docs(p44): American spelling throughout the RFI comment body (refs #650)

docs/comments/2026-09-cms-2026-2377-palliative-rfi-draft.md

@@ -13,7 +13,7 @@
 
 We write in response to the Request for Information on Community-Based Palliative Care at 91 FR 43949–43950. CMS asks that commenters "support your statements with peer-reviewed evidence or evidence from your institution with sufficient detail for review" (91 FR 43949 (¶942); repeated at 91 FR 43950 (¶944)). We took that literally. Rather than an opinion letter, we conducted a systematic evidence review to a written protocol, scoped to the five questions CMS asks, and report its results here — including where the evidence does not support an answer.
 
-The review identified **4,999 records** — eight uncapped, relevance-sorted PubMed E-utilities searches (one per RFI part, plus three standards-anchored searches supporting the code family proposed in part (d)), forward citation-chasing from five anchor publications including CMS's own two serious-illness-identification citations, and the grey-literature standards and RFI-cited sources themselves. **4,997 were screened** at title and abstract, **2,642 excluded** there, **2,355 assessed for full-text eligibility**, **401 excluded** on full text, **1,429 unresolved** because no full text was retrievable, and **525 studies included**. Each included study carries a structured extraction record — design, population, setting, eligibility basis, instrument and measured performance, service elements, outcomes, effect size, risk-of-bias notes — and every citation below is drawn from it. Two limitations we state up front: screening was performed by a single reviewer using a locally hosted language model against written criteria, with uncertain decisions adjudicated in an independent second pass, rather than by two human reviewers; and the 1,429 unresolved full texts are 61 percent of records reaching full-text assessment. We note below where that second limitation bites. Counts per part, exclusion reasons and the protocol pointer are in Appendix B.
+The review identified **4,999 records** — eight uncapped, relevance-sorted PubMed E-utilities searches (one per RFI part, plus three standards-anchored searches supporting the code family proposed in part (d)), forward citation-chasing from five anchor publications including CMS's own two serious-illness-identification citations, and the gray-literature standards and RFI-cited sources themselves. **4,997 were screened** at title and abstract, **2,642 excluded** there, **2,355 assessed for full-text eligibility**, **401 excluded** on full text, **1,429 unresolved** because no full text was retrievable, and **525 studies included**. Each included study carries a structured extraction record — design, population, setting, eligibility basis, instrument and measured performance, service elements, outcomes, effect size, risk-of-bias notes — and every citation below is drawn from it. Two limitations we state up front: screening was performed by a single reviewer using a locally hosted language model against written criteria, with uncertain decisions adjudicated in an independent second pass, rather than by two human reviewers; and the 1,429 unresolved full texts are 61 percent of records reaching full-text assessment. We note below where that second limitation bites. Counts per part, exclusion reasons and the protocol pointer are in Appendix B.
 
 We answer the five questions in CMS's order. We do not comment on the Medicare Hospice Benefit or on maintenance versus comfort-focused dialysis; CMS directs those elsewhere (91 FR 43949 (¶938)), and we have kept them out. Where hospice research is the nearest available evidence for a question about community-based palliative care — which is the case throughout part (a) — we say so and treat it as analogy, as CMS's own framing invites: "While hospice care is not palliative care, in related disciplines and sites of care we are very interested in how to better address fraud, waste, and abuse" (91 FR 43949 (¶941)).
 
@@ -25,13 +25,13 @@ We answer the five questions in CMS's order. We do not comment on the Medicare H
 >
 > "**Where should CMS focus on potential fraud, waste, and abuse in community-based palliative care?** Please support your statements with peer-reviewed evidence or evidence from your institution with sufficient detail for review." — 91 FR 43949 (¶942)
 
-**Recommendation.** Focus on four loci, each observable from data CMS already holds and none requiring new clinician documentation: (1) **enrollment and disenrollment churn** — the community analogue of the hospice live-discharge signal, measured as the rate at which beneficiaries are started on and then stopped from the service without death, transfer, or documented change in goals; (2) **ownership form and operating margin as risk stratifiers**, prioritising review rather than triggering it; (3) **the initiating trigger and the practitioner-to-auxiliary-personnel work ratio**, the guardrails CMS itself raises at 91 FR 43938 ¶8–43939 ¶1; and (4) **equity monitoring of the integrity program itself**. We also urge CMS to treat beneficiary protection — abuse and neglect in the home — as an integrity domain, not only billing integrity.
+**Recommendation.** Focus on four loci, each observable from data CMS already holds and none requiring new clinician documentation: (1) **enrollment and disenrollment churn** — the community analog of the hospice live-discharge signal, measured as the rate at which beneficiaries are started on and then stopped from the service without death, transfer, or documented change in goals; (2) **ownership form and operating margin as risk stratifiers**, prioritizing review rather than triggering it; (3) **the initiating trigger and the practitioner-to-auxiliary-personnel work ratio**, the guardrails CMS itself raises at 91 FR 43938 ¶8–43939 ¶1; and (4) **equity monitoring of the integrity program itself**. We also urge CMS to treat beneficiary protection — abuse and neglect in the home — as an integrity domain, not only billing integrity.
 
 **Evidence.** Hospice operating margin is positively associated with the rate of live discharge (Dolin (2017), PMID 28679817), and the same group's qualitative work found providers describing both financial motivation *and* fear of CMS audit as drivers of enrollment and disenrollment (Dolin (2017), PMID 28323079). That second finding is the design constraint: a blunt integrity program produces defensive under-enrollment as well as reduced over-enrollment. Problematic live-discharge patterns are identifiable from administrative data alone across 3,028 hospice programs (P < 0.001) (Teno (2015), PMID 26004403), with a stable national baseline from 2000 to 2012 (Prsic (2016), PMID 27171543). For-profit ownership is associated with episodes exceeding one year (Sengupta (2014), PMID 23744976) — a risk-stratifier finding, not a fraud finding, and we ask CMS to use it that way.
 
-The evidence also contains a cautionary null: a specific hypothesised abuse mechanism, that increased nursing home hospice use would substitute for nursing assistant staffing, was tested across 10,759 facilities and not found (estimate 0.79, 95% CI 0.373–1.211) (Tyler (2011), PMID 21958012). Plausible fraud mechanisms are not always real ones, and CMS should hold its own integrity hypotheses to the evidentiary standard it asks of commenters here.
+The evidence also contains a cautionary null: a specific hypothesized abuse mechanism, that increased nursing home hospice use would substitute for nursing assistant staffing, was tested across 10,759 facilities and not found (estimate 0.79, 95% CI 0.373–1.211) (Tyler (2011), PMID 21958012). Plausible fraud mechanisms are not always real ones, and CMS should hold its own integrity hypotheses to the evidentiary standard it asks of commenters here.
 
