-weights the signals against a larger population, so the leads reaching your field improve over the course of the launch.
Does this keep running after launch?
Yes. Each refresh links new patient signals to the providers treating them and delivers a specific set of patient leads for your field to go after, flowing into Salesforce as part of managed services. Your field sees the new leads without an export, an email, or a request to an analyst.
We also refine the knowledge workbench with your team as the treated population grows, so the leads reflect what the newest patients taught the model rather than a signal set frozen at go-live.
Already licensing claims data?
The signals for your next patients are probably in it. Let's find out