End-of-Life Care for Older Adults with Aggressive Non-Hodgkin Lymphoma

Abstract

Background:

Aggressive non-Hodgkin lymphoma (NHL) commonly affects older adults and is often treated with intensive therapies. Receipt of intensive therapies and absence of a clear transition between the curative and palliative phases of treatment yield prognostic uncertainty and risk for poor end-of-life (EOL) outcomes. However, data regarding the EOL outcomes of this population are lacking.

Methods:

We conducted a retrospective analysis of adults ≥65 years with aggressive NHL treated with systemic therapy at Massachusetts General Hospital from April 2000 to July 2020 who subsequently died. We abstracted patient and clinical characteristics and EOL outcomes from the medical record. Using multivariable logistic regression, we examined factors associated with hospitalization within 30 days of death and hospice utilization.

Results:

Among 91 patients (median age = 75 years; 37.4% female), 70.3% (64/91) were hospitalized, 34.1% (31/91) received systemic therapy, and 23.3% (21/90) had an intensive care unit admission within 30 days of death. The rates of palliative care consultation and hospice utilization were 47.7% (42/88) and 39.8% (35/88), respectively. More than half of patients (51.6%, 47/91) died in a hospital or health care facility. In multivariable analysis, elevated lactic acid dehydrogenase was associated with risk of hospitalization within 30 days of death (odds ratio [OR] 3.61, p = 0.014). Palliative care consultation (OR 4.45, p = 0.005) was associated with a greater likelihood of hospice utilization, whereas hypoalbuminemia (OR 0.29, p = 0.026) was associated with a lower likelihood of hospice utilization.

Conclusions:

Older adults with aggressive NHL often experience high health care utilization and infrequently utilize hospice care at the EOL. Our findings underscore the need for interventions to optimize the quality of EOL care for this population.

Introduction

Aggressive non-Hodgkin lymphoma (NHL) represents the most common lymphoid malignancy and often affects older adults.¹ Treatment of aggressive NHL usually includes intensive chemotherapy or chemoimmunotherapy, which is potentially curative but commonly results in substantial toxicities and high health care utilization.^2,3 Moreover, older adults with aggressive NHL have high prevalence of frailty and high comorbidity burden that augment their risk of toxicity with intensive treatment.^4,5 Thus, aggressive NHL is a prevalent and highly problematic disease in the older adult population.

Despite the significant risk of toxicity, older adults often experience favorable outcomes with intensive treatment of aggressive NHL, with a significant proportion of patients achieving long-term survival.⁶ However, not all patients have such favorable results, and others may experience progressive disease with a more limited prognosis.^3,6–9 Currently, clinicians cannot reliably predict which patients will tolerate treatment and achieve cure with therapy or experience unfavorable results, thereby leaving patients unsure about their future health. Consequently, older adults with aggressive NHL experience significant prognostic uncertainty and are at risk for intensive health care utilization and poor outcomes at the end of life (EOL), such as low rates of hospice use.¹⁰ Clinicians need information characterizing EOL outcomes in this unique population to guide discussions and enhance informed decision making for patients; however, studies describing EOL outcomes in older adults with aggressive NHL are lacking.

Therefore, in this study we sought to describe EOL outcomes in older adults treated for aggressive NHL and identify factors associated with important EOL outcomes, including hospitalization in the last 30 days of life and hospice utilization. We hypothesized that older patients receiving systemic therapy for aggressive NHL would experience high health care utilization, such as high rates of hospitalization, receipt of systemic therapy, and intensive care unit (ICU) admission in the last 30 days of life as well as low rates of hospice utilization. We also hypothesized that palliative care consultation would be associated with a greater likelihood of hospice utilization.

Methods

Study design

We conducted a retrospective analysis of deceased adult patients age ≥65 years with aggressive NHL receiving initial systemic treatment with chemotherapy or chemoimmunotherapy at the Massachusetts General Hospital (MGH) between April 2000 and July 2020. We excluded patients with an indolent histology and those who were not deceased during the eligible time period. We also excluded patients not receiving any systemic therapy, given our interest in assessing the EOL outcomes of those receiving treatment. We identified the eligible cohort through the MGH Research Patient Data Registry (RPDR) database, which stores clinical data for 6.5 million individuals who receive their care at Mass General Brigham providers in Massachusetts. We used aggressive NHL diagnosis codes to identify patients treated for aggressive NHL, which we confirmed by manual chart review. We received approval for this study from the Dana-Farber/Harvard Cancer Center Institutional Review Board.

