End-of-Life Care Intensity Among Adolescent and Young Adult Patients With Cancer in Kaiser Permanente Southern California

Importance  Cancer is the leading disease-related cause of death among adolescents and young adults (AYAs), but little is known about the care that AYA patients with cancer receive at the end of life (EOL).

Objective  To evaluate the intensity of EOL care among AYA patients with cancer.

Design, Setting, and Participants  Cross-sectional study of Kaiser Permanente Southern California (KSPC) cancer registry data and electronic health records for 663 AYA patients with either stage I to III cancer and evidence of cancer recurrence or stage IV cancer at diagnosis. All patients were treated within KSPC, an integrated health care delivery system, and died between 2001 and 2010 before age 40 years (age range at time of death, 15-39 years).

Main Outcomes and Measures  (1) Chemotherapy use in the last 14 days of life; (2) intensive care unit (ICU) care in the last 30 days of life; (3) more than 1 emergency department (ED) visit in the last 30 days of life; (4) hospitalization in the last 30 days of life; and (5) a composite measure of medically intensive EOL care including any of the aforementioned measures.

Results  Eleven percent of patients (72 of 663) received chemotherapy within 14 days of death. In the last 30 days of life, 22% of patients (144 of 663) were admitted to the ICU; 22% (147 of 663) had more than 1 ED visit; and 62% (413 of 663) were hospitalized. Overall, 68% of patients (449 of 663) received at least 1 medically intensive EOL care measure.

Conclusions and Relevance  Most AYA patients received at least 1 form of medically intensive EOL care. These findings suggest the need to better understand EOL care preferences and decision making in this young population.

Adolescent and young adult (AYA) patients with cancer, defined by the National Cancer Institute as those aged 15 to 39 years, experience cancer at a unique life stage, when their peers are on a trajectory of identity formation and establishment of a life path through education, employment, and the development of social and family ties. Patients in this wide age range share the experience of a cancer diagnosis during a time of major social, developmental, and psychological transitions. For those AYA patients who experience cancer as a terminal illness, the contrast with their life expectations and the experiences of their peer group is particularly great.

Previous work has called for comprehensive attention to medical and psychosocial needs for AYA patients with cancer at the end of life (EOL).^1-4 Yet we know very little about the EOL care that these young patients receive. Existing work has focused on the development of tools for EOL care planning,^5,6 on psychological distress,^7-9 and on understanding adolescent patients’ wishes as they make cancer treatment decisions.¹⁰ In addition, a single-center study in France evaluated care among 45 AYA cancer decedents and found high rates of symptoms and aggressive measures.¹¹ However, we do not know how generalizable this work is to other centers or to young people in the United States. The development of optimal tools for EOL care delivery in this population will depend on a better understanding of the care such patients receive.

Earle and colleagues^12-14 have developed a set of EOL care measures focused on care intensity in the last month of life and propose benchmarks for optimal EOL care, suggesting that intensive EOL care should be rare. Adult patients who recognize that they are dying usually do not wish to receive aggressive EOL measures.¹⁵ However, young people may feel differently about the tradeoffs that are worthwhile for another day. As a result, rates of aggressive EOL care in AYA patients should be considered less normative, or reflective of a “right” rate of intensive measures, and more as a window into what is likely to be a complex story about patient preferences, clinician feelings and behavior, and EOL decision making in this group of young patients.

We used health care utilization data linked to cancer registry information to identify a cohort of decedents aged 15 to 39 years who had received cancer care within Kaiser Permanente Southern California (KPSC), a multicenter health plan and care delivery system that serves 3.7 million patients in California, and died between the years 2001 and 2010. Rates of intensive EOL care measures, including chemotherapy, hospitalizations, emergency department (ED) visits, and intensive care unit (ICU) care, were evaluated, along with patient factors associated with use of EOL intensive care measures.