-On equity: among 2,629 hospice patients with dementia, live discharge was substantially more likely for African American (aOR 2.42, 95% CI 1.34–4.38) and Hispanic (aOR 2.99, 95% CI 1.81–4.94) beneficiaries (Luth (2020), PMID 31750935). A churn-based integrity signal must be risk-adjusted and monitored for disparate impact or it will penalise the programs serving the populations with the least access. Conversely, concerns about health care fraud have themselves limited appropriate use of palliative and hospice services in nursing homes (Zerzan (2000), PMID 11074779) — integrity signals have an access cost that belongs in the design. Jones (2000), PMID 11009798, remains the clearest articulation of compliance-plan risk areas in this space; LeSage (2015), PMID 24848665, documents the post-discharge re-enrollment pattern a churn measure would capture.
+On equity: among 2,629 hospice patients with dementia, live discharge was substantially more likely for African American (aOR 2.42, 95% CI 1.34–4.38) and Hispanic (aOR 2.99, 95% CI 1.81–4.94) beneficiaries (Luth (2020), PMID 31750935). A churn-based integrity signal must be risk-adjusted and monitored for disparate impact or it will penalize the programs serving the populations with the least access. Conversely, concerns about health care fraud have themselves limited appropriate use of palliative and hospice services in nursing homes (Zerzan (2000), PMID 11074779) — integrity signals have an access cost that belongs in the design. Jones (2000), PMID 11009798, remains the clearest articulation of compliance-plan risk areas in this space; LeSage (2015), PMID 24848665, documents the post-discharge re-enrollment pattern a churn measure would capture.
 
 On beneficiary protection: nine candidate quality indicators for abuse and neglect in home-based primary and palliative care were developed, all but one rated high in validity and feasibility by a technical expert panel (Sheehan (2016), PMID 27787878). This benefit is delivered in the home, to beneficiaries with impaired function and stressed caregivers; an integrity framework addressing only billing patterns will miss the harm that matters most.
 
@@ -47,19 +47,19 @@ On beneficiary protection: nine candidate quality indicators for abuse and negle
 >
 > "**If eligibility is restricted to those beneficiaries with a terminal prognosis, is there evidence to suggest a reasonable interval (that is, less than 1 year of life expectancy as in some States' Medicaid programs)?**" — 91 FR 43950 (¶946)
 
-**Recommendation.** **No.** Eligibility should not be restricted to certification of a likely life-expectancy duration. The evidence is consistent: neither clinician judgement nor validated instruments nor administrative data can certify an interval accurately enough to gate a Part B benefit, and the errors are not random — they fall hardest on non-cancer diagnoses and on beneficiaries from minoritised groups. If CMS nonetheless adopts a prognostic criterion, the evidence supports **12 months rather than 6**, as **one of several alternative qualifying pathways**, attested by the **billing practitioner alone** rather than two physicians, and **never conditioned on forgoing disease-directed treatment**.
+**Recommendation.** **No.** Eligibility should not be restricted to certification of a likely life-expectancy duration. The evidence is consistent: neither clinician judgment nor validated instruments nor administrative data can certify an interval accurately enough to gate a Part B benefit, and the errors are not random — they fall hardest on non-cancer diagnoses and on beneficiaries from minoritized groups. If CMS nonetheless adopts a prognostic criterion, the evidence supports **12 months rather than 6**, as **one of several alternative qualifying pathways**, attested by the **billing practitioner alone** rather than two physicians, and **never conditioned on forgoing disease-directed treatment**.
 
 **Clinician prediction is not certification-grade.** Clinician predictions of survival assessed in 1,833 patients in the Prognosis in Palliative care Study II proved adequate for conversation, not for categorical certification (Stone (2022), PMID 35421168); clinician prediction and the Palliative Prognostic Score were each unreliable at the individual-patient level in advanced cancer (Hui (2016), PMID 27372208; Hui (2020), PMID 31564218); and a systematic review and meta-analysis of palliative prognostic scores reaches the same conclusion across the pooled literature (Yoong (2024), PMID 38366659).
 
 **The surprise question does not export across diseases.** Meta-analysis across 25,718 patients found overall accuracy of 74.8 percent (95% CI 68.6–80.5) with poor positive predictive value (White (2017), PMID 28764757), which a more recent systematic review and meta-analysis combining the surprise question with clinician-predicted prognosis confirms (Gupta (2024), PMID 38925876). Disease-specific validations show the spread: 93.4 percent sensitivity against 37.5 percent specificity for 12-month mortality in 378 patients with end-stage liver disease (Homann (2025), PMID 39952927); poor accuracy for 90-day mortality in 100 patients with decompensated cirrhosis (Selesky (2025), PMID 40658791); sensitivity 0.95 against specificity 0.35 for an end-of-life state determination in 315 patients (Arahata (2023), PMID 36636714); and AUC 0.68, sensitivity 55 percent, specificity 74 percent on external validation of the ProPal-COPD tool in 523 patients (Broese (2022), PMID 36579356).
 
-**The administrative operationalisation is worse.** The most directly Medicare-relevant study in our part (b) set examined the "less than six-month prognosis" item in the Minimum Data Set for 20,397 U.S. nursing home residents with cancer and found **94.2 percent specificity but 13.7 percent sensitivity** against actual survival (Vu (2024), PMID 38560743). The existing administrative six-month prognosis field correctly identifies fewer than one in seven of the people it is meant to identify. CMS should not build eligibility on a data element with that operating characteristic.
+**The administrative operationalization is worse.** The most directly Medicare-relevant study in our part (b) set examined the "less than six-month prognosis" item in the Minimum Data Set for 20,397 U.S. nursing home residents with cancer and found **94.2 percent specificity but 13.7 percent sensitivity** against actual survival (Vu (2024), PMID 38560743). The existing administrative six-month prognosis field correctly identifies fewer than one in seven of the people it is meant to identify. CMS should not build eligibility on a data element with that operating characteristic.
 
 **Statistical models do better, but still not well enough to gate.** AUROC 0.861 (95% CI 0.856–0.867) for an explainable EHR model in 5,926 patients with stage 3–4 solid cancer (Zhuang (2024), PMID 38769564); AUC 0.82 for a 1-year mortality index externally validated in 215,731 community-dwelling adults aged 65 and over (Frechman (2025), PMID 40298901); AUC 0.79, adjusted hazard ratio 31.02 (95% CI 27.44–35.08), for language-model prediction of hospice eligibility in 9,872 patients with dementia (McCoy (2025), PMID 41216867); AUC 0.751 for 1-year mortality in 591 patients with multisystem chronic illness (Bretos-Azcona (2022), PMID 32581004); and the RESPECT tool derived and validated for home care users (Hsu (2021), PMID 34226263). These are respectable population-targeting statistics. None is certification-grade for an individual, and no included study claims otherwise.
 
