CMS Home Infusion Therapy Suppliers: The Federal Record of a New Medicare Benefit

Some of the most demanding medicine in the country is now delivered at a kitchen table: a bag of antibiotics, a course of chemotherapy, an immune-globulin drip, a night of parenteral nutrition—run into a vein or under the skin in a patient's own home instead of a hospital ward. For decades Medicare paid for the drug and the pump but not cleanly for the nurse who hangs the bag and teaches the family. That changed on January 1, 2021, when Medicare's permanent home infusion therapy benefit took effect. CMS keeps the record of who is qualified to provide it: roughly 324 home infusion therapy suppliers, one enrollment each—the supply map of a benefit only a few years old.

This article covers what the home infusion therapy supplier dataset is and how it differs from the drug and equipment records people usually associate with infusion; what home infusion therapy actually is—the therapies, the components, and the difference between the drug, the equipment, and the professional services the benefit pays for; the genuinely new statutory history, from the 21st Century Cures Act of 2016 that created the permanent benefit through the Bipartisan Budget Act of 2018 temporary transitional payment that bridged the gap until the permanent benefit took effect on January 1, 2021; the accreditation requirement that is the reason the enrolled-supplier count is so small and concentrated; the schema and the identifiers that anchor it; how the file joins to the other CMS provider-enrollment datasets; a Python workflow that pulls the suppliers, maps them by state, and compares the count against population; and the caveats—a brand-new benefit, a supplier that is not a service-location count, and enrollment that is not access—that every analyst must internalize before drawing conclusions from a dataset this young.

What the dataset is

CMS enrolls providers and suppliers before it pays them. Every entity that bills Medicare—a hospital, a physician group, a durable-medical-equipment company, a home health agency—must first complete the Provider Enrollment, Chain, and Ownership System (PECOS) process, supplying its legal identity, its location, its ownership, and its supplier type, and receiving in return the identifiers that let it submit claims. The home infusion therapy supplier file is the public extract of that record for one specific, new supplier type: not the drugs infused, not the pumps and supplies, not the patients treated, but the administrative fact of which entities are enrolled and qualified to bill Medicare as a home infusion therapy supplier for the professional services that accompany home-administered infusion drugs.

That makes this a supply-side dataset, and a small one. It answers the question “where can a Medicare patient get home infusion therapy services?” rather than “how many patients receive it?” or “what was infused?” In our database it is stored as the table cms_home_infusion, with the grain of one row per enrolled supplier and roughly 324 suppliers—a deliberately small census, because, as the accreditation section explains, qualifying to bill as a home infusion therapy supplier is a high bar that few entities clear. (Because a single supplier can list several service areas, the raw published file carries more lines than that; the supplier count is the distinct enrollment, not the row.) Because enrollment is the front door to billing, the file is the cleanest available map of the formal Medicare home-infusion-services footprint: an entity that is not in it is, by definition, not billing Medicare under this new benefit. The columns capture the supplier's identity, its location, and its enrollment status:

enrollment_id        -- PECOS enrollment identifier for the supplier
ccn_or_npi           -- CMS / National Provider Identifier for the supplier
provider_type        -- home infusion therapy (HIT) supplier
organization_name    -- the supplier's legal / doing-business-as name
address, city        -- physical street location of the supplier
state                -- two-letter state / territory code
zip_code             -- ZIP of the supplier location
enrollment_status    -- enrollment status (e.g. approved / active)
incorporation_date   -- date the organization was incorporated
proprietary_nonprofit -- ownership character flag where reported
practice_location    -- service / practice location of record

The enrollment_id is the load-bearing column. The PECOS enrollment identifier is the persistent key under which the supplier's enrollment application, its reassignments, and its disclosed ownership live, and it is the join key into the rest of the CMS enrollment universe. The ccn_or_npi field carries the CMS Certification Number or National Provider Identifier that ties the supplier to claims and to the other provider files. The provider_type value is what defines membership in this table at all—every row is a home infusion therapy supplier, the supplier type created specifically to administer the new benefit. The location columns (state, zip_code, address) are what make the file a map; joined to population and geography, they convert a short administrative list into a picture of where home infusion services are available and where patients must look elsewhere. The enrollment_status field is the filter that separates suppliers currently able to bill from those that have dropped out—the difference, in a dataset this small, between counting paper and counting the live network.

