Percentage of patients, aged 18 years and older, who died with cancer admitted to the ICU in the last 30 days of life
Measure Specs
General Information
Intensive Care Unit (ICU) admissions in the final days of life among patients with cancer are widely regarded as the result of aggressive end of life (EOL) care and as a marker of failure in advance care planning or late-stage palliative care integration. Patients dying in an ICU have higher levels of physical and emotional distress compared to patients dying at home or in hospice (National Comprehensive Cancer Network [NCCN], 2026); caregivers and family members also perceive better EOL care when their loved ones avoid ICU admissions within 30 days of death, have earlier hospice enrollment, and death occurs outside of the hospital (Wright et al., 2016). The seminal Dying in America report states that a palliative approach often offers the best chance of maintaining the highest possible quality of life for those living with advanced serious illness (Institute of Medicine [IOM], 2015) and proposes, as a core component to quality end-of-life care, to offer palliative care services and a personalized revision of the care plan, as well as access to services based on the changing needs of the patient and family (IOM, 2015). Community-based or home-based palliative care services have been associated with a reduced need for end-of-life emergency department (ED) visits, reduced length and frequency of hospitalizations, and fewer ICU admissions and in-hospital deaths (NCCN, 2026). One matched cohort study evaluated Medicare patients with metastatic cancer also found that patients who received a palliative care consult spent approximately $2,000 less in healthcare costs versus those with no palliative care consultation; patients had even more savings when a palliative care consultation was more than four weeks prior to death (Sheridan et al., 2021).
The goal of Percentage of Patients who Died with Cancer Admitted to the Intensive Care Unit in the Last 30 Days of Life (lower score – better) is to, alongside ASCO’s suite of EOL measures, highlight performance trends over time and encourage timely enrollment in palliative care that focuses on symptom management, rather than low utility and aggressive treatments, among people dying with cancer. This results in a reduction of aggressive interventions leading to ICU visits, ED visits, hospitalizations, improved symptom control and quality of life, and ultimately improved patient, family, and caregiver satisfaction. Timely enrollment in palliative care also reduces resource utilization costs and aligns with the Medicare Payment Advisory Commission’s (MedPAC) goal to reduce high-intensity, low-value care at the end of life by promoting hospice and palliative care (MedPAC, 2025).
As noted later in this form, these end-of-life measures were initially developed in 2003 as clinical indicators for healthcare systems using existing administrative data. Over time these measures were developed as registry measures, but this ICU measure was removed from MIPS in CY 2023 due to a limited patient population that was capturable. As part of the 2023-2024 ASCO EOL Measures Technical Expert Panel (TEP) work, the panel updated this measure and specified it at the practice level, using claims data to aid with more consistent and reliable case identification with no extra administrative burden.
Note that ASCO’s EOL Measures TEP emphasized that performance is not expected to be perfect on this quality measure. A margin of error should be expected to account for scenarios such as patient and family preferences, barriers to palliative care and hospice access, and sudden patient decline.
References:
- Institute of Medicine. (2015). Dying in America: Improving Quality and Honoring Individual Preferences Near The End of Life. National Academies Press. https://doi.org/10.17226/18748
- Medicare Payment Advisory Commission. (2025, March). Report to the Congress: Medicare payment policy. https://www.medpac.gov/wp-content/uploads/2025/03/Mar25_MedPAC_Report_To_Congress_SEC.pdf
- National Comprehensive Cancer Network. (2026). NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines®): Palliative Care (Version 1.2026). https://www.nccn.org/professionals/physician_gls/pdf/palliative.pdf
- Sheridan, P. E., LeBrett, W. G., Triplett, D. P., Roeland, E. J., Bruggeman, A. R., Yeung, H. N., & Murphy, J. D. (2021). Cost Savings Associated with Palliative Care Among Older Adults with Advanced Cancer. American Journal of Hospice and Palliative Medicine, 38(10), 1250–1257. https://doi.org/10.1177/1049909120986800
- Wright, A. A., Keating, N. L., Ayanian, J. Z., Chrischilles, E. A., Kahn, K. L., Ritchie, C. S., Weeks, J. C., Earle, C. C., & Landrum, M. B. (2016). Family Perspectives on Aggressive Cancer Care Near the End of Life. JAMA, 315(3), 284–292. https://doi.org/10.1001/jama.2015.18604
Claims data is a type of administrative data generated every time a healthcare provider submits a request for payment to an insurance payer. Claims data is highly standardized and for the purposes of this measure captures the exact location and dates of service.
Numerator
Patients who were admitted to the ICU in the last 30 days of life
Patients who were admitted to the ICU in the last 30 days of life
Guidance: To count towards the numerator, there should be an ICU revenue code on the inpatient claims form and the inpatient discharge date should be within the last 30 days of the patient’s life. The inpatient discharge date is inclusive of the 30-day timeframe. Step down units are not included in this measure.
Numerator Criteria:
Inpatient Type of Bill:
111- Hospital Inpatient (Including Medicare Part A) admit through discharge
121- Hospital Inpatient (Medicare Part B Only) admit through discharge
851- Specialty Facility Critical Access Hospital: admit through discharge
AND
Most Recent Inpatient Discharge Date
AND
(Date of death minus Most Recent Inpatient Discharge Date) < 30 days
AND
Revenue code
Intensive Care Unit (ICU):
0200 - General
0201 - Surgical
0202 - Medical
0203 - Pediatric
0204 - Psychiatric
0206 - Intermediate ICU
0207 - Burn Care
0208 - Trauma
0209 – Other
OR
Coronary Care Unit (CCU):
0210 - General
0211 - Myocardial Infarction
0212 - Pulmonary Care
0213 - Heart Transplant
0214 - Intermediate CCU
0219 - Other Coronary CCU
Denominator
Patients, aged 18 years and older, who died with cancer
Patients, aged 18 years and older, who died with cancer.
Denominator Criteria (Eligible Cases):
- Patients aged ≥ 18 years at the start of the measurement period:
(Start Date of the Measurement Period minus Date of Birth) ≥ 18 years
AND
- At least two outpatient encounters that meet the following criteria:
2a) Place of Service: 02, 05, 07, 10, 11, 19, 22, 49, 50, 71, 72
AND
2b) Professional Service code:
98000, 98001, 98002, 98003, 98004, 98005, 98006, 98007, 98008, 98009, 98010, 98011, 98012, 98013, 98014, 98015, 99202, 99203, 99204, 99205, 99212, 99213, 99214, 99215, 99242, 99243, 99244, 99245, 99495, 99496, 99441, 99442, 99443
(NOTE: Encounters coded with telehealth modifier GQ, GT, or 95 are allowed for both visits.)
AND
2c) Service Date <during Measurement Period>
AND
2d) Diagnosis code for Cancer
(See tab “CancerDx” in “EOLMeasures_Coding”)
AND
Date of death <during Measurement Period>
Guidance:
For physician/group reporting, attribution of patients to an oncology practice is based on the presence of at least two outpatient visit claims for the patient with that practice. The two outpatient encounters required in the denominator are outpatient visits that occur on different calendar days.
To be eligible in the denominator, patients must have continuous coverage during the measurement period.
A cancer diagnosis code must appear within the top 3 diagnosis positions on an outpatient visit claim that meets the denominator encounter requirement.
Exclusions
This quality measure has no denominator exclusions, but does have denominator exceptions. See below for details:
Denominator exceptions:
Patients admitted to the ICU due to complications from 1) receipt or in process of receipt of bone marrow or peripheral blood stem cell transplant (transplant status) in the last 60 days of life, or 2) receipt or in process of receipt of CAR T cell therapy in the last 60 days of life
Denominator exception details:
Denominator Exceptions Criteria:
Receipt or in process of receipt of bone marrow or peripheral blood stem cell transplant
Code: (See tab “BoneMarrowStemCellTransplant” in “EOLMeasures_Coding” file)
AND
Service/Procedure Date or Claim from Date (for ICD-10-CM transplant status code)
AND
(Date of death minus Service/Procedure Date or Claim from Date) < 60 days
OR
Receipt or in process of receipt of CAR T cell therapy
Code: (See tab “CARTCellTx” in “EOLMeasures_Coding” file)
AND
Service/Procedure Date
AND
(Date of death minus Service/Procedure Date) < 60 days
Measure Calculation
See attached.
This measure is not stratified.
Minimum of five (5) patients.
Supplemental Attachment
Point of Contact
COPYRIGHT:
The Measure is not a clinical guideline, does not establish a standard of medical care, and has not been tested for all potential applications.
The Measure, while copyrighted, can be reproduced and distributed, without modification, for noncommercial purposes, e.g., use by health care providers in connection with their practices. Commercial use is defined as the sale, license, or distribution of the Measures for commercial gain, or incorporation of the Measure into a product or service that is sold, licensed or distributed for commercial gain.
Commercial uses of the Measure require a license agreement between the user and the American Society of Clinical Oncology (ASCO) and prior written approval of ASCO. Contact [email protected] for licensing this measure. Neither ASCO nor its members shall be responsible for any use of the Measure.
ASCO encourages use of the Measures by other health care professionals, where appropriate.
THE MEASURE AND SPECIFICATIONS ARE PROVIDED "AS IS" WITHOUT WARRANTY OF ANY KIND.
© 2026 American Society of Clinical Oncology. All Rights Reserved.
