Palliative Care Involvement in Relation to Rapid Response Events: A Focus on Care Intensity at the End of Life

Abstract

OBJECTIVES:

To examine current practice patterns of palliative care involvement in relation to rapid response (RR) events and to evaluate its association with ICU utilization and intensity of care received before death.

DESIGN:

Retrospective cohort study of patients who underwent a RR event.

SETTING:

General medical, cardiology, oncology, and observation patients admitted to a tertiary care academic medical center in the United States.

PATIENTS:

Adults (≥ 18 yr) with an RR event.

INTERVENTIONS:

None.

MEASUREMENTS AND MAIN RESULTS:

A total of 903 RR events occurred in 616 unique patients. The palliative care team was involved in 110 of 616 (17.9%) patients. Palliative care (PC) involvement was not associated with differences in repeat RR events or ICU transfer rates following an event. In-hospital mortality or hospice discharge occurred in 32.0% (197/616). Palliative care was involved in 36.5% (72/197) of these patients. PC involvement was associated with lower intensity of care within 24 hours before death after adjustment for age, sex, Charlson Comorbidity Index, pre-RR code status, and RR trigger (adjusted odds ratio [aOR], 0.26; 95% CI, 0.14–0.47; p < 0.001; Brant test, p = 0.44). This finding remained across sensitivity analyses for oncology status, primary service line, ICU transfer, and advanced directive documentation. PC consultation occurred only before RR in 16 of 110 (14.5%), only after RR in 63 of 110 (57.3%), and both before and after in 31 of 110 (28.2%). In timing-stratified ordinal regression, PC consultation only after RR (aOR, 0.19; 95% CI, 0.10–0.38; p < 0.001) or both before and after RR (aOR, 0.22; 95% CI, 0.09–0.56; p = 0.002) was associated with less-intense care, while pre-RR PC alone was not (aOR, 1.06; 95% CI, 0.33–3.35; p = 0.93).

CONCLUSIONS:

Palliative services may be underutilized in the RR patient population, a group with a high in-hospital mortality rate; when palliative care is consulted, it may be associated with a reduction in the intensity of interventions at the end of life.

KEY POINTS.

Question: How does palliative care involvement, and its timing relative to rapid response (RR) events, influence patient outcomes?

Findings: In this retrospective cohort of 616 patients undergoing RR events, palliative care was consulted for 17.9% of patients. The in-hospital mortality or hospice discharge was high (32.0%). Although palliative team involvement did not impact ICU transfer rate, it was associated with a decrease in the intensity of care received before death (adjusted odds ratio, 0.26).

Meaning: Operationalizing the RR event as a structured trigger for palliative care consultation may optimize palliative care utilization and improve goal-concordant care.

Since the launch of the 100,000 lives campaign, rapid response (RR) teams have been widely implemented to manage acute deterioration events outside of the ICU (1). Studies evaluating their effect on inpatient cardiac arrests and patient mortality have shown mixed results, but consistently note that many RR events occur in patients near the end of life (2, 3). This recognition has shifted the focus of RR teams from a purely rescue-oriented approach to one that also emphasizes timely goals-of-care discussions and the delivery of goal-concordant care (4, 5). Although palliative care (PC) involvement during admission is associated with increased hospice enrollment (6), its impact on the intensity of care following RR events—such as cardiopulmonary resuscitation (CPR), intubation, invasive procedures, and vasopressor use—remains poorly understood (7).

In this study, our primary objective was to describe the frequency, timing, and outcomes of PC consultations among patients undergoing RR events. Our secondary objective was to evaluate the association between PC involvement and intensity of care received in the 24 hours before death.

MATERIALS AND METHODS

This retrospective cohort study identified adult patients (> 18 yr) admitted to general medical, oncology, cardiology, and observation units who experienced an RR event between March 1, 2024, and August 31, 2024, at a 900-bed adult tertiary academic medical center in the Midwest. RR events were identified through hospital paging system audits and were validated using documentation in the electronic medical record. Canceled RR events and alerts that did not result in a full RR were excluded.

Data were extracted via manual chart review from the electronic medical records by authors J.A., C.N., K.T., M.L., and D.D. Data elements extracted included patient demographics, comorbidities, PC documentation dates, code status, and advance directives on file at the time of RR, ICU transfer, repeat RR events, as well as patient death or discharge to hospice. Indications for RR events were characterized as mutually exclusive groups (respiratory, blood pressure, arrhythmia, neurologic change, and multiple triggers). Timing of PC involvement was defined relative to the index RR and was categorized into three groups: before RR only, both before and after RR, and after RR only.