 **Gating on prognosis produces churn and inequity.** Re-enrollment after live discharge is documented (LeSage (2015), PMID 24848665); the disenrollment burden falls disproportionately on Black and Hispanic beneficiaries with dementia (Luth (2020), PMID 31750935); and a policy analysis concludes the hospice eligibility and election structure does not prepare the system to meet the need it serves (Wallace (2015), PMID 26161945). A prognosis-gated palliative benefit would import these failure modes wholesale.
 
-**If CMS gates anyway: the interval.** Twelve months, not six. Nearly every instrument in the included set validated for community identification is calibrated to a one-year horizon: SPICT (Piers (2021), PMID 34059507), ProPal-COPD (Broese (2022)), the multisystem chronic illness model (Bretos-Azcona (2022)), RESPECT (Hsu (2021)), the Epic End of Life Care Index (Frechman (2025)), and the population-based functional screen of Ankuda (2021), PMID 33506111. A randomised trial successfully enrolled 381 participants using "estimated life expectancy of between 1 month and 1 year," showing the interval is operationally workable (Kutner (2015), PMID 25798575). A 12-month framing also places the service upstream of hospice rather than in competition with it — which is the population CMS says it wants to reach.
+**If CMS gates anyway: the interval.** Twelve months, not six. Nearly every instrument in the included set validated for community identification is calibrated to a one-year horizon: SPICT (Piers (2021), PMID 34059507), ProPal-COPD (Broese (2022)), the multisystem chronic illness model (Bretos-Azcona (2022)), RESPECT (Hsu (2021)), the Epic End of Life Care Index (Frechman (2025)), and the population-based functional screen of Ankuda (2021), PMID 33506111. A randomized trial successfully enrolled 381 participants using "estimated life expectancy of between 1 month and 1 year," showing the interval is operationally workable (Kutner (2015), PMID 25798575). A 12-month framing also places the service upstream of hospice rather than in competition with it — which is the population CMS says it wants to reach.
 
 **What we decline to address.** We decline to endorse any particular State Medicaid interval as evidence-based. No study in our included set compares alternative interval thresholds head to head in a Medicaid or Medicare population, and we will not manufacture support for a number the literature has not tested.
 
@@ -77,9 +77,9 @@ On beneficiary protection: nine candidate quality indicators for abuse and negle
 
 **Recommendation.** **Both** — as alternative qualifying pathways, not cumulative requirements. Eligibility should be met by satisfying **any one** of: (i) a documented deficit in two or more activities of daily living, or documented instrumental-ADL decline; (ii) a positive screen on a brief validated caregiver-strain instrument; (iii) a qualifying serious-illness diagnosis with any functional impairment; or (iv) the 12-month prognostic pathway from part (b). Burden is avoided not by lowering the evidentiary bar but by reusing assessments already collected, accepting attestation of a single met criterion, and paying for the assessment inside the service.
 
-**Function is measurable and predictive at population scale.** ADL-based functional disability, screened across 30,885 adults aged 65 and over in the National Health and Aging Trends Study, predicted both mortality and hospitalisation — the two outcomes this service is meant to change (Ankuda (2021), PMID 33506111). Barthel and Lawton–Brody measurement in 201 patients with advanced respiratory disease identified predictors of increasing ADL disability (OR 1.41, 95% CI 1.14–1.74, p = 0.002) (Fettes (2024), PMID 38073190). Simpler measures work too: usual gait speed predicted one-year mortality in 142 hospitalised patients aged 75 and over (AUC 0.748, p = 0.006; HR 7.456, 95% CI 1.58–35.19) (Van de Vyver (2020), PMID 33234124), and in COPD is accurate, simple, reliable and inexpensive (Karpman (2014), PMID 25473277). The Minimum Data Set 3.0 — already mandated by CMS — supports identification of functional decline in long-term care (Anderson (2012), PMID 22784699), and the Barthel Index has estimated palliative need across 2,028 community residents (Chandra (2022), PMID 36447509).
+**Function is measurable and predictive at population scale.** ADL-based functional disability, screened across 30,885 adults aged 65 and over in the National Health and Aging Trends Study, predicted both mortality and hospitalization — the two outcomes this service is meant to change (Ankuda (2021), PMID 33506111). Barthel and Lawton–Brody measurement in 201 patients with advanced respiratory disease identified predictors of increasing ADL disability (OR 1.41, 95% CI 1.14–1.74, p = 0.002) (Fettes (2024), PMID 38073190). Simpler measures work too: usual gait speed predicted one-year mortality in 142 hospitalized patients aged 75 and over (AUC 0.748, p = 0.006; HR 7.456, 95% CI 1.58–35.19) (Van de Vyver (2020), PMID 33234124), and in COPD is accurate, simple, reliable and inexpensive (Karpman (2014), PMID 25473277). The Minimum Data Set 3.0 — already mandated by CMS — supports identification of functional decline in long-term care (Anderson (2012), PMID 22784699), and the Barthel Index has estimated palliative need across 2,028 community residents (Chandra (2022), PMID 36447509).
 
-**Caregiver strain is measurable with short validated instruments.** The eight-item CAREPAL-8 was validated in 232 caregivers (Ullrich (2024), PMID 39095830) and is under multicentre longitudinal validation in 510 more (Haufschild (2026), PMID 42536973); the Modified Caregiver Strain Index has adequate psychometric properties in 200 palliative caregivers (Ramadasa (2023), PMID 37924086) and reached Cronbach's α of 0.91 in its traditional Chinese form in 125 family carers (Cheng (2021), PMID 34627225); the Zarit Burden Interview has good properties in this population (Quadir (2025), PMID 39510815) and serves as a primary endpoint in outpatient palliative trials (Kluger (2019), PMID 30779960). These are instruments of eight to twenty-two items. They are not burdensome.
+**Caregiver strain is measurable with short validated instruments.** The eight-item CAREPAL-8 was validated in 232 caregivers (Ullrich (2024), PMID 39095830) and is under multicenter longitudinal validation in 510 more (Haufschild (2026), PMID 42536973); the Modified Caregiver Strain Index has adequate psychometric properties in 200 palliative caregivers (Ramadasa (2023), PMID 37924086) and reached Cronbach's α of 0.91 in its traditional Chinese form in 125 family carers (Cheng (2021), PMID 34627225); the Zarit Burden Interview has good properties in this population (Quadir (2025), PMID 39510815) and serves as a primary endpoint in outpatient palliative trials (Kluger (2019), PMID 30779960). These are instruments of eight to twenty-two items. They are not burdensome.
 