Clinical information

We abstracted information from the electronic health record (EHR) through a comprehensive chart review conducted by a trained research coordinator (N.H.M.) under the supervision of an oncologist (P.C.J.). We collected patients' demographics, Eastern Cooperative Oncology Group (ECOG) performance status (date closest to the date when the final systemic therapy before death was received), comorbidity score (Charlson comorbidity index excluding the patients' lymphoma diagnosis),¹¹ lymphoma diagnosis, date of diagnosis, date of treatment initiation, stage at diagnosis, baseline serum lactic acid dehydrogenase (LDH) and albumin (closest to date of diagnosis), initial and last systemic therapy received, number of systemic therapies, and toxicities of therapy (grade 3–5 nonhematologic toxicity as defined by Common Terminology Criteria for Adverse Events version 5.0).

EOL outcomes

We determined patients' place of death, cause of death, palliative care and hospice utilization, and hospice length of stay using the EHR and the Social Security Death Index. We determined the location of palliative care consultation (inpatient, outpatient, or both), the primary reason for consultation (symptom management, goals of care, both symptom management and goals of care, or other reason), and whether palliative care was longitudinal versus a one-time consultation or single-hospitalization consultation. We also determined whether patients were hospitalized (yes vs. no), received systemic therapy (yes vs. no), or were admitted to the ICU (yes vs. no) within 30 days of death. The majority of patients who received lymphoma therapy received their health care utilization and EOL care within our system. In addition, the clinical team at these institutions obtain information regarding health care utilization and EOL outcomes at other institutions, which are scanned into the EHR to maintain data quality.

Statistical analysis

We used descriptive statistics to summarize patients' sociodemographic and clinical characteristics and to characterize EOL outcomes. We examined the association of systemic therapy within 30 days of death with hospitalization and ICU admission within 30 days of death using Fisher's exact test. We used multivariable logistic regression to examine the associations between patient demographics and clinical factors and binary outcomes of interest (hospitalization within 30 days of death and hospice utilization). We examined factors associated with these outcomes given their importance as acceptable EOL quality metrics in patients with hematologic malignancies.^10,12 We first conducted univariate analyses to assess the association between patient demographic (age, gender, marital status, and race) and clinical (lymphoma subtype, ECOG performance status, stage, prior therapies received and number of systemic therapies, grade 3 to 5 nonhematologic toxicity [yes or no], hypoalbuminemia [<3.5 g/dL], elevated LDH [≥250 U/L], comorbidity score, disease response, and palliative care consultation) factors with the binary outcomes of interest. We included variables with a p-value <0.10 from the univariate analyses in the multivariable models, as suggested by literature on model-building.^13,14 When examining hospitalization within 30 days of death and hospice utilization, we included age, gender, and disease subtype into the model a priori, as these variables have known associations with EOL outcomes and hospice utilization,^10,15 and also included date of diagnosis (dichotomized to before vs. on or after January 1, 2013) to account for potential changes in palliative care and hospice use over time. All reported p-values are two-sided with a p-value <0.05 considered statistically significant. We performed statistical analyses using Stata version 14.2 (StataCorp, College Station, TX).

Results

Study participants

Table 1 describes the sociodemographic and clinical characteristics of the patients (n = 91) in this study. The median age was 75 (range 65–99) years, and most patients were male (57/91, 62.6%), white (86/91, 94.5%), and married/with a life partner (56/91, 61.5%). Most (50/89, 56.2%) had an ECOG performance status of 0 or 1. The most common lymphoma diagnosis was de novo diffuse large B cell lymphoma/grade 3B follicular lymphoma (64/91, 70.3%), and patients had received a median of 2 total systemic therapies (range 1–8). Overall, 70.3% (64/91) of patients had advanced stage disease, and the median Charlson comorbidity index score pretreatment was 0 (range 0–6). The median pretreatment albumin was 3.7 g/dL (range 1.7–4.9), and 60.2% (53/88) of patients had an LDH >250 U/L.

Table 1.

