One in five hospices has no quality score, and half of them are not told why

Medicare publishes the Hospice Care Index for 5,411 of 6,852 certified hospices. The remaining 1,441 have no score. Recent certification explains 587 of them. The other 854 operated throughout the full measurement window.

Why a blank score is a finding

A hospice is chosen quickly and under pressure. The decision is often made in a hospital corridor within a day of a terminal prognosis, by a family with no prior knowledge of the field and by a discharge planner working a caseload. Care Compare exists for exactly that moment. It is the only comparison tool most families will ever see.

The Hospice Care Index is the summary number on that page. It condenses ten claims-based indicators into a single score from zero to ten, covering staffing, visit patterns, spending, and live discharge behavior. It is designed to be read in a hurry by someone who will never read the technical manual.

For one in five hospices, that number is absent. The page loads, the provider is listed, and where the score would be there is nothing. A reader has no way to distinguish a hospice that is small and new from one that is small and troubled, and no way to tell which of those situations they are looking at.

The measure matters beyond consumer choice. In June 2026 the Department of Justice charged a hospice owner in the Central District of California in a $27.7 million scheme. The alleged mechanism turned on this family of metrics directly. According to the charges, the defendant enrolled patients who were not terminally ill, understood that the percentage of patients discharged from hospice alive is monitored as a fraud indicator, and paid a funeral home employee between$1,000 and $3,000 per record for the identities of recently deceased Medicare beneficiaries in order to bill for them and reduce his outlier data metrics. Charges are allegations and the defendant is presumed innocent.

That case establishes only one thing for present purposes, and this paper claims nothing further from it. Live discharge behavior is treated as an integrity signal by people whose job is detecting fraud. Four of the ten Hospice Care Index indicators are denominated on live discharges. It is therefore worth asking how well the public version of that signal covers the field it is meant to describe.

The answer is that it covers 79.0% of it.

This paper follows the same argument as Working Paper No. 01, which examined nursing home staffing and found that a published quarterly average erased a within-week pattern that mattered operationally. The failure here is different in kind. The Hospice Care Index does not average away a signal. It declines to produce one, for a fifth of the field, and the public explanation for that silence is often unusable.

Data

Two files from the CMS Provider Data Catalog, plus one reference table.

Retrieved July 28, 2026.

The unit of analysis is the CMS Certification Number. Both files carry it, and the join is one to one after the provider file is reduced to the overall index row. Certification numbers were zero-padded to six characters before joining, because the two files differ in leading-zero handling.

All Hospice Care Index measure codes carry the same measurement window, January 1, 2023 through December 31, 2024. This was read from the file rather than assumed, and it is uniform across all thirty-one index measure codes.

6,852 certified hospices appear in both files. No providers were excluded. Territories are retained, including Guam, Puerto Rico, the Northern Mariana Islands and the Virgin Islands.

The provider file is in long format. Each provider contributes roughly sixty-seven rows, one per measure code. The ten index indicators each appear three times, as an observed value, a denominator, and a percentile. The composite score appears once, as measure code H_012_00_OBSERVED.

Known limitations of the source

Two spellings of the same measure. The provider file contains both Burdensome transitions, Type 1 (% live discharges) and Burdensome transitions, Type 1(% live discharges), differing by one space, and the same for Type 2. Any grouping keyed on the measure name will split these two indicators in half. All work here keys on measure code, which is clean.

Footnote vocabulary is shared across settings. The crosswalk covers every Care Compare provider type. Codes 17 through 26 concern hospital star ratings, Veterans Health Administration facilities, Department of Defense hospitals, and Maryland payment waivers. Hospice records draw from this same list. Two of the codes appearing on hospice index scores describe conditions that do not arise for a claims-based measure.

Certification date is a certification date. It records Medicare certification, not ownership, not the current operator, and not continuity of management. A provider certified in 2009 and sold in 2023 appears here as a long-tenured provider. Nothing in the public files corrects for this.

Scores are point-in-time. This is a single refresh. This paper does not track providers across refreshes and makes no claim about trend.

Method

The analysis is deliberately simple. Every step below is a filter, a join, or a count, so that a reader who disagrees can name the step.

Step 1. Identify scored providers. Select rows where measure code equals H_012_00_OBSERVED. Convert the score column to a number, stripping thousands separators. Providers with a numeric value are scored. Providers whose value does not parse are unscored. This yields 5,411 scored and 1,441 unscored, summing to the full population.

Step 2. Attach provider attributes. Join the general information file on certification number to attach certification date, ownership type and state.

Step 3. Test the exposure explanation. Compare certification date against the window start of January 1, 2023. A provider certified after that date had partial exposure and would be expected to lack sufficient cases. A provider certified before it had the full window available. Cross-tabulate against scored status.

