CMS Home Health Cost Report Data: A CCN-Level PE Screen

Executive Summary

A public CMS cost-report row supports an initial screening fact about a home health agency only when the claim is scoped to one provider, one period and one payer view. This report treats the CMS Certification Number (CCN) as the unit of analysis. The CMS public use file (PUF) for home health agency (HHA) cost reports is described as "organized by CMS Certification Number" ^[1], and CMS states that it holds a subset of the variables in the Healthcare Cost Report Information System (HCRIS) ^[2]. The file is unedited, so outlier, partial-period and duplicate handling falls to the analyst.

The release picture is less tidy than a "latest year" label suggests. A harvested copy of the CMS catalog record listed a 2023 vintage ^[3]. The live catalog read on 2026-10-20 lists a vintage titled 2024-01-01 ^[1]. The separate HCRIS page showed new data "updated 9/30/2026, released 10/06/2026" ^[4]. A July 2026 release notice reported for that page could not be located, so this report does not rely on it. The PUF methodology states a July 17, 2025 HCRIS snapshot ^[2], which does not explain files published in late 2026.

Scope drives the numbers. MedPAC reported a 21.2 percent aggregate fee-for-service (FFS) Medicare margin for freestanding HHAs in 2024 ^[5], against a 5.0 percent all-payer margin ^[5]. KFF reported that 55 percent of eligible Medicare beneficiaries were in Medicare Advantage in 2026 ^[6], and MedPAC found only about 8 percent of agencies serve FFS Medicare beneficiaries exclusively ^[5]. Use all-payer net patient revenue for the revenue screen below; a cost report row is never a consolidated platform revenue figure.

Identity and time are the other two controls. A CCN persists through a standard change of ownership because the provider agreement is assigned to the new owner ^[7], but a corporate stock sale is not a change of ownership at all ^[7]. CMS ownership files hold only currently active enrollments and do not identify the ultimate parent ^[8]. Cost reporting periods may run from 28 to 458 days ^[9], so the report supplies a worksheet that normalizes a flow as value / report days x 365, labels it annualized, and shows 180, 365 and 400 day sensitivities. It closes with a three-way adjudication: clears, fails or stays unresolved.

Introduction and Background

This report is written for origination analysts at private equity firms who screen US Medicare-certified home health agencies. It answers a narrow operational question: when does a public CMS cost-report row support an initial utilization or financial fact, which provider does that row belong to, and when should the mandate criterion be left unresolved? It does not value an agency, rate care quality, or estimate the consolidated revenue of a platform. Docket, a deal-origination and acquisition target research platform, publishes this as general method guidance; nothing here describes a specific engagement or target list.

The query "cms home health cost report data" is ambiguous, because three products answer to similar names. The first is the CMS HHA cost report PUF on data.cms.gov, a curated, agency-level file. The second is the raw HCRIS extract on cms.gov, published as one zipped file for each fiscal year. The third is New York State's Home Care Cost Report, a separate state product required of agencies operating certified home health agencies, licensed home care services agencies and fiscal intermediaries ^[10]. The state report covers calendar year 2025 ^[10], captures costs regardless of payor source ^[10], and identifies entities by operating certificate rather than CMS CCN ^[10]. Nothing in this report applies to it.

The Medicare cost report exists to determine Medicare program costs. The form instructions state that the report gathers information to determine costs for services rendered to Medicare beneficiaries ^[11]. Regulation requires annual cost reports from providers ^[12], and CMS says Medicare-certified institutional providers submit one to a Medicare Administrative Contractor (MAC). The strengths are provider-level identifiers, visit counts, cost centers and financial statements. The limits are a Medicare-oriented structure, a long lag, and no consolidated view of an owner's other entities.

The market context sets the stakes: MedPAC counted over 12,000 HHAs certified to participate in Medicare in 2024 ^[5], so a screen that mistakes a provider-scoped Medicare row for a platform total could misstate size by a wide margin.

Sources are primary, and where they conflict the conflict is shown rather than assumed away. Anything touching legal treatment, accounting classification or a transaction decision should be reviewed by qualified counsel or a healthcare reimbursement accountant; this report cites official guidance but has no professional review.