Limited proprietary coding is contained in the Measure specifications for user convenience. Users of proprietary code sets should obtain all necessary licenses from the owners of the code sets. ASCO disclaims all liability for use or accuracy of any third party codes contained in the specifications.
CPT® contained in the Measure specifications is copyright 2004-2026 American Medical Association. LOINC® copyright 2004-2026 Regenstrief Institute, Inc. This material contains SNOMED Clinical Terms® (SNOMED CT®) copyright 2004-2026 International Health Terminology Standards Development Organisation. ICD-10 copyright 2026 World Health Organization. All Rights Reserved.
Neha Agrawal
Alexandria, VA
United States
Neha Agrawal
ASCO
Alexandria, VA
United States
Importance
Evidence
In the United States, cancer is the second leading cause of death overall and the leading cause of death among people younger than 85 years (Siegel et al., 2026). It is projected that in 2026 there will be approximately 2.1 million new cancer cases and over half a million cancer deaths (Siegel et al., 2026). While individual patients have their own preferences that can change over time, consistently across various populations, most patients nearing end of life wish to die at home (Gomes et al., 2013). Hospitalizations, ED visits, and ICU stays in the last 30 days of life have been repeatedly associated with poor quality end-of-life care, as reported by family caregivers (Ersek et al., 2017, Wright et al., 2016, and Christian et al., 2021). In one national study, family caregivers gave significantly higher EOL care quality ratings for care at home under hospice, than to EOL care received in the ICU, hospital palliative care unit, hospice inpatient unit, or residential hospice (Zhu et al., 2024). And cancer-directed therapy received near the end of life continues to be associated with more hospitalizations, ED visits, and ICU stays (Garg et al., 2024 & Adelson et al., 2024).
As noted above, the goal of Percentage of Patients who Died with Cancer Admitted to the Intensive Care Unit in the Last 30 Days of Life (lower score – better) is to, alongside ASCO’s suite of EOL measures, highlight performance trends over time and encourage timely enrollment in palliative care that focuses on symptom management, rather than low utility and aggressive treatments, among people dying with cancer. This results in a reduction of aggressive interventions leading to ICU visits, ED visits, and hospitalizations, improved symptom control and quality of life, and ultimately improved patient, family, and caregiver satisfaction. Timely enrollment in palliative care also reduces resource utilization costs and aligns with MedPAC’s goal to reduce high-intensity, low-value care at the end of life by promoting hospice and palliative care (MedPAC, 2025). Studies show that the integration of palliative care into the cancer care continuum improves patient outcomes in many ways, including quality of life, symptoms intensity, and end-of-life care (NCCN, 2026). As noted above, the seminal NAM report Dying in America states that a palliative approach often offers the best chance of maintaining the highest possible quality of life for those living with advanced serious illness (Institute of Medicine [IOM], 2015) and proposes, as a core component to quality end-of-life care, to offer palliative care services and a personalized revision of the care plan, as well as access to services based on the changing needs of the patient and family (IOM, 2015). Community-based or home-based palliative care services have been associated with a reduced need for end-of-life emergency department visits, reduced length and frequency of hospitalization, and fewer ICU admissions and in-hospital deaths (NCCN, 2026). A systematic review found that patients who receive advance care planning or palliative care interventions have less ICU admissions and reduced ICU length of stay (Khandelwal et al., 2015). A recent systematic review looked at peer-reviewed observational/experimental advance care planning and cancer patient-specific studies published between 1990-2022 and found that across ~33,500 patients advance care planning was associated with significantly lower odds of chemotherapy, intensive care, hospital admissions, hospice use fewer than seven days, hospital death, and aggressive care composite measures (Levoy et al., 2023).
ASCO and NCCN palliative care guidelines contain the following recommendations:
- The oncology team should consider <palliative care> consultation for patients with limited anticancer treatment options due to lack of access to anticancer therapy; advanced disease process; multiple/severe comorbid conditions; rapidly progressive functional decline; and/or persistently poor performance status. Additional criteria include…frequent emergency department visits or hospital admissions; need for ICU-level care (Category 2A) (NCCN, 2026).
- In general, patients with weeks to days to live (eg, dying patients) and comfort-oriented goals should discontinue all treatments not directly contributing to patient comfort. Intensive palliative care focusing on symptom management should be provided in addition to preparation for the dying process. Referral for hospice care should be placed, if not already done (Category 2A) (NCCN, 2026).
- Clinicians should assess and cultivate prognostic awareness and engage in advance care planning with patients and their families to ensure patient-centered care plans (Category 2A) (NCCN, 2026).
- Clinicians should refer patients with advanced solid tumors and hematologic malignancies to specialized interdisciplinary palliative care teams that provide inpatient and outpatient care early in the course of disease, alongside active treatment of their cancer (Moderate, Strong) (Sanders et al., 2024).
Definitions of Categories of Evidence and Ratings:
- Category 2A: Based upon lower-level evidence, there is uniform NCCN consensus (≥85% support of the Panel) that the intervention is appropriate. Note there are no Category 1 recommendations within NCCN’s guidelines on Palliative Care.
- Strong Strength of Recommendation: In recommendations for an intervention, the desirable effects of an intervention outweigh its undesirable effects. In recommendations against an intervention, the undesirable effects of an intervention outweigh its desirable effects. All or almost all informed people would make the recommended choice for or against an intervention
- Moderate Quality of Evidence: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different
In addition to the literature, ASCO received positive feedback on measure importance via its May 2025 public comment; almost all respondents strongly agreed or agreed with the following statements across all of ASCO’s updated EOL measures: “I believe this measure captures what it intends to capture, i.e., promoting early palliative care among dying patients and reducing aggressive interventions at the end of patients’ lives” and “I believe this measure differentiates good from poor quality care among providers of healthcare services.”
References:
- Adelson, K. B., Canavan, M., Niu, J., Zhao, H., Nortje, N., Xiang, J. J., Giordano, S. H., & Cheng, L. (2024). Systemic anti-cancer treatment and healthcare utilization at end of life: A SEER Medicare analysis. JCO Oncology Practice, 20(10_suppl), 276. https://doi.org/10.1200/OP.2024.20.10_suppl.276
- Christian, T. J., Hassol, A., Brooks, G. A., Gu, Q., Kim, S., Landrum, M. B., & Keating, N. L. (2021). How Do Claims-Based Measures of End-of-Life Care Compare to Family Ratings of Care Quality?. Journal of the American Geriatrics Society, 69(4), 900–907. https://doi.org/10.1111/jgs.16905
- Ersek, M., Miller, S. C., Wagner, T. H., Thorpe, J. M., Smith, D., Levy, C. R., Gidwani, R., Faricy-Anderson, K., Lorenz, K. A., Kinosian, B., & Mor, V. (2017). Association between aggressive care and bereaved families' evaluation of end-of-life care for veterans with non-small cell lung cancer who died in Veterans Affairs facilities. Cancer, 123(16), 3186–3194. https://doi.org/10.1002/cncr.30700
- Garg, V., Ruiz Buenrostro, A., Heuniken, K., Bagnarol, R., Yousef, M., Sajewicz, K., Dhanju, S., Wentlandt, K., Kuruvilla, J., Lheureux, S., Zimmermann, C., & Hannon, B. (2024). Novel Systemic Anticancer Therapy and Healthcare Utilization at the End of Life: A Retrospective Cohort Study. Cancer medicine, 13(23), e70450. https://doi.org/10.1002/cam4.70450
- Gomes, B., Calanzani, N., Gysels, M., Hall, S., & Higginson, I. J. (2013). Heterogeneity and changes in preferences for dying at home: a systematic review. BMC palliative care, 12, 7. https://doi.org/10.1186/1472-684X-12-7
- Institute of Medicine. (2015). Dying in America: Improving Quality and Honoring Individual Preferences Near The End of Life. National Academies Press. https://doi.org/10.17226/18748
- Khandelwal, N., Kross, E. K., Engelberg, R. A., Coe, N. B., Long, A. C., & Curtis, J. R. (2015). Estimating the effect of palliative care interventions and advance care planning on ICU utilization: A systematic review. Critical Care Medicine, 43(5), 1102–1111. https://doi.org/10.1097/CCM.0000000000000852
- Levoy, K., Sullivan, S. S., Chittams, J., Myers, R. L., Hickman, S. E., & Meghani, S. H. (2023). Don't throw the baby out with the bathwater: Meta-analysis of advance care planning and end-of-life cancer care. Journal of Pain and Symptom Management, 65(6), e715–e743. https://doi.org/10.1016/j.jpainsymman.2023.02.003
- National Comprehensive Cancer Network. (2026). NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines®): Palliative Care (Version 1.2026). https://www.nccn.org/professionals/physician_gls/pdf/palliative.pdf
- Sanders, J. J., Temin, S., Ghoshal, A., Alesi, E. R., Ali, Z. V., Chauhan, C., Cleary, J. F., Epstein, A. S., Firn, J. I., Jones, J. A., Litzow, M. R., Lundquist, D. M., Mardones, M. A., Nipp, R. D., Rabow, M. W., Rosa, W. E., Zimmermann, C., & Ferrell, B. R. (2024). Palliative Care for Patients with Cancer: ASCO guideline Update. Journal of Clinical Oncology, 42(19), 2336–2357. https://doi.org/10.1200/JCO.24.00542
- Siegel, R. L., Kratzer, T. B., Wagle, N. S., Sung, H., & Jemal, A. (2026). Cancer statistics, 2026. CA: A Cancer Journal for Clinicians, 76(1), Article e70043. https://doi.org/10.3322/caac.70043