Box Section Ref ID

This study used linked cancer registry and electronic health record data within the KPSC health plan to capture data on EOL care among AYA decedents with cancer. An integrated managed care organization, KPSC provides comprehensive health services to approximately 3.7 million racially/ethnically and socioeconomically diverse members who are broadly representative of residents in Southern California.¹⁶ A number of clinical databases are maintained by KPSC, including membership, diagnosis, procedures, pharmacy/infusions, utilization, outside claims, and cancer registry, all linkable with a unique member identifier. The KPSC’s Surveillance, Epidemiology, and End Result (SEER)-affiliated cancer registry contains data on all patients who were diagnosed and/or treated for a new cancer since 1988. The quality of the cancer registry data are assured by the SEER standard and is audited by SEER staff on a regular basis. Pediatric oncology patients at KPSC receive care within network such that electronic health records are available across the AYA population.

The institutional review boards for KPSC and the Dana-Farber Cancer Institute approved this study, and requirements for patient consent were waived.

To evaluate EOL care in the AYA population, we sought to identify a cohort of patients who died anticipated deaths such that EOL care planning would have been appropriate. Unanticipated deaths due to treatment toxic effects and non–cancer-related deaths, while important, may offer less opportunity for prospective EOL care planning; as such, our goal was to exclude such deaths whenever possible. To identify patients who died anticipated deaths, we used KPSC’s cancer registry and other electronic health records to form 2 retrospective cohorts of decedents who had either (1) stage IV cancer at diagnosis, so that prognosis was limited from the time of diagnosis, or (2) stage I to III (nonmetastatic) disease at diagnosis, with evidence of cancer recurrence before death (Figure).

The stage IV cohort included KPSC patients who (1) had stage IV cancer at diagnosis according to KPSC’s cancer registry; (2) died between the years 2001 and 2010; (3) were aged 15 to 39 years at time of death; (4) were diagnosed at least 30 days prior to death, so that EOL care indicators were evaluable; and (5) were enrolled in the health plan during the month of death, so that EOL care indicators could be found in available records. The cohort was formed using individual patient-level data from KPSC, including cancer diagnosis and stage; dates of birth, diagnosis, and death; and health plan enrollment dates.

The stage I to III cohort included patients with early-stage disease at diagnosis and evidence of cancer recurrence; patients were eligible for inclusion if they (1) had stage I to III or nonstaged cancer at diagnosis according to KPSC’s cancer registry; (2) died between the years 2001 and 2010; (3) were aged 15 to 39 years at time of death; (4) were diagnosed at least 30 days prior to death so that EOL care indicators were evaluable; (5) were enrolled in the health plan during the month of death; and (6) had 1 of 2 possible indicators of cancer recurrence: either new metastases, as indicated by International Classification of Diseases, Ninth Revision (ICD-9)codes for secondary malignant neoplasm of other organs (197.0-197.8, 198.0-198.82, 198.89),¹⁷ or receipt of more than 1 chemotherapy regimen, as indicated by chemotherapy administration with a gap of more than 90 days between episodes of administration, a method previously used in SEER-Medicare to ascertain recurrence.¹⁸

To validate our methods of patient ascertainment, we performed in-depth medical record review for 111 cohort patients, including 54 patients with stage IV disease and 57 patients with stage I to III disease and recurrence, to ensure that our selection strategy appropriately identified patients who died anticipated deaths. For ease of review, patients were selected for in-depth medical record review if they died during the era in which electronic physician notes were available at KPSC (year 2007 and beyond). Records were selected at random, and medical record review was performed by 2 of us (K.C. and O.S.) after training (by J.W.M. and C.R.C.), using an abstraction instrument developed for this study. Review was limited to the last 30 days of life. Patients were considered to have died anticipated deaths if the record included specific references in the last 30 days of life to a poor prognosis, incurable or progressive cancer, or EOL care planning including hospice. Statements relating to this determination were abstracted in full for review by the study team; final assignments were by consensus based on abstracted statements.

Descriptive data were collected from the health plan’s electronic health records and registry data, including dates of birth, diagnosis, and death; sex; race/ethnicity; census block education and income level; cancer type; and stage at diagnosis. Previously developed measures of EOL care intensity^12,13,19 were adapted for use in KPSC’s electronic health records: (1) chemotherapy within 14 days of death; (2) ICU care within 30 days of death; (3) more than 1 ED visit within 30 days of death; and (4) hospitalization within 30 days of death.