Key Changes in the Evidence Landscape

Change 1: A curated CCN-level file now sits beside the raw HCRIS extract

The two CMS products differ in structure. The PUF description says it provides select measures from the annual cost report, including cost and charges by cost center in total and for Medicare, Medicare settlement data and financial statements ^[2]. The raw HCRIS extract is organized differently. Each fiscal-year zip contains the Report (Rpt), Numeric (Nmrc) and Alpha-Numeric (Alphnmrc) files ^[4]. CMS describes the Rpt file as carrying the report identification fields, including the status of the cost report ^[13], and the Rpt_Rec_Num as the unique identifier that links the tables ^[13].

ResDAC, a research data assistance center for Medicare data, notes that the Rpt file has one record for each cost report ^[14], that the numeric and alpha-numeric files contain no facility identifier ^[14], and that each of their rows holds a single value from a single report ^[14]. An analyst who wants a flat CCN table can take the PUF; one who needs report status or worksheet cells outside the PUF must work from the raw files.

The PUF's identifier field, rpt_rec_num, is "the report number assigned by CMS to each specific cost report" ^[15]. That is a report-level key, not a CCN-level key, so more than one row per CCN is possible.

Change 2: Release dates and year labels no longer line up

Three dates need to be kept apart, and this report could not fully reconcile them.

• Catalog vintage: a harvested copy of the CMS catalog record titled the latest distribution "2023-12-04" ^[3] and showed a modified date of 2026-04-13 ^[3]. The live catalog read on 2026-10-20 lists a vintage titled 2024-01-01 ^[1] with a file named CostReport_HHA_2024.csv. The same dataset identifier appears in both records ^{[3] [1]}, so a rolling API address may have been repointed to a newer vintage.
• HCRIS release: the general CMS cost reports page banner reads "updated 9/30/2026, released 10/06/2026" ^[4], and its table lists the CMS-1728-20 series for fiscal years 2020 through 2026 ^[4]. A dedicated FY2026 data-files page exists ^[16]. A July 2026 notice reported for this page was not found in the page this research opened, so it is treated as unverified.
• Methodology snapshot: the PUF methodology PDF says its data match the HCRIS CSV dataset as of July 17, 2025 ^[2]. An earlier methodology document carries the same "identical" language. A snapshot taken in July 2025 would predate reports filed for later periods, and the documents do not say whether the newest files were refreshed from a later snapshot.

The year label is also not a calendar year of operations. ResDAC explains that CMS fiscal-year cost report files are usually defined on the federal fiscal year ^[17]. The first record returned by the PUF's live API on 2026-10-20 had a fiscal period beginning 2023-10-13 and ending 2023-12-31 ^[18], which fits a federal-fiscal-year-2024 file by begin date. That fit is an inference, not a CMS statement. The practical rule: never use a file's year label as the period; use the begin and end dates.

The resolution this report recommends is empirical: record the download date, row count and set of report numbers, then compare that set to the raw HCRIS release for the same fiscal year.

Change 3: Two form generations overlap in 2020 and 2021

CMS-1728-20 applies to cost reporting periods beginning on or after January 1, 2020, and ending on or after December 31, 2020 ^[19]. The raw 1728-94 data set runs from FY1994 ^[20] while the 1728-20 data set begins in FY2020 ^[19]. The general page's table lists the older series ending in 2021 ^[4], so fiscal years 2020 and 2021 can appear under both forms. A multi-year screen therefore has to record the form version on every row.

Change 4: Ownership joins now depend on separate, current-only enrollment files

CMS publishes Provider Enrollment, Chain and Ownership System (PECOS) extracts for HHAs. The catalog describes one as enrollment information on HHAs currently enrolled in Medicare ^[1] and another as ownership information on the same population. The guidance document says the files capture only currently active enrollments as of the PECOS data version ^[8], so a historical ownership trail has to be rebuilt from successive snapshots.

Resolving a CCN to an Agency, a Legal Entity and a Platform

What a CCN identifies

A CCN has six digits. The first two are the state code and the last four identify the facility type ^[21]. The State Operations Manual lists 3100-3199 as the Home Health Agency series, with continuation ranges 7000-8499 and 9000-9799 ^[22]. Medicaid-only (title XIX-only) providers are identified by a separate numbering scheme ^[21]. HHA branches carry the parent's CCN with modifications ^[21], which means the six-character CCN is a parent or subunit identifier, not a location identifier.