Patient Characteristics

Characteristic	N = 91	%
Age at diagnosis (years), median (range)	74.6 (65–99)
Female gender	34	37.4
White race	86	94.5
Relationship status
Married/life partner	56	61.5
Single	8	8.8
Divorced/legally separated	8	8.8
Widowed	18	19.8
Other/unknown	1	1.1
Lymphoma subtype
DLBCL (de novo or transformed)/grade 3B follicular lymphomaa	64	70.3
HGBCL with MYC and BCL2 and/or BCL6 rearrangement	9	9.9
T cell lymphoma	6	6.6
Other	12	13.2
ECOG performance status
0–1	50	54.9
2–4	39	42.9
Unknown	2	2.2
Stage
Limited	27	29.7
Advanced	64	70.3
LDH ≥250 (U/L)b	53	60.2
Pretreatment albumin (g/dL), median (range)b	3.7 (1.7–4.9)
Charlson comorbidity index score, median (range)	0 (0–6)
Lines of therapy, median (range)	2 (1–8)

^a Richter's transformation was classified under other.

^b Three patients with missing data.

DLBCL, diffuse large B cell lymphoma; ECOG, Eastern Cooperative Oncology Group; HGBCL, high-grade B cell lymphoma; LDH, lactic acid dehydrogenase.

EOL outcomes

Table 2 depicts the EOL outcomes of the patients in our cohort. Overall, 70.3% (64/91) of patients were hospitalized in the last 30 days of life, 34.1% (31/91) received systemic therapy in the last 30 days of life, and 23.3% (21/90) had an ICU admission in the last 30 days of life. More patients receiving systemic therapy in the last 30 days of life were hospitalized in the last 30 days of life (30/31, 96.8%) compared with those not receiving systemic therapy in the last 30 days of life (34/60, 56.7%; p < 0.001). Among those receiving systemic therapy in the last 30 days of life, 33.3% (10/30) had an ICU admission in the last 30 days of life versus 18.3% (11/60) of those who did not receive systemic therapy in the last 30 days of life, but this was not statistically significant (p = 0.123). Fewer than half (42/88, 47.7%) of patients had a palliative care consultation, and a minority (21/88, 23.9%) received palliative care >30 days before death. Of those receiving palliative care consultations, the majority (33/42, 78.6%) occurred exclusively in the inpatient setting, and most (32/42, 76.2%) were seen either as a one-time consultation or followed during a single inpatient hospitalization. Symptom management (24/42, 57.1%) was the most common reason for palliative care consultation, followed by both symptom management and goals of care (11/42, 26.2%) and goals of care (4/42, 9.5%). Only 39.8% (35/88) received hospice services, with 80.7% (71/88) having a hospice length of stay ≤7 days. Among hospice enrollees, the median length of stay on hospice was 7 days (range 0–117). Among all patients, 51.6% (47/91) died in hospital, rehab, or nursing home, 23.1% (21/91) died at home, 11.0% (10/91) died in inpatient hospice or a hospice house, and 14.3% (13/91) had an unknown or other place of death. The most common cause of death was cancer progression (44/91, 48.4%) followed by infection or cancer treatment complication (20/91, 22.0%).

Table 2.

End-of-Life Outcomes

End-of-life outcome	N = 91	%
Hospitalization within 30 days of death	64	70.3
Systemic therapy within 30 days of death	31	34.1
ICU admission within 30 days of deatha	21	23.3
Receipt of a palliative care consultationb	42	47.7
Receipt of palliative care >30 days before deathb	21	23.9
Longitudinal palliative care	10	11.0
Location of palliative care consultation (N = 42)
Inpatient	33	78.6
Outpatient	5	11.9
Both inpatient and outpatient	4	9.5
Primary reason for palliative care consultation (N = 42)
Symptom management	24	57.1
Goals of care	4	9.5
Both symptom management and goals of care	11	26.2
Other	3	7.1
Receipt of hospice servicesb	35	39.8
Hospice length of stay >7 daysb	17	19.3
Place of death
Home	21	23.1
Hospital, rehab, or nursing home	47	51.6
Inpatient hospice or hospice house	10	11.0
Unknown/other	13	14.3
Cause of death
Cancer progression	44	48.4
Infection or cancer treatment complication	20	22.0
Other causes	16	17.6
Unknown	10	11.0

^a One patient with missing data.

^b Three patients with missing data.

ICU, intensive care unit.