Data on the prevalence of intensive EOL care were generated as the percentage of decedents experiencing each EOL care measure. A summary measure indicating receipt of any of the 4 measures of intensive EOL care was also calculated, and χ² tests were used to compare differences in the proportions of patients receiving each measure according to cohort (stage IV vs stage I to III with recurrence). Logistic regression was used to examine associations between the different forms of intensive EOL care. Log binomial regression was used to evaluate associations between receipt of intensive EOL care and patient characteristics (eg, age at diagnosis and death, patient race/ethnicity, and diagnosis), presented as prevalence ratios. Due to poor model convergence, relationships between ICU care and associated factors were evaluated using Poisson regression. Bivariate analyses evaluated associations between each measure and patient characteristic; adjusted models were adjusted for all other patient characteristics. Analyses were performed for each intensive EOL care measure and for our summary measure of receipt of any intensive EOL care. Data are shown for the summary measure and for ICU care in the last 30 days of life, as a relatively rare event that signifies high-intensity care. Results using other intensive EOL measures are reported in eTable 1, eTable 2, and eTable 3 in the Supplement.

We identified 663 patients aged 15 to 39 years at time of death who died between the years 2001 and 2010 after receiving cancer care in KPSC (Table 1) and who met our criteria for patients likely to have died anticipated deaths. The cohort included 282 patients with stage I to III disease and evidence of cancer recurrence, and 381 patients with stage IV disease at diagnosis. A limited medical record review performed in 111 cohort patients confirmed the presence of statements in the medical record indicating that death was anticipated in 98% (109 of 111). Half of cohort patients were non-Hispanic white; 11% were black; 29% Hispanic; and 11% identified as Asian/other. The most common cancer diagnosis was gastrointestinal cancer (17%); other common diagnoses included breast cancer (15%), genitourinary cancers (11%), leukemia (14%), and lymphoma (10%).

In the combined cohort, including patients with stage I to III cancer with recurrence and patients with stage IV cancer, 11% of patients received chemotherapy within 14 days of death (Table 2); 22% received care in the ICU; and 22% had more than 1 visit to the ED in the last 30 days of life. Most patients (62%) were hospitalized in the last month of life, with slightly higher rates of hospitalizations among the stage IV cohort (66%) than among the stage I to III with recurrence (58%) (P = .04). Overall, 68% of patients received at least 1 of the measures of intensive EOL care, again with slightly higher rates in the stage IV cohort (71% stage IV vs 63% stage I to III with recurrence; P = .03).

Thirty percent of patients received only 1 intensive EOL measure; 29% received 2. Only 1% of patients received all 4 forms of intensive EOL care. Several intensive measures were correlated; patients who were hospitalized within the last month of life were more likely to receive chemotherapy in the last 14 days of life (odds ratio [OR], 2.29; 95% CI, 1.29-4.10) and ICU care (OR, 5.66; 95% CI, 3.35-9.55) and have more than 1 ED visit (OR, 13.44; 95% CI, 6.70-26.96) in the last month of life. In addition, ICU care was associated with higher odds of chemotherapy use (OR, 2.11; 95% CI, 1.25-3.57). However, ED visits were not associated with chemotherapy use (OR, 1.19; 95% CI, 0.68-2.11) or ICU care (OR, 1.35; 95% CI, 0.88-2.07).

Using leukemia as the reference cancer category, and adjusting for other clinical and patient characteristics, we found that patients with bone or soft-tissue cancers (relative risk [RR], 0.41; 95% CI, 0.24-0.72; P = .002) or breast cancer (RR, 0.50; 95% CI, 0.26-0.95; P = .04) were less likely to receive care in the ICU in the last month of life, as were patients who lived in the highest-income census tract (vs the lowest) (RR, 0.59; 95% CI, 0.38-0.93; P = .02). Compared with non-Hispanic whites, Asian patients received slightly more ICU care in adjusted models (RR, 1.57; 95% CI, 1.03-2.39; P = .04). Age and year of death were not associated with EOL ICU care (Table 3).