The Care Compare provider file lists every HHA registered with Medicare ^[23], adds a coarse control type rather than a named owner ^[24] and the original certification date.

What continuity across ownership changes means

CCN continuity is the default but not a guarantee. Under 42 CFR 489.18, the existing provider agreement is automatically assigned to the new owner ^[7]. The State Operations Manual says CMS does not change a CCN merely because the institution has been sold ^[21]. Three exceptions matter to a screen.

• Stock deals: transfer of corporate stock, or a merger of another corporation into the provider corporation, does not constitute a change of ownership ^[7]. The CCN and legal entity stay unchanged while the economic owner changes.
• Initial-enrollment restriction: CMS implemented a temporary nationwide HHA and hospice enrollment moratorium on May 13, 2026, for six months. It covers initial applications, including non-exempt changes in majority ownership; applications submitted during the moratorium will be denied ^[25]. Applications received by the Medicare contractor before implementation are excluded ^[26]. Have qualified counsel confirm the transaction-specific treatment and current moratorium status before assuming a new Medicare enrollment is available.
• The 36-month rule: if majority ownership of an HHA changes by sale within the regulatory window, the provider agreement and billing privileges do not convey to the new owner ^[27]. The regulation requires both the current and the prospective owner to submit enrollment applications ^[27] and provides a cost-report exception where the HHA submitted two consecutive years of full cost reports since initial enrollment or the last change in majority ownership, whichever is later; low-utilization and no-utilization reports do not qualify as full reports ^[28]. Have qualified counsel review the exception against the specific transaction history. "Change in majority ownership" is defined by acquiring more than a 50 percent direct ownership interest ^[29].

A change of ownership also ends a cost reporting period. A provider that changes ownership files a cost report beginning with the first day not included in a previous period ^[30], which generates a short final report for the seller's period. The manual adds that an accepting owner also assumes the former owner's liabilities ^[21]. For a screen, this means a CCN's history can contain two entities' cost reports, and the Worksheet S-2 question asking whether the HHA changed ownership before the period began ^[11] is a flag to read, not ignore.

Linking the CCN to a proposed platform

The curated PUF should not be treated as a verified ultimate-parent mapping. Full reports can contain home-office identification (Worksheet S-2 Part I line 17), related-party ownership/control relationships (Worksheet A-8-1 Part II), and new-owner information submitted with a CHOW report ^[9]. Reconcile these clues with PECOS and transaction records before assigning a platform. CMS states that cost reports can be linked to PECOS-based ownership files using the Medicare provider ID, the CCN ^[8]. The ownership file lists associates with active ownership interest or managing control, and regulation requires providers to give contractors information identifying the parties that submit program cost reports ^[12]. Three limits apply.

• One entity, several CCNs: most associate identifiers map to a single CCN, but some map to several ^[8]. CMS also distinguishes a provider defined by its tax identification number from CCN-defined facilities.
• No ultimate parent: subsidiaries of owners are not identified in these data ^[8]. A platform name therefore needs separate evidence such as filings or transaction announcements.
• Formatting and gaps: PECOS files keep leading zeros that non-PECOS files may drop ^[8], and the files may omit a small number of enrollments with data quality issues.

Peer-reviewed work shows how researchers have handled the join; an earlier study linked cost reports to quality data using the Medicare provider number ^[31]. A JAMA Health Forum study identified 749 unique HHAs acquired by private equity between 2006 and 2024 ^[32]. The authors verified transactions against public sources ^[32], obtained each agency's CCN from Provider of Services files ^[32] and note that smaller or unreported transactions may be missing ^[32]. MedPAC makes the same caution: ownership counts may reflect analytic criteria unique to each analysis ^[33]. Ownership evidence is a second source, with its own error rate, and the join should be recorded as "proposed" until that evidence is reviewed.

Who is in the file and who is not

Form CMS-1728-20 must be used by all freestanding HHAs ^[9]. HHAs that are part of a hospital healthcare complex use the hospital form ^[11], and those in a skilled nursing facility complex use the SNF form ^[11]. The older form notes that it is not used by hospital-based or SNF-based HHAs ^[34]. MedPAC reports that freestanding HHAs were about 95 percent of agencies, and it excludes facility-based agencies from margin analysis because their cost structures can include parent-facility overhead ^[5]. Hospital-based HHAs are therefore not visible in the HHA file even when they are real Medicare-certified targets.

