Abstract
Context:
Although U.S. palliative care programs have substantial differences in their implementation, whether this heterogeneity impacts patient outcomes is unknown.
Objectives:
To determine if palliative care program characteristics are associated with differences in end-of-life quality metrics for patients with metastatic cancer.
Methods:
Retrospective cohort study of patients with metastatic cancer who received care from programs that participated in the National Palliative Care Registry, 2018–2019. Multilevel regression was used to examine the association between individual program characteristics and outcomes including use of hospice, hospice enrollment ≥ 3 days, use of intensive care (ICU) in the last 30 days of life, and use of chemotherapy in the last 14 days of life.
Results:
The cohort was comprised of 33,015 patients who received care from 235 palliative care programs. Program maturity was the only characteristic associated with a difference in any outcome. Patients who received care from mature programs were more likely to use hospice (adjusted hazard ratio (aHR) 1.15 [1.06–1.25], for 5–10 years vs. < 5 years; aHR 1.18 [1.09–1.29] for > 10 years vs. < 5 years), and were also more likely to have hospice enrollment ≥ 3 days (aHR 1.18 [1.08–1.31] for 5 – 10 years vs. < 5 years; aHR 1.22 [1.11–1.34] for > 10 years vs.< 5 years).
Conclusion:
Palliative care program characteristics largely were not associated with differences in end-of-life quality metrics for patients with metastatic cancer. Further work is needed to better understand why program maturity may be associated with improved outcomes.
Keywords: palliative care, hospices, neoplasms, critical care
Since 2016, the American Society of Clinical Oncology (ASCO) guidelines have endorsed early integration of specialist palliative care for patients with advanced cancer, a recommendation based on randomized controlled trials demonstrating improvements in quality of life and symptom burden for patients.1 Given the interest in using specialist palliative care to improve the value of care for patients with cancer as well as other serious illnesses, the availability of palliative care services has grown, with 83% of hospitals with over 50 beds now having palliative care programs.2 This growth of palliative care has necessarily led to increased heterogeneity in the characteristics of palliative care teams, where programs have substantial differences in their composition, resources and operational characteristics.3,4 For instance, in a voluntary survey of palliative care programs, approximately 80% of programs had a physician or nurse practitioner, but only 68% reported having a social worker, and 56% had a chaplain.5 Differences in team attributes and operational characteristics have not been well-studied, but existing data suggest that some characteristics of palliative care programs may be associated with a higher level of effectiveness in real-world settings.6,7 In ecological studies, increased staffing has been associated with greater penetration of the palliative care service (defined by the number of consults seen over the number of total admissions) and a shorter reported time to initial consultation.7 Programs with palliative care-certified staff, higher staffing-to-bed ratios and an interdisciplinary team have also performed better on hospital-level metrics for end-of-life care, but effects varied across different types of hospitals (e.g. large versus small, non-profit versus for-profit status).6 However, these studies have limitations, as they measured outcomes on a hospital-level, and did not examine a cohort of patients who actually received palliative care services.
Heterogeneity between programs may be of particular relevance for a complex interdisciplinary intervention like specialist palliative care. Despite the importance of specialist palliative care delivery and the focus on improving the end-of-life period for patients with advanced cancer, there has been little examination of program-level factors that may impact a palliative care team’s effectiveness. Thus, in this study, we used a novel dataset to explore whether differences in palliative care program characteristics were associated with differences in end-of-life quality metrics for patients with metastatic cancer who received specialist palliative care.
