Life-Supporting Treatment Limitations in Patients Who Die Within 48 Hours After ICU Admission: A French, Multicenter, Observational, Exploratory Study

Abstract

IMPORTANCE:

The occurrence of death shortly after ICU admission raises concerns about the appropriateness of providing intensive care to frail patients—many of whom are subsequently subject to decisions to limit life-supporting treatment limitation (LST-L). The proportion of patients who die early and are affected by such limitations remains unknown.

OBJECTIVES:

The primary objective was to determine the proportion of patients with a decision of LST-L among patients who died within 48 hours after ICU admission. We also conducted analyses to identify variables associated with LST-L and collected staff perceptions.

DESIGN, SETTING, AND PARTICIPANTS:

A retrospective, observational, multicenter study with data collected immediately after the patient’s death, according to predefined criteria. The study was conducted in 12 ICUs in France. Consecutive patients who died within 48 hours of ICU admission during the study period, in 2022–2023, were included. LST-L decisions were not guided by protocols but were at the discretion of the attending intensivists.

MAIN OUTCOMES AND MEASURES:

Of 1615 patients admitted to the participating ICUs during the study period, 100 died (6.2%) within 48 hours, including 62 with LST-L.

RESULTS:

In the LST-L group, age was significantly older (72 yr [64–77.8 yr] vs. 63 yr [59.0–69.8 yr]; p = 0.002), Charlson Comorbidity Index significantly higher (5.5 [2.0–8.0] vs. 4.0 [2.0–5.0]; p < 0.001), and management less invasive compared with the full-care group. By multivariable analysis, male patients were less likely to have LST-L decisions (odds ratio, 0.35; 95% CI, 0.13–0.93; p = 0.03). Most physicians, but a smaller proportion of nurses, perceived LST-L decisions as consensual. For 28 of 100 patients, the intensivist retrospectively deemed the ICU admission not the most suitable option. Patient wishes were rarely considered when making LST-L decisions. Time-limited trials were rarely used. Two-thirds of LST-L decisions were made during on-call hours.

CONCLUSIONS AND RELEVANCE:

Deaths occurring shortly after ICU admission were usually preceded by LST-L decisions. Efforts are needed to better consider patients’ wishes and to strengthen communication between ICU physicians and nursing staff, to ensure appropriate care—even when patients’ wishes are unknown and alternatives to ICU admission are not straightforward. Such rare and sometimes unforeseeable cases may also reflect unspoken preferences of patients or their families.

KEY POINTS

Question: What is the proportion of patients (excluding those admitted for palliative care or organ donation) who have a decision to limit life-supporting treatment (LST) among those who die early following their admission to intensive care?

Findings: Early deaths within 48 hours were uncommon (6.2% of admissions) and were preceded by LST limitation in 62% of cases.

Meaning: The rapid death of a patient after admission to intensive care is uncommon and usually occurs after a treatment limitation decision.

Decisions to limit life-supporting treatment (LST) are common in ICUs. A vast majority of all ICU deaths occur after LST limitation (LST-L) (1–3). LST-L can range from withholding some LSTs to withdrawing all LSTs (4). The goal is to avoid nonbeneficial care or care that does not align with the patient’s wishes (5–8). Among ICU patients with LST-L, 20–25% survive hospital discharge (9, 10).

A recent Canadian study showed that patients who die within 48 hours of ICU admission account for 3.3% of all ICU admissions and up to 30% of all ICU deaths (11). Factors associated with early ICU mortality include a higher Sequential Organ Failure Assessment (SOFA) score, older age, admission due to neurologic conditions, and admission following emergency surgery or trauma (12–16).

Early mortality after ICU admission may seem to challenge the appropriateness of ICU admission decisions, which rest on acute illness severity and potential reversibility, current and expected future quality of life, comorbidities and frailty, patient’s wishes about the level of care, and considering available resources (principle of distributive justice) (17). Age is considered but is not in itself a decisive criterion (18–23). Early death in the ICU can cause moral distress among staff, who may feel the admission was an inappropriately high level of care (24), although denial of ICU admission may cause moral distress in families and patients.

In some cases, early death is expected because the reason for admission was either potential organ donation or palliative/comfort care only (25, 26). However, the original aim of this study was to explore the interplay between ICU admission decisions made with curative intent, early LST-L decisions, and deaths occurring shortly after admission, within a multicenter framework.

The primary objective of this study was to determine if LST-L orders exist for patients who die within 48 hours after ICU admission and, if so, to what extent. The secondary objectives were to identify variables associated with LST-L in this specific population and to assess the perceptions of staff members.

MATERIALS AND METHODS

Design and Patients

All procedures were conducted in accordance with the ethical standards of the institutional research committee (Nantes University Hospital, Research Department) and with the 1975 Helsinki Declaration and its later amendments. The protocol was approved by the appropriate ethics committee on January 11, 2022 (No. CE SRLF 22-002) and the study was registered with the French Data Protection Authority (CNIL No. 2226700 v 0). Patients were included after death, and their consent to participate in the study was therefore not sought. Families were informed about the study orally and in writing. If family members were unwilling to have the patient participate, the patient’s data were removed from the database.