What home infusion therapy is

Home infusion therapy is the administration of drugs and biologicals intravenously—into a vein—or subcutaneously—under the skin—in a patient's home, rather than in a hospital, an infusion center, or a physician's office. The therapies span a wide clinical range. The most common is intravenous antibiotic, antifungal, and antiviral therapyfor serious infections that need weeks of medication but no longer need a hospital bed. Others include immune globulin for immune deficiencies and certain neurologic conditions, chemotherapy and other antineoplastic agents for cancer, inotropic therapy for advanced heart failure, pain management, hydration, and—among the most complex—parenteral nutrition, in which a patient who cannot absorb food through the gut is fed a precisely formulated nutrient solution directly into the bloodstream. What unites them is that the medication must be delivered slowly and continuously through a catheter or an infusion pump, often over hours, and often for weeks or months at a stretch.

The single most important conceptual point about the dataset follows from how Medicare divides up what home infusion involves. A home infusion episode has three distinct components, paid in three different ways. First is the infusion drug itself—the antibiotic, the immune globulin, the nutrition solution. Second is the equipment and supplies: the external infusion pump, the tubing, the catheters, and the related items, which Medicare has long covered under the durable medical equipment (DME) benefit. Third—and this is the genuinely new piece—are the professional services: the skilled nursing visit that establishes the infusion, the training of the patient and family to run the pump and care for the line, the monitoring and assessment, and the coordination of care. The home infusion therapy benefit pays for that third component, the professional services furnished on the days the drug is administered. The cms_home_infusion table is therefore the record of the suppliers qualified to furnish and bill those professional services—not the drug companies, not the pump vendors, but the entities standing up the clinical service that makes home infusion safe.

A genuinely new benefit: from Cures to permanence

The reason this dataset is worth a deep dive is that it documents the supply side of a benefit that is, by the standards of Medicare, brand new. For most of Medicare's history there was a gap: Medicare covered the infusion drug and the DME pump and supplies, but the professional services that make home infusion clinically possible—the nursing, the training, the monitoring—fell into the cracks between benefit categories. A patient could get the drug and the pump shipped to the house, but the program had no clean, dedicated way to pay for the skilled service of starting the therapy and keeping it safe. Closing that gap took two acts of Congress and a multi-year ramp.

The first step was the 21st Century Cures Act of 2016. Among its many provisions, Cures created the new Medicare home infusion therapy benefit itself—directing CMS to pay separately for the professional services associated with home infusion drugs and setting the permanent benefit to take effect on January 1, 2021. It is the statutory origin of the entire home infusion therapy supplier category: the law that established the benefit and the “qualified home infusion therapy supplier” enrollment type that fills this dataset.

Because the permanent benefit could not stand up overnight, the Bipartisan Budget Act of 2018 bridged the gap. It established a temporary transitional payment for eligible home infusion suppliers, running from January 1, 2019 through the end of 2020— a placeholder that began paying for the service while CMS finished designing and implementing the permanent structure the Cures Act had called for. The transitional payment ended the day before the permanent benefit began, so the two acts dovetail: the Cures Act created the permanent benefit and set its January 1, 2021 start date, and the Budget Act paid the bridge in between.

That start date is the load-bearing fact of the dataset: January 1, 2021 is the moment the permanent benefit—and with it the formal “qualified home infusion therapy supplier” enrollment category—came fully into existence. Everything in cms_home_infusion postdates it. This is what distinguishes the file from the mature, decades-old enrollment files for hospitals or physicians: it is not a stable census of a long-settled supplier population but a snapshot of a population that, in dataset terms, started from zero only a few years ago and is still forming. The growth of the supplier base since the start of 2021—how fast new suppliers enrolled, where they appeared first, and whether the count is still climbing—is therefore one of the most interesting things the data can show, in a way that simply is not available for older, fully built-out benefits.

Accreditation: why the supplier count is small

The most striking feature of the dataset to anyone used to other CMS files is its size. There are tens of thousands of enrolled DME suppliers and home health agencies; there are only a few hundred qualified home infusion therapy suppliers. The roughly 324 suppliers are not an artifact of incomplete data—they reflect a genuinely small and concentrated supplier base, and the reason is the qualification standard.

A qualified home infusion therapy supplier must be accredited by a CMS-recognized accrediting organizationand must meet the program's standards before it can enroll and bill for the professional services. Accreditation is not a formality; it is a substantive demonstration that the supplier can safely furnish skilled infusion nursing, train patients and caregivers, monitor therapy, ensure pharmacy and clinical competencies, and manage the considerable safety risks of administering high-acuity drugs outside a clinical facility. CMS recognizes a defined set of accrediting organizations to perform this review against published standards, and only after a supplier earns that accreditation can it qualify. The effect is a high barrier to entry that filters the universe down sharply: the entities that clear it tend to be the larger, established infusion and specialty-pharmacy operators with the clinical infrastructure to meet the standards, not small or casual entrants. That is precisely why the count is small, why it is concentrated among sophisticated suppliers, and why each enrolled supplier in the file represents a meaningful, vetted node of capacity rather than a name on a long list. It is also why, for this benefit, the supply question is real and pressing: when the qualified-supplier population numbers in the hundreds nationally, geographic gaps in coverage are not hypothetical.

The schema and the identifiers that anchor it

The fields are best understood through the identifiers that make the file joinable to the rest of the CMS data ecosystem. The enrollment ID is the spine. It is the PECOS-side identifier under which the supplier's enrollment record, its reassignments, and its disclosed ownership live, and it is the key that lets an analyst follow a home infusion therapy supplier from this snapshot into the broader provider-enrollment and ownership files. Because the same enrollment ID grammar runs across the CMS enrollment universe, the small home infusion file is not an island—it wires directly into the same graph as every other enrolled provider type.

The CMS Certification Number or National Provider Identifieris the second key, and it points toward claims and certification. The CCN is the structured identifier CMS assigns to certified institutional providers; the NPI is the universal provider identifier that appears on claims. Whichever applies, it is the field that ties the supplier to the records of what it actually billed and to its certification status. The enrollment status field records where the supplier stands in its lifecycle—approved and active, versus deactivated or revoked—which, in a file of only a few hundred suppliers, is the crucial distinction between the number of suppliers that have ever enrolled and the number that can currently bill. The organization name and address fields supply the human-readable identity and the geography, and the proprietary-versus-nonprofit character, where reported, hints at the ownership structure the ownership files spell out in full. The discipline the schema demands is the familiar one: the keys (enrollment ID, CCN/NPI) are for joining, the descriptors (name, location, ownership character) are for grouping and mapping, and the status is for filtering down to the suppliers that are actually live—and in a brand-new benefit that filter matters more, not less, because the population is still churning as it forms.

Joining to the other CMS provider-enrollment data

The home infusion therapy supplier file is most valuable not in isolation but as one small, new node in the connected CMS provider graph, and the enrollment ID and CCN/NPI are what wire it in. Three joins matter most.

The first is to the broader provider-enrollment record. The same enrollment ID that keys this file keys the wider Medicare enrollment files, where a supplier's reassignments and the individual clinicians who bill under it can be related. Joining the home infusion supplier to the individual-clinician record—through the National Provider Identifiers that reassign to the supplier's enrollment—lets an analyst ask who actually furnishes the service: how many nurses and other clinicians a given home infusion therapy supplier bills through, and whether a supplier covering a wide rural geography does so with depth or with a thin staff stretched across distance.

The second join is to the ownership files. CMS's all-owners disclosures, keyed by the same enrollment ID, expose who owns and controls an enrolled supplier. Because qualified home infusion suppliers skew toward larger, established infusion and specialty-pharmacy operators, the ownership join is where consolidation becomes visible: whether the few hundred suppliers nationally trace back to a still smaller number of corporate parents, and whether the for-profit and roll-up dynamics documented across the rest of post-acute and home-based care are present in this new benefit from the start. With a supplier population this small, even modest concentration at the ownership level can mean that a handful of companies effectively define national access.

The third join is to the rest of the home-based care landscape—and to geography and need. Home infusion therapy sits alongside the home health benefit, hospice, and DME as part of the larger map of care delivered where people live, and the same identifiers let an analyst relate a home infusion supplier to the home health agencies and DME suppliers operating in the same markets. Joined to census population and to geography, the supplier locations place each enrollment in demographic context—urban versus rural, the states and regions with several suppliers versus the ones with one or none. This is the join that turns the file from a short list into an access map and supports the policy question the data exists to answer: in a benefit only a few years old, has the supply reached the patients who need it, or are there whole states where a Medicare patient who would benefit from home infusion has no qualified supplier nearby?