 **The combination outperforms either alone.** The most directly applicable study we found validated criteria for identifying populations who may benefit from home-based primary care using the National Health and Aging Trends Study linked to Medicare claims (n = 21,727); those criteria combine **serious illness, functional impairment, and social isolation** — precisely the construct CMS describes at ¶947 — and are computable from data CMS already holds (Salinger (2025), PMID 39404637). Functional impairment and caregiver capacity are entangled and should be assessed together (Reckrey (2023), PMID 37349637, n = 2,521). Multi-domain complexity scales perform well: PALCOM, validated in 283 advanced cancer patients (Viladot (2023), PMID 37627210) and, pooled across 607, discriminating clinical instability at OR 4.3 (95% CI 1.86–9.73, p < 0.001) (Tuca (2024), PMID 38730696). Descriptions of advanced multimorbidity converge on combined rather than single-domain definitions (Bowers (2025), PMID 40110541).
 
@@ -87,7 +87,7 @@ On beneficiary protection: nine candidate quality indicators for abuse and negle
 
 **On CMS's own citations.** CMS cites Kelley (2017) at footnote 113 and the Kelley and Bollens-Lund denominator paper at footnote 114 (91 FR 43950 (¶948), (¶949)). Both are in our evidence base (Kelley (2017), PMID 26990009; Kelley and Bollens-Lund (2018), PMID 29125784) and both resolved as `uncertain / unavailable-fulltext` because we could not retrieve full text through an open-access route. We therefore rely on them as CMS does — for the framing of the denominator problem — rather than asserting extracted findings we did not verify.
 
-**Avoiding burden (¶953) — three mechanisms.** First, **reuse existing instruments**: ADL items are already collected in the MDS 3.0, OASIS, and the Annual Wellness Visit health risk assessment, and any should satisfy the functional criterion without re-administration. Second, **accept attestation of a single met criterion** on the claim rather than a scored instrument or narrative justification — any one of the criteria above identifies a population with elevated mortality and utilisation, so requiring all of them buys precision CMS does not need. Third, **pay for the assessment inside the service**: the PCM-1 assessment below *is* the eligibility determination, and making it a billable first service rather than an unpaid gate removes the incentive to under-identify and produces an auditable record as a by-product of care.
+**Avoiding burden (¶953) — three mechanisms.** First, **reuse existing instruments**: ADL items are already collected in the MDS 3.0, OASIS, and the Annual Wellness Visit health risk assessment, and any should satisfy the functional criterion without re-administration. Second, **accept attestation of a single met criterion** on the claim rather than a scored instrument or narrative justification — any one of the criteria above identifies a population with elevated mortality and utilization, so requiring all of them buys precision CMS does not need. Third, **pay for the assessment inside the service**: the PCM-1 assessment below *is* the eligibility determination, and making it a billable first service rather than an unpaid gate removes the incentive to under-identify and produces an auditable record as a by-product of care.
 
 ---
 
@@ -105,43 +105,43 @@ On beneficiary protection: nine candidate quality indicators for abuse and negle
 
 **Recommendation.** Differentiate on five axes, and establish a **distinct three-code palliative care management family** rather than adding conditions to APCM. The differentiators are: (1) a **named interdisciplinary core team** — physician or advanced practice clinician, registered nurse or advanced practice nurse, social worker, and chaplain or spiritual care professional — rather than undifferentiated "clinical staff"; (2) **the home as an expected site of service**, with a stated home-visit expectation; (3) **24-hour access to a clinician with access to the care plan**, not merely a triage line; (4) **the caregiver as a unit of assessment and service**, not as an informant; and (5) **a reassessment cadence measured in weeks to months**, driven by symptom change rather than an annual cycle.
 
-**The element set, and home visits.** A synthesis of scalable principles of community-based high-value care for seriously ill individuals identifies interdisciplinary team-based care, 24/7 access and responsiveness, concurrent palliative care alongside disease-directed treatment, targeting to high-risk patients, integrated medical and social supports, and caregiver support (Parikh (2017), PMID 27687915). CMS's ¶958 list is close to this; the literature supports adding **concurrent care** and **integrated social supports**, and making **caregiver assessment** a service element rather than a documentation item. Hughes (2023), PMID 37740895, reviews community palliative programme *components* specifically — the unit of analysis CMS's question needs. On the home: in a population-based cohort of 231,047 people, a physician home visit versus none carried an odds ratio of 1.95 (95% CI 1.91–2.00) on the utilisation outcomes studied, the largest delivery-model effect in our part (d) set (Brown (2021), PMID 33884934); and a claims-derived taxonomy of Medicare-funded home-based clinical care gives CMS a ready mechanism to distinguish home- from clinic-delivered care management in its own data (Ankuda (2023), PMID 36747175).
+**The element set, and home visits.** A synthesis of scalable principles of community-based high-value care for seriously ill individuals identifies interdisciplinary team-based care, 24/7 access and responsiveness, concurrent palliative care alongside disease-directed treatment, targeting to high-risk patients, integrated medical and social supports, and caregiver support (Parikh (2017), PMID 27687915). CMS's ¶958 list is close to this; the literature supports adding **concurrent care** and **integrated social supports**, and making **caregiver assessment** a service element rather than a documentation item. Hughes (2023), PMID 37740895, reviews community palliative program *components* specifically — the unit of analysis CMS's question needs. On the home: in a population-based cohort of 231,047 people, a physician home visit versus none carried an odds ratio of 1.95 (95% CI 1.91–2.00) on the utilization outcomes studied, the largest delivery-model effect in our part (d) set (Brown (2021), PMID 33884934); and a claims-derived taxonomy of Medicare-funded home-based clinical care gives CMS a ready mechanism to distinguish home- from clinic-delivered care management in its own data (Ankuda (2023), PMID 36747175).
 