Full reports are also not universal. One peer-reviewed study restricted its sample to HHAs large enough to be required to submit full Medicare cost reports ^[31]; in that study's 2014 to 2017 data, those were 68 percent of the roughly 12,000 agencies nationally ^[31]. A missing CCN row is not evidence of a missing agency. CMS adds that a provider absent from the public files is most likely there because HCRIS rejected the submission ^[13].

Field Dictionary and the CCN-Level Worksheet

Defining the claim before the variable

Every screen should begin with a one-line claim in four parts: the entity (one CCN), the period (begin and end dates), the scope (Medicare line or total) and the measure type (utilization, cost, charges or a financial-statement line). (Hypothetical Example) "Agency X has $4 million of revenue" fails all four tests; "CCN 000001 reported total visits of N for the period 2023-10-13 to 2023-12-31" passes.

Variables that carry the most risk

• Visits, total versus Medicare: the Medicare visit field is defined as total program visits rendered during the entire cost reporting period ^[15], and the all-payer field as total visits for the same period ^[15]. The form instructs Medicare patients only in the Medicare columns ^[9] and total HHA visits and census in columns 7 and 8 ^[9].
• Total cost: the dictionary defines the field as total cost for the facility ^[15]. The cost comes from Worksheet A, the trial balance of expense accounts ^[9].
• Medicare cost: this is a calculated field. Medicare cost of services is derived by multiplying the average cost per visit by Medicare visits ^{[15] [9]}, so it is an allocation, not a separately booked amount.
• Gross and net patient revenue: gross revenue comes from the HHA accounting books ^[15], and net patient revenue is gross revenue minus allowances and discounts ^[15]. The allowances include provisions for bad debts and contractual adjustments.
• Managed care revenue: revenue from a managed care program is entered in the "other" column, not the Medicare column ^[9]. Medicare Advantage revenue is therefore outside the Medicare column.
• Medicare payment: the nearest Medicare payment fields are the Worksheet D PPS payment lines, and they cover only payment amounts associated with the period's completed episodes ^[15].
• Balance sheet items: total assets are a point-in-time sum and are not flows, so they are never annualized.

The dictionary used for this analysis is the October 2024 PDF; the current dictionary page is a JavaScript application that could not be rendered, so field names should be confirmed against the downloaded file's header. The live API returned 201 columns with keys such as "Provider CCN" ^[18] and "Total HHA Medicare Program Cost". Dictionary labels and API keys differ slightly (the dictionary writes "Avg. Cost Per Visit"; the API omits the periods). No report-status or amendment field appeared among the 201 keys, so status requires the raw HCRIS Rpt file.

The worksheet

Table 1 below is the reference asset: one row per CCN-report observation, with the columns an analyst should populate before any number is quoted.

The table repeats a discipline that the dictionary alone does not enforce. Reported and derived values live in separate columns, and an unresolved legal-entity match blocks any platform-level statement. The status row is the weakest link: the codes are assigned by the contractor, not the provider ^[9], and they exist in the raw Rpt file rather than in the PUF's 201 columns.

Periods, Amendments, Missing Rows and Normalization

Report status and which row to trust

Worksheet S defines the cost report status code the contractor enters: 1 as submitted, 2 settled without audit, 3 settled with audit, 4 reopened and 5 amended ^[9]. A provider-entered counter records how many times an amended report was resubmitted ^[9], and a Notice of Program Reimbursement date must be present for statuses 2, 3 or 4 ^[9]. Regulation allows amended reports to be permitted or required as determined by CMS ^[30], and the contractor keeps the as-filed electronic report unaltered ^[30]. A report counts as filed only when the contractor accepts it ^[30]. A 2004 GAO study noted that a large portion of the reports it used had not been settled for final payment ^[35], a reminder that settlement status changes how a number should be described.

CMS documents a "highest status wins" rule only for hospital data. The hospital page says its files contain the highest level of cost report status for all reported fiscal years ^[36] and that a reopened report replaces a final settled one. The HHA pages opened for this research do not repeat the rule. The safe treatment is to treat the supersession rule as unconfirmed for HHAs, keep every report row in the worksheet and mark status explicitly.

Partial years, long years and duplicates

The CMS edit specification requires a period greater than 27 days and less than 459 days ^[9]. Cost reports generally cover a consecutive 12-month period ^[30], with shorter periods for terminations and changes of ownership ^[30]. The live PUF illustrates the risk: the first record returned was an 80-day report running from 2023-10-13 to 2023-12-31 ^[18]. CMS flags initial and final reports for a CCN on Worksheet S ^[9], so the first and last periods of an agency are expected to be irregular.

Duplicates and overlaps need rules before analysis. A mirror of CMS's hospital production notes states that one provider can submit two or more cost reports for the same year ^[37], with fiscal-year changes or ownership changes as the cause ^[37]. A 2004 GAO study excluded cost reports with incomplete or questionable data ^[35] and removed duplicates, 113 in 2001 and 76 in 2002 ^[35]. A peer-reviewed home health study excluded agencies with less than a complete year of cost data ^[31] and trimmed the top and bottom 1 percent of the total cost distribution ^[31], retaining 20,571 observations (78 percent of its initial sample) ^[31]. When fiscal years did not match calendar years, it built calendar-year variables as a weighted average from two cost reports ^[31]. These are method precedents. This report recommends logging each exclusion with its reason and never silently dropping a row.

Normalizing a flow: annualized, not reported

A flow variable (visits, cost, charges, revenue, net income) can be normalized with one formula:

annualized value = reported value / report days x 365

Report days equal the end date minus the begin date plus one, an analyst convention to be stated in the worksheet. The result must be labelled "annualized (analyst-derived)", never "reported". The formula assumes activity is spread evenly across the period, which is weak for a short or ramp-up period and invalid for balance sheet items. Table 2 applies it to fictional numbers that do not describe any real agency.

(Hypothetical Example) Fictional figures: a reported net patient revenue of $3,600,000 and 12,000 Medicare visits, each placed under three different period lengths.

The same reported number can imply $7.3 million or $3.3 million of annual revenue depending on the period, a 2.2-fold spread. The 180-day case is the dangerous one, because errors in the days count or a seasonal peak compound by a factor above two, while for a 400-day report a screen that skips normalization overstates the year. If the reported period is 80 days, as in the live PUF's first record ^[18], the factor is 4.56 and the result should usually be left as a fact about the interval, not an annual figure.

Provider-Scoped Data Versus Consolidated Platform Revenue

A cost report belongs to a provider, not to the group that owns it. The CCN is a parent or subunit identifier, a legal entity can hold several CCNs, and the same investor can control agencies that file under different CCNs. Table 3 contrasts the two objects.

No single cost report field substitutes for the right-hand column: the report supplies a lagged provider-level observation, and the platform figure requires information the report never collects.

Why the two must not be equated, as a checklist:

• Different entity: a CCN is a facility identifier, and a platform may hold many CCNs plus entities that file no Medicare cost report.
• Different payer view: Medicare columns exclude Medicare Advantage and commercial revenue, and MedPAC notes the cost report cannot give Medicare Advantage margins separately ^[33].
• Different basis: Medicare cost lines are partly allocated by a cost-per-visit method ^[9].
• Different completeness: hospital-based agencies and rejected reports are absent.
• Different ownership clock: the owner may have changed after the report's period, and a stock deal leaves no trace in the CCN.
• Different assurance: CMS says authenticated information is accurate only as of the point in time of validation and verification ^[4].

Data Analysis and Evidence

What the scope gap looks like in published aggregates

MedPAC's March 2026 report is the clearest published demonstration of why scope matters, and its source is the same family of data: Medicare home health cost report files ^[5]. For freestanding agencies in 2024, the aggregate FFS Medicare margin was 21.2 percent ^[5] while the all-payer margin was 5.0 percent ^[5]. The distribution is wide: the 2024 FFS Medicare margin ran from 5.7 percent at the 25th percentile to 31.0 percent at the 75th ^[5]. Hospital-based agencies had a negative FFS Medicare margin that year and sit outside the headline figure ^[5]. MedPAC projected a 19 percent FFS Medicare margin for 2026 ^[5] and recommended a 7 percent reduction in the payment rate for 2027 ^[5].

An agency's Medicare-only margin is therefore not its business margin, and its position in the distribution is more informative than the national average, since the quartile spread is about 25 points. Cost per full 30-day period rose just 0.2 percent in 2024 ^[5].

Payer mix and the Medicare subset

The Medicare subset is shrinking relative to total activity. MedPAC reported that FFS Medicare's share of revenue for the average freestanding HHA fell to 49 percent in 2022 ^[33], while average total revenue rose to $4.8 million ^[33]. KFF reported that Medicare Advantage covered 35.2 million of 64.2 million beneficiaries with both Parts A and B in 2026 ^[6]. MedPAC found only 8 percent of agencies serve FFS beneficiaries exclusively ^[5].

A Medicare-only row captured roughly half of the average freestanding agency's revenue as of 2022; no later equivalent was found. MedPAC also found that both Medicare Advantage encounter data and OASIS are incomplete sources of MA home health use ^[38].

Payment-rule context

Payment rules shift what a period looks like. The statutory payment unit has been the 30-day period for periods beginning on and after January 1, 2020 ^[39]; the CY2026 final rule estimated a net decrease in aggregate payments of 1.3 percent, or $220 million ^[40], and CMS proposed a 2.4 percent aggregate increase for CY2027 ^[41]. A multi-year screen therefore compares periods under different payment parameters.

Counting agencies and checking coverage

Counts vary by source and definition. MedPAC counted over 12,000 certified agencies in 2024 ^[5], up 1.5 percent ^[5], from the Provider of Services file ^[5], while the CY2026 rule used 11,904 active HHAs for burden estimates ^[39]. The provider-level Medicare utilization PUF covers calendar years 2014 to 2024 ^[42] and is based on FFS claims ^[42], so a reconciliation of cost-report rows to claims-based rows by CCN is a useful cross-check, with the expectation that the two will not match exactly.

Private equity activity is part of the landscape: MedPAC cited an outside estimate that 5.7 percent of Medicare HHAs were owned by private equity firms in 2023 ^[33], and the JAMA Health Forum count of 749 acquired agencies spans 2006 to 2024 ^[32]. MedPAC also cites a transaction-share estimate that rose from about 20 percent in 2013 to about 50 percent in 2021 ^[33]. These are descriptive counts with acknowledged measurement limits ^[33]. BLS projects 18 percent growth in home health and personal care aide employment from 2025 to 2035 ^[43].

Implementation Considerations and Process Changes

From extraction to adjudication

A repeatable process has seven steps.

1. State the claim. Write the entity, period, scope and measure type before opening any file.
2. Choose the file. Use the PUF for a flat CCN table and the raw HCRIS extract when status or other worksheet cells matter. Record release date, report year and form version.
3. Resolve identity. Match the CCN to Care Compare and PECOS, then to any proposed platform using separate ownership evidence.
4. Define the variable. Pull the dictionary definition and mark Medicare or total scope and cost or charges.
5. Handle periods and amendments. Compute report days, keep raw and normalized values, and log status.
6. Compare years. Line up consecutive reports by CCN and flag gaps, overlaps and form changes.
7. Adjudicate. Assign one of three outcomes with a written reason.

A provisional threshold and the three outcomes

Assume a fictional mandate criterion: annualized all-payer net patient revenue of at least $5 million. The structure of the evidence suggests an asymmetric rule.

• Clears: the all-payer net patient revenue field, once annualized for a full, settled or as-submitted period with a matched legal entity, meets the threshold. State the period and label it annualized if normalized.
• Fails: the all-payer field for a full period with a matched entity falls below the threshold and no later period is available. Because the cost report is lagged, "fails" should read "failed as of the report period".
• Unresolved: the evidence is Medicare-only, partial-period, status unknown, entity unmatched, or the years contradict each other. For this screen, require all-payer net patient revenue to clear or fail the $5 million test. Use the total-facility net patient revenue field, defined as gross patient revenue less allowances and discounts ^[15]; do not substitute Medicare-only revenue, costs, charges, gross revenue or PPS payments.