For all patients who died in the hospital or were discharged to hospice, a review was conducted to assess the intensity of care received before death. Intensity of care at death was defined as one of four escalating groups: 1) hospice enrollment, 2) comfort-focused care without invasive measures (intubation, dialysis initiation, central line insertion, vasopressor administration, extracorporeal membrane oxygenation), 3) receiving invasive measures but not CPR, and 4) attempted CPR. Patients were classified into the highest level of care received in the 24 hours immediately before their death.

We performed descriptive statistics to characterize event rates. Categorical variables were summarized as frequencies and compared using chi-square or Fisher exact testing. Continuous variables were reported as mean values with sd, while skewed distributions were reported as median with interquartile range (IQR) and compared using Welch’s t-test and Mann-Whitney U test, respectively. Additionally, ordinal logistic regression models were constructed to examine how PC involvement relative to the RR event was associated with the intensity of care received before death.

This study was reviewed and approved on August 14, 2024, by the Northwestern University Institutional Review Board (IRB; STU00222589: Rapid Response Protocol and Outcomes Review at a Single Academic Medical Center). Informed consent was waived for patients included in this study and was approved by the IRB. The procedures for this study were followed in accordance with the ethical standards of the responsible committee on human experimentation and with the Helsinki Declaration of 1975.

RESULTS

Between March 1, 2024, and August 31, 2024, 2126 events were identified. Of those, 903 RR activations resulted in events for 616 unique patients. Excluded events included 850 on nonmedical units, 358 that did not result in an RR, and 15 duplicate entries.

Among all RR patients, the mean age was 63 years (sd = 17), and 320 (51.9%) were female. The primary service line was predominantly general medicine (401, 65.1%), followed by oncology (124, 20.1%), observation (50, 8.1%), and cardiology (41, 6.7%). The mean Charlson Comorbidity Index (CCI) was 5.2 (sd = 2.8). ICU consultation occurred in 632 of 903 (70.0%) instances, with 442 of 903 (49.0%) requiring ICU transfer.

The PC team was involved in 110 of 616 (17.9%) RR patients during their hospitalization. The indication for PC consultation in the group was symptom management in 64 of 110 (58.2%) and assistance in advanced care planning in 79 of 110 (71.8%). Patients for whom PC was consulted had a higher CCI (5.8 ± 2.5 vs. 5.1 ± 2.8; p = 0.007), were more likely to have a solid malignancy diagnosis (49.1% vs. 20.4%; p < 0.001), were more likely to have multi-trigger RR events (27.3% vs. 17.0%; p = 0.02) and had an overall higher mortality or discharge to hospice rate (65.5% vs. 24.7%; p < 0.001). Patient code status and advanced directives available at the time of the RR event were not statistically different. PC team involvement was not associated with a change in ICU transfer rate following RR (p = 0.20) or the likelihood of repeat RR within 72 hours (p =1.00; Table 1).

TABLE 1.

Baseline Characteristics of the Rapid Response Cohort Stratified by Palliative Care Consultation During Hospitalization (N = 616)

Characteristics	Palliative Care Consultation (N = 110)	No Palliative Care Consultation (N = 506)	p
Age (yr)	64 ± 17	63 ± 17	0.58
Sex (female)	63 (57.3%)	257 (50.8%)	0.26
Service line
Medicine	61(55.5%)	340 (67.2%)
Oncology	45 (40.9%)	79 (15.6%)	< 0.001
Cardiology	3 (2.7%)	38 (7.5%)
Observation	1 (0.9%)	49 (9.7%)
Trigger for RR
Respiratory	38 (34.5%)	137 (27.1%)	0.15
Blood pressure	15 (13.6%)	115 (22.7%)	0.05
Arrythmia	12 (10.9%)	58 (11.5%)	1.00
Neurologic change	7 (6.4%)	50 (9.9%)	0.33
Othera	6 (5.5%)	53 (10.5%)	0.15
> 2 reasons	30 (27.3%)	86 (17.0%)	0.02
No trigger documented	2 (1.8%)	7 (1.4%)	0.67
Code status at time of RR
Full code	89 (80.9%)	444 (87.7%)	0.08
DNR	21 (19.1%)	62 (12.3%)	0.08
Advance directives at time of RR
HCPOA	45 (40.9%)	166 (32.8%)	0.13
POLST	7 (6.4%)	28 (5.5%)	0.91
Living will	2 (1.8%)	5 (1.0%)	0.61
Any of the above	49 (44.5%)	183 (36.2%)	0.12
Comorbidities
Charlson Comorbidity Index	5.8 ± 2.5	5.1 ± 2.8	0.007
Malignancy
Solid only	54 (49.1%)	103 (20.4%)	< 0.001
Liquid only	17 (15.5%)	66 (13.0%)	0.61
Both	4 (3.6%)	10 (2.0%)	0.29
Solid or bone marrow transplantation	8 (7.3%)	72 (14.2%)	0.07
Outcomes
ICU transfer rate	61 (55.5%)	244 (48.2%)	0.20
Repeat RR within 72 hr	9 (8.2%)	40 (7.9%)	1.00
Inpatient mortality or hospice enrollment rate	72 (65.5%)	125 (24.7%)	< 0.001

DNR = do not resuscitate, HCPOA = health care power of attorney, POLST = physician orders for life-sustaining treatment, RR = rapid response.