-**Programme effect on utilisation and cost.** Cost reductions: roughly $12,000 per patient across 651 patients in an ACO-embedded home programme, p = 0.0002 (Lustbader (2017), PMID 27574868); $619 per patient per month, 95% CI −1,124 to −114, p = 0.0404, across 746 (Yosick (2019), PMID 30950679); $5,669 (95% CI 4,736–6,602) and $4,606 (95% CI 3,221–5,990) for Advanced Illness Management and $2,122 (95% CI 575–3,670) for a second Innovation Award model, all p < 0.001 (Ruiz (2017), PMID 30480116); $24,643 per member, a 16.7 percent reduction (Gordon (2022), PMID 34637346); and further plan-level results across 24,882 members (Bower (2024), PMID 38271576) and 1,453 Medicare Advantage beneficiaries (Cassel (2016), PMID 27590922). Utilisation: against hospice-only comparators in 7,177 patients, relative risks of 0.80 (95% CI 0.77–0.83) for hospitalisation and 0.68 (95% CI 0.62–0.74) for skilled nursing facility stays (Wang (2019), PMID 30830695). We flag the counterweight honestly: a systematic review restricted to **randomised** trials in adult non-cancer patients found the approach **cost-saving or cost-neutral**, observational studies reporting larger savings than trials do (Janke (2024), PMID 39098890). CMS should expect this benefit to pay for itself, not to generate net savings.
+**Programme effect on utilization and cost.** Cost reductions: roughly $12,000 per patient across 651 patients in an ACO-embedded home program, p = 0.0002 (Lustbader (2017), PMID 27574868); $619 per patient per month, 95% CI −1,124 to −114, p = 0.0404, across 746 (Yosick (2019), PMID 30950679); $5,669 (95% CI 4,736–6,602) and $4,606 (95% CI 3,221–5,990) for Advanced Illness Management and $2,122 (95% CI 575–3,670) for a second Innovation Award model, all p < 0.001 (Ruiz (2017), PMID 30480116); $24,643 per member, a 16.7 percent reduction (Gordon (2022), PMID 34637346); and further plan-level results across 24,882 members (Bower (2024), PMID 38271576) and 1,453 Medicare Advantage beneficiaries (Cassel (2016), PMID 27590922). Utilisation: against hospice-only comparators in 7,177 patients, relative risks of 0.80 (95% CI 0.77–0.83) for hospitalization and 0.68 (95% CI 0.62–0.74) for skilled nursing facility stays (Wang (2019), PMID 30830695). We flag the counterweight honestly: a systematic review restricted to **randomized** trials in adult non-cancer patients found the approach **cost-saving or cost-neutral**, observational studies reporting larger savings than trials do (Janke (2024), PMID 39098890). CMS should expect this benefit to pay for itself, not to generate net savings.
 
-**Timely follow-up after ED visit or discharge.** A transitional model of weekly home visits or calls for four weeks then monthly follow-up produced a readmission relative risk of 0.55 (95% CI 0.35–0.88) in 84 randomised patients with end-stage heart failure (Wong (2016), PMID 26969631); meta-analysed randomised transitional-care trials in serious illness show a standardised mean difference of 0.20 (95% CI 0.08–0.33) at moderate-to-very-low certainty (Di Nitto (2025), PMID 40389925). Saunders (2019), PMID 31201875, reviews acute-to-community transitions but found most constituent studies at high risk of bias, so we do not rest weight on its pooled direction. On the ED link: the EMPallA randomised trial of 500 older adults reported a 9.6-point quality-of-life difference (95% CI 5.9–13.3) (Schmucker (2021), PMID 34247588), and Stoltenberg (2022), PMID 36518882, describes an integrated geriatric palliative consultation service built into the ED. This element is well supported and should be required.
+**Timely follow-up after ED visit or discharge.** A transitional model of weekly home visits or calls for four weeks then monthly follow-up produced a readmission relative risk of 0.55 (95% CI 0.35–0.88) in 84 randomized patients with end-stage heart failure (Wong (2016), PMID 26969631); meta-analyzed randomized transitional-care trials in serious illness show a standardized mean difference of 0.20 (95% CI 0.08–0.33) at moderate-to-very-low certainty (Di Nitto (2025), PMID 40389925). Saunders (2019), PMID 31201875, reviews acute-to-community transitions but found most constituent studies at high risk of bias, so we do not rest weight on its pooled direction. On the ED link: the EMPallA randomized trial of 500 older adults reported a 9.6-point quality-of-life difference (95% CI 5.9–13.3) (Schmucker (2021), PMID 34247588), and Stoltenberg (2022), PMID 36518882, describes an integrated geriatric palliative consultation service built into the ED. This element is well supported and should be required.
 
-**24-hour access and out-of-hours response.** Goodrich (2024), PMID 38381768, synthesises systematic-review evidence on out-of-hours utilisation of community palliative services; a programme combining coordination centres, an out-of-hours telephone line and discharge in-reach nurses across 3,594 patients produced odds ratios of 0.20 (95% CI 0.17–0.27) and 0.33 (95% CI 0.21–0.50) for hospital death in its two areas (p ≤ 0.001) (Purdy (2015), PMID 24838731), and a transmural care pathway raised death at the preferred place (adjusted OR 3.14, 95% CI 1.49–6.62, p = 0.003) among 711 older adults (Van Doorne (2023), PMID 37533107).
+**24-hour access and out-of-hours response.** Goodrich (2024), PMID 38381768, synthesises systematic-review evidence on out-of-hours utilization of community palliative services; a program combining coordination centers, an out-of-hours telephone line and discharge in-reach nurses across 3,594 patients produced odds ratios of 0.20 (95% CI 0.17–0.27) and 0.33 (95% CI 0.21–0.50) for hospital death in its two areas (p ≤ 0.001) (Purdy (2015), PMID 24838731), and a transmural care pathway raised death at the preferred place (adjusted OR 3.14, 95% CI 1.49–6.62, p = 0.003) among 711 older adults (Van Doorne (2023), PMID 37533107).
 
-**Team composition beyond the physician.** Across 574,567 high-intensity home-based primary care users in national Medicare claims, nurse practitioner-delivered care performed comparably on twelve quality measures in three domains — direct evidence that supervision should not be physician-only (Perloff (2024), PMID 39291622). Core competencies for palliative social workers support the social worker as a defined team member rather than optional auxiliary personnel (Yao (2025), PMID 40490802); community health workers delivering monthly home visits were evaluated across 903 patients (Britt (2019), PMID 30541333) and economically across four cancer centres (Monton (2025), PMID 41193212); nurse-led rural COPD models are reviewed in Poston (2025), PMID 40724712.
+**Team composition beyond the physician.** Across 574,567 high-intensity home-based primary care users in national Medicare claims, nurse practitioner-delivered care performed comparably on twelve quality measures in three domains — direct evidence that supervision should not be physician-only (Perloff (2024), PMID 39291622). Core competencies for palliative social workers support the social worker as a defined team member rather than optional auxiliary personnel (Yao (2025), PMID 40490802); community health workers delivering monthly home visits were evaluated across 903 patients (Britt (2019), PMID 30541333) and economically across four cancer centers (Monton (2025), PMID 41193212); nurse-led rural COPD models are reviewed in Poston (2025), PMID 40724712.
 
 **Electronic care plans and coordination.** This is the thinnest element and we say so. Voruganti (2017), PMID 28720558, piloted a web-based collaborative-care tool in 48 patients with advanced cancer; Fathi (2016), PMID 27217093, developed an interdisciplinary team communication framework and quality metrics for home-based medical care practices, a record that resolved as `uncertain / unavailable-fulltext` and which we cite for scope only. CMS should require an electronic care plan available at transitions of care because it is a prerequisite for the coordination elements that *are* well evidenced, not because the artefact itself has been shown to change outcomes.
 
