CMS Change of Ownership: The Federal Record of Hospital and Nursing-Home M&A

When a hospital or a nursing home changes hands, the buyer rarely starts over. Instead it usually steps into the seller's shoes — keeping the same Medicare provider number, inheriting the same compliance history, and assuming the same outstanding liabilities — through a formal Medicare process called a change of ownership. CMS records each of those transfers, and the result is roughly 5,900 changes of ownership: one row per transaction, keyed to the facility's CMS Certification Number and effective date, naming the buyer and the seller. It is the transaction-level ledger of American healthcare consolidation, the moving picture that the static ownership snapshots only ever show a single frame of.

This article covers what the change-of-ownership dataset is and how the CHOW process works; the defining feature that separates a CHOW from a new enrollment — the assume-the-provider-agreement default and the right to reject assignment; why inheriting the provider number means inheriting its compliance history and its liabilities; the kinds of providers that file CHOWs, with hospitals and skilled nursing facilities at the center; the wave of consolidation the data exposes, from hospital mergers to the private-equity and real-estate-investment-trust acquisition of nursing homes and multi-facility chain roll-ups; how the CHOW record joins to the provider-ownership and enrollment datasets to follow a facility through successive changes of hands; a Python workflow that pulls the CHOWs from data.cms.gov, tallies them by provider type and year, and surfaces the most acquisitive buyers; and the caveats — entity-resolution difficulty, transaction-only scope, and reporting lag — that every analyst must internalize before drawing conclusions about who is buying what.

What the dataset is

A change of ownership — CHOW, in the universal shorthand of Medicare enrollment — is the formal process that occurs when an existing Medicare provider is sold or transferred to a new owner. A community hospital is acquired by a regional system; a skilled nursing facility is bought by a chain operator; a freestanding facility is taken over by a private-equity-backed holding company. In each case the facility keeps operating, the same patients keep coming through the door, and Medicare keeps paying — but the legal entity that owns and is responsible for the facility has changed. The CHOW is how Medicare is told that the change has happened and how the program decides what to do about it.

CMS publishes the record of these transactions as the Change of Ownership data on data.cms.gov, and in our database it is stored as the table cms_provider_chow, comprising roughly 5,900 hospital and skilled-nursing changes of ownership. The grain is one row per transaction: a single facility that has been sold three times over two decades contributes three rows, each with its own effective date, buyer, and seller. The columns identify the facility, the parties, and the timing of the transfer:

ccn                  -- CMS Certification Number of the facility transferred
provider_type        -- hospital, skilled nursing facility, or other category
provider_name        -- the name of the facility at the time of the CHOW
state                -- the state in which the facility operates
seller_name          -- the prior owner / transferor of the provider agreement
buyer_name           -- the new owner / transferee assuming the agreement
effective_date       -- the date the change of ownership took effect
disposition          -- assigned (agreement assumed) vs rejected (new enrollment)

The load-bearing column is the ccn — the CMS Certification Number, the persistent identifier Medicare assigns to a certified provider facility. The CCN is the key that makes the whole dataset analytically valuable, because it does not change when ownership does: a facility sold from one chain to another keeps the same CCN, so the CCN is what lets an analyst stitch successive CHOWs into a single facility's ownership timeline and, crucially, join that timeline to the facility's certification, quality, and compliance records, which are also keyed by CCN. The seller_name and buyer_name are the transaction's parties — the transferor and the transferee of the provider agreement — and the effective_date is the moment the transfer took legal effect. Together those three are what turn a static list of who owns what into a record of who bought what, from whom, and when. The disposition distinguishes the ordinary CHOW, in which the buyer accepts assignment of the existing agreement, from the case in which the buyer rejects it — the distinction the next section turns on.

The CHOW process and the assume-the-agreement default

The single most important fact about a CHOW — the thing that makes it a distinct category rather than just paperwork — is what happens to the Medicare provider agreement. The provider agreement is the contract under which a facility participates in Medicare: it is what entitles the facility to bill the program and bind it to the program's conditions of participation. When a facility undergoes a change of ownership, Medicare's default rule is that the existing provider agreement is automatically assignedto the new owner. The buyer does not negotiate a fresh agreement; it inherits the seller's, by operation of the regulation, unless it takes affirmative steps to do otherwise.