Methods
Patients and Data Collection
This study was approved by the Columbia University Medical Center Institutional Review Board (AAA-T2256) and adheres to STROBE guidelines.8 This study utilized a novel dataset comprised of patient-level data from Medicare Claims, 2017–2019; data on palliative care program characteristics from the National Palliative Care Registry (NPCR),9 a multi-center national registry of palliative care programs that provides comparative performance feedback to participating programs; data on hospital characteristics from the American Hospital Association (AHA) Annual Survey; and prospectively collected data on National Provider Identifier (NPI) numbers for clinicians from palliative care programs. Some programs provided services for more than one hospital, and patients had to have at least one visit in a hospital with a program that participated in the NPCR. Details on the creation of this dataset have been published previously.10
We created a cohort of patients with a diagnosis of metastatic cancer (identified using ICD-10 codes C77.x, C78.x, C79.x, C80.x, C7B.x) between January 1, 2018 to December 31, 2019 who received at least one specialist palliative care visit in any care setting (i.e. inpatient or outpatient),11 and for whom we had complete information about their end-of-life period. We chose to only include patients with metastatic cancer because all patients would be considered to have advanced disease and thus, be appropriate for palliative care. Given existing inaccuracies in identifying specialist palliative care delivery in population-level data,12,13 patients were counted as having received specialist palliative care if the rendering NPI of a physician bill matched an NPI in our prospectively collected and verified list of palliative care clinicians. Patients with claims solely from palliative care clinicians who provided other services (e.g. surgery, critical care) were also counted as having received specialist palliative care if the claim contained the diagnosis code Z51.5 indicating delivery of palliative care; otherwise, these patients were excluded from the analysis.
Exposures
The NPCR collects data from participating palliative care programs annually using an online questionnaire with approximately 70 questions to obtain detailed information on program characteristics; the data pertain to all care settings in which a program operates. We used the Consolidated Framework for Implementation Research (CFIR), an Implementation Science framework that broadly identifies factors across multiple levels that may influence implementation of an intervention,14 to guide selection of key variables to evaluate for inclusion. The CFIR is informed by empirical literature and includes 5 broad domains: 1) intervention characteristics, 2) outer setting, 3) inner setting, 4) characteristics of the individuals involved, and 5) implementation process. We considered variables measuring various team attributes and operational factors of palliative care programs and retained variables that had less than 5% missingness on a patient-level. Included variables are listed in Table 1. Continuous variables (e.g. full-time equivalent (FTE) of palliative care staff) were normalized to account for the number of overall hospital admissions and categorized into quartiles. Program maturity (years in existence) was categorized into 3 levels (< 5 years, 5–10 years, > 10 years). All other variables were dichotomized (for full details of variable definitions, see Supplemental Table 1).
Table 1.
Palliative Care Program and Hospital Characteristics
| All Patients (N = 33,015) | All Program/Hospitals (N = 281) | |
|---|---|---|
| Operational Characteristics a | ||
| Total FTE/10000 admissions, N (%) | ||
| 0th – 25th percentile | 8,059 (24.4%) | 69 (24.56%) |
| 25th – 50th percentile | 8,157 (24.7%) | 68 (24.2%) |
| 50th – 75th percentile | 7,955 (24.1%) | 68 (24.2%) |
| 75th – 100th percentile | 8,008 (24.3%) | 68 (24.2%) |
| Worked collaboratively with ICU team | 13,987 (42.4%) | 113 (40.2%) |
| Worked collaboratively with EM team | 5,922 (17.9%) | 41 (14.6%) |
| Fully integrated with hospice | 12,749 (38.6%) | 108 (38.4%) |
| Use of standardized screening criteria | 16,825 (51.0%) | 138 (49.1%) |
| Bereavement plan | 252 (0.8%) | 1 (0.4%) |
| Quality improvement plan | 292 (0.9%) | 2 (0.7%) |
| Satisfaction survey plan | 15,154 (45.9%) | 115 (40.9%) |
| Team wellness plan | 24,235 (73.4%) | 188 (66.9%) |
| Education plan | 252 (0.8%) | 1 (0.4%) |
| Team Attributes a | ||
| Interdisciplinary team | 17,970 (54.4%) | 119 (42.3%) |
| Teaching team | 8,861 (26.8%) | 39 (13.9%) |
| Joint Commission Certification | 5,798 (17.6%) | 48 (17.1%) |
| Years in existence till 2018, median (IQR) | ||
| < 5 yr | 4,387 (13.3%) | 56 (19.9%) |
| 5 – 10 yr | 11,373 (34.5%) | 100 (35.6%) |
| > 10 yr | 17,255 (52.3%) | 125 (44.5%) |
| Penetration (%), median (IQR) | 4.9 (3.9–6.6) | 4.9 (3.6–6.9) |
| Frequency of consults completed within the first 24 hours of referral | ||
| Rarely | 815 (2.5%) | 20 (7.1%) |
| Sometimes | 3,913 (11.9%) | 78 (27.8%) |
| Often | 21,383 (64.8%) | 120 (42.7%) |
| Always | 6,152 (18.6%) | 51 (18.1%) |
| Frequency of consults completed within the first 48 hours of admission | ||
| Rarely | 1,746 (5.3%) | 8 (2.8%) |
| Sometimes | 9,885 (29.9%) | 35 (12.5%) |
| Often | 13,321 (40.3%) | 182 (64.8%) |
| Always | 6,648 (20.1%) | 47 (16.7%) |
| Hospital features b | ||
| Teaching status | 15,354 (46.5%) | 80 (28.5%) |
| Bed size | ||
| <= 400 beds | 12,755 (38.6%) | 166 (59.1%) |
| > 400 beds | 20,260 (61.4%) | 115 (40.9%) |
| Total bed size, median (IQR) | 519 (308 – 812) | 349 (214 – 588) |
| Total admissions, median (IQR) | 28,160 (15,629 – 37,439) | 18,225 (10,238 – 29,426) |
| Total surgeries, median (IQR) | 19,115 (11,188 – 32,445) | 12,979 (8,422 – 22,118) |
| Total full-time physicians with privileges, median (IQR) | 31 (0 – 163) | 16 (0 – 88) |
| Total full-time nurses, median (IQR) | 1,005 (543 – 2,043) | 619 (326 – 1,147) |
FTE, full-time equivalent.
Data sourced from the National Palliative Care Registry.
Data sourced from the American Hospital Association Annual Survey.
Outcomes
We examined four end-of-life quality metrics as outcomes for patients with cancer: use of hospice, hospice enrollment ≥ 3 days, use of intensive care (ICU) in the last 30 days of life, and use of chemotherapy in the last 14 days of life.15,16 The follow-up period began at the time of the first PC visit and continued until patients had a study outcome or died. Because patients could be eligible for hospice at any time, we used time-to-hospice outcome to account for differential length of follow-up. For ICU and chemotherapy use, because patients could not be eligible for the outcome until a specific period before the end of life, we used binary outcomes. For any given outcome, if patients had the outcome prior to the first palliative care visit, they were excluded from analyses for that specific outcome.
Statistical analysis
We summarized patient characteristics of the cohort and palliative care program characteristics. We used multilevel Cox proportional hazards models for time-to-event outcomes and multilevel logistic regressions for binary outcomes to examine the association between individual palliative program characteristics and relevant end-of-life outcomes, where hospital was modeled as a random intercept to account for clustering of patients within hospitals. Patients were assigned to the hospital for which they had the most palliative care physician claims; if there was a tie, then they were assigned to the hospital from the earliest claim. For all regression analyses, we adjusted for patient characteristics measured at the first palliative care visit, including socio-demographics (age, sex, race and ethnicity, dual eligibility status); number of Elixhauser comorbidities (0, 1–3, ≥4); primary cancer type (lung, breast, genitourinary, gastrointestinal, other); healthcare use in 12 months prior to the first palliative care visit including all inpatient, outpatient and ED visits, and severe acute illness episodes (hospitalization with sepsis, ICU use or mechanical ventilation). We also included hospital-level variables including teaching status; quartile of number of beds; hospital type (non-federal government, not-for-profit, for-profit, federal government), and quartile of total admissions, quartile of total surgical procedures, quartile of full-time physicians, and quartile of full-time nurses.
Some program characteristics may affect outcomes directly (e.g. relationship with a hospice), while others (e.g. FTE of staff) may have both a direct effect and an indirect effect through affecting process measures like penetration (defined as the number of consults over the number of admissions for a given hospital), or time to consultation completion (Figure 1). For characteristics that we hypothesized could have an indirect effect, we planned to conduct mediation analyses following methods specified by Barron and Kenny,17 only if the program characteristic was significantly associated with the outcome.
Figure 1. Conceptual Diagram of Palliative Care Program Characteristics and Hypothesized Effects.
We examined characteristics of palliative care programs and their association with end-of-life quality metrics. A priori, certain characteristics were hypothesized to have only direct effects on outcomes, while others were hypothesized to potentially have direct and indirect effects mediated through process measures such as program penetration (defined as the number of patients seen over the total number of hospital admissions for a program) and reported time to consultation. Hospital and patient-level variables were also included in analyses as confounders.
Because of the number of exposures and outcomes being examined, we adjusted for multiple comparisons using Holm’s procedure.18 Database management and statistical analysis were performed using SAS 9.4 (SAS institute, Cary, NC) and R version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria).
Results
We had 33,015 patients with metastatic cancer who received specialist palliative care from 235 palliative care programs in 281 hospitals (Figure 2). Patients had a median age of 74.1 years, 49.5% were female, 2.5% were Asian or Pacific Islander, 13.5% were Black, 0.5% were North American Native, 77.6% were White, and 4.0% were of Hispanic ethnicity. The three most common primary sites of cancer were gastrointestinal (25.6%), genitourinary (24.3%) and lung (23.5%). Patients had high levels of comorbidity, with 44.7% of patients having 1–3 and 31.2% having ≥4 Elixhauser comorbidities. Patients had a median of 2 inpatient visits (interquartile range (IQR) 1–3) and 10 outpatient visits (IQR 4–22) in the 12 months prior to their first specialist palliative care visit. Patients had a median number of 2 palliative care visits (IQR 1–3) and 76.6% of initial visits occurred in the inpatient setting. Patients were relatively evenly distributed around the U.S. (Northeast 25.7%, South 33.9%, Midwest 19.8%, West 16.5%) (Table 2).
Figure 2.
Flow diagram of study subjects.
Table 2.
Characteristics of Patients who Received Specialist Palliative Care
| All Patients (n = 33,015) | |
|---|---|
| Age, Median (IQR) | 74.1 (68.6–80.6) |
| Age, N (%) | |
| < 65 | 3,762 (11.4%) |
| 65–69 | 6,642 (20.1%) |
| 70–74 | 7,427 (22.5%) |
| 75–79 | 6,253 (18.9%) |
| 80–84 | 4,462 (13.5%) |
| >= 85 | 4,469 (13.5%) |
| Female, N (%) | 16,332 (49.5%) |
| Race and Ethnicity, N (%) | |
| Asian/Pacific Islander | 809 (2.5%) |
| Black | 4,449 (13.5%) |
| North American Native | 159 (0.5%) |
| Other/Unknown | 683 (2.1%) |
| White | 25,609 (77.6%) |
| Hispanic Ethnicity | 1,306 (4.0%) |
| Dual Eligibility, N (%) | 7,502 (22.7%) |
| Number of Elixhauser Comorbidities, N (%) | |
| 0 | 7,972 (24.1%) |
| 1–3 | 14,758 (44.7%) |
| >=4 | 10,285 (31.2%) |
| Primary Cancer Type, N (%) | |
| Lung | 7,761 (23.5%) |
| Breast | 3,088 (9.4%) |
| Genitourinary | 8,028 (24.3%) |
| Gastrointestinal | 8,467 (25.6%) |
| Other | 5,671 (17.2%) |
| Prior Acute Care Conditions, N (%) | |
| Sepsis | 9,230 (28%) |
| ICU use | 10,392 (31.5%) |
| Mechanical ventilation | 3,102 (9.4%) |
| Prior Healthcare Utilization, median (IQR) | |
| Inpatient visits | 2 (1–3) |
| Outpatient visits | 10 (4–22) |
| ED visits | 1 (0–2) |
| Number of Palliative Care Visits, median (IQR) | 2 (1–3) |
| Inpatient Initiation of Palliative Care, n% | 25,286 (76.6%) |
| Region | |
| Northeast | 8,487 (25.7%) |
| South | 11,195 (33.9%) |
| Midwest | 6,538 (19.8%) |
| West | 5,443 (16.5%) |
IQR, interquartile range
With respect to operational characteristics of programs, patients received care from programs that had a median FTE of PC clinicians of 2.8 per 10,000 admissions (IQR 1.9–4.2) and a median penetration rate (defined as the number of patients seen over the total number of hospital admissions annually) of 4.9% (IQR 3.9%-6.6%). Over half of patients received care from programs that self-reported completion of consults within the first 24 hours of referral (Always 18.6%, Often 64.8%) or within the first 48 hours of admission (Always 20.1%, Often 40.3%) and used standardized screening criteria to identify patients in need of palliative care services (51.0%). Less than 50% of patients received care from programs that reported collaborative working relationships in the ICU (42.4%) or emergency department (17.9%); only 38.6% of patients received care from a program that was fully integrated with a hospice. Less than half of patients received care from programs that used a satisfaction survey (45.9%), and few patients received care from programs that used an education plan (0.8%), quality improvement plan (0.9%) or bereavement plan (0.8%). For team attributes, over half of patients received care from programs that had been in existence for over 10 years (52.3%), had interdisciplinary teams (54.4%), and used a team wellness plan (73.4%), while less than half of patients received care from programs with Joint Commission certification (17.6%) or teaching teams (26.8%) (Table 1).
Of patients included in the cohort, 64.5% used hospice, 52.9% were enrolled in the hospice for at least 3 days,14.2% used intensive care in the last 30 days of life, and 1.5% received chemotherapy in the last 14 days of life. The median time to hospice use was 15 days (95% confidence interval (CI): 14 – 15), and the median time to hospice enrollment for ≥ 3 days was 16 days (95% CI: 15 – 16). The only palliative care program characteristic that was significantly associated with an end-of-life outcome was program maturity. Patients who received care from mature programs were more likely to use hospice (adjusted hazard ratio (aHR) 1.15, 95% confidence interval (CI) 1.06–1.25, adjusted p-value 0.06 for 5–10 years vs. < 5 years; aHR 1.18, 95% CI 1.09–1.29, adjusted p-value 0.01 for > 10 years vs. < 5 years;), and were also more likely to enroll in hospice for at least 3 days (aHR 1.18, 95% CI 1.08–1.3, adjusted p-value 0.01 for 5–10 years vs. < 5 years; aHR 1.22, 95% CI 1.11–1.34, adjusted p-value 0.002 for > 10 years vs. < 5 years). There were no associations between any palliative care program characteristics and use of ICU or chemotherapy at the end of life (Table 3). Because we did not identify significant associations for any variables that were postulated to potentially have indirect effects, we did not conduct any prespecified mediation analyses.
Table 3.
Associations between Palliative Care Program Characteristics and Outcomes
| Program characteristics | Hospice enrollment | Hospice enrollment >= 3 days | ICU utilization in the last 30 days of life | Chemotherapy in the last 14 days of life | ||||
|---|---|---|---|---|---|---|---|---|
| Hazard Ratio (95% CI) | Adjusted p-value* | Hazard Ratio (95% CI) | Adjusted p-value* | Odds Ratio (95% CI) | Adjusted p-value* | Odds ratio (95% CI) | Adjusted p-value* | |
| Characteristics likely to have direct effect only | ||||||||
| Any bereavement plan, Y vs. N | 0.70 (0.42–1.16) | 1 | 0.54 (0.32–0.91) | 1 | 0.81 (0.37–1.79) | 1 | 1.6 (0.52–4.98) | 1 |
| Any education plan, Y vs. N | 0.7 (0.42–1.16) | 1 | 0.54 (0.32–0.91) | 1 | 0.81 (0.37–1.79) | 1 | 1.6 (0.52–4.98) | 1 |
| Any quality improvement plan, Y vs. N | 0.86 (0.6–1.23) | 1 | 0.74 (0.5–1.11) | 1 | 0.84 (0.43–1.65) | 1 | 1.25 (0.43–3.58) | 1 |
| Any satisfaction survey plan, Y vs. N | 0.92 (0.86–0.98) | 0.5 | 0.9 (0.85–0.96) | 0.2 | 1.02 (0.91–1.15) | 1 | 1.06 (0.86–1.31) | 1 |
| All integrated hospices were functioning as one administrative entity, or under the same hospital/health system, Y vs. N | 1.00 (0.94–1.07) | 1 | 1.01 (0.94–1.08) | 1 | 1 (0.88–1.13) | 1 | 1.19 (0.94–1.5) | 1 |
| Joint Commission Certification, Y vs. N | 1.06 (0.98–1.16) | 1 | 1.05 (0.97–1.15) | 1 | 1.01 (0.87–1.18) | 1 | 1.16 (0.89–1.51) | 1 |
| Years in existence till 2018, 5 – 10 yr vs. < 5 yr | 1.15 (1.06–1.25) | 0.06 | 1.18 (1.08–1.30) | 0.01 | 1.04 (0.88–1.23) | 1 | 1.13 (0.83–1.55) | 1 |
| Years in existence till 2018, > 10 yr vs. < 5 yr | 1.18 (1.09–1.29) | 0.01 | 1.22 (1.11–1.34) | 0.002 | 1.03 (0.88–1.22) | 1 | 1.01 (0.74–1.38) | 1 |
| Interdisciplinary team, Y vs. N | 0.94 (0.89–0.99) | 1 | 0.95 (0.89–1.00) | 1 | 0.97 (0.87–1.09) | 1 | 1.04 (0.83–1.31) | 1 |
| Teaching team, Y vs. N | 0.99 (0.91–1.08) | 1 | 0.98 (0.89–1.07) | 1 | 1.16 (0.99–1.35) | 1 | 1.09 (0.81–1.45) | 1 |
| Characteristics that may have indirect and direct effects | ||||||||
| Standardized screening criteria, Y vs. N | 1.09 (1.03–1.16) | 0.1 | 1.07 (1–1.13) | 1 | 1.16 (1.04–1.29) | 0.11 | 1.25 (1.01–1.55) | 0.7 |
| Any team wellness plan, Y vs. N | 0.96 (0.9–1.03) | 1 | 0.97 (0.9–1.06) | 1 | 1.1 (0.96–1.26) | 1 | 1.11 (0.84–1.45) | 1 |
| Worked collaboratively with all EM teams, Y vs. N | 1.04 (0.96–1.12) | 1 | 1.03 (0.94–1.11) | 1 | 0.88 (0.76–1.02) | 1 | 0.8 (0.6–1.08) | 1 |
| Worked collaboratively with all ICU teams, Y vs. N | 0.98 (0.93–1.04) | 1 | 0.99 (0.93–1.05) | 1 | 0.97 (0.86–1.08) | 1 | 1.06 (0.85–1.31) | 1 |
| Total FTE/10000 admissions, Q2 vs. Q1 | 0.99 (0.91–1.07) | 1 | 1.02 (0.94–1.11) | 1 | 0.97 (0.83–1.13) | 1 | 1.13 (0.83–1.55) | 1 |
| Total FTE/10000 admissions, Q3 vs. Q1 | 0.99 (0.92–1.07) | 0.3 | 0.99 (0.92–1.07) | 1 | 1.13 (0.97–1.3) | 1 | 1.01 (0.74–1.38) | 1 |
| Total FTE/10000 admissions, Q4 vs. Q1 | 1.04 (0.97–1.12) | 0.3 | 1.02 (0.95–1.1) | 1 | 1.03 (0.87–1.23) | 1 | 1.22 (0.87–1.72) | 1 |
P-value adjustment was done using Holm’s stepdown Bonferroni. For m tests in total, we first ranked the tests by p-values in ascending order, then rejected the kth test’s null hypothesis if the kth p-value was no more than 18. The level of significance (α) was set at 0.05.
Discussion
In a cohort of patients with metastatic cancer who received specialist palliative care, we found that except for program maturity, palliative care program characteristics were not associated with differences in end-of-life outcomes. Thus far, there has been limited examination of how differences in team attributes and operational characteristics of palliative care programs may be associated with patient outcomes. A study examining program-level variation in end-of-life outcomes for patients with metastatic cancer who received palliative care did not find that program characteristics were significantly associated with above or below average performance.19 In contrast, an ecological study that merged self-reported hospital data with data from the Dartmouth Atlas demonstrated that high levels of staffing alone was associated with decreased healthcare utilization in the last 6 months of life; other characteristics needed to be combined (e.g. presence of an interdisciplinary team and palliative care certification) to observe the same association.6 However, this ecological study was not limited to patients who received palliative care and did not include any measure of palliative care delivery, which may account for the difference in results.
We did find that program maturity (years in existence) was associated with hospice outcomes, where patients that received care from programs that had been in existence for more than 5 years were approximately 20% more likely to use hospice or have hospice enrollment ≥ 3 days. To our knowledge, this is a novel finding and deserves further investigation to better understand if this is a causal relationship, and if so, what the active ingredient(s) in program maturity might be. One possibility is that successful integration of specialist palliative care requires relationship building between teams. In a qualitative study examining clinician perspectives on using an integrated specialist palliative care model in oncology,20 the importance of trust was noted but specific ways to develop trust were not elicited. A different qualitative study examining adoption of specialist palliative care in the ICU demonstrated that the development of trust (which was essential to generating buy-in) centered on four different aspects of the relationship between palliative care and ICU clinicians.21 Thus, it may be that maturity is a reflection of the time that it takes to develop collaborative relationships and trust, as relationships between palliative care and primary team clinicians are likely to improve over time.22 If this finding is confirmed, a more in-depth understanding of why program maturity is associated with better outcomes may help programs may achieve improved outcomes.
While we found that most characteristics of palliative care programs were not associated with end-of-life quality metrics, these data should not be interpreted to mean that the way a program is implemented does not matter. In a national study of palliative care programs, programs that reported higher levels of staffing also reported higher levels of penetration (defined as the number of patients seen by the palliative care team/the total number of hospital admissions).7 Because we only included patients who received specialist palliative care in our cohort, we were unable to assess the impact of palliative care program characteristics on improving a program’s reach. Our findings suggest that if access to palliative care is equal, then program characteristics (except for maturity) are not associated with end-of-life quality metrics. Further studies are needed to determine whether these program characteristics make a difference with respect to the number of patients seen by a program or time to consultation. Additionally, the fact that program maturity, a complex latent construct likely comprising multiple factors, was the sole characteristic significantly associated with outcomes also raises the possibility that we have an incomplete understanding of how palliative care programs function to affect outcomes, and that we may not be measuring what matters most with respect to palliative care program implementation.
Our study has limitations. First, we relied on self-reported characteristics of palliative care programs; actual characteristics may differ, although there was no incentive for programs to falsely report their data. To our knowledge, the NPCR represents the most comprehensive survey of palliative care program characteristics in the U.S., but there may be important characteristics that we did not evaluate. Second, we examined end-of-life outcomes for patients with cancer, as these have been deemed important for the quality of cancer care. However, these outcomes are not solely affected by the work of palliative care teams and do not represent all outcomes of importance to programs (e.g. patient-reported quality of life). Third, we examined each program characteristic individually, but it may be that particular combinations of characteristics (e.g. high level of staffing and an interdisciplinary team) are necessary to observe an association with outcomes. Fourth, we used data from a cohort of patients with metastatic cancer who received care from a palliative care program that participated in the NPCR, limiting the generalizability of the results to other centers and other populations of patients with serious illness. Fifth, because of the observational nature of the study, our findings should not be taken to be causal in nature, and should be viewed as hypothesis-generating. Lastly, demonstrated relationships with outcomes should not necessarily be the only factor guiding how we choose to implement and deliver care;23 preferences of patients, clinicians and other interested parties should also be considered.
In a large national cohort of patients with metastatic cancer who received specialist palliative care, program maturity was the only palliative care program characteristic that was significantly associated with differences in end-of-life quality metrics. Future work should focus on replicating and identifying mechanisms underlying this finding, determining whether palliative care program characteristics are associated with other outcomes, and elucidating whether other program factors that may impact a program’s ability to affect outcomes should be measured.
Supplementary Material
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Contributor Information
May Hua, Department of Anesthesiology, College of Physicians and Surgeons, Department of Epidemiology, Mailman School of Public Health, Columbia University.
Ling Guo, Department of Anesthesiology, College of Physicians and Surgeons, Columbia University.
Shuang Wang, Department of Biostatistics, Mailman School of Public Health, Columbia University.