-**Initiating trigger, and telehealth.** Early palliative identification across 2,370 patients produced a hazard ratio of 1.69 (95% CI 1.56–1.82) for subsequent palliative service use (Mittmann (2020), PMID 32005036); with Salinger (2025) in part (c), this supports **registry- or risk-based identification** as a legitimate initiating trigger, and we recommend against a face-to-face initiating visit within a fixed window as the *only* permitted trigger — the question CMS raises at 91 FR 43938 ¶8–43939 ¶1. On modality: a review of telehealth integration into community-based palliative care across seven studies reports functional status, hospitalisation, psychological distress and quality-of-life outcomes but notes most constituent trials carried some concerns of bias and two were at high risk from attrition (Haroen (2025), PMID 41230403). Telehealth should be explicitly permitted for the follow-up code and explicitly insufficient on its own to satisfy the home-visit expectation.
+**Initiating trigger, and telehealth.** Early palliative identification across 2,370 patients produced a hazard ratio of 1.69 (95% CI 1.56–1.82) for subsequent palliative service use (Mittmann (2020), PMID 32005036); with Salinger (2025) in part (c), this supports **registry- or risk-based identification** as a legitimate initiating trigger, and we recommend against a face-to-face initiating visit within a fixed window as the *only* permitted trigger — the question CMS raises at 91 FR 43938 ¶8–43939 ¶1. On modality: a review of telehealth integration into community-based palliative care across seven studies reports functional status, hospitalization, psychological distress and quality-of-life outcomes but notes most constituent trials carried some concerns of bias and two were at high risk from attrition (Haroen (2025), PMID 41230403). Telehealth should be explicitly permitted for the follow-up code and explicitly insufficient on its own to satisfy the home-visit expectation.
 
 **What should not be included.** We recommend against three requirements: **a minimum monthly staff-time threshold as the sole basis for payment** (no included evaluation identifies a time threshold as the active ingredient, and such thresholds are part of the complexity CMS says it wants to simplify, 91 FR 43938 ¶5); **a requirement that the beneficiary forgo disease-directed treatment** (concurrent care is one of Parikh's scalable principles, and part (a) shows what happens when a benefit forces that trade-off); and **physician-only supervision**, for the reasons in Perloff (2024). Singer (2016), PMID 27533892, surveying populations and interventions across the field, supports breadth of population rather than narrowing by intervention type.
 
 ### The concrete structure: a three-code palliative care management family
 
-These are **draft descriptors offered for comment, not valued codes**: nothing here models RUC valuation, RVU inputs, or budget neutrality. Each is crosswalked to three professional standards — the Center to Advance Palliative Care (CAPC) programme standards, the National Association of Social Workers (NASW) *Practice Standards for Serious Illness Care: Hospice and Palliative Social Work*, and The Joint Commission (TJC) Palliative Care (PAL) performance measure set — so that every required element is an existing professional-standards requirement rather than an invention of this comment.
+These are **draft descriptors offered for comment, not valued codes**: nothing here models RUC valuation, RVU inputs, or budget neutrality. Each is crosswalked to three professional standards — the Center to Advance Palliative Care (CAPC) program standards, the National Association of Social Workers (NASW) *Practice Standards for Serious Illness Care: Hospice and Palliative Social Work*, and The Joint Commission (TJC) Palliative Care (PAL) performance measure set — so that every required element is an existing professional-standards requirement rather than an invention of this comment.
 
 **PCM-1 — Palliative care screening and initial assessment.**
 
 > Palliative care screening and comprehensive initial assessment of a patient with serious illness, by a member of the palliative care core interdisciplinary team; including validated symptom screening (pain, dyspnea), assessment of functional status and caregiver strain, psychosocial and spiritual screening, and documentation of screening results in the medical record; first 60 minutes
 
-*Standards:* CAPC screening-criteria consensus (Weissman and Meier (2011), PMID 21133809) and daily symptom-assessment metric domain (Weissman, Morrison and Meier (2010), PMID 19922199); NASW Standard 3 (Assessment); TJC PAL-01, PAL-02, PAL-03, and the core-team face-to-face population definition. *PFS analogues:* 99497; 99495. *Answers parts (b), (c), (d).*
+*Standards:* CAPC screening-criteria consensus (Weissman and Meier (2011), PMID 21133809) and daily symptom-assessment metric domain (Weissman, Morrison and Meier (2010), PMID 19922199); NASW Standard 3 (Assessment); TJC PAL-01, PAL-02, PAL-03, and the core-team face-to-face population definition. *PFS analogs:* 99497; 99495. *Answers parts (b), (c), (d).*
 
 **PCM-2 — Palliative care planning.**
 
 > Development and documentation of a patient-centered palliative care plan by the interdisciplinary team with patient and caregiver participation; including goals of care, treatment preferences, advance directive reconciliation, and shared decision-making; recorded as an electronic care plan available at transitions of care, including discharge
 
-*Standards:* CAPC patient-centered goals-of-care metric domain (2010 consensus, PMID 19922199); NASW Standards 4 (Intervention/Treatment Planning) and 7 (Documentation); TJC PAL-04 and PAL-05. *PFS analogues:* 99497/99498; the care-plan element of 99490. *Answers part (d).*
+*Standards:* CAPC patient-centered goals-of-care metric domain (2010 consensus, PMID 19922199); NASW Standards 4 (Intervention/Treatment Planning) and 7 (Documentation); TJC PAL-04 and PAL-05. *PFS analogs:* 99497/99498; the care-plan element of 99490. *Answers part (d).*
 
 **PCM-3 — Palliative care follow-up.**
 
 > Palliative care follow-up management, per calendar month: symptom reassessment and care-plan revision; continuity with a designated team member; access to timely clinical support; coordination with treating physicians; patient and caregiver education and support; and timely follow-up after emergency department visit or facility discharge
 
-*Standards:* CAPC transitions-across-care-sites and caregiver-support metric domains (2010 consensus, PMID 19922199); NASW Standards 8 (Interdisciplinary Teamwork) and 6 (Empowerment and Advocacy); TJC core-team continuity underlying the PAL population. *PFS analogues:* G0556–G0558; 99490/99487. *Answers parts (d) and (e).*
+*Standards:* CAPC transitions-across-care-sites and caregiver-support metric domains (2010 consensus, PMID 19922199); NASW Standards 8 (Interdisciplinary Teamwork) and 6 (Empowerment and Advocacy); TJC core-team continuity underlying the PAL population. *PFS analogs:* G0556–G0558; 99490/99487. *Answers parts (d) and (e).*
 
 The service elements in PCM-3 deliberately track CMS's own list at ¶958 — continuity with a designated team member, access to timely clinical support, comprehensive symptom and caregiver assessment, electronic care plans, coordination with treating physicians, patient and caregiver education, and timely follow-up after ED visit or discharge — and the crosswalk shows each is also a professional-standards requirement. The elements we would add are **concurrent care with disease-directed treatment**, **integrated medical and social supports**, and an **expected home-visit cadence**, each supported above.
 
