The Leading Indicator for the Safety Net

A standing instrument for the finances of safety-net care.

One trajectory engine, run market by market on public filings — IRS Form 990, Medicare cost reports, and the SAMHSA facility census — with confidence intervals and an exposure-weighted policy shock model. Built and maintained by Third Horizon.

Operating margin · Chicago hospital cohort · FY2016–2030Live
cost-report actuals projection · 80% CI COVID

Coverage

Market by market

The engine deploys city by city. Chicago is live; the next wave of metropolitan markets is in preparation.

Live · data flowing Planned

How Bellwether Works

Data & methods, real-world expertise, and AI

01 / Data & Methods

Public financials, one engine

Every cohort runs on the same trajectory model — IRS Form 990 filings, Medicare cost reports, and the SAMHSA treatment-facility census, with COVID years controlled out and an 80% confidence interval on every projection.

02 / Real-World Expertise

Exposure measured, not assumed

Each organization’s grant-cliff and Medicaid/program exposure is measured directly from its filings, so the policy shock engine weights every lever on real data — not a flat assumption.

03 / AI

Ask the data directly

Jane, the platform assistant, answers questions about any cohort or provider in plain language and points back to the underlying filings and methodology.

Methodology & Provenance

The credibility is in showing the work.

Bellwether is a reference instrument, not an advocacy tool. Every figure traces to a public source, every projection carries its interval, and the method is published alongside the result.

01Source. IRS Form 990 financials, Medicare cost reports, and the SAMHSA treatment-facility census — public, auditable filings.
02Model. A single trajectory engine across hospitals, FQHCs, and behavioral health, with COVID years controlled out.
03Uncertainty. Projections shown as a median with an 80% confidence interval (bootstrap), never a single false-precise line.
04Shocks. Policy levers scale by each organization’s measured exposure, so sensitivity is grounded in its own financials.

Start with a market.

Open a live cohort dashboard, or read how the engine works.