Abstract
Background: It is not clear whether use of specialty palliative care consults and “comfort measures only” (CMO) order sets differ by type of intensive care unit (ICU). A better understanding of palliative care provided to these patients may help address heterogeneity of care across ICU types.
Objectives: Examine utilization of specialty palliative care consultation and CMO order sets across several different ICU types in a multihospital academic health care system.
Design: Retrospective cohort study using Washington State death certificates and data from the electronic health record.
Setting/Subjects: Adults with a chronic medical illness who died in an ICU at one of two hospitals from July 2013 through December 2018. Five ICU types were identified by patient population and attending physician specialty.
Measurements: Documentation of a specialty palliative care consult during a patient's terminal ICU stay and a CMO order set at time of death.
Results: For 2706 eligible decedents, ICU type was significantly associated with odds of palliative care consultation (p < 0.001) as well as presence of CMO order set at time of death (p < 0.001). Compared with medical ICUs, odds of palliative care consultation were highest in the cardiothoracic ICU and trauma ICU. Odds of CMO order set in place at time of death were highest in the neurology/neurosurgical ICU.
Conclusion: Utilization of specialty palliative care consultations and CMO order sets varies across types of ICUs. Examining this variability within institutions may provide an opportunity to improve end-of-life care for patients with chronic, life-limiting illnesses who die in the ICU.
Keywords: comfort measures, intensive care unit, palliative care, utilization
Introduction
Patients admitted to the intensive care unit (ICU) with an underlying chronic, life-limiting illness are at a relatively high risk of death, and many have substantial palliative care needs.1 These palliative care needs may be addressed by ICU clinicians providing primary palliative care, by palliative care clinicians providing specialty palliative care, or by some combination thereof.2 Although the role of specialty palliative care has been well established during the past decade, with consultations occurring earlier in the trajectory of a patient's illness and for a broader set of diagnoses,3–5 it is unclear how patterns of specialty palliative care utilization have evolved in the critical care environment. Although different ICU services may have different cultural practices around decision making at the end of life,6 the need for specialty palliative care has been recognized in both medical ICUs (MICUs) and surgical ICUs (SICUs).7–10 There is limited literature examining variability in end-of-life decision making and patterns of utilization for palliative care consultation by type of ICU service.6,11–13 Variability in involvement of specialty palliative care in the ICU may be problematic if it results in significant differences in the quality of care provided, particularly at the end of life.
In addition to the increasing use of specialty palliative care consultations, there has also been growing recognition that aggressive resuscitative measures are not appropriate or desired by many seriously ill patients nearing the end of life.14–16 Differences in the quality of primary palliative care (palliative care provided by clinicians who are not palliative care specialists) and the willingness to involve specialty palliative care may influence the application of intensive interventions at the end of life.17 During the past decade, efforts to improve the quality of palliative and end-of-life care in the ICU have included implementation of specific order sets related to withdrawing or withholding of life-sustaining treatments,18–20 often referred to as “comfort measures only” (CMO) order sets. There is a paucity of literature on the frequency of use of these order sets as well as on variability in their use across types of ICUs.
We conducted a retrospective cohort study with the objective of examining the variability in utilization of specialty palliative care consults and CMO order sets between several different ICU types in a multihospital academic health care system. To reduce patient heterogeneity, our patient population was composed of chronically ill patients who died during their ICU stay.
Methods
Setting and study population
The study population consisted of adult patients (age 18 years and older) with a chronic medical illness who died in Washington State from July 2013 through December 2018 in an ICU at University of Washington Medical Center (UWMC) or Harborview Medical Center (HMC). These patients were identified by using data from Washington State Death Certificates and from UW Medicine electronic health records (EHR). Chronic medical illness was defined by documentation in the EHR of the presence of at least one of the nine Dartmouth Atlas chronic conditions during the 24 months before death. These conditions are cancers with poor prognosis (primary malignancies with poor prognoses, leukemias, and metastatic disease), chronic lung disease, coronary artery disease, congestive heart failure, peripheral vascular disease, chronic renal failure, severe chronic liver disease, diabetes with end-organ damage, and dementia.21 The presence of a chronic medical condition was based on ICD-10 codes abstracted from the EHR (see Supplementary Appendix SA1 for a complete list of ICD-10 codes). Location of death was determined by EHR designation of the patient's physical location in an ICU bed in one of the two designated hospitals at the time of death. ICU type was determined by the primary service listed in the EHR at the time of death. To focus on patients who were most appropriate for new palliative care consults during their terminal ICU stay, we excluded 58 patients who had been seen by specialty palliative care clinicians before their ICU admission. University of Washington affiliated hospitals have an electronic order set titled “Comfort Measures Only,” which can be used when the care team, family, and patient have made the decision to withhold or withdraw life-sustaining measures and instead focus on optimizing comfort during a patient's last days. Widespread EHR documentation of this order set began in July 2013, thereby establishing the start of the study period. The study period ended on December 31, 2018—the last date for which death certificates were available at the time of analyses. The university institutional review board assessed this study as not involving human subjects, because all patients were deceased, and a waiver of Health Insurance Portability and Accountability Act (HIPAA) consent was approved, as required by Washington State law.