- Wright, A. A., Keating, N. L., Ayanian, J. Z., Chrischilles, E. A., Kahn, K. L., Ritchie, C. S., Weeks, J. C., Earle, C. C., & Landrum, M. B. (2016). Family Perspectives on Aggressive Cancer Care Near the End of Life. JAMA, 315(3), 284–292. https://doi.org/10.1001/jama.2015.18604
- Zhu, E., McCreedy, E., & Teno, J. M. (2024). Bereaved Respondent Perceptions of Quality of Care by Inpatient Palliative Care Utilization in the Last Month of Life. Journal of general internal medicine, 39(6), 893–901. https://doi.org/10.1007/s11606-023-08588-4
Measure Impact
To reiterate, the goal of Percentage of Patients who Died with Cancer Admitted to the Intensive Care Unit in the Last 30 Days of Life (lower score – better) is to highlight performance trends over time and encourage timely enrollment in palliative care among people dying with cancer. This results in a reduction of aggressive interventions leading to ICU visits, ED visits, and hospitalizations, improved symptom control and quality of life, and ultimately improved patient, family, and caregiver satisfaction. There is a plethora of evidence from the literature that early palliative care also reduces costs at the end of life for the patient and healthcare system at large. Community-based or home-based palliative care services have been associated with a reduced need for end-of-life emergency department visits, reduced length and frequency of hospitalization, and fewer ICU admissions and in-hospital deaths (NCCN, 2026). Davis et al. analyzed health insurance data to determine the impact of palliative care on “aggressive end of life” and found that palliative care <90 days before death was associated with increased costs while palliative care consults >90 days before death lowered cost (P < .0001); completed advanced directives reduced cost by ~$4000 per patient (2023). Cheung et al. evaluated a cohort of patients who died of cancer between 2005-2009 and found that patients who received “aggressive end of life care” incurred 43 percent higher costs than those managed non-aggressively and that early and repeated palliative care consults were associated with reduced mean per-patient costs (2015). Starr et al. conducted a systematic review to determine the impacts of advance care planning and goals-of-care discussions on healthcare utilization, costs, and place of death; researchers found that EOL discussions are associated with lower healthcare costs in the last 30 days of life (median $1,048 vs. $23,482; p < .001); lower likelihood of acute care at EOL [Odds Ratios (OR) ranging 0.43 to 0.69]; lower likelihood of intensive care at EOL (ORs ranging 0.26 to 0.68); lower odds of chemotherapy near death (ORs 0.41, 0.57); lower odds of emergency department use and shorter length of hospital stay; greater use of hospice (ORs ranging 1.79 to 6.88); and greater likelihood of death outside the hospital (2020). The American Cancer Society summarized several key studies and review articles that examine the impact of palliative care on overall patients costs and found that palliative care either reduces overall costs to the patient or is cost neutral, while improving the patient’s quality of life (2022).
References:
- American Cancer Society Cancer Action Network. (2022, November 22). Palliative Care: Key studies on cost savings. https://www.fightcancer.org/sites/default/files/palliative_care_effects_on_costs_11-18-22_update.pdf
- Cheung, M. C., Earle, C. C., Rangrej, J., Ho, T. H., Liu, N., Barbera, L., Saskin, R., Porter, J., Seung, S. J., & Mittmann, N. (2015). Impact of aggressive management and palliative care on cancer costs in the final month of life. Cancer, 121(18), 3307–3315. https://doi.org/10.1002/cncr.29485
- Davis, M. P., Vanenkevort, E. A., Elder, A., Young, A., Correa Ordonez, I. D., Wojtowicz, M. J., Ellison, H., Fernandez, C., Mehta, Z., Behm, B., Digwood, G., & Panikkar, R. (2023). The Financial Impact of Palliative Care and Aggressive Cancer Care on End-of-Life Health Care Costs. The American journal of hospice & palliative care, 40(1), 52–60. https://doi.org/10.1177/10499091221098062
- Medicare Payment Advisory Commission. (2025, March). Report to the Congress: Medicare payment policy. https://www.medpac.gov/wp-content/uploads/2025/03/Mar25_MedPAC_Report_To_Congress_SEC.pdf
- National Comprehensive Cancer Network. (2026). NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines®): Palliative Care (Version 1.2026). https://www.nccn.org/professionals/physician_gls/pdf/palliative.pdf
- Starr, L. T., Ulrich, C. M., Corey, K. L., & Meghani, S. H. (2019). Associations Among End-of-Life Discussions, Health-Care Utilization, and Costs in Persons With Advanced Cancer: A Systematic Review. The American journal of hospice & palliative care, 36(10), 913–926. https://doi.org/10.1177/1049909119848148
There are a few existing quality measures that relate to planning for end of life and palliative care; however, these are either not specific to cancer patients or are specified only for a specific EMR environment (in the case of the QCDR measure). Below are the existing measures:
CBE # | Measure Name | Description |
|---|---|---|
| 3665 | Ambulatory Palliative Care Patients Experience of Feeling Heard and Understood | The percentage of top-box responses among patients aged 18 years and older who had an ambulatory palliative care visit and report feeling heard and understood by their palliative care clinician and team within 2 months (60 days) of the ambulatory palliative care visit. |
| 326 | Advance Care Plan | Percentage of patients aged 65 years and older who have an advance care plan or surrogate decision maker documented in the medical record or documentation in the medical record that an advance care plan was discussed but the patient did not wish or was not able to name a surrogate decision maker or provide an advance care plan. |
| N/A | ALS Patient Care Preferences | Percentage of patients diagnosed with Amyotrophic Lateral Sclerosis (ALS) who were offered assistance in planning for end of life issues (e.g., advance directives, invasive ventilation, lawful physician-hastened death, or hospice) or whose existing end of life plan was reviewed or updated at least once annually or more frequently as clinically indicated (i.e., rapid progression). |
| N/A | PIMSH1: Oncology: Advance Care Planning in Metastatic Cancer Patients | Percentage of patients with metastatic (stage 4) cancer who have a documented Advance Care Planning discussion in the first 6 months after metastatic diagnosis to inform treatment decisions and end-of-life care. |
Note that there is the existence of OP-35: Admissions and Emergency Department (ED) Visits for Patients Receiving Outpatient Chemotherapy, however this is specified and tested at the outpatient hospital level, rather than clinician level.
Taken together, ASCO’s suite of end-of-life measures provides a more comprehensive picture of the quality of end-of-life care among patients with cancer who are dying. Lastly, ASCO has developed claims-based versions of its EOL measures to assist with more consistent and reliable case identification, with no added administrative burden of data collection for the measure implementer.
ASCO’s end-of-life quality measures were originally developed in 2003 using a patient-centered methodology to capture outcomes meaningful to those with advanced illness (Earle et al., 2003). This included:
- Focus groups consisting of patients with incurable cancer and family members of deceased patients. These participants identified and vetted potential EOL quality measures to ensure they reflected patient-centered priorities.
- Expert Consensus: A multidisciplinary expert panel applied a modified Delphi approach to rank the importance and meaningfulness of potential measures based on the focus group input. Measures that did not resonate with patient and family values (such as those focused solely on economic efficiency) were excluded.
- Literature searches.
ASCO has continued to integrate the patient and caregiver voice into the current versions of these measures:
- Expert Panel Participation: A family caregiver representative served as a formal member of the 2023–2024 ASCO EOL Expert Panel, providing direct input during the review and updating of the measures.
- May 2025 Public Comment Period: Following the updates, ASCO held a public comment period to ensure the measures remained valuable to stakeholders. Of the respondents, which included a patient representative, a significant majority agreed that the measures effectively differentiate between high- and low-quality care and assess what they intend to assess (i.e., quality of end-of-life care).
References:
Earle, C. C., Park, E. R., Lai, B., Weeks, J. C., Ayanian, J. Z., & Block, S. (2003). Identifying potential indicators of the quality of end-of-life cancer care from administrative data. Journal of Clinical Oncology: Official Journal of the American Society of Clinical Oncology, 21(6), 1133–1138. https://doi.org/10.1200/JCO.2003.03.059
Performance Gap
ASCO evaluated performance on the "Percentage of Patients who Died with Cancer Admitted to the Intensive Care Unit (ICU) in the Last 30 Days of Life" measure. Using combined data from January 2023 to December 2024, the study included 563 practices that met the minimum requirement of five eligible patients (refer to Sections 1.26 and 5.1.1 for details on sample size and testing).
Because this measure tracks high-intensity care near the end of life, a lower percentage reflects stronger performance in aligning clinical interventions with patient goals. Results spanned the full spectrum from 0% to 100%, illustrating extreme variance in how terminal patients are managed across different care settings. The mean performance on the measure was 37% (±19%) with a 95% confidence level of ±2%; the median was slightly higher at 38%.
The distribution is slightly positively skewed at 0.22, suggesting a relatively balanced spread across the cohort with a minor lean toward higher-than-average utilization. Notably, the mode was 40%, indicating that for many individual entities, the most frequent outcome was that four out of ten terminal patients were admitted to the ICU in their final month. These results show that while the highest-performing entities achieved 0%, the lowest-performing entities (100%) had a rate that is 163% (2.63 times) higher than the median. For practices performing above the median, these findings highlight a critical opportunity to strengthen palliative care integration and discuss end-of-life preferences earlier in the disease trajectory.