This is the central test. If the coverage gap were an artifact of recent entry, it would sit almost entirely in the certified-after group. It does not. 854 of the 1,441 unscored providers, or 59.3%, were certified before the window opened.

Step 4. Examine certification cohorts. Group by certification year and compute the share scored. The 2021 and 2022 cohorts are the informative ones, because both were certified before the window and therefore had complete exposure to it.

Step 5. Decompose the gap by footnote. Every unscored provider carries exactly one footnote row on the composite measure code. Count providers by footnote value, then join the crosswalk for definitions. The four footnote conditions sum to 1,441 with no remainder, which confirms the decomposition is complete rather than approximate.

A note on counting. Footnote counts on the ten individual indicators appear three times larger than the provider count, because each indicator contributes an observed, a denominator, and a percentile row. Counts reported here use the composite measure code, which appears once per provider, so no division is required.

Step 6. Describe composition. Cross-tabulate scored status by ownership type and by state. Report states with at least fifty certified hospices.

Step 7. Describe the scored population. Among providers with a score, report the distribution across the zero to ten scale, and the association between score and provider age at window start.

What the method does not do

It fits no model, estimates no effect, and tests no hypothesis about cause. Every result is a count, a share, or a cross-tabulation. Where an association appears, such as index score rising with provider tenure, this paper reports the association and stops.

Findings

One in five hospices has no score

Of 6,852 certified hospices, 5,411 carry an HCI overall score and 1,441 do not. The unscored share is 21.0%.

The obvious explanation is recent certification: a hospice that entered the measurement window partway through has fewer cases and may not reach the threshold for scoring. That explanation covers 587 of the 1,441. The other 854 were certified before January 1, 2023, meaning they were operating for the full two years the index covers. Their absence from the scored population reflects case volume, not tenure.

Scoring rates fall as certification volume rises

The 2021 and 2022 cohorts had complete exposure to the measurement window. Their scoring rates of 72.4% and 53.6% confirm that a hospice can have been operating for two or more years and still produce no index score.

The spike in new certifications is descriptive, not interpretive. Annual new certifications ran roughly 100 to 250 through 2019, then 379, 738, and 771 in 2020, 2021, and 2022. This paper reports the pattern and assigns no cause to it.

For-profit hospices are scored 16 points less often

For-profit hospices are scored 82.0% of the time. Non-profit hospices are scored 98.1% of the time.

Among states with at least 50 certified hospices, Nevada is 47.3% unscored, California 33.8%, and Texas 30.3%. California holds 30% of all certified hospices in the file and 48% of all unscored ones.

These associations are reported without causal claims. A state with more recently certified for-profit hospices would produce both patterns for unrelated reasons.

The score distribution is compressed

Among the 5,411 scored providers, the index clusters at the top of its range. No provider scores below 3. The effective range in the current refresh is 3 to 10, not 0 to 10.

86.0% of scored providers score 8 or higher. 38.6% are tied at exactly 10.

This compression matters for how the index functions as a comparison tool. A measure where more than a third of providers share the maximum score cannot distinguish among them. A discharge planner choosing between two hospices that both score 10 receives no information about which performed better on any individual indicator.

Score rises monotonically with provider age

Among scored providers, mean HCI rises with the number of years a provider was certified before the measurement window opened.

The association is reported as an association. It is consistent with multiple explanations, including selection effects (providers that perform poorly close earlier), learning effects (established providers develop better processes over time), and unmeasured characteristics that correlate with both tenure and performance.

A sampling footnote on a measure with no sampling step

Every unscored provider carries exactly one footnote on the composite measure code, and the four conditions sum to 1,441 with no remainder.

Footnote 1 is the expected condition and is self-explanatory. Too few cases means no score. That accounts for 682 providers.

Footnote 8 describes a condition from the hospital-acquired infection literature. A confidence interval lower bound for an infection count is not a concept that arises for a claims-based quality index. It appears on 58 hospice index scores with no apparent applicability.

Footnote 2 is the one that requires explanation. The Hospice Care Index is explicitly described by CMS as comprising ten indicators calculated from Medicare claims data. CMS confirms that the data source for claims-based measures is Medicare claims data already collected and submitted to CMS, and that hospices with claims data are 100% compliant with this data source requirement because no additional submission is required. There is no sampling step in the HCI calculation. A sampling footnote cannot describe the reason this measure was not computed for 677 hospices.

The Care Compare footnote vocabulary is shared across every provider setting on the platform: acute hospitals, skilled nursing facilities, home health agencies, hospices, and others. A footnote that is valid in one setting is available in all. Footnote 2 was designed for survey-based or sampled measures. It was applied here, at scale, to a fully enumerated claims-based measure. The reason a family looking up one of those 677 hospices sees no score is not recoverable from the published footnote.