The UWMC is a quaternary care center for the surrounding regions and has 479 acute care beds and 79 intensive care beds. The adult ICUs at UWMC include a MICU (which also treats neurology patients), a SICU, a cardiac ICU (CCU), a cardiothoracic ICU (CT ICU), and an oncology/bone marrow transplant ICU (BMT ICU). HMC is the county safety-net hospital that is operated by the University of Washington. It is the only Level 1 Trauma Center in Washington State and serves five surrounding states. HMC has 413 acute care beds and 89 intensive care beds. The adult ICUs present at HMC and included in this study are a trauma/surgical ICU (TICU), a neurology/neurosurgery ICU (NICU), a MICU, and a CCU. Table 1 lists the specialties of the attending physicians who staff each ICU at each institution.
Table 1.
Intensive Care Units and Physician Specialties by Study Site
| UW Medicine study sites | ICU type | Specialty of staffing attending physicians |
|---|---|---|
| UWMC-M | MICU | Pulmonary/critical care |
| BMT ICU | Pulmonary/critical care | |
| SICU | Anesthesia/critical care, pulmonary/critical care | |
| CCU | Cardiology | |
| CT ICU | Anesthesia/critical care, cardiothoracic surgery, emergency/critical care | |
| HMC | TICU | Anesthesia/critical care, emergency/critical care, trauma surgery, pulmonary/critical care |
| NICU | Anesthesia/critical care, emergency/critical care, neurology/critical care, pulmonary/critical care | |
| MICU | Pulmonary/critical care | |
| CCU | Cardiology (with ventilator management done by Pulmonary Consult Team) |
BMT ICU, bone marrow transplant intensive care unit; CCU, cardiac intensive care unit; CT ICU, cardiothoracic intensive care unit; HMC, Harborview Medical Center; ICU, intensive care unit; MICU, medical intensive care unit; NICU, neurologic intensive care unit; SICU, surgical intensive care unit; TICU, trauma/surgical intensive care unit; UWMC-M, University of Washington Medical Center—Montlake.
Variables
Our outcomes included EHR documentation of a new specialty palliative care consult during a patient's terminal ICU stay and documentation of a “Comfort Measures Only” order set by the time of death. The predictor was ICU type, grouped into categories based on patient population and the specialties of attending physicians, which resulted in five categories across the seven ICUs: (1) MICU/CCU/BMT ICU; (2) HMC TICU; (3) UW SICU; (4) CT ICU; and (5) Neurology/Neurosurgical ICU (NICU).
The MICU and BMT ICU were initially grouped, because the same core group of clinicians work in both ICUs at UWMC. The CCU was added to limit the number of distinct ICU categories after analysis showed that the CCU was not significantly different than the MICU in terms of placement of palliative care consults (after adjusting for confounders). Additional rationale for this grouping includes a common background in internal medicine for clinicians providing care in these ICUs. This combination of ICUs was chosen to be the reference group, as it contained the largest number of study participants (n = 1617). The SICU at UWMC primarily cares for post-operative patients who underwent elective procedures and recent transplant patients. The TICU at HMC primarily cares for trauma patients with fewer patients post-elective surgery. Due to their very different patient populations, these ICUs were separated for the analysis.
Seven variables were selected a priori as potential confounders of the associations between the predictor and our outcomes: age, gender, level of education (eight ordinal categories), racial/ethnic minority status (white non-Hispanic vs. minority), cancer diagnosis, dementia diagnosis, and number of comorbidities. The variables “level of education” and “racial/ethnic minority status” were the two variables with the most missing data, at 10.6% and 0.8% of patients, respectively.
Statistical analysis
Descriptive statistics were used to characterize patients in each of the five ICU groups. Associations between the five ICU types and each outcome were tested by using multipredictor logistic regression models with robust standard errors. ICU type was modeled with four dummy indicators, using the MICU/CCU/BMT ICU group as the reference group. Preliminary tests were used to identify confounders of associations of ICU type with each outcome. A variable was retained as a possible confounder if its addition to the unadjusted model changed the coefficient for any of the ICU-type indicators by 10% or more. Then, to preserve the maximum sample size, we did an additional confounder test for the variables that had missing data. If either education or racial/ethnic minority status remained as a possible confounder, we compared the model with only the other confounders to the model that included the variable with missing data. If adding the variable changed the coefficient for any ICU-type indicator by 10% or more, we retained it in the final set of covariates; otherwise, we retained only the confounders that had no missing data.
Confounder testing
Of the two potential confounders that included missing data, only education showed preliminary evidence of confounding. However, when education was added to a model that included the remaining confounders as covariates, the coefficients for ICU type changed <6%, suggesting that much of the apparent confounding by education overlapped with confounding by other variables. Therefore, the final models for both outcomes were based on the full set of 2706 patients and included covariate adjustment for patient age, cancer diagnosis, and number of comorbidities.
Results
We identified 2706 eligible decedents who died in an ICU during the study period (Table 2). The median age at death was 66 years. The sample was predominantly white (75.9%), and the majority were male (63.1%). Most of the sample had less than a college degree (71.7% of those whose education level was known). The most common comorbidity was heart failure (45.1%) and the least common was dementia (5.6%). The median number of comorbidities per patient was 2, with higher medians observed in patients in the CT ICU and CCU (3 for both). Overall, 36.1% of the cohort received a palliative care consultation before death and 72.2% died with a “comfort measures only” order set in place. A minority of patients (8.3%) were full code at the time of death.
Table 2.
Sample Characteristics by Intensive Care Unit
| Characteristic | ICU type |
Total sample, n = 2706 | ||||||
|---|---|---|---|---|---|---|---|---|
| MICU, n = 937 | BMT ICU, n = 330 | CCU, n = 350 | NICU, n = 447 | SICU, n = 101 | TICU, n = 318 | CT ICU, n = 223 | ||
| Age at deatha | ||||||||
| Mean (SD) | 61.9 (14.6) | 57.6 (14.9) | 67.5 (14.3) | 69.7 (14.0) | 63.7 (16.0) | 69.9 (13.9) | 61.2 (15.4) | 64.4 (15.1) |
| Median (IQR) | 63.0 (17.2) | 60.3 (20.2) | 68.0 (16.4) | 71.6 (18.0) | 66.6 (21.1) | 72.5 (18.3) | 64.8 (18.8) | 66.0 (19.2) |
| Female | 375 (40.0) | 139 (42.1) | 111 (31.7) | 171 (38.3) | 42 (41.6) | 91 (28.6) | 69 (30.9) | 998 (36.9) |
| Race | ||||||||
| Unreported | 4 (0.4) | 1 (0.3) | 5 (1.4) | 4 (0.9) | 0 (0.0) | 4 (1.3) | 3 (1.3) | 21 (0.8) |
| White | 666 (71.1) | 279 (84.5) | 264 (75.4) | 337 (75.4) | 79 (78.2) | 251 (78.9) | 177 (79.4) | 2053 (75.9) |
| Minority | 267 (28.5) | 50 (15.2) | 81 (23.1) | 106 (23.7) | 22 (21.8) | 63 (19.8) | 43 (19.3) | 632 (23.4) |
| Black | 89 (9.5) | 12 (3.6) | 25 (7.1) | 22 (4.9) | 5 (5.0) | 19 (6.0) | 14 (6.3) | 186 (6.9) |
| Native American | 27 (2.9) | 3 (0.9) | 11 (3.1) | 13 (2.9) | 2 (2.0) | 17 (5.3) | 3 (1.3) | 76 (2.8) |
| Asian | 89 (9.5) | 19 (5.8) | 27 (7.7) | 47 (10.5) | 7 (6.9) | 15 (4.7) | 17 (7.6) | 221 (8.2) |
| Pacific islander | 7 (0.7) | 3 (0.9) | 4 (1.1) | 5 (1.1) | 1 (1.0) | 3 (0.9) | 2 (0.9) | 25 (0.9) |
| Hispanic | 33 (3.5) | 8 (2.4) | 5 (1.4) | 12 (2.7) | 1 (1.0) | 6 (1.9) | 2 (0.9) | 67 (2.5) |
| Mixed race | 22 (2.4) | 5 (1.5) | 9 (2.6) | 7 (1.6) | 6 (5.9) | 3 (0.9) | 5 (2.2) | 57 (2.1) |
| Education | ||||||||
| 8th Grade or less | 40 (4.3) | 3 (0.9) | 13 (3.7) | 29 (6.5) | 3 (3.0) | 14 (4.4) | 6 (2.7) | 108 (4.0) |
| Some high school, no diploma | 83 (8.9) | 7 (2.1) | 22 (6.3) | 29 (6.5) | 4 (4.0) | 28 (8.8) | 12 (5.4) | 185 (6.8) |
| High school graduate or equivalent | 286 (30.5) | 60 (18.2) | 94 (26.9) | 152 (34.0) | 24 (23.8) | 98 (30.8) | 64 (28.7) | 778 (28.8) |
| Some college, no degree | 157 (16.8) | 47 (14.2) | 69 (19.7) | 62 (13.9) | 20 (19.8) | 56 (17.6) | 41 (18.4) | 452 (16.7) |
| Associate's degree | 76 (8.1) | 31 (9.4) | 26 (7.4) | 27 (6.0) | 11 (10.9) | 23 (7.2) | 17 (7.6) | 211 (7.8) |
| Bachelor's degree | 118 (12.6) | 96 (29.1) | 50 (14.3) | 75 (16.8) | 19 (18.8) | 32 (10.1) | 39 (17.5) | 429 (15.9) |
| Master's degree | 52 (5.5) | 38 (11.5) | 25 (7.1) | 23 (5.1) | 8 (7.9) | 17 (5.3) | 13 (5.8) | 176 (6.5) |
| Doctoral/professional degree | 25 (2.7) | 20 (6.1) | 10 (2.9) | 5 (1.1) | 4 (4.0) | 7 (2.2) | 9 (4.0) | 80 (3.0) |
| Unreported | 100 (10.7) | 28 (8.5) | 41 (11.7) | 45 (10.1) | 8 (7.9) | 43 (13.5) | 22 (9.9) | 287 (10.6) |
| Comorbiditiesb | ||||||||
| Cancer with poor prognosis | 192 (20.5) | 302 (91.5) | 25 (7.1) | 64 (14.3) | 51 (50.5) | 21 (6.6) | 24 (10.8) | 679 (25.1) |
| Chronic lung disease | 345 (36.8) | 61 (18.5) | 81 (23.1) | 107 (23.9) | 22 (21.8) | 96 (30.2) | 69 (30.9) | 781 (28.9) |
| Coronary artery disease | 306 (32.7) | 76 (23.0) | 283 (80.9) | 188 (42.1) | 32 (31.7) | 144 (45.3) | 146 (65.5) | 1175 (43.4) |
| Congestive heart failure | 370 (39.5) | 100 (30.3) | 300 (85.7) | 144 (32.2) | 22 (21.8) | 122 (38.4) | 163 (73.1) | 1221 (45.1) |
| Peripheral vascular disease | 136 (14.5) | 29 (8.8) | 108 (30.9) | 56 (12.5) | 19 (18.8) | 111 (34.9) | 96 (43.0) | 555 (20.5) |
| Severe chronic liver disease | 300 (32.0) | 19 (5.8) | 28 (8.0) | 29 (6.5) | 22 (21.8) | 29 (9.1) | 17 (7.6) | 444 (16.4) |
| Diabetes with end-organ damage | 222 (23.7) | 48 (14.5) | 109 (31.1) | 119 (26.6) | 28 (27.7) | 71 (22.3) | 39 (17.5) | 636 (23.5) |
| Chronic renal failure | 333 (35.5) | 67 (20.3) | 131 (37.4) | 95 (21.3) | 30 (29.7) | 94 (29.6) | 73 (32.7) | 823 (30.4) |
| Dementia | 68 (7.3) | 5 (1.5) | 17 (4.9) | 33 (7.4) | 6 (5.9) | 19 (6.0) | 3 (1.3) | 151 (5.6) |
| No. of comorbiditiesb | ||||||||
| Mean (SD) | 2.4 (1.3) | 2.1 (1.1) | 3.1 (1.3) | 1.9 (1.1) | 2.3 (1.3) | 2.2 (1.2) | 2.8 (1.4) | 2.4 (1.3) |
| Median (IQR) | 2 (2) | 2 (2) | 3 (2) | 1 (2) | 2 (2) | 2 (2) | 3 (2) | 2 (2) |
| Year of death | ||||||||
| 2013 (July through December) | 99 (10.6) | 27 (8.2) | 14 (4.0) | 36 (8.1) | 8 (7.9) | 22 (6.9) | 21 (9.4) | 227 (8.4) |
| 2014 | 166 (17.7) | 66 (20.0) | 59 (16.9) | 95 (21.3) | 20 (19.8) | 73 (23.0) | 34 (15.2) | 513 (19.0) |
| 2015 | 193 (20.6) | 60 (18.2) | 76 (21.7) | 75 (16.8) | 16 (15.8) | 65 (20.4) | 51 (22.9) | 536 (19.8) |
| 2016 | 171 (18.2) | 68 (20.6) | 57 (16.3) | 95 (21.3) | 23 (22.8) | 59 (18.6) | 42 (18.8) | 515 (19.0) |
| 2017 | 159 (17.0) | 54 (16.4) | 77 (22.0) | 71 (15.9) | 15 (14.9) | 55 (17.3) | 29 (13.0) | 460 (17.0) |
| 2018 | 149 (15.9) | 55 (16.7) | 67 (19.1) | 75 (16.8) | 19 (18.8) | 44 (13.8) | 46 (20.6) | 455 (16.8) |
| ICU length of stay | ||||||||
| Mean (SD) | 10.4 (16.9) | 7.7 (11.6) | 10.5 (19.7) | 7.5 (10.7) | 14.7 (31.8) | 10.1 (13.2) | 16.0 (24.6) | 10.2 (17.2) |
| Median (IQR) | 5.0 (10.0) | 4.0 (7.0) | 5.0 (9.0) | 4.0 (7.0) | 6.0 (12.0) | 6.0 (10.0) | 8.0 (16.0) | 5.0 (10.0) |
| Code status at death | ||||||||
| Full code | 46 (4.9) | 19 (5.8) | 63 (18.0) | 11 (2.5) | 12 (11.9) | 19 (6.0) | 55 (24.7) | 235 (8.3) |
| DNR/DNI or DNI | 731 (78.0) | 251 (76.1) | 246 (70.3) | 392 (87.7) | 68 (67.3) | 231 (72.6) | 129 (57.8) | 2048 (75.7) |
| DNR/intubation OK | 131 (14.0) | 57 (17.3) | 16 (4.6) | 42 (9.4) | 11 (10.9) | 45 (14.2) | 18 (8.1) | 320 (11.8) |
| Status not documented | 29 (3.1) | 3 (0.9) | 25 (7.1) | 2 (0.4) | 10 (9.9) | 23 (7.2) | 21 (9.4) | 113 (4.2) |
| Palliative care consultation | 313 (33.4) | 161 (48.8) | 133 (38.0) | 73 (16.3) | 34 (33.7) | 136 (42.8) | 127 (57.0) | 977 (36.1) |
| CMO orders | 700 (74.7) | 258 (78.2) | 182 (52.0) | 372 (83.2) | 68 (67.3) | 234 (73.6) | 140 (62.8) | 1954 (72.2) |
Unless otherwise specified, each cell includes the number of patients (% of column total) with the characteristic.
All distributions of age, number of comorbidities, and length of stay were significantly non-normal, based on both the Kolmogorov-Smirnov and Shapiro-Wilk tests, making the median/IQR better representations of central tendency and dispersion.
Based on ICD codes specified in Dartmouth Atlas with a few codes for lung disease and dementia excluded.
CMO, comfort measures only; DNI, do-not-intubate order; DNR, do-not-resuscitate order; IQR, interquartile range; SD, standard deviation.
Palliative care consultation
The patient's ICU type at death was significantly associated with odds of palliative care consultation during the patient's terminal ICU stay (p < 0.001; Table 3 and Fig. 1). When compared with the MICU/CCU/BMT ICU group, odds of receiving a palliative care consultation before death were highest in the CT ICU (odds ratio [OR] 2.30; 95% confidence interval [CI] 1.72–3.07; p < 0.001) and trauma ICU (OR 1.73; 95% CI 1.33–2.25; p < 0.001) and lowest in the Neurology/Neurosurgical ICU (OR 0.44; 95% CI 0.34–0.59; p < 0.001).
Table 3.
Association of Intensive Care Unit Type and End-of-Life Outcomes
| Outcome | Predictors | Confounders | Β | OR | 95% CI | p |
|---|---|---|---|---|---|---|
| Palliative care | ICU type | <0.001 | ||||
| MICU, CCU, BMT ICU | 0.000 | 1.000 | ||||
| TICU | 0.546 | 1.726 | 1.326–2.248 | <0.001 | ||
| SICU | −0.191 | 0.826 | 0.533–1.280 | 0.392 | ||
| CT ICU | 0.832 | 2.297 | 1.719–3.068 | <0.001 | ||
| NICU | −0.813 | 0.444 | 0.335–0.587 | <0.001 | ||
| Cancer | 0.420 | 1.522 | 1.256–1.845 | <0.001 | ||
| No. of comorbidities | 0.169 | 1.185 | 1.110–1.264 | <0.001 | ||
| Age | −0.021 | 0.979 | 0.974–0.985 | <0.001 | ||
| Comfort-measures-only orders | ICU type | <0.001 | ||||
| MICU, CCU, BMT ICU | 0.000 | 1.000 | ||||
| TICU | 0.160 | 1.174 | 0.886–1.555 | 0.264 | ||
| SICU | −0.237 | 0.789 | 0.509–1.223 | 0.290 | ||
| CT ICU | −0.251 | 0.778 | 0.576–1.050 | 0.101 | ||
| NICU | 0.692 | 1.997 | 1.513, 2.637 | <0.001 | ||
| Cancer | 0.342 | 1.408 | 1.145–1.730 | 0.001 | ||
| No. of comorbidities | −0.065 | 0.937 | 0.877–1.002 | 0.058 | ||
| Age | 0.008 | 1.008 | 1.002, 1.014 | 0.011 |
Results for each outcome were based on a robust logistic regression model based on 2706 patients who died in an ICU. Each model included the predictor and outcome of interest, along with the three variables that confounded the association between ICU type and outcome (given in italics).
CI, confidence interval; OR, odds ratio.
FIG. 1.
Proportion of patients who received specialty palliative care consultation during their terminal ICU stay by unit type. *Reference group. **Statistically significantly different from reference group (p < 0.001). HMC, Harborview Medical Center; ICU, intensive care unit; UWMC-M, University of Washington Medical Center—Montlake.
“Comfort measures only” order set
ICU type was also significantly associated with presence of the “comfort measures only” order set at time of death (p < 0.001; Table 3 and Fig. 2). When compared with the MICU/CCU/BMT ICU group, odds of having “comfort measures only” orders were highest in the Neurology/Neurosurgical ICU (OR 2.00; 95% CI 1.51–2.64; p < 0.001). Differences between the other ICU types and the reference group were not statistically significant.
FIG 2.
Proportion of patients who had a CMO order set in place at time of death by unit type. *Reference group. **Statistically significantly different from reference group (p < 0.001). CMO, comfort measures only.
Discussion
In this retrospective cohort study, we assessed the variability of specialty palliative care consultations and CMO order set use at the end of life across types of ICUs. There is growing recognition of the importance of palliative care across various ICU types, especially in units underrepresented in the current literature, such as CCUs, trauma ICUs, and neurosurgical ICUs.7,9,10 Although patient characteristics differ across ICU type by diagnosis, prognosis, and treatment plan, they also share certain features. By focusing on patients with chronic, life-limiting illness who died during the ICU stay, we hoped to highlight a patient population with more pronounced palliative care needs, regardless of the initial indication for or location of ICU admission. Although prior studies have shown variability in end-of-life practices in ICUs both nationally22–25 and internationally,26 none have looked at differences in care across types of ICUs nor at the use of palliative care consults and CMO order sets as a means of understanding variations in end-of-life care across ICU types. Our results demonstrated significant variation in the use of specialty palliative care and CMO order sets by ICU type.
When compared with our group of MICUs (MICU/CCU/BMT ICU), the odds of having a palliative care consult placed during a terminal ICU stay were higher in the CT ICU and TICU. These findings are at odds with our hypothesis; we expected that surgically oriented ICUs would use less specialty palliative care. Prior research has shown that when surgeons have primary responsibility for patients, a “covenantal” ethic is likely to be followed in which defeating death is seen as the primary goal.27 Also, prior surgical literature on barriers to palliative consultations in trauma ICUs noted the fear of being seen as “giving up.”12 Finally, historically, surgical services and their attendings have been found to have comparatively lower rates of specialty palliative care consultation.11,13
We offer several potential explanations for our unexpected finding of relatively higher odds of palliative care consultation in the trauma and CT ICUs as compared with the MICUs. First, over the past decade, there has been growing awareness and utilization of palliative care consultations in general,4,5 including an increasing number of publications on the utility and underutilization of palliative care in surgical and trauma services.8,10,12,28 This growing awareness of and improved perception of palliative care among surgeons, who may have less time to invest in long palliative care conversations, may have led to a higher rate of specialty palliative care consultations in these surgically oriented ICUs. It is also possible that clinicians in MICUs are more comfortable providing primary palliative care and are less likely to request specialty palliative care consultation than their surgical colleagues. When using MICUs as the reference group, as we did here, this difference would appear as a higher odds of palliative care consultation among surgical services. Of note, the same increased odds of specialty palliative care consultation were not seen in the UWMC SICU, possibly because this service is traditionally staffed by a higher proportion of Pulmonary/Critical Care trained providers than the other surgically oriented ICUs. Finally, the UW Medicine system has historically been an early adopter of palliative care and has developed programs to enhance primary palliative care in the ICU.29,30 Although this may be a system-level factor that influenced our findings, it would not explain why SICUs have proportionally higher consults than other ICUs.
We also found lower odds of palliative care consultation and higher odds of “comfort measures only” order set utilization in the neurology/neurosurgical ICU. Previous studies have demonstrated high palliative care needs in this patient population.7 One possible reason for less involvement of specialty palliative care would be that cases at the extreme (e.g., severe traumatic brain injury or large devastating strokes) are easier to prognosticate and may reduce the need for complex goals-of-care discussions that necessitate the skills of specialty palliative care providers. Prior work has suggested that for patients who present with sudden advanced illness, providers were less likely to consider a palliative care consult, especially for patients with “imminent death early in patient course.”31 Another institutionally specific factor is that the neurology/neurosurgical ICU utilizes senior (nonoperating) neurosurgeons to assist with family communication and goals-of-care discussions. This unique model may decrease the subsequent need for specialty palliative care consultation. Similar models may be worth considering at other institutions where specialty palliative care consultation capacity is limited.
This study has several important limitations. First, decedents were drawn from a single health care system in a single geographical region where efforts to increase primary and specialty palliative care involvement have been robust and sustained. However, there are many other institutions across the United States who have made significant strides in the integration of palliative care into their hospitals. Second, study inclusion was determined by both location of death (defined as the patient's physical location in an ICU bed at the time of death) and the primary service listed in the EHR at the time of death. Due to the way the data were coded in the EHR, it is possible that some patients included in this study were physically located in an ICU, but not cared for by an ICU team (i.e., recently transferred to a floor service but still physically located in an ICU bed). Although this type of misclassification may have occurred, we suspect it to be minimal in impact and distributed relatively evenly across the various ICU types. Third, this study examines decedents and might raise concerns about a decedent bias.32 However, since our research question focuses specifically on end-of-life care in the ICU, this is less of a concern. Finally, we do not have data on the exact time CMO orders or palliative care consults were placed in relation to the time of death of a patient. This would be important to know, as consults or order sets placed immediately before death may not have been as helpful to the patient as those placed earlier on in their ICU stay.
Conclusions
This study shows that practice patterns for utilization of specialty palliative care consultations and “comfort measures only” order sets vary across different types of ICUs, even when the patient population is limited to chronically ill patients who died in the ICU. Although some of this variation may be related to the different comorbidities and reasons for ICU admission present in the different types of ICUs, it is likely that individual provider practices, unique unit cultures, changing attitudes toward specialty palliative care, and increasing perceived and real expertise in primary palliative care play an important role in the observed differences.11 This type of culture- and clinician-driven variability may not necessarily indicate poor quality of care. The quality of end-of-life care as delivered by different ICUs will necessarily reflect factors at the institutional level that drive and determine care. Additional study of these topics in a variety of geographic regions would be helpful to assess regional or national trends. Hospital systems should be encouraged to examine their own system for variability in palliative care and CMO order set usage. Examining reasons for this variability may provide insight into opportunities to improve end-of-life care in the ICU for patients with chronic, life-limiting illness.
Supplementary Material
Funding Information
This work was supported by the National Institutes of Health (Grants T32HL07287 [J.D.L.] and K23HL144830 [N.K.]) and the Cambia Health Foundation.
Author Disclosure Statement
No competing financial interests exist.
Supplementary Material
References
- 1. Fitzsimons D, Mullan D, Wilson JS, et al. : The challenge of patients' unmet palliative care needs in the final stages of chronic illness. Palliat Med 2007;21:313–322 [DOI] [PubMed] [Google Scholar]
- 2. Nelson JE, Bassett R, Boss RD, et al. : Models for structuring a clinical initiative to enhance palliative care in the intensive care unit: A report from the IPAL-ICU Project (Improving Palliative Care in the ICU). Crit Care Med 2010;38:1765–1772 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Dumanovsky T, Augustin R, Rogers M, et al. : The growth of palliative care in U.S. hospitals: A status report. J Palliat Med 2016;19:8–15 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Hughes MT, Smith TJ: The growth of palliative care in the United States. Annu Rev Public Health 2014;35:459–475 [DOI] [PubMed] [Google Scholar]
- 5. Schoenherr LA, Bischoff KE, Marks AK, et al. : Trends in hospital-based specialty palliative care in the United States from 2013 to 2017. JAMA Netw Open 2019;2:e1917043. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Baggs JG, Norton SA, Schmitt MH, et al. : Intensive care unit cultures and end-of-life decision making. J Crit Care 2007;22:159–168 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Creutzfeldt CJ, Engelberg RA, Healey L, et al. : Palliative care needs in the neuro-ICU: Crit Care Med 2015;43:1677–1684 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Mosenthal AC, Weissman DE, Curtis JR, et al. : Integrating palliative care in the surgical and trauma intensive care unit: A report from the Improving Palliative Care in the Intensive Care Unit (IPAL-ICU) Project Advisory Board and the Center to Advance Palliative Care. Crit Care Med 2012;40:1199–1206 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Romano’ M: The role of palliative care in the cardiac intensive care unit. Healthcare (Basel) 2019;7:30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Toevs CC: Palliative medicine in the surgical intensive care unit and trauma. Surg Clin North Am 2011;91:325–331 [DOI] [PubMed] [Google Scholar]
- 11. Kross EK, Engelberg RA, Downey L, et al. : Differences in end-of-life care in the ICU across patients cared for by medicine, surgery, neurology, and neurosurgery physicians. Chest 2014;145:313–321 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Karlekar M, Collier B, Parish A, et al. : Utilization and determinants of palliative care in the trauma intensive care unit: Results of a national survey. Palliat Med 2014;28:1062–1068 [DOI] [PubMed] [Google Scholar]
- 13. Rodriguez R, Marr L, Rajput A, Fahy BN: Utilization of palliative care consultation service by surgical services. Ann Palliat Med 2015;4:194–199 [DOI] [PubMed] [Google Scholar]
- 14. Prendergast TJ, Luce JM: Increasing incidence of withholding and withdrawal of life support from the critically ill. Am J Respir Crit Care Med 1997;155:15–20 [DOI] [PubMed] [Google Scholar]
- 15. Prendergast TJ, Claessens MT, Luce JM: A national survey of end-of-life care for critically ill patients. Am J Respir Crit Care Med 1998;158:1163–1167 [DOI] [PubMed] [Google Scholar]
- 16. Rubin EB, Buehler A, Halpern SD: Seriously ill patients' willingness to trade survival time to avoid high treatment intensity at the end of life. JAMA Intern Med 2020;180:907–909 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Lustbader D, Pekmezaris R, Frankenthaler M, et al. : Palliative medicine consultation impacts DNR designation and length of stay for terminal medical MICU patients. Palliat Support Care 2011;9:401–406 [DOI] [PubMed] [Google Scholar]
- 18. Bailey FA, Williams BR, Woodby LL, et al. : Intervention to improve care at life's end in inpatient settings: The BEACON trial. J Gen Intern Med 2014;29:836–843 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Walling AM, Ettner SL, Barry T, et al. : Missed opportunities: Use of an end-of-life symptom management order protocol among inpatients dying expected deaths. J Palliat Med 2011;14:407–412 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Walling AM, Brown-Saltzman K, Barry T, et al. : Assessment of implementation of an order protocol for end-of-life symptom management. J Palliat Med 2008;11:857–865 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Iezzoni LI, Heeren T, Foley SM, et al. : Chronic conditions and risk of in-hospital death. Health Serv Res 4;29:435–460 [PMC free article] [PubMed] [Google Scholar]
- 22. Hart JL, Harhay MO, Gabler NB, et al. : Variability among US intensive care units in managing the care of patients admitted with preexisting limits on life-sustaining therapies. JAMA Intern Med 2015;175:1019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. DeCato TW, Engelberg RA, Downey L, et al. : Hospital variation and temporal trends in palliative and end-of-life care in the ICU. Crit Care Med 2013;41:1405–1411 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Barnato AE, Herndon MB, Anthony DL, et al. : Are regional variations in end-of-life care intensity explained by patient preferences?: A study of the US Medicare population. Med Care 2007;45:386–393 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Quill CM, Ratcliffe SJ, Harhay MO, Halpern SD: Variation in decisions to forgo life-sustaining therapies in US ICUs. Chest 2014;146:573–582 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Romain M, Sprung CL: End-of-life practices in the intensive care unit: The importance of geography, religion, religious affiliation, and culture. Rambam Maimonides Med J 2014;5:e0003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Cassell J, Buchman TG, Streat S, et al. : Surgeons, intensivists, and the covenant of care: Administrative models and values affecting care at the end of life. Crit Care Med 2003;31:1263–1270 [DOI] [PubMed] [Google Scholar]
- 28. Suwanabol PA, Reichstein AC, Suzer-Gurtekin ZT, et al. : Surgeons' perceived barriers to palliative and end-of-life care: A mixed methods study of a surgical society. J Palliat Med 2018;21:780–788 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Curtis JR, Treece PD, Nielsen EL, et al. : Integrating palliative and critical care: Evaluation of a quality-improvement intervention. Am J Respir Crit Care Med 2008;178:269–275 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Curtis JR, Nielsen EL, Treece PD, et al. : Effect of a quality-improvement intervention on end-of-life care in the intensive care unit: A randomized trial. Am J Respir Crit Care Med 2011;183:348–355 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Tilden LB, Williams BR, Tucker RO, et al. : Surgeons' attitudes and practices in the utilization of palliative and supportive care services for patients with a sudden advanced illness. J Palliat Med 2009;12:1037–1042 [DOI] [PubMed] [Google Scholar]
- 32. Bach PB, Schrag D, Begg CB: Resurrecting treatment histories of dead patients: A study design that should be laid to rest. JAMA 2004;292:2765–2770 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.