Mean Performance Score by Decile (Percentage of Patients who Died with Cancer Admitted to the Intensive Care Unit (ICU) in the Last 30 Days of Life)
Overall | Min | Decile 1 | Decile 2 | Decile 3 | Decile 4 | Decile 5 | Decile 6 | Decile 7 | Decile 8 | Decile 9 | Decile 10 | Max | |
| Mean Performance Score | 37% | 0% | 7% | 16% | 22% | 30% | 36% | 40% | 44% | 50% | 58% | 72% | 100% |
| Number of Entities | 563 | 1 | 57 | 56 | 56 | 57 | 56 | 56 | 57 | 56 | 56 | 56 | 1 |
| Number of Persons / Encounters / Episodes | 15,685 | 5 | 1,168 | 1,962 | 1,479 | 1,878 | 1,833 | 1,925 | 2,572 | 1,192 | 975 | 701 | 5 |
Care Gaps
Closing Care Gaps
A recent systematic review and meta-analysis looked at aggressive EOL cancer care among ~2.7 million patients across 129 studies and found that aggressive EOL care is a common global practice (Ma et al., 2024):
- Repeated hospital admissions (>1) in the last 30 days of life: 17.9%.
- Repeated emergency room visits (>1) in the last 30 days of life: 14.8%.
- Intensive care unit (ICU) stays in the last 30 days of life: 14.4%.
- Hospice enrollment less than 3 days before death: 14.4%.
- Chemotherapy in the last 14 days of life: 11.6%.
Additionally, of the studies using a composite score, more than half of the patients experienced at least one measure of aggressive care at the end of their lives. The research also showed that patients with hematologic malignancies were significantly more likely to receive aggressive care, including higher rates of late hospice enrollment, ICU stays, and chemotherapy in the last weeks of life, compared to those with solid tumors (Ma et al., 2024).
There is evidence of gaps in care among lower-level studies as well. A retrospective review evaluated 2,844 patients in a multistate, community-based hospital network with stage IV NSCLC and found that only 8 percent of patients were referred for outpatient palliative care (Meggyesy et al., 2022).
References:
- Ma, Z., Li, H., Zhang, Y., Zhang, L., Huang, G., Zhang, Y., Shi, L., Liu, W., An, Z., & Guan, X. (2024). Prevalence of aggressive care among patients with cancer near the end of life: a systematic review and meta-analysis. EClinicalMedicine, 71, 102561. https://doi.org/10.1016/j.eclinm.2024.102561
- Meggyesy, A. M., Buehler, K. E., Wilshire, C. L., Chiu, S. T., Chang, S.-C., Rayburn, J. R., Gilbert, C. R., & Gorden, J. A. (2022). Utilization of palliative care resource remains low, consuming potentially avoidable hospital admissions in stage IV non-small cell lung cancer: A community-based retrospective review. Supportive Care in Cancer, 30(12), 10117–10126. https://doi.org/10.1007/s00520-022-07364-0
Feasibility
Feasibility
Data Generation and Availability
As this is a claims-based measure, the data elements required are routinely generated during the delivery of care as part of the standard billing and reimbursement cycle. These data are 100% available in electronic sources via HIPAA-standard electronic data interchange (EDI) transactions.
Data Structure
All required data elements are housed in structured fields within the claims database. The measure does not rely on unstructured data or free-text clinical notes, ensuring high consistency and ease of extraction.
Inaccuracies and Missing Data
As with all administrative data, this measure is primarily susceptible to inaccuracies related to "claims lag" (the delay between service delivery and claim finalization) and potential coding omissions where secondary diagnoses may not be captured if they do not impact reimbursement. Additionally, while rare in structured claims, "missing data" may occur if a provider fails to populate non-mandatory fields.
Data Integrity and Auditability
Data integrity is maintained through the payer’s internal validation and adjudication engines. This process is highly rigorous, involving extensive financial and clinical reconciliation to ensure fiscal accuracy and adherence to billing regulations. Furthermore, the data is fully auditable; each record is linked to a unique Claim Control Number (CCN) and provider National Provider Identifier (NPI), allowing for a direct reconciliation path back to the original clinical source of truth should a discrepancy be detected.
Annual Specification Maintenance and Data Mapping
For ongoing maintenance, changes to measure specifications (such as annual ICD-10 or CPT code updates) are managed through a standard yearly mapping review process. These updates impact the data structure by requiring new code strings to be added to the measure logic, but they do not affect the overall availability of the electronic data.
Costs and Administrative Burden
As a claims-based measure, there is negligible administrative burden and no direct implementation cost for the measured entities. Data collection is "passive," utilizing administrative claims that are already generated as part of the standard billing and reimbursement cycle. No manual data abstraction, registry reporting, or additional data entry is required.
Impact on Clinical Workflow and Interaction
This measure has no impact on clinician workflow, diagnostic thought processes, or the patient-physician interaction. Because the data is captured retrospectively through existing ICD-10 and CPT codes, clinicians do not need to modify their documentation habits or navigate additional EHR alerts. This ensures that the clinician's focus remains entirely on clinical decision-making rather than the reporting process.
Barriers and Stakeholder Feedback
ASCO continues to monitor feedback from stakeholders; to date, there have been no concerns regarding implementation burden or patient confidentiality, as the measure relies on de-identified administrative data. Potential barriers are limited to the inherent limitations of claims data, such as claims lag or coding variability. However, because the measure specifications rely on standardized, mandated code sets, these barriers are mitigated by the existing high-compliance environment of healthcare billing.
Data collection for this measure is conducted in strict accordance with HIPAA Privacy and Security Rules. Confidentiality is maintained because the measure utilizes de-identified administrative claims data sourced from private payers. Direct identifiers are removed or masked prior to the data being made available for measure calculation, ensuring that the analysis remains focused on clinical patterns rather than individual identities.
To mitigate the risk of re-identification in small patient populations (the "Small N" problem), a minimum threshold of five (5) patients is suggested for performance reporting. This recommended suppression guideline is intended to prevent "deductive disclosure," where an individual's identity could potentially be inferred from a very small data set or outlier results. By suggesting this minimum volume, the measure balances the need for transparent reporting with the highest standards of patient privacy.
Finally, confidentiality risks associated with patient surveys are not applicable to this measure, as it relies entirely on administrative claims data with no direct patient interaction or survey-based data collection.
Based on the feasibility results, no major structural changes were necessary for the core data elements. As the data requisite for measure calculation are routinely generated during the delivery of care, feasibility considerations were effectively validated during the formulation of the measure specifications. The final specifications focus on the use of high-fidelity, structured billing codes to ensure the measure remains reliable and easily implementable across all participating federal and private payers.
Proprietary Information
As the world’s leading professional organization for physicians and others engaged in clinical cancer research and cancer patient care, American Society of Clinical Oncology, Inc. (“Society”) and its affiliates1 publishes and presents a wide range of oncologist‐approved cancer information, educational and practice tools, and other content. The ASCO trademarks, including without limitation ASCO®, American Society of Clinical Oncology®, JCO®, Journal of Clinical Oncology®, Cancer.Net™, QOPI®, QOPI Certification Program™, and Conquer Cancer®, are among the most highly respected trademarks in the fields of cancer research, oncology education, patient information, and quality care. This outstanding reputation is due in large part to the contributions of ASCO members and volunteers. Any goodwill or commercial benefit from the use of ASCO content and trademarks will therefore accrue to the Society and its respective affiliates and further their tax‐exempt charitable missions. Any use of ASCO content and trademarks that may depreciate their reputation and value will be prohibited.
ASCO does not charge a licensing fee to not-for-profit hospitals, healthcare systems, or practices to use the measure for quality improvement, research or reporting to federal programs. ASCO encourage all of these not-for-profit users to obtain a measure library license so ASCO can:
- Keep users informed about measure updates and/or changes
- Learn from measure users about any implementation challenges to inform future measure updates and/or changes
- Track measure utilization (outside of federal reporting programs) and performance rates
ASCO has adopted the Council of Medical Specialty Society’s Code for Interactions with Companies (https://cmss.org/wp-content/uploads/2026/04/CMSS-Code-for-Interactions-…), which provides guidance on interactions with for‐profit entities that develop produce, market or distribute drugs, devices, services or therapies used to diagnose, treat, monitor, manage, and alleviate health conditions. The Society’s Board of Directors has set Licensing Standards of American Society of Clinical Oncology ( https://cdn.bfldr.com/KOIHB2Q3/as/bsrth8mwgbsrpvrsbt6gxqb/2023-ASCO-Lic…) to guide all licensing arrangements.
In addition, ASCO has adopted the Council of Medical Specialty Society’s Policy on Antitrust Compliance (https://cmss.org/statements/cmss-policy-on-antitrust-compliance/), which provided guidance on compliance with all laws applicable to its programs and activities, specifically including federal and state antitrust laws, including guidance to not discuss, communicate, or make announcements about fixing prices, allocating customers or markets, or unreasonably restraining trade.