Two hypotheses the data ruled out

Two hypotheses were stated before the analysis ran and did not survive contact with the data. Both are recorded here because they describe real features of the index that matter beyond this paper.

Rejected: the suppression shelter. The initial hypothesis was that a hospice suppressing its live discharge rate would drop out of four indicators denominated on live discharges while retaining the other six, producing an artificially inflated composite. This is false. Of 1,385 providers missing at least one live discharge indicator, zero carry an overall HCI score. Missing any indicator removes the provider from the composite entirely. The index handles this correctly.

Demoted: the zero floor as an integrity signal. Three of the four live discharge indicators have a 10th percentile of exactly zero, with 978, 970, and 2,386 providers reporting zero respectively. The initial reading was that this represented a measurement blind spot at the low end. On inspection this is primarily a small-denominator artifact. The median denominator across all four indicators is 48 live discharges, meaning a single event is roughly 2%, and zero is an ordinary outcome for a small provider. Government hospices, the smallest group at 70 providers, show the highest zero rates, which is consistent with the small-denominator explanation. This observation is retained as a precision limitation, not an integrity signal.

What this does and does not support

Supported.

• 1,441 of 6,852 certified hospices have no published HCI score in the current data refresh.
• 854 of those were certified before the measurement window opened.
• Unscored share is 21.0% nationally, 47.3% in Nevada, 33.8% in California, and 30.3% in Texas.
• For-profit providers score at 82.0% against 98.1% for non-profit.
• Among scored providers, 86.0% score 8 or higher and 38.6% score exactly 10, with no provider scoring below 3.
• Mean HCI rises monotonically with provider age at window start.
• Footnote 2, describing a sampling condition, appears on 677 hospices carrying no HCI score, despite the HCI being a fully enumerated claims-based measure for which no sampling step exists.
• Footnote 8, describing an infection confidence interval, appears on 58 hospice index records where it has no applicable meaning.
• The suppression shelter hypothesis was tested against the data and rejected.
• The zero floor on live discharge indicators is primarily a small-denominator artifact, not an integrity signal.

Not supported.

• No claim about fraud prevalence. No provider is named or ranked.
• No claim that an unscored hospice is deficient. Absence of a score reflects case volume or window exposure in most cases, and most small hospices are ordinary.
• No causal claim about why for-profit providers score lower or why certain states have more unscored providers. The associations are stated as associations.
• No claim that the geographic concentration of unscored providers relates to enforcement activity. The resemblance to known enforcement geography was noted and deliberately left unconnected. That relationship was not tested, and the data cannot test it.
• No evaluation of hospice care quality. The HCI is assessed here as a public information signal, which is a narrower question than whether it measures care well.
• No audit of CMS internal detection. The Health Care Fraud Unit’s non-public analytics are outside scope and appear substantially more capable than anything published. The argument concerns the public measure only.
• The footnote finding does not identify an error in the index calculation. It identifies a mismatch between the vocabulary used to describe why a score is absent and the actual conditions under which the index operates.

Reproduction

Two files from the CMS Provider Data Catalog and one footnote crosswalk, all public and freely available: fm2n-hjj6, yje5-53sw, and y9us-9xdf.

The repository contains three scripts, run in sequence:

1. fetch_hospice_pdc.py: discovers and downloads the source files and prints the schema, confirming that per-provider HCI values are published at the indicator level.
2. analyze_hci.py: extracts the overall score distribution, the four live discharge indicators, and the footnote decomposition.
3. analyze_coverage.py: joins certification date and ownership type to the scored and unscored split, runs the cohort exposure test, and produces the geographic breakdown.

Python 3.9, pandas, requests. No dependencies beyond the standard scientific stack.

One part of this paper is a document argument rather than a computation and should be evaluated on those terms. The claim that footnote 2 does not describe a condition applicable to the HCI rests on the HCI measure specifications, specifically the FY 2022 final rule and the February 2026 HQRP Current Measures List. CMS’s own compliance guidance confirms that hospices are automatically considered fully compliant on HCI submission because the measure is calculated entirely from claims data already on file, with no separate submission step and therefore no sampling step for footnote 2 to describe.

Data: CMS Hospice Provider Data and Hospice General Information, retrieved July 28, 2026. 6,852 certified hospices. Code and data snapshots: zatara-research-data/03-hospice-coverage-gaps.

@techreport{zatara2026paper3,
  title       = {Coverage gaps in the Hospice Care Index},
  author      = {Zatara, David},
  year        = {2026},
  institution = {Independent Researcher},
  type        = {Working Paper},
  number      = {3}
}