Analytical uses

A national, geocoded census of the qualified home infusion therapy supplier base—small, new, and still forming—supports a distinctive set of analyses that no drug-side or claims dataset can.

Access mapping is the most immediate and most important use. Because each enrollment carries a location, an analyst can place every qualified supplier on the map, count suppliers by state, and normalize by population to compute suppliers per million residents. With a national population in the hundreds, the headline finding is almost always about gaps: the states and rural regions with few suppliers or none at all, where a patient who could safely receive antibiotics or nutrition at home instead faces a longer hospital stay or a daily drive to an infusion center because no qualified supplier serves the area. For a young benefit, this map is the single most policy-relevant product the data yields.

Tracking growth since 2021 exploits the benefit's youth directly. Because the qualified-supplier category did not exist before January 1, 2021, successive snapshots of the file show the network being built in real time: how quickly suppliers enrolled after the permanent benefit took effect, which states gained capacity first, and whether the count is still climbing, has plateaued, or has begun to thin. That trajectory is a leading indicator of whether the benefit is taking hold as Congress intended. Comparing supply against need goes a step further: joined to the populations most likely to benefit—older adults, patients with the conditions home infusion treats—and to the rest of the home-based care map, the supplier file shows where capacity is aligned with demand and where it is not.

Finally, the consolidation picture uses the ownership join to ask whether a benefit barely a few years old is already concentrated. Counting the corporate parents behind the qualified suppliers, and watching how that concentration moves as the network grows, reveals whether home infusion access nationally rests on a broad base of independent suppliers or on a handful of large operators—a structural question that matters precisely because, in a new benefit with a small supplier count, the answer is set early and shapes access for years.

Python workflow: suppliers by state against population

The script below pulls the home infusion therapy supplier file from CMS's data.cms.gov API, maps the suppliers by state, and computes two of the core metrics: the count of suppliers per state, and suppliers per million residents—the access metric that matters most for a benefit this new and this concentrated. No API key is required for the CMS data. Because the CMS datastore re-versions its files and the column names drift between releases, the script discovers the working state and key column names at runtime rather than hard-coding them, and resolves the dataset identifier from the catalog. It also counts distinct enrollment IDs rather than raw rows, because a single supplier can list several service areas. The population join uses the Census American Community Survey API.

import requests, pandas as pd

# CMS Provider Data / enrollment files on data.cms.gov -- no API key
# required for public data. CMS publishes the Medicare enrollment record
# for home infusion therapy (HIT) suppliers: the suppliers qualified to
# bill for the professional services that accompany home-administered
# infusion drugs. The file is small -- a few hundred suppliers -- because
# qualification requires accreditation by a CMS-recognized organization.
BASE = "https://data.cms.gov/data-api/v1/dataset"

# Replace with the current home infusion therapy supplier dataset UUID
# from data.cms.gov. CMS re-versions its files, so resolve the working
# UUID from the catalog page if a request 404s.
HIT_DATASET_UUID = "REPLACE_WITH_CURRENT_HIT_SUPPLIER_UUID"


def fetch_all(uuid, page_size=2000):
    # Page through the datastore endpoint until a short page ends it.
    rows, offset = [], 0
    while True:
        params = {"size": page_size, "offset": offset}
        r = requests.get(f"{BASE}/{uuid}/data", params=params, timeout=120)
        r.raise_for_status()
        page = r.json()
        if not page:
            break
        rows.extend(page)
        if len(page) < page_size:
            break
        offset += page_size
    return pd.DataFrame(rows)


def _find(cols, *needles):
    # First column whose name contains all needles (case-insensitive).
    for c in cols:
        u = c.upper()
        if all(n.upper() in u for n in needles):
            return c
    return None


# --- Pull the supplier file --------------------------------------------
hit = fetch_all(HIT_DATASET_UUID)
hit.columns = [c.strip() for c in hit.columns]
state_col = _find(hit.columns, "STATE") or _find(hit.columns, "ST")
key_col = (_find(hit.columns, "ENROLL", "ID") or _find(hit.columns, "CCN")
           or _find(hit.columns, "NPI"))
# One supplier can carry several service-area rows, so always count
# DISTINCT enrollment IDs, never raw rows.
print(f"Loaded {hit[key_col].nunique():,} home infusion therapy suppliers")