Scientific Acceptability
Testing Data
Testing for this measure was conducted using administrative claims data from two primary, high-volume sources: a major national federation of independent commercial health insurers and The US Oncology Network (USON)/McKesson claims databases. The data utilized provided comprehensive national geographic coverage across urban, suburban, and rural regions. The initial data sets identified 1,327 reporting entities from the national health insurance federation and 9 large-scale practices within the USON/McKesson network.
The testing period for both data sources encompassed the timeframe from January 1, 2023, to December 31, 2024. This two-year window ensures the measure was validated against the most current coding standards and clinical practice patterns, providing a stable and contemporary baseline for analysis.
There are specific differences in the sample sizes used for various aspects of testing, driven by the clinical sensitivity of the measure and the statistical requirements of the analysis:
Performance Gap and Validity Testing
Threshold: Entities with a minimum of five (5) patients meeting the measure denominator.
Sample Size: 563 reporting entities (derived from the national insurance federation and USON/McKesson administrative claims data, Jan 1, 2023 – Dec 31, 2024).
Rationale: This is an end-of-life admittance to the ICU measure. Because admittance to the ICU is a critical quality indicator in oncology, ASCO determined that "each patient matters" in the assessment of care delivery. A lower threshold of N ≥ 5 was utilized to prioritize “Visibility over Volatility.”
- Health Equity: Utilizing a broader inclusion criteria prevents quality "blind spots" in community-based or rural oncology settings where EOL events are significant but may occur less frequently. High thresholds would effectively penalize these practices by making their care invisible to the measure.
- Clinical Significance: While smaller denominators inherently have a higher standard error, the clinical priority of preventing admittance to the ICU outweighs the statistical risk of "noisy" scores at the individual entity level.
Reliability Testing
Threshold: Entities with a minimum of twelve (12) patients meeting the measure denominator.
Sample Size: 271 reporting entities (a subset of the longitudinal claims data described above).
Rationale: A higher threshold of N ≥ 12 was applied specifically for reliability testing to maintain psychometric rigor.
- Signal-to-Noise Ratio: For a measure’s reliability coefficient (Beta) to be stable, there must be enough patient volume to distinguish true clinical variation from random statistical noise.
- Instrument Validation: Using these 271 higher-volume entities ensures that the reliability of the "measurement instrument" itself is validated on a stable data set before being applied to the broader clinical population.
Exclusions and Risk Adjustment
No other differences in data sources or timeframes were utilized for exclusions or risk-adjustment testing.
For performance gap and validity testing, the sample included 554 entities from a national federation of independent commercial insurers and nine USON/McKesson practices. For reliability testing, this cohort was refined to 262 national federation entities and nine USON practices. The resulting testing group represents a national cross-section of oncology care across all 50 U.S. states, including independent practices, hospital-affiliated groups, and integrated delivery networks. By applying an N ≥ 5 threshold for the performance gap analysis, the study successfully captured diverse clinical settings - ranging from high-volume academic hubs to rural providers where hospice access may be limited.
Selection was based on the availability of structured claims data, ensuring a sample that reflects the national oncology landscape without regional payer bias. The inclusion of USON practices highlights the experience of community oncologists, while the variety of practice sizes supports the "Each Patient Matters" philosophy, ensuring an equitable assessment of hospice utilization across all provider tiers.
Data Source and Sampling
The descriptive statistics for this measure were derived from The US Oncology Network (USON)/McKesson database for the period of January 1, 2023, to December 31, 2024. The sample includes 2,655 unique patients who met the denominator criteria. No sampling was used; 100% of the nine (9) high-volume USON practices met the minimum threshold of N ≥ 5, ensuring the data reflects the entire eligible population from this clinical source.
Representativeness
As a national network of independent community-based practices, the USON cohort serves as a robust proxy for the broader oncology population. This demographic distribution provides high-fidelity insight into intensive care utilization patterns at the end of life, mirroring the diversity found in national administrative claims data.
Descriptive Statistics: Patient Population (N = 2,655)
| Race | Number (n) | Percentage (%) |
| White | 2,371 | 89.3% |
| Hispanic | 102 | 3.8% |
| Black | 69 | 2.6% |
| Other or Unknown | 56 | 2.1% |
| Asian / Pacific Islander | 41 | 1.5% |
| American Indian / Alaska Native | 16 | 0.6% |
| Grand Total | 2,655 | 100% |
| Gender | Number (n) | Percentage (%) |
| Female | 1,327 | 50.0% |
| Male | 1,323 | 49.8% |
| Other or Unknown | 5 | 0.2% |
| Grand Total | 2,655 | 100% |
Reliability
Person or Encounter Level (Data Element) Testing
End-of-life (EOL) care is a foundational quality metric in oncology, and administrative claims data serve as the primary vehicle for its assessment. Admission to an Intensive Care Unit (ICU) represents a "hard" event in administrative records. Unlike nuanced clinical symptoms, an ICU stay triggers specific facility billing requirements, most notably Revenue Center Codes (typically 020x series), that are required for the facility to receive the higher reimbursement rate associated with critical care. Because these codes are the primary trigger for payment, they are rarely missing or erroneously coded.
Consequently, ASCO utilized foundational evidence to establish the reliability of the numerator, denominator, and numerator exclusions:
- Denominator (Identification of Cancer Decedents): Research confirms that administrative claims are highly reliable for identifying the patient population. Studies using diagnostic codes to identify cancer patients have shown a Positive Predictive Value (PPV) of up to 99.68% (Shin et al., 2019).
- Numerator (ICU Admission in the Last 30 Days): Earle et al. (2005) evaluated the accuracy- defined as percent agreement within +/- 1 day - specifically for the indicator of ICU admission in the last month of life. By comparing Medicare claims from 48,906 cancer decedents against a clinical gold standard of 150 medical records, the accuracy for this numerator was calculated as 0.95.
References
- Earle, C. C., Neville, B. A., Landrum, M. B., Souza, J. M., Weeks, J. C., Block, S. D., Grunfeld, E., & Ayanian, J. Z. (2005). Evaluating claims-based indicators of the intensity of end-of-life cancer care. International Journal for Quality in Health Care, 17(6), 505–509. https://doi.org/10.1093/intqhc/mzi061
- Shin, D. W., Cho, J. H., Kim, S. Y., Guallar, E., & Cho, J. (2019). Validation of Administrative Big Database for Colorectal Cancer Searched by International Classification of Disease 10th Codes. Journal of Cancer, 10(15), 3381–3387. https://doi.org/10.7150/jca.30454
Accountable Entity Level (Measure Score) Testing
An assessment of the measure's reliability was performed through the utilization of signal-to-noise analysis, a method that determines the precision of the actual construct in comparison to the random variation. The signal-to-noise ratio is determined by calculating the ratio of between-unit variance to total variance. This analysis provides valuable insight into the measure's reliability and its ability to produce consistent results by describing how well one can confidently distinguish the performance of one clinician group from another.
Based on the hierarchical modeling approach for provider profiling, the following steps were taken:
- Data Aggregation: Patient-level data were captured as binary (pass/fail) events and aggregated to the clinician group level to determine the numerator and denominator for each practice.
- Model Selection: We utilized a Beta-Binomial model, which is the natural fit for estimating the reliability of simple pass/fail rate measures.
- Variance Partitioning: The model partitioned the total observed variability in practice scores into two components: between-unit variance (the "signal," or true differences in practice quality) and within-unit variance (the "noise," or random sampling error).
- Reliability Calculation: For each clinician group, a reliability coefficient (R) was calculated using the ratio of the estimated provider-to-provider variance to the sum of the provider-to-provider variance and the binomial error variance (p(1-p)/n).
- Threshold Application: The analysis focused on identifying the stability of these scores for practices meeting a minimum patient count of 12.
Reliability was assessed at the clinician group level using a beta-binomial signal-to-noise analysis to partition total observed variation into true performance differences and random sampling error. Based on a comprehensive sample of 271 clinician groups and 13,562 patient encounters, the analysis yielded a system-wide mean reliability coefficient of 0.658. All practices in the sample were included in the calculation, with individual group reliability scores ranging from a minimum of 0.452 to a maximum of 0.976. While lower-volume entities in the first four deciles averaged below 0.60, the mean reliability crossed the target threshold starting at Decile 5 (0.603) and reached 0.927 in the tenth decile, confirming that the measure successfully achieves the >0.60 benchmark as patient volume increases.
Person or Encounter Level (Data Element) Testing
An accuracy value of 0.95 for the numerator and a PPV of nearly 1.00 for the denominator indicates an exceptionally high level of agreement between administrative datasets and clinical reality. A 95% agreement rate confirms that the risk of misclassifying a patient's ICU status is minimal. For the purposes of national endorsement, this level of precision proves that administrative claims are a highly reliable instrument for monitoring ICU utilization at the end of life, ensuring that provider performance is assessed on a transparent, verifiable, and scientifically sound basis.