When factors associated with receipt of any intensive measure were evaluated and leukemia again serving as referent, bone or soft-tissue cancer (adjusted prevalence ratio [APR] 0.60; 95% CI, 0.38-0.96; P = .03) and gastrointestinal cancer (APR, 0.77; 95% CI, 0.63-0.94; P = .01) were associated with lower use of intensive EOL care (Table 4).

More than two-thirds of the young patients with cancer in our study received at least 1 form of intensive EOL care. Rates of intensive measures among AYA patients exceed proposed desirable benchmarks among older adults,¹³ which include, for example, fewer than 10% of patients using chemotherapy in the last 14 days of life, fewer than 4% of patients with more than 1 ED visit in the last month of life, and fewer than 4% admitted to the ICU in the last month of life. In addition, rates for ED and ICU utilization among AYA patients exceed those of Medicare decedents with advanced cancer in the last month of life (<10% for both intensive measure among the Medicare decedents).¹⁴ However, our population received inpatient care at rates similar to those reported in elderly patients; recent data suggest that more than 60% of Medicare cancer patients are hospitalized in the last month of life, rates comparable to those we have seen.¹⁹ Although population-based pediatric data for the measures we have used are somewhat limited, previous work outside the United States also notes widespread use of aggressive EOL measures among children with cancer, suggesting that use of intensive measures may extend to the youngest patients as well.^20,21

The real question, which may be more important than aggregate rates of intensive measures, is how EOL decision making unfolded for the individual patients we studied. What did these patients understand about their illness and prognosis? What kind of care did they want? How much were these patients willing to tolerate to prolong life? And, given our knowledge that young people may base their decisions on the needs of their loved ones,¹⁰ who else were these patients considering as they made choices for care? Although adult patients who know they are dying usually do not want to receive aggressive care,¹⁵ which is associated with poorer quality of life near death,^22,23 we do not know whether AYA patients feel the same way. High rates of intensive EOL measures in this population may not be a failure of communication or palliative care but might reflect very different values for EOL care in these young people compared with older adults.

Intensive EOL measures were correlated, suggesting that use of one form of intensive care, such as chemotherapy, can lead to a need for others, such as hospital and ICU care. However, some correlation is explained by variable construction, as with ICU care, which is also considered in-hospital care. In other cases, correlations represent linked processes of care, such as ED visits that precede hospitalizations.

We found few measured patient characteristics that were reliably associated with use of intensive EOL measures, with the exception of cancer type. The presence of some relatively weak associations combined with the number of relationships tested also suggests that such findings might be best considered exploratory. Nonetheless, patients with bone or soft-tissue cancers and gastrointestinal cancers used intensive EOL measures less than patients with leukemia, similar to findings reported previously.²⁴ This may reflect the poorer prognoses of sarcomas and gastrointestinal cancers from diagnosis, especially for patients diagnosed with advanced-stage disease. Patients with leukemia may also experience more rapid progression from the time of relapse, as well as symptoms such as bleeding that may be more likely to lead to emergency and inpatient care, even when such care is used primarily for palliative purposes. Also of note, some findings in our population differed from those reported in older adults,^12,25 especially the lack of changes in care intensity over time and the lack of pronounced racial and ethnic differences in care intensity.

Just as in older adults, in whom the Earle measures of intensive EOL care were originally used,^12,13 these measures have limitations. Most importantly, we do not know the intent of the care delivered. As noted, ED visits and hospitalizations may have been used for symptom palliation when supportive care services already in place were insufficient to manage challenging symptoms. In addition, these measures were defined retrospectively from the time of death, whereas care is delivered prospectively. Prognostication is difficult, and clinicians and patients are not always able to reliably recognize when death is likely to occur and make plans for care accordingly. Finally, these measures incorporate a global perception that high rates of aggressive care are undesirable, but they do not account for preferences of individual patients. This last issue, as we have noted, is likely to be particularly salient to dying young people.