^aOther indications include nurse concern, arterial blood gas laboratory result outside of the ICU, transportation to the ICU, syncope, or fall. Each represented > 5% of the total indications.

Continuous variables are presented as mean ± sd and compared using Welch’s t-test. Categorical variables are presented as n (%) and compared using χ² or Fisher exact test.

The in-hospital mortality or hospice discharge rate was 32.0% (197/616). The PC team was involved in 72 of 197 (36.5%) of these patients. Among the 197 patients, 21 (10.7%) died following CPR, 88 (44.7%) died following invasive measures, 53 (26.9%) died with comfort-focused care, and 35 (17.8%) were discharged to hospice. PC was involved in 28.6%, 21.6%, 50.9%, and 57.1% of these cases, respectively (Fig. 1). Using ordinal regression, when controlled for age, sex, and CCI, code status at the time of RR, and indication for RR, PC team involvement was associated with a lower care intensity before death (aOR, 0.26; 95% CI, 0.14–0.47; p < 0.001). The proportional-odds assumption was supported by Brant’s test (p = 0.44; Supplement Table S1, https://links.lww.com/CCX/B686). Additional sensitivity analyses were performed to understand the impact of adding the presence of advance directive documentation at the time of RR, patient malignancy status, primary treating service line, and ICU transfer to the model as covariates. With each sensitivity analysis, PC consultation remained significantly associated with lower care intensity before death (Supplement Tables S2–S5, https://links.lww.com/CCX/B686).

Figure 1.

Intensity of care received before death for patients with and without palliative care involvement. Patients with palliative care involvement were more likely to transition to hospice or comfort-focused care before death (**p < 0.001). DNR = do not resuscitate.

Timing of PC involvement varied, with 16 of 110 (14.5%) occurring only before RR, 63 of 110 (57.3%) only after the RR, and 31 of 110 (28.2%) both before and after the RR (Fig. 2). The median interval from RR event to PC team involvement was +1 day post RR (IQR, –1 to 6 d), with 18.2% (20/110) of initial consultative notes occurring greater than +1 week after the RR event. Characteristics of the patients were similar regardless of consultation time. Indication for consultation differed by timing, with more patients in the before and after group being consulted for symptom management and fewer for advanced care planning. Intensity of care received was similar when stratified by indication for PC consultation. (Supplement Table S6, https://links.lww.com/CCX/B686). There was no difference between rates of ICU transfer (p = 0.92), rates of repeat RR events within 72 hours (p = 0.61), or in-hospital mortality or discharge to hospice (p = 0.64) depending on timing of consultation (Table 2).

Figure 2.

Timing of any palliative care team consultation and progress notes relative to the index rapid response event in the study cohort. RR = rapid response. RRT = rapid response team activation.

TABLE 2.

Baseline Characteristics, Consultation Details, and Hospital Outcomes of Palliative Care-Consulted Patients Stratified by Timing of Consultation (N = 110)

Characteristics	PC Consult Before RR Only (N = 16)	PC Consult After RR Only (N = 63)	PC Consult Before and After RR (N = 31)	p
Age (yr)	62.1 ± 15.4	66.5 ± 17.8	60.2 ± 16.9	0.13
Charlson Comorbidity Index	6.7 ± 2.3	5.6 ± 2.5	5.7 + 2.5	0.36
Service line
Medicine	9 (56%)	34 (54%)	18 (58%)	0.93
Oncology	6 (38%)	26 (41%)	13 (42%)	0.95
Cardiology	1 (6%)	2 (3%)	0 (0%)	0.44
Observation	0 (0%)	1 (2%)	0 (0%)	0.69
Code status at time of RR
Full code	12 (75%)	53 (84%)	24 (77%)	0.60
DNR	4 (25%)	10 (16%)	7 (23%)	0.60
Characteristics of PC consultations
Timing of initial consultation (median days from RR event)	–4 (–10 to –1)	5 (2–10)	–1 (–2 to 0)	< 0.001
Duration of PC involvement (median days)	6 (1–20)	3 (0–7)	12 (8–22)	< 0.001
Outcomes
ICU transfer rate	8 (50%)	38 (60%)	15 (48%)	0.92
Repeat RR within 72 hr	1 (6%)	5 (8%)	3 (10%)	0.61
Inpatient mortality or hospice enrollment rate	12 (75%)	41 (65%)	19 (61%)	0.64

DNR = do not resuscitate, PC = palliative care, RR = rapid response.