(Hypothetical Example) Fictional CCN 000001 files two reports. Report A covers 2023-01-01 to 2023-12-31 (365 days) with net patient revenue of $4,200,000, status 3 (settled with audit). Report B covers 2024-01-01 to 2024-06-30 (182 days) with $2,400,000, status 1 (as submitted), after a change in majority ownership. Annualizing B gives $2,400,000 / 182 x 365 = $4,813,187. Neither report clears the $5 million threshold, but the partial year and the ownership change both weaken comparison. The adjudication is "unresolved, reviewer follow-up": request the next full-year report, confirm the legal entity with PECOS and obtain management accounts. A fail would be premature.

Reviewer follow-up triggers

• Status gap: a PUF row with no status information, since the PUF has no status field among its 201 keys.
• Period irregularity: a period outside roughly 350 to 380 days (an analyst heuristic, not a CMS rule).
• Ownership flag: Worksheet S-2 indicates a change of ownership or the 36-month rule may apply ^[27].
• Scope mismatch: the only field available is a Medicare line while the criterion is all-payer.

Documentation standards

Individual reports can be requested from the MACs through the Freedom of Information Act ^[44]. When the screen is used in a deal memo, write the sentence the evidence supports, for example (hypothetical wording): "CCN 000001 reported total visits of N for the period ending 2023-12-31 (PUF vintage 2024-01-01, downloaded 2026-10-20); legal-entity match unresolved."

Implications and Future Directions

A cost-report fact is a dated observation of one provider, and the screen's value comes from the discipline around it. Three changes in the data environment deserve attention.

First, the vintage and identifier behavior of the PUF may continue to shift: the catalog now lists 2020 through 2024. Pinning downloaded files and recording hashes protects the audit trail. Second, analysts should check the CMS page for the current release statement before citing it, since this research found a banner date of 10/06/2026 but not the July 2026 notice. Third, payment parameters are in motion, and CY2027 proposed adjustments ^[41] will show up in cost reports only after the lag described above.

Research platforms, including Docket's Triage, Scout and Audit agents, are among the tools origination teams use for extraction, identity matching and logging. The same adjudication rules should apply to manual and automated work, and no tool can supply ownership evidence the public files do not contain ^[8].

Open gaps include whether HHA files apply the hospital page's "highest status wins" rule, the snapshot date behind the newest PUF files, whether CMS will publish an HHA change-of-ownership dataset like its hospital series ^[1], and a current estimate of Medicare's share of agency revenue. Until resolved, "unresolved" is the honest default.

Frequently Asked Questions (FAQs)

How is the CMS HHA cost report public use file used?

Download the file for the relevant vintage, keep the CCN as text, compute report days from the begin and end dates, and keep the dictionary definition beside each field. The PUF is a subset of HCRIS variables.

How does a home health CCN lookup for cost report data work?

Match the six-character CCN in the PUF to the Care Compare listing and to PECOS enrollment. Watch for leading zeros.

What fields are on Form CMS-1728-20 and the older CMS-1728-94?

CMS-1728-94 covers earlier periods ^[20]. On the current form, worksheets include S (status), S-2, S-3 (visits and census), A (trial balance) ^[9], C, D, F (balance sheet) and F-1 (revenue and expenses); the worksheets must be consistent with the HHA financial statements ^[9]. The status codes sit on Worksheet S.

How should a home health agency acquisition target be evaluated with this data?

Use it for a provider-scoped, dated fact, then request management accounts for anything the file does not hold. Financial due diligence on a home health agency begins with the entity match and the period, and the cost report never replaces management accounts.

Conclusion

A CMS home health cost report row supports an initial screening fact when four things are written down: the CCN, the fiscal period with its day count, the Medicare or total scope of the field, and the vintage of the file. It does not support a consolidated revenue claim, a valuation, or a quality judgment.

The timing evidence is unsettled in ways an analyst can manage but not remove: the catalog lists a 2024 vintage beside 2020 through 2023, the HCRIS page showed a 10/06/2026 release, and the methodology PDF cites a July 17, 2025 snapshot. These are recorded as a discrepancy to be resolved by comparing the downloaded file with the raw release. The scope evidence is firmer: Medicare-only and all-payer views diverge widely in published aggregates ^[5], and this screen requires all-payer net patient revenue for its revenue threshold. The CCN persists through a standard change of ownership ^[7], but the public files do not name the ultimate parent. When any link in the chain is missing, the recommended default is to mark the mandate criterion unresolved and attach the reviewer follow-up that would resolve it.