References
- Earle, C. C., Neville, B. A., Landrum, M. B., Souza, J. M., Weeks, J. C., Block, S. D., Grunfeld, E., & Ayanian, J. Z. (2005). Evaluating claims-based indicators of the intensity of end-of-life cancer care. International Journal for Quality in Health Care, 17(6), 505–509. https://doi.org/10.1093/intqhc/mzi061
- Shin, D. W., Cho, J. H., Kim, S. Y., Guallar, E., & Cho, J. (2019). Validation of Administrative Big Database for Colorectal Cancer Searched by International Classification of Disease 10th Codes. Journal of Cancer, 10(15), 3381–3387. https://doi.org/10.7150/jca.30454
Accountable Entity Level (Measure Score) Testing
The results support a strong inference of reliability, as the overall mean coefficient of 0.658 indicates that real performance variation is the primary driver of observed scores rather than random chance. With a mean performance score of 37.2%, the measure avoids "topping out" effects, which preserves the necessary provider-to-provider variation to distinguish between practice styles. The stability of performance scores across reliability deciles - narrowly ranging from 31.5% to 43.2% - demonstrates that the high reliability in the upper deciles is a function of increased patient denominators and reduced sampling error rather than biased outliers. While the minimum reliability was 0.452, the measure's progression to a maximum of 0.976 confirms it provides a precise and repeatable performance signal, especially for entities meeting the 12-patient minimum where the signal effectively distinguishes one group from another.
Percentage of Patients who Died with Cancer Admitted to the Intensive Care Unit (ICU) in the Last 30 Days of Life
Overall | Min | Decile 1 | Decile 2 | Decile 3 | Decile 4 | Decile 5 | Decile 6 | Decile 7 | Decile 8 | Decile 9 | Decile 10 | Max | |
| Reliability | 0.658 | 0.452 | 0.46 | 0.536 | 0.564 | 0.558 | 0.603 | 0.646 | 0.708 | 0.762 | 0.824 | 0.927 | 0.976 |
| Mean Performance Score | 37.2% | 40.7% | 38.4% | 38.9% | 34.3% | 37.2% | 35.2% | 40.0% | 38.3% | 37.3% | 41.2% | 31.5% | 43.2% |
| Number of Entities | 271 | 17 | 28 | 27 | 27 | 27 | 27 | 27 | 27 | 27 | 27 | 27 | 1 |
| Number of Persons / Encounters / Episodes | 13,562 | 204 | 347 | 392 | 439 | 500 | 605 | 774 | 990 | 1,334 | 2,047 | 6,134 | 710 |
Percentage of Patients who Died with Cancer Admitted to the Intensive Care Unit (ICU) in the Last 30 Days of Life
Overall | Min | Decile 1 | Decile 2 | Decile 3 | Decile 4 | Decile 5 | Decile 6 | Decile 7 | Decile 8 | Decile 9 | Decile 10 | Max | |
| Reliability | 0.658 | 0.407 | 0.438 | 0.492 | 0.534 | 0.581 | 0.626 | 0.665 | 0.721 | 0.774 | 0.832 | 0.927 | 0.984 |
Validity
Person or Encounter Level (Data Element) Testing
End-of-life (EOL) care is a foundational quality metric in oncology, and administrative claims data serve as the primary vehicle for its assessment. Admission to an Intensive Care Unit (ICU) represents a "hard" event in administrative records. Unlike nuanced clinical symptoms, an ICU stay triggers specific facility billing requirements - most notably Revenue Center Codes (typically 020x series) - that are required for the facility to receive the higher reimbursement rate associated with critical care. Because these codes are the primary trigger for payment, they are rarely missing or erroneously coded.
Consequently, ASCO utilized foundational evidence to establish the validity and scientific acceptability of the numerator and denominator:
- Denominator (Identification of Cancer Decedents): Research confirms that administrative claims are highly valid for identifying the intended patient population. Validation studies using ICD-10 codes to identify cancer patients have demonstrated a sensitivity of 100% and a specificity of 98.86% (Shin et al., 2019). This ensures the measure accurately targets patients with a confirmed cancer diagnosis while minimizing the risk of including non-cancer decedents.
- Numerator (ICU Admission in the Last 30 Days): Earle et al. (2005) evaluated the performance of claims-based indicators against the clinical gold standard of medical record review. For the "ICU admission in the last month of life" measure, the administrative claims data demonstrated a sensitivity of 0.87 and a specificity of 0.97.
References
- Earle, C. C., Neville, B. A., Landrum, M. B., Souza, J. M., Weeks, J. C., Block, S. D., Grunfeld, E., & Ayanian, J. Z. (2005). Evaluating claims-based indicators of the intensity of end-of-life cancer care. International Journal for Quality in Health Care, 17(6), 505–509. https://doi.org/10.1093/intqhc/mzi061
- Shin, D. W., Cho, J. H., Kim, S. Y., Guallar, E., & Cho, J. (2019). Validation of Administrative Big Database for Colorectal Cancer Searched by International Classification of Disease 10th Codes. Journal of Cancer, 10(15), 3381–3387. https://doi.org/10.7150/jca.30454
Accountable Entity Level (Measure Score) Testing
To evaluate the validity of the measure set at the accountable entity level, we conducted a convergent and divergent validity analysis using Pearson’s product-moment correlation coefficient (r).
Steps Conducted:
- Data Aggregation: Performance rates were calculated for each measure at the accountable entity level.
- Hypothesis Formulation: We hypothesized that measures reflecting high-intensity care (Systemic Cancer-Directed Therapy [SCDT], ICU Admissions, ED/Obs Visits, and Multiple Hospitalizations) would show positive correlations with one another. Conversely, we hypothesized these measures would show a negative correlation with the Hospice Enrollment measure, as hospice utilization represents a transition toward comfort-oriented care.
- Correlation Calculation: Pearson’s r was calculated for all pairs of measures within the set to determine the linear relationship between performance rates.
- Significance Testing: Two-tailed p-values were calculated for each pair to determine the statistical significance of the associations.
Refer to the attached Validity-Results.zip file for details.
Person or Encounter Level (Data Element) Testing
A specificity of 0.97 for the numerator is exceptionally high, indicating that the claims data is highly reliable at identifying patients who did not have an ICU stay, effectively eliminating the risk of false-positive penalties for providers. A sensitivity of 0.87 confirms that the vast majority of ICU admissions recorded in clinical charts are successfully captured via Revenue Center Codes in the administrative record. These values prove that administrative claims are a scientifically sound and highly valid instrument for monitoring the use of intensive care at the end of life.
References
- Earle, C. C., Neville, B. A., Landrum, M. B., Souza, J. M., Weeks, J. C., Block, S. D., Grunfeld, E., & Ayanian, J. Z. (2005). Evaluating claims-based indicators of the intensity of end-of-life cancer care. International Journal for Quality in Health Care, 17(6), 505–509. https://doi.org/10.1093/intqhc/mzi061
- Shin, D. W., Cho, J. H., Kim, S. Y., Guallar, E., & Cho, J. (2019). Validation of Administrative Big Database for Colorectal Cancer Searched by International Classification of Disease 10th Codes. Journal of Cancer, 10(15), 3381–3387. https://doi.org/10.7150/jca.30454
Accountable Entity Level (Measure Score) Testing
- Hypothesized Relationships: The empirical results strongly support the conceptual framework of the measure set. All indicators of high-intensity end-of-life care demonstrated positive correlations with each other, confirming they are capturing related facets of aggressive medical utilization.
- Validation Rationale: Divergent validity was confirmed by the Hospice Enrollment measure, which exhibited a statistically significant negative correlation with every high-intensity care indicator. Notably, the strongest positive associations were found between the two treatment strata (SCDT 14d and 30d, r = 0.627) and between outpatient and inpatient acute transitions (ED/Obs and Greater than 1 Hospitalization, r = 0.489).
- Statistical Significance: Every correlation in the matrix reached statistical significance (p < 0.01), providing robust evidence that these measures move in the hypothesized directions at the entity level.
Risk Adjustment
The decision to maintain unadjusted performance scores for these end-of-life measures is rooted in the philosophy that quality palliative care and clinical stewardship represent universal standards that should not fluctuate based on patient complexity. Unlike outcomes heavily influenced by biological variance, metrics such as systemic therapy administration and ICU utilization reflect direct clinical decision-making and provider agency; therefore, risk adjustment could inadvertently "normalize" aggressive care by suggesting that medical complexity justifies a departure from palliative best practices. Furthermore, because these measures are calculated using a denominator of patients who have already deceased, the cohort is inherently characterized by high clinical risk, making additional adjustment statistically redundant and potentially misleading. By prioritizing unadjusted data, ASCO maintains a transparent view of the raw clinical reality, ensuring that gaps in service and health inequities remain visible rather than being masked by statistical smoothing. Ultimately, this approach upholds the principle that every patient, regardless of their diagnosis or comorbidities, deserves a timely transition to hospice and a coordinated, comfort-focused end-of-life experience.
Use & Usability
Use
As the measure steward, ASCO is committed to the broad implementation of this measure across the national quality landscape. We are in ongoing consultations with CMS regarding its inclusion in programs such as MIPS, PCHQR and IQR. Our roadmap includes finalizing the technical specifications required for federal uptake while simultaneously promoting the measure for use in private payer quality initiatives, VBPs and ASCO’s own quality improvement portfolio.
1. Target Populations
The measure is applicable to adult patients (aged 18 and older) with a confirmed diagnosis of cancer.
2. Accountable Entities
Accountability is attributed at the level of the Oncology Physician Group Practice (PGP) or individual Clinician Group/Practice.
- Attribution Logic: Patients are attributed to the entity that provides the plurality of oncology-related services or manages the "episode of care" (e.g., the 6-month period following the start of chemotherapy).
- Responsibility: The entity is held accountable for the patient’s clinical outcomes, resource utilization (e.g., avoidable ER visits), and adherence to evidence-based pathways.