Adolescents and young adults are likely to experience many different and highly individual influences as they make EOL decisions. This wide age range includes teens, whose parents make decisions on their behalf, and young adults, who may consider the feelings of partners and their own young children, across a background of wide differences in educational attainment, personal life stage, employment, and financial independence. Although we do not fully know how their decision-making priorities differ from those of older adults, previous work suggests that AYAs are especially interested issues of legacy and the emotional lives of those they leave behind^10,26 but may be less invested in planning for use of medical treatments,²⁶ an issue which could in part underlie our findings. Although we did not find notable differences in use of intensive EOL measures by age, more nuanced data could help us to understand ways decision-making processes differ across the age spectrum and in turn influence treatment goals. Similarly, the setting of care delivery and availability of age-appropriate hospice and palliative services may also dictate the care that is possible.

Predicting mortality for AYA patients can also be challenging; clinicians may therefore be more likely to favor life-prolonging measures in the setting of clinical uncertainty. Clinicians may also have different perspectives about when it is appropriate to use EOL interventions for AYA patients, and clinician practice with respect to prognosis discussions and EOL care planning is also likely to vary, just as it does for physicians of older adults.²⁷ While our data did not allow us to explore this issue, EOL conversations can return the focus of care for dying young patients to individual needs and preferences and thus offer an opportunity to optimize patient-centered EOL care.

Our data sources had inherent limitations. We could not capture care received by dually insured patients outside of the KPSC network and therefore may have underestimated aspects of EOL care for some patients. However, even if care outside of the network were a significant contributor to findings, we would expect actual rates of intensive EOL care to be even higher than those we have found, which are already striking. In addition, we could not capture patient-level data on education and income and were not able to evaluate other EOL care measures such as use of hospice care because data on hospice care could not be readily assessed electronically. Data on use of surgical interventions and duration of chemotherapy would offer added insight into care patterns and should be prioritized in future work. Finally, although our findings reflect care in a large cohort of patients who received care at a number of institutions in Southern California, they are limited to insured patients at select institutions and may not reflect care for the wider US population, including the poor and uninsured. Nonetheless, to our knowledge, our cohort is larger than that of any existing study on EOL care in AYA patients and includes substantial racial and ethnic diversity. As such, our study offers important insights into the care of these young patients as they approach death.

We identified high rates of intensive EOL care, especially hospitalizations, among AYA patients with cancer in the KPSC health care system. Available data offered limited ability to identify which patients are at high risk for such care. Almost certainly, there is more to this story. In particular, we do not know how many patients recognized their poor prognosis and yet elected to receive life-prolonging care, a choice that would differentiate these young patients from older adults. However, our data provide a starting point for understanding patterns of care and ultimately defining optimal EOL care in this young population. Ongoing work should focus on understanding EOL care needs and preferences in this young population.

Accepted for Publication: May 11, 2015.

Corresponding Author: Jennifer W. Mack, MD, MPH, Dana-Farber Cancer Institute, 450 Brookline Ave, Boston, MA 02215 (jennifer_mack@dfci.harvard.edu).

Published Online: July 9, 2015. doi:10.1001/jamaoncol.2015.1953.

Author Contributions: Drs Chen and Chao had full access to the data in the study and take full responsibility for the integrity of the data and the accuracy of data analysis.

Study concept and design: Mack, Cooper, Chao.

Acquisition, analysis, or interpretation of data: Mack, Chen, Cannavale, Sattayapiwat, Chao.

Drafting of the manuscript: Mack, Chao.

Critical revision of the manuscript for important intellectual content: Mack, Chen, Cannavale, Sattayapiwat, Cooper, Chao.

Statistical analysis: Chen, Chao.

Obtained funding: Mack.

Administrative, technical, or material support: Cannavale, Sattayapiwat.

Study supervision: Mack, Cooper, Chao.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was funded by grant U24CA171524 from the Cancer Research Network/National Cancer Institute (CRN/NCI).

Role of the Funder/Sponsor: The CRN/NCI was not involved in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication.