Temporal trends and determinants of inpatient palliative care utilization among hospitalized patients with acute myeloid leukemia: a retrospective cross-sectional study

Abstract

Background

Acute myeloid leukemia (AML) remains an aggressive malignancy, despite advancements in diagnosis and treatment. Early integration of palliative care has been shown to alleviate symptom burden and enhance quality of life among patients and their caregivers. However, palliative care remains underutilized, especially in hematologic malignancies. This study aims to determine the rate of inpatient palliative care utilization among hospitalized patients with AML and to identify predictors of inpatient palliative care utilization.

Methods

This retrospective study utilized data extracted from the National Inpatient Sample database from 2016 to 2020. Adult patients diagnosed with acute myeloid leukemia (AML) and recipients of palliative care services were identified through the International Classification of Diseases, 10th Revision (ICD-10) coding system. Sociodemographic, hospital-related, and clinical characteristics were summarized using frequencies and percentages. Binary logistic regression analysis was performed to determine independent predictors of inpatient palliative care utilization.

Results

There were 121,892 hospital admissions of patients with a primary diagnosis of AML. Of these, 460 patients (0.3%) received inpatient palliative care services. Among those receiving palliative care, most were aged ≥ 65 years (59.8%), male (53.3%), White (83.1%), and privately insured (40.2%). In multivariable analysis, older age (OR 1.57; 95% CI 1.23–2.01; p < 0.001), private insurance (OR 1.43; 95% CI 1.33–1.53; p < 0.001), receipt of chemotherapy or stem cell therapy (OR 4.13; 95% CI 3.28–5.20; p < 0.001), and inpatient mortality (OR 8.80; 95% CI 6.86–11.28; p < 0.001) were independently associated with increased palliative care utilization. While non-White race (OR 0.70; 95% CI 0.61–0.81; p < 0.001), treatment in certain hospital regions (OR 0.75; 95% CI 0.67–0.83; p < 0.001), and shorter length of stay (OR 0.60; 95% CI 0.47–0.76; p < 0.001) were associated with lower utilization.

Conclusion

Inpatient palliative care remains markedly underutilized among patients with acute myeloid leukemia. Sociodemographic, economic, hospital, and clinical factors significantly influence its utilization. Addressing these disparities through targeted interventions and institutional policies may enhance palliative care integration, potentially improving symptom management, quality of life, and overall care outcomes for patients with AML.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12904-025-01949-2.

Background

In the United States, acute myeloid leukemia (AML) is the most common type of leukemia in adults, particularly among those 65 years or older [1]. Although the age-adjusted incidence rates for AML cases have remained stable, the incidence of the disease is expected to rise due to the increasing aging population [2]. In the last decade, there have been considerable advances in the therapeutic landscape of AML. Still, despite these advances, it remains an aggressive malignancy, with a 5-year relative survival rate of just 32.9% [2, 3].

Patients with AML face a significant symptom burden and decline in their quality of life due to the nature of the disease and the effects of treatment, especially when receiving intensive chemotherapy or cellular therapy. These patients are at a high risk for infections, life-threatening hemorrhage, and organ dysfunction, which lead to higher rates of hospitalization and intensive care admissions and are a significant cause of morbidity and mortality [4, 5]. Beyond the physical burden of the disease, these patients and their caregivers often experience substantial emotional and psychological distress associated with symptom burden and prognostic uncertainty [6].

Integrating palliative care in cancer patients, especially early in the disease course, has been shown to reduce symptom burden from the disease, improve mental and emotional well-being, and improve end-of-life outcomes for patients with advanced disease [7–9]. However, compared to solid malignancies, palliative care services are infrequently utilized in hematologic malignancies, including AML [10, 11]. Beyond provider and disease-related factors, socio-demographic, socio-economic, and hospital-level factors also influence the utilization of palliative care in hematologic malignancies [12]. A study by Jackson et al. found age, insurance type, and race to be predictors of palliative care utilization among hospitalized patients with multiple myeloma [13]. While studies have been conducted to determine hospital-associated and sociodemographic factors associated with palliative care utilization among specific types of hematologic malignancies, to the best of our knowledge, there remains a dearth of population studies examining these factors among hospitalized patients with AML [13, 14].

Therefore, this study aims to determine the prevalence of palliative care utilization in patients hospitalized with acute myeloid leukemia and examine the sociodemographic, socioeconomic, and hospital-related factors that predict its use in the United States, using data from the National Inpatient Sample (NIS) database.

Methods

Study design and population

This was a retrospective study, where we obtained data from the National Inpatient Sample (NIS), part of the Healthcare Cost and Utilization Project’s collection of databases and tools. It is the largest publicly available all-payer inpatient healthcare database, created to provide regional and national estimates in the U.S. for inpatient utilization, access, costs, quality, and outcomes [15]. The dataset includes approximately 7 million hospital stays and 35 million hospitalizations annually [15]. Researchers and policymakers depend on the NIS to produce nationwide data on healthcare utilization, costs, quality, and outcomes. The NIS protects patient confidentiality by no longer including state and hospital identifiers. Data from 2016 to 2020 were used for this analysis.

Adult patients diagnosed with a primary diagnosis of acute myeloid leukemia were identified using ICD-10 codes for AML: C92.00, C92.01, C92.50, C92.51, C92.60, C92.61, C92.62, C92.A0, C92.A1, C92.A2, C92.Z0, C92.Z1, C92.Z2, C93.00-C93.02, C93.90-C93.92, C93.Z0-C3.Z2, C94.00-C94.02, C94.20-C94.22 (please see supplementary file for a complete description of the ICD codes).

Dependent variable

The dependent variable was palliative care utilization. This was identified using the ICD-code Z51.5, which is termed Encounter for Palliative Care. This variable was classified as Yes and No for patients who had an encounter for palliative care versus patients without a palliative care encounter, respectively.

Independent variable

Sociodemographic, racial, clinical, and hospital-level factors were analyzed as independent variables that could influence palliative care use among hospitalized patients with AML: Age (65 years and older), Sex, Race/Ethnicity (White, African American, Hispanic, Native Americans, Others), Insurance type (Medicare, Medicaid, Private, Others), Mode of admission (elective vs. non-elective), Hospital region (Midwest, Northwest, South, West), Length of Stay (five days or less vs. more than five days), Charlson comorbidity index (low, medium, high), receipt of chemotherapy or stem cell therapy and in-hospital mortality.

Statistical analysis

All statistical analyses were performed on data from the National Inpatient Sample (NIS) spanning 2016 to 2020. To ensure nationally representative estimates and adequately account for the NIS’s complex survey design—encompassing stratification, clustering, and weighting—sampling weights provided by the Healthcare Cost and Utilization Project (HCUP) were applied. Descriptive statistics summarized sociodemographic, hospital-level, and clinical characteristics, with categorical variables expressed as weighted frequencies and percentages, while continuous variables were expressed as mean and standard deviation. Chi-square was used to compare differences in the frequency of palliative care utilization across the sociodemographic and economic variables. Logistic regression analyses identified predictors of palliative care utilization, presenting results as odds ratios (ORs) with 95% confidence intervals (CIs). All statistical tests were two-tailed, with p-values < 0.05 indicating statistical significance. Analyses were conducted using IBM SPSS Statistics, Version 30.0 (IBM Corp., Armonk, NY).

Results

Sociodemographic and hospital characteristics

Table 1 presents the sociodemographic characteristics of inpatient hospitalizations among patients diagnosed with a primary diagnosis of acute myeloid leukemia (AML) between 2016 and 2020. During this period, a total of 121,895 hospital admissions were recorded. Palliative care services were utilized in fewer than 1% of these admissions (n = 460; 0.3%). Fig. 1 is a bar chart representing the frequency of palliative care utilization from 2016 to 2020, showing that the lowest frequency of utilization was in 2018.

Table 1.

Comparison of patient and hospital characteristics by palliative care utilization N = 121,895

Variables n (%)	Palliative care utilization		X2	p-value
	Yes n = 460	No, n = 121,435
Age			58.9	< 0.0001
Mean age (mean ± SD)	67 ± 16	58 ± 17
Greater than 65	275 (59.8)	51,085 (42.1)
Less than 65	185 (40.2)	70,355(57.9)
Sex			0.05	0.83
Male	245 (53.3)	6555 (54.0)
Female	215 (46.7)	55,815 (46.0)
Race			75.3	< 0.001
White	370 (83.1)	82,410 (70.1)
Black	60 (13.5)	12,170 (10.3)
Hispanic	5(1.1)	11,945(10.2)
Asian or Pacific	5(1.1)	4865(4.1)
Other	5(1.1)	6245 (5.3)
Median household income			6.55	0.088
0-25th percentile (lowest income quartile)	115 (26.1)	28,680 (24.0)
26-50th percentile (second income quartile)	90 (20.5)	30,105 (25.2)
51-75th percentile (third income quartile)	110 (25.0)	30,440 (25.5)
76-100th percentile (Highest income quartile)	125 (28.4)	30,130 (25.2)
Primary payer			5.06	0.168
Medicare	135 (29.3)	50,705 (41.8)
Medicaid	60 (13.0)	17,280 (14.2)
Private insurance	185 (40.2)	46,475 (38.3)
Self-pay	15 (3.3)	2800 (2.3)
No charge	0 (0.0)	390 (0.3)
Other	65(14.1)	3615 (3.0)
Hospital region			101.15	< 0.001
Northeast	155 (33.7)	27,335 (22.5)
Mid-west	40 (8.7)	24,800 (20.4)
South	235 (51.1)	48,435(39.9)
West	30 (6.5)	20,870 (17.2)
Type of Hospital			29.86	< 0.001
Teaching	305 (78.2)	87,220 (87.4)
Non-teaching	85 (21.8)	12,560 (12.6)
Mode of admission			2.199	0.138
Elective	155 (65.2)	37,395 (30.8)
Non-elective	300 (33.7)	83,825 (69.0)
Length of Stay			102.2	< 0.001
LOS < 5	350(76.1)	68,645 (56.5)
LOS ≥ 5	110 (23.9)	52,790 (43.5)
Charlson co-morbidity severity index			5.77	0.06
No comorbidity	0 (0)	0 (0)
Mild (1–2)	240 (52.2)	63,600 (52.4)
Moderate (3–4)	150 (32.6)	43,480 (35.8)
Severe (> 5)	70 (15.2)	14,360 (11.8)
Received chemotherapy or stem cell therapy			389.4	< 0.001
Yes	320(69.6)	34,090 (28.1)
No	140(30.4)	87,350 (71.9)
Disposition on discharge			1113.67	< 0.001
Discharged	280 (60.9)	115,400 (95.1)
Died during hospitalization	180 (39.1)	5995 (4.9)

X2 chi-square, LOS length of stay, N total sample size, n subgroup sample size

Fig. 1.

Prevalence of palliative care use in the sample of AML patients from 2016 to 2020

.

Most patients who received palliative care were aged over 65 years (n = 275; 59.8%) and predominantly male (n = 245; 53.3%), with a mean age of 67 ± 16 years. Most patients identified as White (n = 370; 83.1%) and were insured through Private Insurance (n = 185; 40.2%). Most had a median household income above the 75th percentile (n = 125; 28.4%).

Among those who received palliative care, the highest proportion were hospitalized in the Southern United States (n = 235; 51.1%), as illustrated in Fig. 2, and were predominantly admitted to teaching hospitals (n = 305; 78.2%). A substantial proportion had a mild comorbidity burden, as indicated by the severity index (n = 240; 52.2%), and the majority experienced hospital stays exceeding five days (n = 350; 76.1%). In-hospital mortality among this group was less than 50% (n = 180; 39.1%).

Fig. 2.

Geographic distribution of palliative care utilization across the four U.S. hospital regions. The South accounted for the highest proportion (51.1%) of hospitalizations with palliative care. [16]

Older age, White race, Southern region, shorter hospital stays, receipt of chemotherapy/SCT, and in-hospital mortality were all significantly associated with higher palliative care utilization (p < 0.001). Socioeconomic indicators (income, insurance), admission type, sex, and comorbidity burden showed no significant differences.

Predictors of palliative care utilization among AML patients

Multivariable logistic regression analysis identified several significant predictors of palliative care utilization (Table 2). Patients over 65 years of age had higher odds of receiving palliative care compared to younger patients (Adjusted Odds Ratio [AOR] = 1.57; 95% Confidence Interval [CI]: 1.233–2.00; p < 0.001). Non-White patients had slightly lower odds of receiving inpatient palliative care services than White patients (AOR = 0.7; 95% CI: 0.608–0.808; p < 0.001). Patients with non-Medicare were more likely to receive palliative care than those with other payers (AOR = 1.425; 95% CI: 1.325–1.533; p < 0.001). A hospital stay of less than five days was associated with increased odds of palliative care use (AOR = 0.597; 95% CI: 0.468–0.762; p < 0.001). Patients who received chemotherapy or stem cell therapy had over four times the odds of receiving palliative care (AOR = 4.129; 95% CI: 3.280–5.197; p < 0.001). Patients who died during hospitalization were eight times more likely to have received palliative care (AOR = 16.24; 95% CI: 15.55–16.95; p < 0.001). Regional variation was observed, with patients hospitalized in the Midwest, South, and West having lower odds of receiving palliative care than those in the Northeast (AOR = 0.746; 95% CI: 0.670–0.830; p < 0.001). Co-morbidity burden, sex, median household income, and whether the hospital was a teaching hospital did not influence the odds of receiving palliative care.

Table 2.

Multivariable logistic regression predicting palliative care utilization among hospitalized patients

Variable	B	OR (Exp(B))	p-value	95% CI for OR
Age Category (age ≥ 65)	0.453	1.573	< 0.001	1.233–2.006
Female (vs. Male)	−0.063	0.939	0.564	0.758–1.163
Race (Reference group: White)	−0.355	0.701	< 0.001	0.608–0.808
Primary Payer (Reference group Medicare)	0.354	1.425	< 0.001	1.325–1.533
ZIP Income Quartile	0.027	1.027	0.594	0.931–1.133
Region of Hospital (Reference group: Northeast region)	−0.294	0.746	< 0.001	0.670–0.830
Hospital Teaching Status	−0.227	0.797	0.105	0.608–1.045
Length of Stay	−0.517	0.597	< 0.001	0.468–0.762
Charlson Severity Index	−0.132	0.877	0.102	0.747–1.030
Chemo/SCT	1.418	4.129	< 0.001	3.280–5.197
Died During Hospitalization	2.174	8.795	< 0.001	6.858–11.281

Discussion

In this study, we found that inpatient palliative care services were considerably underutilized among hospitalized patients with AML, with less than 1% receiving such services. Factors such as advanced age, racial background, insurance type, hospital geographic region, length of hospital stay, receipt of chemotherapy or stem cell transplant, and in-hospital mortality rates were significantly associated with inpatient palliative care use.

The underutilization of inpatient palliative care among patients with AML is consistent with existing literature, which reports low utilization of palliative care services among patients with other hematologic malignancies [13, 14, 17]. El-Jawahri et al. highlighted some of these barriers: prognostic uncertainty of hematologic malignancies, a strong curative-intent mindset, and, interestingly, a misconception of the purpose of palliative care among oncologists [18]. With advancements in novel therapeutic strategies for hematologic malignancies, including AML, many of which lead to sustained remission even in cases of relapsed or refractory disease, oncologists may equate palliative care with end-of-life care, fearing they have failed the patient and their families in treating their disease [19]. Similarly, in another study by El-Jawahri et al., 52% of bone marrow transplant physicians believed that palliative care is tantamount to end-of-life care, and 66% indicated that an encounter with the palliative care team could diminish hope in patients and their caregivers [20]. The same authors also highlighted frequent clinical concerns, including that palliative care might signal abandonment, provoke negative reactions to the term “palliative,” or create conflict with disease-directed care [19]. Yet, these patients, due to the treatment complexities of hematologic malignancies and the higher risk of complications such as sepsis, graft-versus-host disease in SCT patients, ICU admissions, and increased inpatient mortality, are the ones who stand to benefit significantly from palliative care services. In a randomized trial, stem cell transplant recipients who received concurrent palliative care reported less depression and fewer post-traumatic stress symptoms compared to those receiving standard care alone [21].

Research has consistently shown that sociodemographic factors influence palliative care utilization. Our findings also reinforce the influence of these factors on palliative care access. Older patients were more likely to receive palliative care services than younger patients, a pattern that has also been observed in prior studies [12, 22]. These findings may reflect the higher prevalence of comorbidities, functional decline, and treatment intolerance in older populations, prompting physicians to focus on comfort care-oriented approaches. However, a study by Colibaseanu et al. revealed that patients over 80 were less likely to receive palliative care services [23]. In that study, it is essential to note that most of these patients passed away over a decade ago, at a time when the awareness and application of palliative care services were underrecognized.

Racial disparities in palliative care utilization were also evident. We found that non-White patients had lower odds of receiving inpatient palliative care services compared to White patients, consistent with similar studies in other hematologic malignancies [13, 14]. These can be attributed to systemic health inequities, disparities in health literacy, access to specialist care services, and differences in cultural perspectives about palliative care [24]. Interestingly, some studies of patients with advanced cancer found that African Americans and Hispanics reported higher odds of inpatient palliative consultation compared to Whites [25, 26]. Although the National Inpatient Database does not include data on disease stage, it is likely that a higher proportion of African American and Hispanic patients probably had advanced metastatic disease necessitating end-of-life care discussions.

Health insurance coverage appears to influence the use of inpatient palliative care services. We found that patients not covered by Medicare were more likely to receive inpatient palliative care services than those on Medicare, consistent with previous studies by Hsieh et al. and Jackson et al. [12, 27]. Although a direct causal relationship is unlikely in the inpatient setting, several explanations may account for this observation. Patients not on Medicare, especially those with private insurance, are probably White, have higher health literacy, and a better understanding of the importance of palliative care, and may be willing to engage in discussions with the palliative care team. However, these findings are inconclusive and thus potentially lay the grounds for further research exploring how insurance-related disparities intersect with other sociodemographic characteristics to influence palliative care utilization.

Beyond sociodemographic characteristics, regional disparities also appear to impact the utilization of inpatient palliative care services significantly. Patients hospitalized in the Northeast region were more likely to receive palliative care services than those in other regions across the U.S. These findings are similar to previous NIS-based studies [13, 28–30]. A status report on the growth of palliative care in the U.S. showed that 88% of hospitals in the New England states offered palliative care services, compared to 43% and 42% in the West South and East South-Central states, respectively [30]. These differences may be attributed to the higher concentration of academic centers and the more robust healthcare infrastructure in the Northeast, which enhances the availability and accessibility of palliative care teams. These findings highlight geographic inequities and the need for targeted efforts to expand palliative care capacity in resource-limited regions.

Clinical factors were also significant predictors of palliative care utilization. Patients receiving chemotherapy or stem cell therapy, inpatient mortality, and hospital length of stay were substantial predictors of inpatient palliative care use, consistent with similar studies, which looked at the relationship between other hematologic malignancies and palliative care utilization [13, 14]. This is expected, as patients receiving intensive therapies often experience substantial symptom burden, such as pain, fatigue, and anxiety, as well as symptom treatment-related complications, such as infection or sepsis [31–33]. Interestingly, we found that patients with a more extended hospital stay had lower odds of receiving palliative care, a counterintuitive finding, similar to Hsieh et al. [12]. One possible explanation is that patients with shorter hospitalizations are often admitted for acute, life-threatening complications, thereby prompting earlier palliative care involvement to facilitate difficult end-of-life discussions with patients and their families. In contrast, patients with prolonged stays, given the prognostic uncertainties, remain on disease-directed therapies, delaying palliative care involvement in these cases. These findings emphasize the need for clearer guidelines and systemic protocols to trigger timely palliative care consultations. Interestingly, palliative care use was linked to higher in-hospital mortality. Although the limitations of the NIS database restrict a detailed assessment of disease severity, patient preferences, and treatment decisions during these hospitalizations, this association likely reflects the pattern of late palliative consultation often initiated for challenging end-of-life discussions, hospice transition, and comfort care measures. Although these discussions are essential and are associated with reduced use of intensive and acute care services, greater hospice enrollment, and lower health care costs, it is best practice to integrate palliative care early, from the time of cancer diagnosis [34]. Beyond end-of-life care, early palliative care integration has been shown to reduce pain and psychological distress and improve overall quality of life in patients living with cancer [35, 36].

A strength of our study is that it is one of the few national analyses to examine inpatient palliative utilization among AML patients, using a large, diverse, and geographically representative database, enhancing the generalizability of our findings to real-world settings. Yet, our study is not without limitations that warrant consideration. The cross-sectional design restricts our ability to follow the patients longitudinally, leaving it unclear whether the patients who did not receive palliative care initially had received these services later in their care, perhaps on an outpatient basis. Secondly, the NIS database lacks granular clinical data, including the AML risk profile, performance status, treatment trajectory, and patient or caregiver preferences, all of which can meaningfully influence palliative care decisions. Third, because the NIS is a discharge-based, rather than patient-based database, we were unable to account for multiple admissions by the same patient, which may have led to over-or underestimation of utilization rates.

For future research, we propose a more rigorous study design to examine the timing of the initial palliative care consultation (whether inpatient or outpatient) relative to cancer diagnosis. Such studies should aim to identify, in addition to sociodemographic and economic factors, provider-related and disease-specific variables that influence the integration of this critical yet often underutilized service within oncology practice.”

Conclusion

Our study highlights significant underutilization of inpatient palliative care in AML and reveals demographic, geographic, and clinical disparities that influence access. Efforts to improve integration must be multifaceted, including educational initiatives to address misconceptions, standardized triggers for referral, culturally sensitive models of care, and expansion of palliative care infrastructure in resource-limited regions.

Abbreviations

AML

Acute Myeloid Leukemia

ICD

10-International Classification of Diseases, 10th Revision Codes

NIS

National Inpatient Sample

Authors’ contributions

LCA conceptualized the research project, performed data analysis, and drafted the manuscript. MI reviewed the manuscript. All authors approved the final version.

Funding

Not applicable.

Data availability

The datasets utilized in this study are obtained from the Healthcare Cost and Utilization Project (HCUP), sponsored by the Agency for Healthcare Research and Quality (AHRQ). Access to the Nationwide Inpatient Sample (NIS) database is subject to restrictions. Researchers interested in accessing the data may request permission by completing the HCUP data use agreement at: [https://www.hcup-us.ahrq.gov] (https:/www.hcup-us.ahrq.gov).

Declarations

Ethics approval and consent to participate

This research utilized de-identified, publicly accessible data from the Healthcare Cost and Utilization Project’s (HCUP) National Inpatient Sample (NIS) database. Consequently, it was exempt from Institutional Review Board (IRB) approval and informed consent requirements, which aligns with the U.S. Department of Health and Human Services’ policy on research involving publicly available, de-identified datasets.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.