Prepared for Agilon Health
"What we can tell you about your business that your own data can't."
Talon AI maintains a 364-million-row federal healthcare data lake built specifically to answer questions that no commercial database currently solves — at the physician level, across six years of Medicare history.
Federal public data, assembled at physician grain.
Every source referenced is U.S. federal public record. No patient-level data. No PHI. No data sharing agreements required. What took years to assemble is available to Agilon today.
What We Can Do for Agilon
Five ways the data lake creates direct value for your platform
Every use case below is executable today against data already in the lake. No data sharing agreements required. No patient-level data involved. All sources are federal public records.
External RAF Validation and Chronic Condition Gap Discovery
Agilon's internal data pipeline covers approximately 85% of membership for RAF baseline visibility. The remaining 15% — plus undercoding across the full population — drove a $73 million revenue miss in 2025. The V28 model transition compounds this: CMS has nearly doubled RAF scores for CKD, and research shows the vast majority of Stage 3 CKD patients are never coded at all.
The lake cross-references six years of Medicare Part D prescribing (153 million rows) against six years of Part B billing (58 million rows) to identify physicians whose patient panels show strong signals for chronic conditions — diabetes, CHF, COPD, CKD, dementia — that are not reflected in current HCC documentation. If a physician is prescribing furosemide and spironolactone at high volume but HCC 85 (congestive heart failure) is missing from their panel coding, that is a quantifiable gap with a dollar amount attached.
Every 0.1 RAF point improvement per member equals approximately $900/year per Medicare Advantage enrollee. At 426,000 MA members, a 0.1 RAF recovery = ~$38 million annually. This analysis runs at the individual physician level, across every market simultaneously.
Proactive PCP Group Targeting — Ranked Prospect Lists Before Your BD Team Calls
Agilon's business development process for identifying new physician group partners is largely relationship-driven and reactive. Groups come to you, or your regional reps surface them through their networks. There is no systematic method to rank every independent PCP group in the country by their expected platform economics before the first conversation.
The lake generates that ranked list. For every independent primary care group in the United States, we score on: Medicare panel size and six-year growth trend, chronic condition burden of their patient population, existing value-based care participation (ACO REACH, PECOS enrollment), confirmed independence from hospital employment, geographic proximity to existing Agilon payer contracts, and OIG exclusion clean status. The result is a prioritized prospect list — the top 50 or 200 independent PCP groups in any state you name — ordered by expected medical margin contribution if successfully onboarded.
Agilon's 2024–2025 losses were directly tied to entering one new market where prior-year data on the physician group did not exist. A pre-ranked target list with six years of Medicare evidence behind each score changes that underwriting decision before you sign.
Pre-Partnership Physician Profile — Six Years of Medicare History Before You Sign
When Agilon recruits a new PCP group, you need to know what their Medicare patient panel actually looks like — volume, chronic condition burden, prescribing patterns, place-of-service distribution. Currently this information either doesn't exist in a structured form or must be purchased from enterprise data vendors at significant cost, without the longitudinal depth the platform model requires.
For any physician group under consideration, the lake produces a complete profile: total Medicare E&M volume and six-year trend, drug utilization by therapeutic class as a proxy for panel chronic burden, place-of-service split (office vs. SNF vs. home visits — which predicts post-acute cost exposure), manufacturer payment relationships via Open Payments (seven years, 82 million rows) that may indicate formulary conflicts, and PECOS enrollment status to confirm independence. This profile can be generated for any physician group in the country, on demand.
Every failed market entry costs Agilon in exits, stranded payer contract costs, and management bandwidth. Pre-entry physician profiles using six years of federal data are a direct underwriting tool — the same logic that drives reinsurance due diligence, applied to physician recruitment.
Specialist Referral Network Mapping — Where Your Medicare Dollars Are Actually Going
Agilon's medical margin depends on controlling specialty spend. Inpatient admissions, specialist utilization, and high-cost episodes are the three categories management has specifically called out as margin levers. But to manage specialist spend, you need to know where your PCPs are sending patients today — which cardiologists, orthopedists, and oncologists are receiving referrals, and whether those specialists are high-cost or low-cost relative to regional benchmarks.
The lake maps this from Part B billing data. Every specialist who billed for a Medicare patient in the same geography as an Agilon PCP can be identified by procedure, place of service, and billing volume across six years. Cross-referencing with HCRIS cost reports and Care Compare facility data reveals whether those specialists are employed by a hospital system driving facility fees or operating independently. The output is a market-by-market referral network map: the highest-cost specialists receiving the most volume from your partner physicians in each geography.
No commercial database answers this at NPI × HCPCS × place-of-service × six-year grain. This analysis is unique to the lake. Identifying and redirecting even a small percentage of high-cost specialist referrals across 426,000 MA members moves medical margin materially.
Part D Cost Trajectory Modeling — Market-by-Market Drug Spend Forecasting
Part D risk is one of Agilon's most significant financial exposures. Management has made reducing Part D exposure below 15% a strategic priority. Planning around Part D requires knowing the drug spend trajectory for the senior population in each of your markets — which drug classes are driving the highest cost per beneficiary, which PCPs are prescribing at unusually high cost relative to peers, and what the GLP-1 trajectory means for 2026 and 2027 cost projections.
The lake has the full Medicare Part D prescriber dataset from 2018 through 2023 — 153 million rows at physician × drug × year grain. The GLP-1 trajectory in the lake: prescriber count grew 124x from 2018 to 2023, claims grew 277x, and Medicare cost grew from $21 million to $8.5 billion. That trajectory is now in Agilon's market population. Understanding it at the physician level, by market, before it shows up in your claims data is the planning advantage this analysis provides.
Part D trend modeling at physician and market grain gives Agilon's actuarial team an external benchmark they currently lack — federal prescribing data at six-year depth, at the individual physician level, no data sharing required, arriving before your payer reconciliation data with its 18-month lag.
What's in the Lake
Federal data sources — active and queryable today
76 active sources across 132 tracked datasets from 12 federal authorities — CMS, FDA, HRSA, OIG, NLM, NIH, CDC, and more. All public domain. No licensing fees. Refreshed on government publication schedules.
12 highest-signal sources shown. Full inventory of 76 active sources available on request.
The Right Question to Start With
One question worth asking in your next conversation
"When you underwrite a new PCP group for the platform — what does your physician profile look like? And what would you pay to have six years of their Medicare billing and prescribing history before you sign them?"
Agilon's 2024–2025 losses were directly tied, in management's own words, to entering a new market where they had no prior-year data on the physician group. That is not a one-time problem. It is a structural gap in how physician group due diligence is performed across the industry — and it is exactly the gap the Talon Data Lake is built to close.
This is not a data licensing pitch. It is a risk reduction pitch. Agilon is a $5.5 billion revenue company targeting breakeven in 2026 after a $296 million EBITDA loss. The problems described above — RAF undercoding, new market underwriting risk, specialist cost opacity, and Part D trajectory — are the same problems management cited publicly as the drivers of that loss. The data to address them exists. It is federal, public, and already assembled.
All data sources referenced are U.S. federal public records. No patient-level data. No PHI. Public domain use.