@@ -159,19 +159,19 @@ The service elements in PCM-3 deliberately track CMS's own list at ¶958 — con
 >
 > "**If no viable MIPS Value Pathway reporting mechanism is found, what are the essential quality elements required for palliative care management? Is sole reporting of ambulatory palliative care patients feeling heard and understood (CBE 3665) sufficient? Should other measures be considered?**" — 91 FR 43950 (¶962)
 
-**Recommendation.** **Yes to a reporting requirement; no to the primary-care MIPS Value Pathway; and no, CBE 3665 alone is not sufficient.** CMS should require quality reporting as a condition of billing the family but develop a **serious-illness MVP** rather than extending the primary-care MVP to this population. In the interim, CMS should require a **five-element minimum set**: one care-experience measure (CBE 3665), one patient-reported symptom measure, one documented-process measure, one claims-computable utilisation outcome, and one caregiver measure.
+**Recommendation.** **Yes to a reporting requirement; no to the primary-care MIPS Value Pathway; and no, CBE 3665 alone is not sufficient.** CMS should require quality reporting as a condition of billing the family but develop a **serious-illness MVP** rather than extending the primary-care MVP to this population. In the interim, CMS should require a **five-element minimum set**: one care-experience measure (CBE 3665), one patient-reported symptom measure, one documented-process measure, one claims-computable utilization outcome, and one caregiver measure.
 
-**Why not the primary-care MVP.** Its measure concepts — blood pressure control, haemoglobin A1c control, cancer screening completion — are the wrong targets for a population in the last year or two of life, and requiring them would penalise good care. In 381 adults randomised with an estimated life expectancy of one month to one year, discontinuing statin therapy produced no excess mortality (23.8 versus 20.3 percent; 90% CI −3.5 to 10.5 percent; p = 0.36), with improved quality of life and cost savings (Kutner (2015), PMID 25798575). Deprescribing is the evidence-based act; a primary-care MVP would score it as a care gap.
+**Why not the primary-care MVP.** Its measure concepts — blood pressure control, hemoglobin A1c control, cancer screening completion — are the wrong targets for a population in the last year or two of life, and requiring them would penalize good care. In 381 adults randomized with an estimated life expectancy of one month to one year, discontinuing statin therapy produced no excess mortality (23.8 versus 20.3 percent; 90% CI −3.5 to 10.5 percent; p = 0.36), with improved quality of life and cost savings (Kutner (2015), PMID 25798575). Deprescribing is the evidence-based act; a primary-care MVP would score it as a care gap.
 
 **On CBE 3665.** We support including the measure and oppose relying on it alone, for three reasons. First, **its measurement properties support use as one measure, not the measure**: the Feeling Heard and Understood scale validated in 608 outpatients showed excellent internal consistency (Cronbach's α = 0.93) but a test–retest intraclass correlation of **0.54**, only moderate stability (Takahashi (2025), PMID 40547401). Second, **we could not verify the measure's own development and testing literature**: the inpatient development study (Gramling (2016), PMID 26596879), the before-and-after benchmarking study (Ingersoll (2018), PMID 29729348), and the RAND MACRA testing summary report (Ahluwalia (2022), PMID 35837526) all reached full-text assessment and all resolved as `uncertain / unavailable-fulltext`. A safeguard whose validation literature is not readily obtainable by the clinicians subject to it is a poor candidate for sole reliance. Third, **an experience measure is only collectable from beneficiaries who can respond**; the nursing-home cancer cohort of Vu (2024), PMID 38560743, and the dementia cohorts of Luth (2020), PMID 31750935, and Nothelle (2022), PMID 35488709, show the scale of that exclusion. A set resting entirely on patient self-report will systematically not measure the sickest beneficiaries.
 
-**Candidate measures.** *Patient-reported symptom burden.* The Integrated Palliative care Outcome Scale (IPOS) is the best-evidenced community candidate: validated in 376 adults across home, outpatient and community settings (α = 0.77, acceptable-to-good test–retest reliability) (Murtagh (2019), PMID 31185804); implemented as measurement-based palliative care across 1,507 community patients (α = 0.78), where staff valued it and perceived improved service quality (Sandham (2022), PMID 35805407); and, in 557 patients, surfacing information physicians had not otherwise known in **69.3 percent** of questionnaires (Müller (2023), PMID 36641450) — it changes what clinicians know, not just what regulators see. Across 458 specialised institutions, 92 percent of PROM users considered them useful for symptom management (p = 0.002) (Ito (2023), PMID 36894802). The Edmonton Symptom Assessment System is a reasonable alternative. One caution: professionals need training and implementation support to use PROMs well (Bausewein (2011), PMID 21507232, n = 663), so CMS should phase this requirement in.
+**Candidate measures.** *Patient-reported symptom burden.* The Integrated Palliative care Outcome Scale (IPOS) is the best-evidenced community candidate: validated in 376 adults across home, outpatient and community settings (α = 0.77, acceptable-to-good test–retest reliability) (Murtagh (2019), PMID 31185804); implemented as measurement-based palliative care across 1,507 community patients (α = 0.78), where staff valued it and perceived improved service quality (Sandham (2022), PMID 35805407); and, in 557 patients, surfacing information physicians had not otherwise known in **69.3 percent** of questionnaires (Müller (2023), PMID 36641450) — it changes what clinicians know, not just what regulators see. Across 458 specialized institutions, 92 percent of PROM users considered them useful for symptom management (p = 0.002) (Ito (2023), PMID 36894802). The Edmonton Symptom Assessment System is a reasonable alternative. One caution: professionals need training and implementation support to use PROMs well (Bausewein (2011), PMID 21507232, n = 663), so CMS should phase this requirement in.
 
-*Documented process.* Piloting the PEACE measures in 593 patients found two both usable and feasible — **screening for physical symptoms on admission** and **documentation of treatment preferences** (Schenck (2014), PMID 24921162) — exactly the elements PCM-1 and PCM-2 require, so they can be computed from care already paid for. The gap is large: among 562 seriously ill hospitalised patients, a comprehensive palliative assessment was completed for 10 percent, against 56 percent of those receiving specialist palliative care (p < 0.001) (Hanson (2012), PMID 22687268). De Roo (2013), PMID 23809769, provides the systematic-review basis for indicator selection.
+*Documented process.* Piloting the PEACE measures in 593 patients found two both usable and feasible — **screening for physical symptoms on admission** and **documentation of treatment preferences** (Schenck (2014), PMID 24921162) — exactly the elements PCM-1 and PCM-2 require, so they can be computed from care already paid for. The gap is large: among 562 seriously ill hospitalized patients, a comprehensive palliative assessment was completed for 10 percent, against 56 percent of those receiving specialist palliative care (p < 0.001) (Hanson (2012), PMID 22687268). De Roo (2013), PMID 23809769, provides the systematic-review basis for indicator selection.
 