R. Sean Morrison, Icahn School of Medicine at Mount Sinai and James J Peters VA, Bronx NY.
References
- 1.Ferrell BR, Temel JS, Temin S, et al. Integration of Palliative Care Into Standard Oncology Care: American Society of Clinical Oncology Clinical Practice Guideline Update. Journal of Clinical Oncology. 2016;35(1):96–112. [DOI] [PubMed] [Google Scholar]
- 2.Growth of Palliative Care in U.S. Hospitals 2022, Snapshot, New York: Center to Advance Palliative Care, 2022. [Google Scholar]
- 3.Pantilat SZ, Kerr KM, Billings JA, Bruno KA, O’Riordan DL. Characteristics of palliative care consultation services in California hospitals. J Palliat Med. 2012;15(5):555–560. [DOI] [PubMed] [Google Scholar]
- 4.Pantilat SZ, O’Riordan DL, Bruno KA. Two steps forward, one step back: changes in palliative care consultation services in California hospitals from 2007 to 2011. J Palliat Med. 2014;17(11):1214–1220. [DOI] [PubMed] [Google Scholar]
- 5.Spetz J, Dudley N, Trupin L, Rogers M, Meier DE, Dumanovsky T. Few Hospital Palliative Care Programs Meet National Staffing Recommendations. Health Affairs. 2016;35(9):1690–1697. [DOI] [PubMed] [Google Scholar]
- 6.Thygeson NM, Wang M, O’Riordan D, Pantilat SZ. Self-Reported California Hospital Palliative Care Program Composition, Certification, and Staffing Level Are Associated with Lower End-of-Life Medicare Utilization. J Palliat Med. 2016;19(12):1281–1287. [DOI] [PubMed] [Google Scholar]
- 7.Dumanovsky T, Rogers M, Spragens LH, Morrison RS, Meier DE. Impact of Staffing on Access to Palliative Care in U.S. Hospitals. J Palliat Med. 2015;18(12):998–999. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Ann Intern Med. 2007;147(8):573–577. [DOI] [PubMed] [Google Scholar]
- 9.National Palliative Care Registry. 2017 Graphs of Hospital Findings: Palliative Care Program Features. National Palliative Care Registry. https://registry.capc.org/metrics-resources/summary-data/. Published 2017. Accessed July 23, 2018. [Google Scholar]
- 10.Hua M, Guo L, Ing C, Lackraj D, Wang S, Morrison RS. Specialist Palliative Care Use and End-of-Life Care in Patients with Metastatic Cancer. J Pain Symptom Manage. 2024. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43(11):1130–1139. [DOI] [PubMed] [Google Scholar]
- 12.Hua M, Li G, Clancy C, Morrison RS, Wunsch H. Validation of the V66.7 Code for Palliative Care Consultation in a Single Academic Medical Center. J Palliat Med. 2017;20(4):372–377. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Cassel JB. What Does “Palliative Care” Represent in Research Using Secondary Data? J Pain Symptom Manage. 2024;68(4):421–428. [DOI] [PubMed] [Google Scholar]
- 14.Damschroder LJ, Aron DC, Keith RE, Kirsh SR, Alexander JA, Lowery JC. Fostering implementation of health services research findings into practice: a consolidated framework for advancing implementation science. Implement Sci. 2009;4(1):50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.PPS-Exempt Cancer Hospital Quality Reporting (PCHQR) Program Measures. https://qualitynet.cms.gov/pch/pchqr/measures. Accessed November 7th, 2024.
- 16.American Society of Clinical Oncology Quality Measures Library. https://old-prod.asco.org/practice-patients/quality-measures/measures-library/national-quality-forum. Accessed November 7th, 2024.
- 17.Baron RM, Kenny DA. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Soc Psychol. 1986;51(6):1173–1182. [DOI] [PubMed] [Google Scholar]
- 18.Holm S A Simple Sequentially Rejective Multiple Test Procedure. Scand J Stat. 1979;6(2):65–70. [Google Scholar]
- 19.Hua M, Guo L, Ing C, Wang S, Morrison RS. Variation in Palliative Care Program Performance for Patients with Metastatic Cancer. J Pain Symptom Manage. 2024. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Tartaglione EV, Vig EK, Reinke LF. Bridging the Cultural Divide Between Oncology and Palliative Care Subspecialties: Clinicians’ Perceptions on Team Integration. The American journal of hospice & palliative care. 2018;35(7):978–984. [DOI] [PubMed] [Google Scholar]
- 21.Hua M, Fonseca LD, Morrison RS, Wunsch H, Fullilove R, White DB. What Affects Adoption of Specialty Palliative Care in Intensive Care Units: A Qualitative Study. J Pain Symptom Manage. 2021;62(6):1273–1282. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Sagin A, Kirkpatrick JN, Pisani BA, Fahlberg BB, Sundlof AL, O’Connor NR. Emerging Collaboration Between Palliative Care Specialists and Mechanical Circulatory Support Teams: A Qualitative Study. J Pain Symptom Manage. 2016;52(4):491–497.e491. [DOI] [PubMed] [Google Scholar]
- 23.Finney JW, Humphreys K, Kivlahan DR, Harris AH. Why health care process performance measures can have different relationships to outcomes for patients and hospitals: understanding the ecological fallacy. American journal of public health. 2011;101(9):1635–1642. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.