Continuous variables are presented as mean ± sd or median (interquartile range) and compared using the Kruskal-Wallis test. Categorical variables are presented as n (%) and compared using χ² or Fisher exact test.

Using ordinal logistic regression controlled for age, sex, CCI, and PC consultation before RR only was not associated with change in care intensity (aOR, 1.06 [0.33–3.35]; p = 0.93). PC consultation after RR only (aOR, 0.19 [0.10–0.38]; p < 0.001), as well as patients seen both before and after RR (aOR, 0.22 [0.09–0.56]; p = 0.002), were associated with less-intense care before death. The proportional-odds assumption was supported by Brant’s test (p = 0.62; Supplement Table S7, https://links.lww.com/CCX/B686).

DISCUSSION

In this single-center study, we corroborate prior literature showing that patients undergoing RR have high mortality (4, 7). Despite this nearly 32% mortality rate among RR patients—substantially higher than the approximately 20% reported for a general medical ICU population (8)—PC consultation rates were comparable in our RR cohort to the typical ICU population (9). Although our data cannot directly establish the adequacy of the currently observed consultation patterns, the discordance between higher in-hospital mortality in the RR cohort and palliative care consultation rates comparable to the ICU population raises the possibility of unmet PC needs in this population.

Prior observational and implementation studies of PC incorporation into RR systems have been associated with fewer ICU transfers, higher hospice referral rates, and improved patient, provider, and family experience (7, 10). Our study corroborates this and suggests that PC team involvement was associated with less-intense forms of care before death, possibly via facilitation of shared decision-making and goal-concordant care delivery.

Our data further suggest that the RR event may function as a clinical inflection point at which patients, families, and treating teams can meaningfully reconsider trajectory and goals. This interpretation is supported by the timing-stratified finding that PC consultations occurring only after the RR were associated with lower-intensity end-of-life care (aOR, 0.19; p < 0.001), whereas consultations occurring only before the RR were not (aOR, 1.06; p = 0.93), suggesting the RR event itself may be a temporally relevant exposure.

Given that PC services are a limited resource, utilizing the findings above, developing an improved integration of PC with RR patients may better align resources with patient need. Possible ways to facilitate this could include an automatic trigger for PC review following RR, standardization of goals-of-care documentation following RR events in high-risk patients, or an electronic health record notification to primary teams for high-risk patients to consider consultation. While many critical care and hospital medicine providers have skills in navigating goals-of-care conversations, a recent multicenter retrospective study found that goals-of-care documentation is present or updated in only 30% of critically ill patients (11). Regardless of who ultimately has the discussion, improvements are needed in the documentation and frequency of these discussions, particularly following clinical decompensation.

Our study has several limitations. First, this is a single-center retrospective study at a large academic medical center, which may limit generalizability. Oncology and transplant patients, who often have complex goals-of-care needs, represent a relatively large proportion of our cohort. Second, there is the possibility of indication bias in the sample. However, PC involvement after RR was associated with less-intense care—regardless of whether PC had been involved earlier—argues against severity-driven selection as the sole explanation and is worth exploring in future prospective and propensity-matched studies. Third, with retrospective observational studies, residual confounding cannot be completely excluded, especially with complex clinical environments and encounters. To account for this, we have performed several sensitivity analyses for potential confounding (ICU transfer, oncology status, RR trigger indication, advanced directives documentation), and none were noted to be significant confounders. Unmeasured variables that we did not adjust for include acute illness severity at the moment of RR (vital signs) or prognostic statements communicated to the patient before or after the RR event. Finally, our study did not capture PC involvement outside of RR events; upstream consultations in the outpatient setting or during prior admission may have preempted the index RR event or independently influenced care trajectories.

Many questions remain regarding understanding the intersection of RR and PC (12). Future research should evaluate whether structured PC triggers tied to RR events improve goal-concordant care while also understanding knowledge gaps and other barriers primary teams face in facilitating difficult conversations about goals-of-care in high-stress environments.

CONCLUSIONS

Patients experiencing RR events have high comorbidity burdens and risk for mortality. This population should serve as a target for ongoing investigations and interventions to ensure timely goal-concordant care. Involvement of PC teams following RR events may influence the intensity of care received by these patients at the end of their life.

ACKNOWLEDGMENTS

We acknowledge the Northwestern Medicine Canning Thoracic Institute Quality Committee and Emergency Management Quality Committees, as well as Clinical Quality Leader Dalya Durst, for their invaluable support.

Supplementary Material

cc9-8-e1489-s001.pdf ^{(196.2KB, pdf)}