-*Claims-computable utilisation outcomes.* These impose no reporting burden because CMS computes them. Place of death and hospice enrolment were validated as end-of-life quality measures across 1.5 million nursing home residents (Mukamel (2012), PMID 22500481). Potentially-inappropriate-end-of-life-care indicators — chemotherapy near death, two or more emergency department visits, two or more admissions, hospitalisations over 14 days, intensive care admission, hospital death — applied across 43,067 patients distinguished specialist from generalist palliative models at an adjusted odds ratio of 0.20 (95% CI 0.15–0.26, p < 0.001) (Boddaert (2022), PMID 33355176). Days alive and out of hospital, and quality-of-life-adjusted days at home, were used in 272 patients with heart failure (Campbell (2018), PMID 29952090); days at home most closely tracks what beneficiaries say they want. Indicators derived for 14,312 seriously ill home care clients from the Resident Assessment Instrument for Home Care reached an odds ratio of 2.77 (95% CI 2.66–2.89) against negative health outcomes (Harman (2019), PMID 30626374) — an existing instrument can carry the set.
+*Claims-computable utilization outcomes.* These impose no reporting burden because CMS computes them. Place of death and hospice enrollment were validated as end-of-life quality measures across 1.5 million nursing home residents (Mukamel (2012), PMID 22500481). Potentially-inappropriate-end-of-life-care indicators — chemotherapy near death, two or more emergency department visits, two or more admissions, hospitalizations over 14 days, intensive care admission, hospital death — applied across 43,067 patients distinguished specialist from generalist palliative models at an adjusted odds ratio of 0.20 (95% CI 0.15–0.26, p < 0.001) (Boddaert (2022), PMID 33355176). Days alive and out of hospital, and quality-of-life-adjusted days at home, were used in 272 patients with heart failure (Campbell (2018), PMID 29952090); days at home most closely tracks what beneficiaries say they want. Indicators derived for 14,312 seriously ill home care clients from the Resident Assessment Instrument for Home Care reached an odds ratio of 2.77 (95% CI 2.66–2.89) against negative health outcomes (Harman (2019), PMID 30626374) — an existing instrument can carry the set.
 
-*Caregiver measures.* Caregiver-reported quality measures were characterised in 391 home hospice patients using the Burden Scale for Family Caregivers, FAMCARE-2 and CEQUEL (Phongtankuel (2020), PMID 32856023); CEQUEL itself was validated in 275 caregivers of advanced cancer patients (Higgins (2013), PMID 23762467). Since caregiver strain is one of the eligibility constructs CMS raises at ¶947 and ¶952, measuring the caregiver outcome closes the loop.
+*Caregiver measures.* Caregiver-reported quality measures were characterized in 391 home hospice patients using the Burden Scale for Family Caregivers, FAMCARE-2 and CEQUEL (Phongtankuel (2020), PMID 32856023); CEQUEL itself was validated in 275 caregivers of advanced cancer patients (Higgins (2013), PMID 23762467). Since caregiver strain is one of the eligibility constructs CMS raises at ¶947 and ¶952, measuring the caregiver outcome closes the loop.
 
 *Equity monitoring.* End-of-life quality metrics among 126,434 Medicare decedents with poor-prognosis cancers differed at minority-serving cancer centers on all four metrics examined (10.3 versus 7.4 percent, p < .01; 32.9 versus 30.4 percent, p = .03; 39.5 versus 37.0 percent, p = .03; 19.4 versus 16.2 percent, p < .01) (Wasp (2020), PMID 31925998), and hospice CAHPS scores across 217,596 patients ranged from 74.9 to 89.5 (Parast (2021), PMID 33469741). Whatever set CMS adopts should be stratified and reported by race, ethnicity, dual-eligibility status and rurality from the first performance year. On selection principles and what is coming, see Kamal (2015), PMID 25827852, on "measuring what matters," and Schlang (2026), PMID 42157151, on patient-reported serious-illness quality measures now in development — the most promising near-term source of measures purpose-built for this population.
 
@@ -191,8 +191,8 @@ The service elements in PCM-3 deliberately track CMS's own list at ¶958 — con
 8. **Do not require** a minimum monthly staff-time threshold as the sole basis for payment, a face-to-face initiating visit as the only permitted trigger, or physician-only supervision.
 9. **Permit registry- and risk-based identification as an initiating trigger** (Mittmann (2020), PMID 32005036; Salinger (2025), PMID 39404637).
 10. **Permit telehealth for follow-up but do not allow it to satisfy the home-visit expectation.**
-11. **Require quality reporting, but develop a serious-illness MVP rather than extending the primary-care MVP**, whose measure concepts would penalise evidence-based deprescribing (Kutner (2015), PMID 25798575).
-12. **Do not rely on CBE 3665 alone.** Require a five-element minimum set: CBE 3665; a patient-reported symptom measure (IPOS or ESAS); two PEACE-derived process measures (symptom screening at initiation; documented treatment preferences); one claims-computable utilisation outcome (days at home, or potentially-inappropriate-end-of-life-care indicators); and one caregiver measure (CEQUEL or equivalent).
+11. **Require quality reporting, but develop a serious-illness MVP rather than extending the primary-care MVP**, whose measure concepts would penalize evidence-based deprescribing (Kutner (2015), PMID 25798575).
+12. **Do not rely on CBE 3665 alone.** Require a five-element minimum set: CBE 3665; a patient-reported symptom measure (IPOS or ESAS); two PEACE-derived process measures (symptom screening at initiation; documented treatment preferences); one claims-computable utilization outcome (days at home, or potentially-inappropriate-end-of-life-care indicators); and one caregiver measure (CEQUEL or equivalent).
 13. **Stratify and publicly report all quality results** by race, ethnicity, dual-eligibility status and rurality from the first performance year.
 14. **Focus program-integrity attention** on enrollment and disenrollment churn, ownership and margin as risk stratifiers, the initiating trigger and practitioner-to-auxiliary work ratio, and equity monitoring of the integrity program itself — and treat abuse and neglect in the home as an integrity domain, using the indicator set of Sheehan (2016), PMID 27787878.
 15. **Build measurement into the benefit from day one.** The clearest finding of this review is that the direct evidence CMS asks for does not yet exist for a service Medicare has not yet created. The remedy is a designed evaluation, not a longer wait.