Factors associated with hospitalization within 30 days of death

In univariate logistic regression, elevated LDH (odds ratio [OR] 3.62, 95% confidence interval [CI]: 1.39–9.43, p = 0.008) and diagnosis in 2013 or later (OR 2.66, 95% CI: 1.24–5.67, p = 0.012) were the only factors significantly associated with the likelihood of hospitalization within 30 days of death. In multivariable models, elevated LDH remained significantly associated with hospitalization within 30 days of death (OR 3.61, 95% CI: 1.30–10.0, p = 0.014) (Table 3).

Table 3.

Factors Associated with Hospitalization within 30 Days of Death

Variable	Odds ratio (95% CI)	SE	p
Age	0.96 (0.89–1.03)	0.04	0.361
Female gender	0.93 (0.32–2.74)	0.51	0.895
DLBCL/grade 3B follicular lymphoma	1.18 (0.37–3.83)	0.71	0.777
LDH ≥250 U/L	3.61 (1.30–10.0)	1.88	0.014
Diagnosis date in or after 2013	2.69 (0.84–8.61)	1.60	0.095

CI, confidence interval; SE, standard error.

Factors associated with hospice utilization

In univariate logistic regression analyses, the only factor that was significantly associated with likelihood of hospice utilization was palliative care consultation (OR 3.43, 95% CI: 1.40–8.40, p = 0.007). Although not statistically significant, hypoalbuminemia (OR 0.44, 95% CI: 0.17–0.94, p = 0.088) was associated with lower likelihood of hospice utilization and thus we included this factor in the multivariable model. In multivariable analysis, palliative care consultation (OR 4.45, 95% CI: 1.58–12.5, p = 0.005) was associated with a greater likelihood of hospice utilization, whereas hypoalbuminemia (OR 0.29, 95% CI: 0.10–0.87, p = 0.026) was associated with a lower likelihood of hospice utilization (Table 4).

Table 4.

Factors Associated with Hospice Referral

Variable	Odds ratio (95% CI)	SE	p
Palliative care consultation	4.45 (1.58–12.5)	2.35	0.005
Age	1.06 (0.99–1.14)	0.04	0.096
Female gender	0.87 (0.29–2.64)	0.49	0.805
DLBCL/grade 3B follicular lymphoma	1.11 (0.36–3.43)	0.64	0.860
Hypoalbuminemia (<3.5 g/dL)	0.29 (0.10–0.87)	0.16	0.026
Diagnosis date in or after 2013	1.81 (0.50–6.54)	1.19	0.363

Discussion

In this study, we found that older adults with aggressive NHL often experience intensive health care utilization at the EOL. Among this cohort of deceased patients with aggressive NHL, >70% were hospitalized and nearly a quarter had an ICU admission in the last 30 days of life. Under 40% received hospice services, whereas <20% had a hospice length of stay greater than seven days, and the majority of patients died in a hospital or health care facility. These findings underscore the intensity of EOL care for older adults with aggressive NHL and highlight the need for interventions to optimize the quality of EOL care in this population.

Although previous study suggests that most patients with cancer would prefer to maximize time at home, especially at the EOL,^16,17 the majority of patients in our cohort were hospitalized near the EOL and died in a hospital or health care facility. Notably, patients diagnosed in 2013 or later experienced higher rates of hospitalizations in the last 30 days of life compared with those diagnosed before 2013, underscoring the intensity of care especially at the EOL in this population. Prior studies in patients with other hematologic malignancies have also demonstrated high health care utilization and low rates of palliative care use.^18–20 Our results add to the literature by extending characterization of EOL outcomes to older adults with aggressive NHL. NHL is highly prevalent in older adults, with potentially curative treatment options that must balance the promise of therapeutic benefit and prolonged survival³ with the risk of significant toxicities and poor EOL outcomes.^2,20,21 This prognostic uncertainty, along with the absence of a clear transition between the curative and palliative phases of treatment likely impact EOL decision making and health care utilization in older adults with aggressive NHL.^20,22 By describing intensity of health care use at the EOL in older adults with aggressive NHL, our results provide important data to guide patient–clinician discussions and the development of interventions aimed at improving the EOL care of this unique geriatric oncology population.