# --- Metric 1: suppliers by state --------------------------------------
by_state = hit.groupby(state_col)[key_col].nunique().sort_values(ascending=False)
print(by_state.to_string())

# --- Metric 2: suppliers per million residents -------------------------
# Census ACS state population (key optional). Join the supplier counts to
# population to get suppliers per million people -- the headline access
# metric for a benefit this new and concentrated.
ACS = "https://api.census.gov/data/2022/acs/acs5"
resp = requests.get(ACS, params={"get": "NAME,B01003_001E", "for": "state:*"},
                    timeout=60).json()
pop = pd.DataFrame(resp[1:], columns=resp[0])
pop["pop"] = pop["B01003_001E"].astype(float)
pop["abbr"] = pop["NAME"].map(lambda n: n)  # map to USPS code in production

access = by_state.rename("suppliers").to_frame().join(
    pop.set_index("NAME")["pop"])
access["per_million"] = access["suppliers"] / (access["pop"] / 1_000_000)
print(access.sort_values("per_million").to_string())
states_with_none = "states with zero enrolled suppliers must be added back as 0"
print(states_with_none)

Two practical notes apply, and both stem from how small the file is. First, the access metric is dominated by zeros: with only a few hundred suppliers spread across fifty states and the territories, many states will have a handful and some will have none, so the most important line in any honest analysis is the one that adds the zero-supplier states back in. Computing suppliers per million only for states that have at least one supplier silently hides exactly the access gaps the analysis exists to find; the script flags this explicitly, because dropping the empty states is the single easiest way to make a coverage problem disappear from a chart. Second, a supplier row is an enrolled entity, not a service-location count: a single qualified supplier may serve a wide geography from one enrollment and may even list several service areas in the file, so the map shows where suppliers are headquartered or enrolled, which is a useful proxy for access but not the same as the area a supplier actually reaches. For a rigorous access study an analyst should pair the supplier locations with the service areas implied by claims and with the home health and DME networks that operate alongside home infusion in the same markets.

Limitations and analytical caveats

The supplier file is the cleanest public census of the qualified Medicare home infusion therapy supplier base, but it carries structural limitations—sharpened by the benefit's youth—that an analyst must internalize before drawing conclusions.

This is a brand-new benefit, and the data is still forming.The qualified-supplier category did not exist before January 1, 2021. The file is therefore not a settled census but a snapshot of a population in its early years, when enrollment is still ramping, suppliers are still entering, and the geographic footprint is still filling in. A gap on the map today may close next year as a supplier enrolls; a count that looks low may simply reflect how recently the benefit began. Conclusions about adequacy should be framed as statements about a specific point in a still-rising curve, not as verdicts on a mature program.

Enrollment is not access, and a supplier is not a service location. The file records that an entity is qualified and enrolled to bill Medicare for home infusion professional services—nothing about how many patients it serves, whether it is accepting new ones, how far it will travel, or what its capacity is. A single enrolled supplier may cover a large region, and a state with one supplier on the map is not necessarily a state with adequate access. Counting enrollments and calling the result access overstates what the file can bear; the enrollment count is the supply ceiling, not the service reality.

It is the professional-services supplier, not the whole infusion picture. The benefit, and therefore this file, is about the professional services— the nursing, training, and monitoring—that accompany home-administered infusion drugs. The drug itself and the DME pump and supplies are covered and recorded separately, under different benefit categories and in different data. An analyst who reads this file as the totality of Medicare home infusion will miss most of the spending and most of the activity, which live on the drug and DME sides. The supplier file is one component of a three-part picture, and it is specifically the newest and smallest component.

The small count makes every record consequential—and every error consequential too. In a file of tens of thousands of rows, a few miscoded or stale records wash out. In a file of a few hundred suppliers, a handful that have closed but remain enrolled, or a few that enrolled but never began billing, can meaningfully move a per-state count or a per-capita rate. The enrollment-status filter is essential, the data should be read as a near-current snapshot rather than a real-time directory, and any state-level claim about access should be checked against the actual identity of the small number of suppliers it rests on.

Held with these caveats in mind, the cms_home_infusion table is a uniquely valuable resource: a system-resolved, geocoded census of the suppliers qualified to bring hospital-grade infusion care into people's homes under a benefit that did not exist a few years ago—the supply map of one of the youngest corners of Medicare, where the question is not how a long-settled program is performing but whether a new one is reaching the patients it was built for, written one enrollment at a time into the federal record.