That automatic-assignment default is precisely what distinguishes a CHOW from a brand-new enrollment. When a wholly new facility seeks to participate in Medicare, it enrolls from scratch: it submits an enrollment application, undergoes an initial survey and certification, and is issued a new provider number with a clean slate. A CHOW is the opposite. The buyer does not get a clean slate; it steps into the existing provider agreement, with the same provider number and everything that attaches to it. The new owner generally must file the appropriate change-of-information enrollment paperwork to update the ownership records, but the agreement itself — the legal relationship with the program — carries over intact. This is why the CHOW exists as a concept at all: it is the mechanism by which a facility's Medicare participation survives the sale of the business that runs it.

The default is a default, not a mandate, and the buyer retains one consequential choice: it may reject assignment of the existing agreement. Rejecting assignment means declining to assume the seller's provider agreement and instead enrolling as a new provider — which entails a new initial certification and, in effect, starting over. Buyers reject assignment when the thing they are inheriting is more burden than benefit: when the facility carries a poor compliance history, an enforcement posture, or potential liabilities the buyer does not want to assume. Rejection is the exception rather than the rule, because for most buyers the value of an existing, billable provider agreement — with no gap in Medicare participation and no fresh initial survey — outweighs the baggage that comes with it. But the choice is real, and the disposition field records which path the transaction took, making the assume-versus-reject decision itself something the data can be used to study.

Inheriting the number means inheriting the history

The consequence of the assume-the-agreement default reaches far beyond billing continuity, and it is the part that makes a CHOW a genuinely weighty transaction rather than a clerical update. When the new owner assumes the existing provider agreement, it assumes everything attached to it: not just the provider number, but the facility's entire compliance history, its outstanding obligations, and its liabilities. The buyer takes the facility as it finds it, warts and all.

Concretely, this means the new owner inherits the compliance and survey history tied to the provider number: the past deficiencies cited on certification surveys, the enforcement actions and plans of correction, the quality-measure track record. A nursing home with a long history of health-and-safety citations does not shed that history when it is sold; the record follows the CCN, and the new owner steps into it. The buyer also assumes the facility's outstanding obligations and liabilities under the program — which can include exposure for prior overpayments, the obligation to satisfy outstanding plans of correction, and successor liability for certain pre-existing program debts. This is the legal logic of successor liability operating through the Medicare enrollment system: by accepting assignment, the buyer accepts the program-side consequences of what the seller did.

For analysis, this inheritance is what gives the CHOW dataset its real power. Because the compliance history travels with the CCN across a change of ownership, the dataset makes it possible to ask a question that matters enormously for patient safety and program integrity: what happens to a facility's quality and compliance after it is acquired? Does a facility's deficiency rate rise or fall once a chain operator or a private-equity buyer takes over? Are buyers acquiring already-troubled facilities, or are the troubles emerging after the acquisition? The CHOW supplies the precise pivot point — the effective date and the change of owner — around which a before-and-after analysis of the facility's CCN-keyed quality record can be built. Without the CHOW, the ownership change is invisible in the quality data; with it, the acquisition becomes an event whose consequences can be measured.

Who files a CHOW: hospitals, SNFs, and the certified-provider universe

CHOWs apply across the universe of certified Medicare providers and suppliers, but this dataset sits at the center of the consolidation story because it concentrates on the two provider types where ownership change carries the greatest stakes: hospitals and skilled nursing facilities (SNFs). These are institutional providers — large, capital-intensive, certified facilities whose participation in Medicare runs through a provider agreement and a CCN, exactly the structure the CHOW process is built around.

On the hospital side, a CHOW is the footprint of the decades-long wave of hospital mergers and acquisitions: standalone community hospitals absorbed into multi-hospital systems, regional systems combining into larger ones, for-profit operators buying nonprofit facilities and vice versa. Each such transfer of a hospital's ownership, where the buyer assumes the provider agreement, leaves a CHOW row keyed to the hospital's CCN. On the skilled nursing facility side, the dynamics are different but the data structure is the same. The nursing-home sector is characterized by frequent ownership churn, by operators that run dozens or hundreds of facilities under common control, and increasingly by financial buyers — private-equity firms and real-estate investment trusts — for whom nursing homes are an asset class. Every time one of these facilities is bought, sold, or moved between affiliated entities, the change of ownership runs through the CHOW process.

The reason hospitals and SNFs dominate is partly definitional and partly substantive. They are the institutional providers whose Medicare participation is anchored to a provider agreement that can be assumed; and they are also the providers where the public-interest stakes of an ownership change are highest, because a hospital or nursing home is not just a business but a piece of community health infrastructure whose ownership affects access, staffing, and quality of care for a defined population. That is why the transaction-level record of who is buying and selling these specific facilities — rather than physician practices or smaller suppliers — is the version of the CHOW data that carries the most analytical weight.

The consolidation wave the data exposes

What the CHOW dataset is really a record of is healthcare consolidation, and it is one of the few federal datasets that captures consolidation as a flow of discrete, dated transactions rather than as a static end-state. Three distinct consolidation dynamics all leave their mark in the CHOW record, and reading them apart is much of the interpretive work.

The first is the hospital merger wave. Over the last several decades the American hospital sector has steadily concentrated, as independent hospitals have been folded into ever-larger health systems through merger and acquisition. Each completed acquisition of a hospital's ownership, where the buyer assumes the provider agreement, registers as a CHOW — making the dataset a transaction-by-transaction chronicle of which systems have been acquiring which hospitals, in which states, and over what stretch of years. Aggregated and ordered by effective date, the hospital CHOWs trace the pace and the geography of hospital-market concentration.

The second, and in recent years the most scrutinized, is the rise of private-equity and real-estate-investment-trust ownership of nursing homes. Private-equity firms acquire nursing-home operators as investments, often financing the purchase with debt and frequently separating the operating business from the underlying real estate — selling the buildings to a real-estate investment trust and leasing them back. These transactions, and the subsequent transfers among affiliated financial entities, show up as CHOWs. The pattern matters because the ownership form has become a subject of intense policy concern: questions about whether financial ownership affects staffing, quality, and resident outcomes turn on first being able to identify which facilities are under such ownership and when they came to be — which is exactly what the CHOW record, joined to the ownership data, supplies.

The third dynamic is the multi-facility chain roll-up. A great deal of the churn in the nursing-home sector is not arms-length sales between unrelated parties but the assembly and reshuffling of chains: an operator acquiring facility after facility to build scale, or moving facilities among the many entities that make up a commonly controlled chain. A buyer that appears across dozens of CHOW records, each adding a facility to its portfolio, is a chain in the act of rolling up. Surfacing those acquisitive buyers — ranking the parties by the number of distinct CCNs they have acquired — is one of the most direct uses of the data, and the one the Python example below performs, with the heavy caveat that doing it well requires resolving the many names and shell entities a single chain operates under into the real controlling party.

Joining to the ownership and enrollment datasets

The CHOW dataset is at its most powerful not in isolation but as the transaction-level complement to two other CMS datasets, and the joins among them are what let an analyst follow a facility through every change of hands and understand the full ownership structure on each side of a deal.

The first and most important join is to the provider-ownership data. CMS publishes detailed ownership datasets — for hospitals, for skilled nursing facilities, and for other provider types — that give, for each facility keyed by CCN, the full roster of owning and managing entities and individuals, their ownership percentages, their roles, and ownership types including direct, indirect, and the financial-ownership flags that identify private-equity and real-estate-investment-trust involvement. Those datasets are static snapshots: they tell you who owns a facility now. The CHOW dataset supplies the missing temporal dimension — the transitions between snapshots. Joining the two by CCN lets an analyst pair each ownership change with the before-and-after ownership structure, turning a buyer name on a CHOW row into the full chain of entities that buyer brought with it, and anchoring the static ownership picture to the dated events that produced it.

The second join is to the provider-enrollment data. The CHOW process runs through Medicare enrollment — the new owner files the change through the enrollment system — and the enrollment datasets carry the broader record of how providers register, the enrollment identifiers, the practice and ownership information collected at enrollment, and the linkages between provider organizations. Joining CHOWs to the enrollment record connects the transaction to the formal enrollment footprint of both the seller and the buyer, helping to resolve the buyer named on a CHOW into the enrolled entity it corresponds to and to relate that entity to the other facilities it has enrolled. Used together — CHOW for the dated transactions, ownership for the structure on each side, enrollment for the formal entity footprint — the three datasets let an analyst trace a facility through its entire succession of owners and reconstruct the portfolios of the chains and financial sponsors that have been assembling them.

Analytical uses

A national, facility-resolved, date-stamped record of provider ownership changes supports a distinctive set of analyses that the static ownership snapshots alone cannot.

Tracking the consolidation curve is the most immediate use. Because every CHOW carries an effective date, an analyst can tally transactions by year and by provider type to chart the tempo of hospital and nursing-home consolidation over time — identifying the years of peak deal-making, the periods of relative quiet, and any structural shifts in the mix between hospital mergers and SNF transfers. The same tally cut by state shows where consolidation is concentrated geographically, an input to questions about local-market concentration and access.