We also examined hospice utilization among older adults with aggressive NHL at the EOL. Prior studies of patients with hematologic malignancies have found low rates of hospice utilization.^10,23,24 Similarly, our results demonstrate low rates of hospice utilization in older adults with aggressive NHL. Multiple factors limit the application and utility of traditional hospice models in this population, such as significant prognostic uncertainty, rapid clinical decline at the EOL, unrealistic expectations from patients and oncologists, delays in hospice conversations until death is imminent, and system-based barriers, including the need for transfusion support.^20,22,25 Future studies should examine whether novel EOL care interventions tailored to the unique needs of older adults with hematologic malignancies can improve their EOL care. Moreover, future study should focus on increasing hospice utilization by removing system-based barriers to hospice utilization such as the lack of transfusion access.²⁶ Notably, in our study, palliative care consultation was the only factor associated with a greater likelihood of hospice utilization. This aligns with prior study demonstrating that early palliative care integration improves the quality of life and care of patients with solid tumors and hematologic malignancies.^27–35 Thus, the integration of early palliative care interventions for high-risk older adults with aggressive NHL while pursuing curative intent therapy represents a potential strategy to enhance their quality of life and care during intensive treatment and at the EOL.³⁶

In this study, we also identified salient baseline clinical factors associated with hospitalization in the last 30 days of life and hospice utilization. We found that elevated LDH was associated with a greater likelihood of hospitalization in the last 30 days of life and hypoalbuminemia was associated with a lower likelihood of hospice utilization. LDH levels and hypoalbuminemia are prognostic factors for aggressive NHL, and hypoalbuminemia correlates with patients' nutritional status and comorbid illnesses; thus, these factors related to a worse prognosis that would be expected to correlate with higher rates of hospice use.^37–39 However, aggressive lymphomas are diseases with high prognostic uncertainty, and clinical decompensation can be rapid, which may contribute to these findings given that prior study has demonstrated delays in hospice conversations.^20,23 Interestingly, disease response to treatment was not associated with likelihood of palliative care consultation or hospice referral, also underscoring the prognostic uncertainty for some patients who respond to treatment. Moreover, because LDH levels and hypoalbuminemia are associated with aggressive disease,^37,38 they may also be associated with symptom burden and risk of hospitalization for symptom management. These baseline clinical factors are often obtained as part of routine care and can identify a subset of patients at higher risk for worse EOL outcomes. Therefore, these factors could potentially be utilized as triggers for additional interventions to optimize EOL care in this population. In addition, these findings can help better inform both patients and clinicians about the potential illness trajectory.

Our study has several limitations worth considering. First, this was a study of patients at a large academic center with limited patient diversity, and thus our findings may not generalize to other patient populations. Importantly, our study included a patient population lacking racial diversity, and future studies should evaluate EOL outcomes in a more diverse patient population. Notably, studies have shown that there are substantial disparities in hospice utilization and the intensity of EOL care among racial and ethnic minorities⁴⁰; thus, our study may underestimate the intensity of health care use at the EOL. Second, in this retrospective analysis we were limited to information about patients' EOL outcomes and health care utilization that were available in the EHR and, therefore, our data may not have fully captured all clinical outcomes. We also lacked data regarding transfusion support, the number of total cycles of systemic therapy received, and reasons for hospitalization in the last 30 days of life. Third, our sample size of deceased patients identified through the RPDR was modest and, therefore, limited the number of covariates we could analyze in multivariable logistic regression models for EOL outcomes; because of this, our analyses could not fully account for all possible confounders. Fourth, we were unable to analyze additional important EOL outcomes such as the number of patients with advance care directives and the number with documented EOL conversations nor were we able to examine trends and practice patterns between providers. Future research studies should examine the filing of advance care directives and frequency of EOL conversations in this patient population.

Conclusion

We demonstrated that older adults with aggressive NHL often experience high health care utilization and utilize palliative care and hospice services infrequently at the EOL. We also identified palliative care consultation as an important factor associated with a greater likelihood of hospice utilization. Our findings underscore the need for novel EOL care interventions in older adults with aggressive NHL to optimize the quality of their EOL care.

Authors' Contributions

All authors made substantial contributions to conception and design, acquisition of data, or analysis and interpretation of data. All were involved in drafting the article or revising it critically for important intellectual content. All provided final approval of the article and agree to be accountable for all aspects of the study.

Funding Information

Leukemia and Lymphoma Society.

Author Disclosure Statement

No competing financial interests exist.