3. Care Settings
The primary care setting is the Outpatient Oncology Clinic, including:
- Community-based oncology practices.
- Hospital Outpatient Departments (HOPDs).
- Infusion Centers.
Note on Care Settings: Clinical oncologists provide care within the outpatient setting; however, this measure set monitors related clinical outcomes across multiple sites of service. Evaluated events include, but are not limited to, inpatient/outpatient hospital infusions, ICU stays, and emergency department encounters stemming from complications of the outpatient treatment.
Usability
There is clear evidence that there are interventions that can be put in place to reduce ICU visits among dying patients, therefore improving the performance on this quality measure. Palliative care is specialized medical care for people living with a serious illness that is focused on providing relief from the symptoms and stress of the illness, and can be provided along with curative treatment (Center to Advance Palliative Care, n.d.). Palliative care reduces avoidable spending and utilization in all healthcare settings and improves the quality of life for the patients it serves (Center to Advance Palliative Care, n.d.). Both ASCO and NCCN guidelines recommend palliative care for the patients this measure addresses:
- The oncology team should consider <palliative care> consultation for patients with limited anticancer treatment options due to lack of access to anticancer therapy; advanced disease process; multiple/severe comorbid conditions; rapidly progressive functional decline; and/or persistently poor performance status. Additional criteria include…frequent emergency department visits or hospital admissions; need for ICU-level care (Category 2A) (NCCN, 2026).
- In general, patients with weeks to days to live (eg, dying patients) and comfort-oriented goals should discontinue all treatments not directly contributing to patient comfort. Intensive palliative care focusing on symptom management should be provided in addition to preparation for the dying process. Referral for hospice care should be placed, if not already done (Category 2A) (NCCN, 2026).
- Clinicians should assess and cultivate prognostic awareness and engage in advance care planning with patients and their families to ensure patient-centered care plans (Category 2A) (NCCN, 2026).
- Clinicians should refer patients with advanced solid tumors and hematologic malignancies to specialized interdisciplinary palliative care teams that provide inpatient and outpatient care early in the course of disease, alongside active treatment of their cancer (Moderate, Strong) (Sanders et al., 2024).
The below outlines the difficulty of the actions described above and how measured entities can overcome those difficulties:
| Action | Difficulty Level | Why it is Difficult | How to Overcome |
| Early Referral to palliative care | Moderate | Shortage of specialist palliative care clinicians and the stigma that palliative care means "giving up." | In addition to physicians, oncology nurses can be positioned to provide primary palliative care and provide increased advance care planning with patients with advanced cancer (NCCN, 2026). The <NCCN> Panel emphasizes the importance of initiating or continuing advance care planning conversations and systematically reviewing advance care plans to ensure ongoing accuracy as illness or situation evolves. To avoid demeaning the value of end-of-life care, palliative and/or hospice care should not be framed as “giving up” but instead refocusing the care plan to achieve a better quality of life (NCCN, 2026) |
| ACP Documentation | High | These conversations are time-intensive and clinicians often lack training in high-stakes communication. | Embed ACP templates in the EHR. Use a "primary care/oncology" shared model where social workers or nurses lead initial GOC discussions. |
References:
- Center to Advance Palliative Care. (n.d.). About Palliative Care. https://www.capc.org/about/palliative-care/
- National Comprehensive Cancer Network. (2026). NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines®): Palliative Care (Version 1.2026). https://www.nccn.org/guidelines/guidelines-detail?id=1454
- Sanders, J. J., Temin, S., Ghoshal, A., Alesi, E. R., Ali, Z. V., Chauhan, C., Cleary, J. F., Epstein, A. S., Firn, J. I., Jones, J. A., Litzow, M. R., Lundquist, D. M., Mardones, M. A., Nipp, R. D., Rabow, M. W., Rosa, W. E., Zimmermann, C., & Ferrell, B. R. (2024). Palliative Care for Patients with Cancer: ASCO guideline Update. Journal of Clinical Oncology, 42(19), 2336–2357. https://doi.org/10.1200/JCO.24.00542
There may be patients and caregivers who prefer ICU care at the end of life and/or do not have access to appropriate outpatient and palliative care. ASCO’s End of Life Measures Technical Expert Panel emphasized that performance is not expected to be perfect on this quality measure. A margin of error should be expected to account for such scenarios and benchmarks set within ASCO Certified will account for such margins.
Comments
Staff Preliminary Assessment
CBE #5591 Staff Assessment
Importance
Strengths:
- A clear logic model is provided, depicting the relationships between inputs (e.g., access to an interdisciplinary palliative care team, electronic health record [EHR] systems capable of tracking Advance Directives and triggering alerts, and formal education for providers), activities (e.g., proactive screening, standardized goals of care discussions, and timely referral), and desired outcomes (e.g., reductions in emergency department [ED] visits, hospitalizations, and intensive care unit [ICU] visits, increase in the percentage of patients enrolled in hospice for at least three days before death, and improved symptom control, comfort, quality of life, and patient and caregiver satisfaction). This model demonstrates how the measure's implementation will lead to the anticipated outcomes.
- If implemented, the developer argued the measure’s anticipated impact on important outcomes, such as improved patient and caregiver satisfaction, reduced aggressive interventions at the end of life, lower health care utilization, and lower costs, is expected to be positive, based on empirical studies.
- The measure is supported by a comprehensive literature review, including clinical practice guidelines with evidence grading of strong/high and high quality empirical studies, demonstrating a clear net benefit in terms of improved outcomes and reduced cost/resource use for patients with advanced cancer who are nearing the end of life.
- Data from January 2023 to December 2024 show a performance gap, with decile ranges from 0.0% to 38.8%, indicating variation in measure performance across the target population.
- The proposed measure addresses a health care need not sufficiently covered by existing measures, offering advantages in terms of a cancer-specific target population, claims-based case identification, reduced implementer burden, and a hospital-level focus that differs from existing end-of-life measures.
- Description of patient input supports the conclusion that the measured intermediate outcome. is meaningful with at least moderate certainty. Patient input was obtained through focus groups, expert consensus, and public comments which included patients, family members, and caregivers.
Limitations:
- Although the developer indicated they had gathered patient and caregiver input through a technical expert panel and public comment period, direct engagement with and feedback from patients themselves is limited.
Rationale:
- This new measure meets all criteria for ‘Met’ for importance due to the significance of the problem it addresses and its significant anticipated impact, its robust evidence base, a documented performance gap in acute care admissions in the final month of life, and well-articulated logic model, making it essential for addressing end-of-life care coordination and support.
- There is at least moderate confidence that the business case is adequate, i.e., the anticipated impacts of the measure on patient outcomes and resource use justify use of the measure.
Closing Care Gaps
Strengths:
- The developer provided evidence of gaps in care related to the measures focus for subgroups, including a literature review and their claim that the measure will help close care gaps by highlighting performance trends in end-of-life care for cancer patients and encouraging timely referral to palliative care is credible.
Limitations:
- While the developer assessed gaps in care for end of life care cancer care, the developer did not clearly provide recommended actions to close care gaps. Note that empirical analysis of gaps in care is not required for initial endorsement.
Rationale:
- This new measure is rated 'Not Met but Addressable' for Closing Care Gaps. While the developer attempted to assess gaps in care for hospital admissions in the last 30 days of life, the developer did not provide recommended actions to close care gaps. This limits the ability to provide a comprehensive understanding of the differences in performance across different populations.
Feasibility Assessment
Strengths:
- All required data elements are routinely generated during care delivery, and required elements are available from digital or electronic sources.
- The developer stated that no feasibility issues were found requiring adjustment of the final measure specifications.
- The developer described the costs and burden associated with data collection and data entry, validation, and analysis. They stated that as a claims-based measure, there is negligible administrative burden and no direct implementation cost for the measured entities associated with the measure.
- The developer described how all required data elements can be collected without risk to patient confidentiality, including suggesting a minimum volume of five patients for performance reporting to prevent the potential re-identification in small patient populations.
- Any fees, licensing, or other requirements to use any aspect of the measure (e.g., value/code set, risk model, programming code, algorithm) are clearly described and justified.
Limitations:
- The fee for entities that are not non-profits is not stated. There is an outstanding question if the fee structure places the measure out of reach for rural and safety net clinicians and group practices.
- The measure specification is embedded in the measure submission, which makes review challenging.
Rationale:
- This new measure meets all criteria for 'Met' for feasibility due to its well-documented feasibility assessment, clear and implementable data collection strategy, and transparent handling of patient confidentiality, burden, licensing, and fees. These factors collectively ensure that the measure can be implemented effectively and sustainably in a real-world health care setting.
- Reading and interpreting the measure specification within the E&M submission is challenging. Committee members will benefit from a pdf attachment of the measure specification
Scientific Acceptability
Strengths:
- None identified.
Limitations:
- The developer did not perform the required reliability testing for this new measure, namely, they did not present valid, current evidence of person/encounter-level (“data element”) reliability testing for all critical data elements. The Earle et al., 2005 article cited is over 20 years old, and there is no record of the Shin et al., 2019 article in the Journal of Cancer.
- Accountable entity-level testing is not required for this new measure, so this observed limitation has no impact on the rating. The developer provided accountable entity-level reliability testing. For a two-year dataset consisting of 13,562 patients across 271 entities less than 70% of the entities have a reliability greater than 0.6. If the reliability were calculated for a one-year period of performance, less than 40% of the entities are estimated to have a reliability greater than 0.6.