Surfacing the most acquisitive buyers exploits the buyer field across many transactions. Ranking buyers by the number of distinct facilities they have acquired identifies the chains, systems, and financial sponsors doing the most roll-up activity, and tracing a single buyer's CHOWs over time reconstructs the chronology of how a portfolio was assembled. This is the analysis that converts a list of individual deals into a map of who controls what, and it is the entry point to studying the largest consolidators.

Measuring quality and compliance around acquisition is the analytic payoff that the inheritance of the provider number makes possible: because the CCN's compliance and quality record survives the change of ownership, the CHOW supplies the event date around which a before-and-after study of a facility's deficiencies, staffing, and outcomes can be constructed — the central method behind the research on whether private-equity and chain ownership change how facilities perform. Finally, detecting facility churn flags the CCNs that have changed hands repeatedly, which is itself a signal worth attention: a facility cycling through owners in quick succession can indicate financial distress, asset-stripping, or a property being passed among affiliated entities, all of which warrant a closer look at the facility and the parties involved.

Python workflow: pulling CHOWs from data.cms.gov

The script below pulls the change-of-ownership records from CMS's public data API on data.cms.gov, then computes three of the core views: changes of ownership by provider type (the hospital-versus-SNF split), changes of ownership by effective year (the consolidation curve), and the most acquisitive buyers (the chains and sponsors rolling up facilities). No API key is required for public data. Because CMS column labels and dataset identifiers shift between releases, the script resolves the working column names by substring at runtime and isolates the dataset id in one place; before running, the current Change of Ownership dataset id should be resolved from the data.cms.gov catalog, and any production use should page through the full result set.

import requests, pandas as pd
from collections import Counter

# CMS Change of Ownership (CHOW) data, published on data.cms.gov.
# No API key is required for public data. CMS serves each dataset
# through a Socrata-style "data-viewer" API and as a flat CSV; the
# stable way to fetch programmatically is the dataset’s data API,
# which paginates JSON rows. The dataset id below is a placeholder --
# resolve the current CHOW dataset id from the data.cms.gov catalog
# (the "Change of Ownership" listing) before running.
CATALOG = "https://data.cms.gov/data-api/v1/dataset"
DATASET_ID = "CHANGE_OF_OWNERSHIP_DATASET_ID"   # resolve from the catalog


def fetch_chows(dataset_id=DATASET_ID, page=5000):
    # Page through the dataset’s data API. CMS returns a JSON array
    # per page; an empty array signals the end of the result set.
    rows, offset = [], 0
    while True:
        url = f"{CATALOG}/{dataset_id}/data"
        r = requests.get(url, params={"size": page, "offset": offset}, timeout=120)
        r.raise_for_status()
        batch = r.json()
        if not batch:
            break
        rows.extend(batch)
        print(f"  fetched {len(rows):,} CHOW records so far...")
        if len(batch) < page:
            break
        offset += page
    return pd.DataFrame(rows)


def col(frame, *needles):
    # CMS column labels vary across releases ("CCN", "CMS Certification
    # Number", "Buyer", "Buyer Name", "Effective Date"). Resolve them by
    # substring rather than hard-coding the exact header.
    for c in frame.columns:
        u = c.upper()
        if all(n.upper() in u for n in needles):
            return c
    return None


df = fetch_chows()
print(f"Total change-of-ownership records: {len(df):,}")

c_ccn   = col(df, "CCN") or col(df, "CERTIFICATION", "NUMBER")
c_type  = col(df, "PROVIDER", "TYPE") or col(df, "PROVIDER", "CATEGORY")
c_buyer = col(df, "BUYER") or col(df, "NEW", "OWNER")
c_eff   = col(df, "EFFECTIVE", "DATE") or col(df, "EFFECTIVE")

# --- 1. CHOWs by provider type --------------------------------------
if c_type:
    print("\nChanges of ownership by provider type:")
    for t, n in df[c_type].fillna("(blank)").value_counts().head(15).items():
        print(f"  {str(t)[:40]:<40} {n:>6,}")

# --- 2. CHOWs by year (the consolidation curve) ---------------------
if c_eff:
    yr = pd.to_datetime(df[c_eff], errors="coerce").dt.year
    print("\nChanges of ownership by effective year:")
    for y, n in yr.value_counts().sort_index().items():
        if pd.notna(y):
            print(f"  {int(y)}  {n:>6,}")

# --- 3. Most acquisitive buyers -------------------------------------
# A buyer that appears across many CCNs is rolling up facilities. This
# is a first-pass tally on the raw buyer string -- real attribution
# needs entity resolution (see the caveats) because one chain files
# under many slightly different names and shell LLCs.
if c_buyer:
    print("\nTop 20 buyers by facilities acquired:")
    for buyer, n in df[c_buyer].fillna("(unknown)").value_counts().head(20).items():
        print(f"  {str(buyer)[:46]:<46} {n:>4,}")

# --- 4. Facilities that changed hands more than once ----------------
if c_ccn:
    churn = df[c_ccn].value_counts()
    repeats = churn[churn > 1]
    print(f"\nFacilities (CCNs) with more than one CHOW: {len(repeats):,}")
    print(f"Most-churned single facility: {int(churn.max())} ownership changes")

Two practical notes apply. First, the most-acquisitive-buyers tally in the script is deliberately a first pass: it counts on the raw buyer string exactly as it appears in the record, which will systematically undercount the largest consolidators because a single chain files under many slightly different names and through a thicket of shell limited-liability companies created for individual facilities or deals. A rigorous version must perform entity resolution — collapsing the many names and shells into the real controlling party — ideally by joining out to the ownership data, where the indirect-owner and chain-affiliation structure is recorded, rather than by string-matching buyer names alone. The script leaves that resolution as the natural next step. Second, for serious work the CHOW data should always be analyzed alongside the CCN-keyed ownership, certification, and quality datasets; the CHOW supplies the dated transaction, but the meaning of that transaction — who really bought the facility, and what happened to it afterward — lives in the datasets it joins to.

Limitations and analytical caveats

The change-of-ownership dataset is the most direct public record of hospital and nursing-home transactions in the Medicare system, but it carries structural limitations that an analyst must internalize before drawing conclusions from it.

Entity resolution is the hard part, and it is essential.The buyer and seller on a CHOW row are recorded as the legal entity to the transaction, and in the nursing-home sector especially that entity is very often a single-purpose limited-liability company created for the deal or the facility rather than the recognizable chain or sponsor behind it. A private-equity firm controlling two hundred facilities may appear in the data under two hundred different entity names, none of which is the firm's. Any count of “facilities acquired by buyer” that runs on the raw name field will therefore badly understate the true concentration of ownership. Resolving the names and shells to their real controlling parties — using the indirect-ownership and chain structure in the ownership data — is not optional polish; it is the precondition for the data telling the truth about who is consolidating the sector.

A CHOW is a transaction record, not a financial one.The dataset records that an ownership change occurred, who the parties were, and when it took effect — but it does not record the price, the deal terms, the debt structure, or the real-estate arrangements that often define the economics of a healthcare acquisition, particularly in private-equity and REIT deals where the separation of the operating business from the building is the whole strategy. The CHOW tells you that a facility changed hands; it does not tell you for how much or on what terms. Those questions require other sources entirely, and treating the CHOW as a complete account of a transaction over-reads what it contains.

There is reporting lag and a definitional boundary on what counts. A change of ownership appears in the data after it is processed through the Medicare enrollment system, so the most recent transactions are systematically under-represented in any snapshot, and recency metrics will understate the latest activity at the leading edge. The dataset is authoritative for established patterns and multi-year trends; it is not a real-time monitor of last quarter's deals. There is also a definitional boundary: the dataset captures transactions that ran through the Medicare CHOW process for the covered provider types — principally hospitals and skilled nursing facilities — and not every economically meaningful change in control of a healthcare business takes that form. A purely corporate change that does not trigger a CHOW, or a transaction among provider types outside the dataset's scope, will not appear, so the CHOW record is a faithful ledger of the changes that run through this specific process rather than of all healthcare M&A.

Assignment is the default but not the whole story.Because most CHOWs proceed by assignment of the existing agreement, the dataset is largely a record of facilities whose provider number and history carried over — but the cases where assignment was rejected, and the facility re-enrolled as a new provider, are exactly the cases where the buyer chose to break the continuity, often to escape the seller's liabilities or compliance baggage. Those rejections are analytically interesting precisely because they are rare, and an analysis that ignores the disposition field, or assumes every CHOW carried the history forward, will miss the deliberate discontinuities that the assume-versus-reject choice creates.

Held with these caveats in mind, the cms_provider_chow table is a uniquely valuable resource: a facility-resolved, date-stamped record of who has bought and sold the country's hospitals and nursing homes, keyed to the same CMS Certification Number that carries each facility's compliance history forward through every change of hands — the transaction-level ledger of a consolidation wave whose consequences for cost, access, and quality are written, facility by facility, in the data it joins to.