Rationale:
- This new measure is rated as ‘Not Met, But Addressable’ for reliability because the data do not meet the requirements for reliability testing indicating potential issues with the consistency and accuracy of the results across different settings and populations. However, the identified limitations are deemed addressable, as the developer may consider providing valid current evidence of person/encounter-level reliability.
Strengths:
- The developer provided adequate evidence of person- or encounter-level (“data element”) validity testing from prior research for this new measure's numerator. The study the developer cited reported sensitivity (87%) and specificity (97%) for the numerator ("Admission to the ICU in the last month of life"; Earle et al., 2005), indicating that claims data can identify patients who were and were not admitted to the ICU in the last 30 days of life, with reasonable certainty.
- The developer also provided results from accountable-entity validity testing in their submission. This testing is not required for new measures and is not considered in the validity rating.
Limitations:
- The study cited by the developer to support the data element validity of the measure's denominator could not be found. A study with a similar title that appears to report the same sensitivity and specificity estimates, evaluated sensitivity and specificity only for only a subset of patients with colorectal cancer (Hwang YJ, Kim N, Yun CY, Yoon H, Shin CM, Park YS, Son IT, Oh HK, Kim DW, Kang SB, Lee HS, Park SM, Lee DH. Validation of Administrative Big Database for Colorectal Cancer Searched by International Classification of Disease 10th Codes in Korean: A Retrospective Big-cohort Study. J Cancer Prev. 2018 Dec;23(4):183-190. doi: 10.15430/JCP.2018.23.4.183. Epub 2018 Dec 30.).
- In addition, the developer indicated the measure has denominator exceptions, specifically, for patients admitted to the ICU due to complications arising from certain cancer treatments (bone marrow/stem cell transplants, chimeric antigen receptor [CAR] T-cell therapy) received in the last 60 days of life. Evidence of validity should be submitted for these data elements as well.
- Finally, the study cited to support data element validity for the numerator is more than 20 years old; if possible, the developer should provide additional context that supports the continued validity of the numerator.
- The developer did not conduct risk adjustment or stratification, but provided the rationale that quality palliative care is a universal standard that should not vary based on patient complexity and that adjustment could normalize departures from palliative care best practices. The developer did not provide supporting literature, a conceptual model, or empirical analysis demonstrating that differences in patient characteristics do not affect measure results or inhibit fair comparisons.
Rationale:
- This maintenance measure is rated as ‘Not Met But Addressable’ for validity; data element validity testing results partially support an inference of validity, suggesting that the measure somewhat accurately reflects performance on quality and can distinguish good from poor performance to a limited extent.
- The developer did not conduct risk adjustment or stratification and provided a rationale for the decision, but did not support the rationale with supporting literature or empirical analysis.
Use and Usability
Strengths:
- The measure is not currently in use, but the developer described a plan for use in payment programs such as the Merit-based Incentive Payment System (MIPS), PPS-Exempt Cancer Hospital Quality Reporting (PCHQR), and Inpatient Quality Reporting (IQR), as well as in private payer quality initiatives, value-based purchasing (VBP) programs, and ASCO’s quality improvement portfolio.
- Attributes of a suitable program for this measure are described, and these include a target population of adult patients with cancer, accountability attributed to the level of Oncology Physician Group Practice (PGP) or individual clinician group/practice, and the care setting being the outpatient oncology clinic.
The developer provided a summary of how accountable entities can use the measure results to improve performance. Specifically, in providing timely referral to hospice care, offering referrals to specialized, interdisciplinary palliative care, and engaging in advance care planning with patients and caregivers. - The developer did not identify any potential unintended consequences. They acknowledged that performance on the measure is not expected to be perfect. A margin of error should be expected to account for factors such as patient preference and lack of access to hospice care.
Limitations:
- None identified
Rationale:
- This new measure is rated ‘Met’ for use and usability because there is a clear plan for use in at least one accountability application, and the measure provides actionable information for improvement. The developer reported that no potential unintended consequences were identified.
Committee Independent Review
Concur with Staff Assessment
Importance
Nice, comprehensive logic model
Closing Care Gaps
Would have like to see analysis related to disparities by rural/urban, race/ethnicity, etc., if feasible. Also missing examples of strategies to reduce care gaps.
Feasibility Assessment
Claims-based measure with data elements readily available
Scientific Acceptability
Reliability <0.7 for most deciles
Submission could have benefited from a more detailed discussion and evidence supporting testing hypothesis
Use and Usability
Appreciate description of how strategies for improving measure performance
5591 Review
Importance
I agree with the staff assessment.
Closing Care Gaps
The developer did not provide recommended actions to close care gaps.
Feasibility Assessment
I agree with the staff assessment: well-documented feasibility assessment, clear and implementable data collection strategy, and transparent handling of patient confidentiality, burden, licensing, and fees.
Scientific Acceptability
As a Patient Partner, I am not a subject matter expert in this area and will rely on the staff assessment.
As a Patient Partner, I am not a subject matter expert in this area and will rely on the staff assessment.
Use and Usability
I agree with the staff assessment that there is a clear plan for use in at least one accountability application, and the measure provides actionable information for improvement.
Summary
As a patient partner, I support this measure.
Overall, supportive of this measure
Importance
The evidence review supporting this measure was through and makes the case for the importance of this measure in terms of patient centered care and healthcare utilization. The logic model is excellent
Closing Care Gaps
The evidence review briefly discusses care gaps, primarily among cancer types (e.g. hematologic malignancies compared to solid tumors). This is a new measure so it is understandable that care gaps are not yet addressed. Will the data capture as proposed allow for evaluation of care gaps across race, gender and SES in the future?
Feasibility Assessment
Claims based measure' no burden to organization. The measure is proprietary to ASCO, but ASCO does not charge a licensing fee to not-for-profit hospitals, healthcare systems, or practices to use the measure for quality improvement, research or reporting to federal programs.
Scientific Acceptability
Defer to Staff Preliminary Assessment
Defer to Staff Preliminary Assessment
Use and Usability
This is a new measure that is not yet in use, but the developers list Payment Programs and Professional Certification/Recognition Programs as potential uses. I think this metric would also be useful for external benchmarking and internal QI. The logic model is detailed enough to assist organizations in the identification of QI opportunities.
ASCO’s End of Life Measures TEP emphasizes that performance on this measure is not expected to be perfect, so I am curious about how it will be used in performance programs (e.g. IQR, MIPS, etc). I do understand that endorsement of the metric is idependent of how it is used, but it does raise questions about potential unintended consequences (e.g. limiting care options for patients who choose agressive treatment, etc) if this metric is included in these programs.
Summary
Overall I am supportive of this measure. I do have 2 questions about the measure specifications:
the numerator description states: "Step down units are not included in this measure" However, the inpatient billing codes listed include:
0206 - Intermediate ICU
0214 - Intermediate CCU
Why are these considered ICUs?
2. The denominator lists "exceptions" not exclsuions. What is the difference between exclsuions and exceptions? (I do agree with BMT and CAR T as exceptions)
Public Comments
Comment on End-of-Life and Hospice Cancer Care Measures
Dear Partnership for Quality Measurement:
On behalf of the more than 5,000 members of the American Academy of Hospice and Palliative Medicine (AAHPM), we appreciate the opportunity to submit comments in response to the Partnership for Quality Measurement (PQM) Spring 2026 Endorsement and Maintenance (E&M) cycle. AAHPM is the professional organization for physicians specializing in Hospice and Palliative Medicine (HPM). Our membership also includes nurses, social workers, spiritual care providers, pharmacists, and other health professionals deeply committed to improving quality of life for the expanding population of patients facing serious illness as well as their families and caregivers. Together, we strive to advance the field and ensure that patients across all communities and geographies have access to high-quality palliative and hospice care.
We appreciate PQM’s review of several measures focused on hospice and end-of-life care for patients with cancer. These measures address a meaningful gap in quality measurement and support patient-centered care at the end of life. We offer the following comments on the individual measures.
Percentage of Patients who Died with Cancer Admitted to the Intensive Care Unit (ICU) in the Last 30 Days of Life (CBE ID 5591)
AAHPM supports endorsement and adoption of this measures. The measure is intended to create incentives for earlier referrals to appropriate palliative care, which has been shown to significantly reduce ICU admissions, emergency department visits, and hospitalizations at the end of life^1 while increasing hospice enrollment and supporting longer hospice lengths of stay. Endorsing this measure reinforces a care model that prioritizes symptom management and patient goals over avoidable, high-intensity interventions in the final weeks of life.
Conclusion
AAHPM appreciates the opportunity to provide comment on these measures and supports adoption of measures that continue to support high-value, patient-centered end-of-life care. Please direct questions or requests for additional information to Katherine Ast, AAHPM Director of Quality and Research, at [email protected].
^
National Comprehensive Cancer Network. (2026). NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines®): Palliative Care (Version 1.2026). https://www.nccn.org/professionals/physician_gls/pdf/palliative.pdf
Public comment: 5591
The Alliance of Dedicated Cancer Centers appreciates the opportunity to submit comments on the ASCO EOL claims measures. The comments are the same for each measure:
In future testing iterations, we continue to support testing risk adjustment methods for these measures using carefully selected covariates to adjust for certain case mix variables.
Thank you.