This retrospective, observational, multicenter study was carried out in five medical and seven medical and surgical ICUs for adults, located in 12 cities in western France. Each ICU included patients over a predefined period of 2 consecutive months, and the study took place between June 2022 and January 2023.

We included consecutive adults who died within 48 hours of ICU admission. Patients admitted for comfort care only or as potential organ donors were excluded, as well as those under guardianship or those who had expressed unwillingness to participate in clinical research.

Study Objectives and Outcomes

The primary outcome was the proportion of patients with LST-L among patients who died within 48 hours after ICU admission. LST-L was defined as a decision to withhold and/or withdraw one or more life-sustaining treatments, regardless of whether the decision was ultimately carried out.

The secondary objectives included identifying variables that were more common in patients with LST-L compared with those without and investigating whether certain variables were associated with a higher risk of being affected by an LST-L decision. Additionally, we examined the LST-L decision-making process and documented the nature of each limitation. Last, we assessed ICU staff members’ perceptions of the care provided to patients. There was no intervention in this study. We did not apply a predefined procedure nor mandate a formal staff meeting to decide on LST-L. However, French law requires a collegial decision involving at least two physicians, and the nursing team if available (27).

Data Collection

Inclusion took place at the time of death, with data collection performed immediately after patient inclusion, based on predefined criteria. For each included patient, the attending intensivists used a standardized form to collect clinical data and data on the LST-Ls.

Clinical data were: organ impairments using the SOFA score, acute illness severity using the Simplified Acute Physiology Score II (SAPS II), admission modality (medical, defined as no surgery performed; scheduled surgery, or emergency surgery), and reason for admission (nature of the acute condition).

The data collected on the LST-Ls were as follows: nature of the limitation decision recorded as withholding and/or withdrawing; type of organ support withheld or withdrawn; individuals involved in the decision recorded as senior intensivist, resident intensivist, nurse, nursing assistant, external consultant, patient, and/or relatives; the time of the decision recorded as weekday vs. on-call hours (defined as night or weekend); and the main reasons for the decision. This last section included the following predefined items: clinical severity (acute phase), frailty and comorbidities, patient’s wishes and/or as reported by their relatives, opinion of an external consultant, failure to respond to intensive care, and terminal illness. Based on an ordinal numerical scale, the clinician assigned a score from 0 to 10 for each item, considering the factors they deemed most significant in the decision (0: no importance; 10: critical importance).

The perceptions of ICU staff regarding patient care were assessed using a questionnaire specifically designed for this study. For each patient, the questionnaire was distributed by the local investigator and completed as soon as possible after the death by all staff members involved in their care (i.e., the attending senior intensivist and/or the resident and/or other healthcare workers in charge of the patient).

The questionnaire included three closed-ended questions (yes/no) to assess whether a consensus was established on the ICU admission decision, on the LST-L order, and whether the care provided in the ICU was deemed optimal.

Senior intensivists were also asked whether they considered ICU admission to have been the most suitable option for the patient and, if not, to provide their reasons.

Statistical Analysis

The patients were divided into two groups depending on whether LST-L decisions were made before death.

Quantitative variables, being non-normally distributed, were described as median (interquartile range) and compared using the Mann-Whitney U test. Qualitative variables were described as n (%) and compared using Pearson chi-square test. Pearson chi-square test was also used to compare LST-L decisions made at ICU admission vs. later during the ICU stay. We performed multivariable logistic regression to evaluate the association between LST-L decisions and several covariates, including sex, age, Charlson Comorbidity Index, SAPS II score, Clinical Frailty Scale (CFS), and on-call period. Variable selection was based on clinical relevance and previous literature. Adjusted odds ratios with the 95% CIs were calculated, and statistical significance was assessed using Wald test. The use of variance inflation factors detected no significant multicollinearity. Analyses were based on complete data.

For the staff-perceptions questionnaire, physicians (senior intensivists and residents) and nursing staff (nurses and nursing assistants) were analyzed in two separate groups. When at least one response for a given patient was available for both groups, agreement between groups was assessed by computing Cohen’s κ coefficient.

When two responses were available for the same patient about whether ICU admission had been the most suitable option, agreement was also evaluated based on Cohen’s κ coefficient.

All analyses were performed using RStudio, Version 4.1.1 (Posit Software, Boston, MA) in the R program.

RESULTS

Study Population

Figure 1 is the patient flowchart. Of the 1615 patients admitted to any of the 12 ICUs during the study inclusion period, 129 were included. Among them, 29 were excluded due to death after 48 hours (n = 5), potential organ donor status (n = 17), or admission for end-of-life care only (n = 7). No patient was excluded due to unwillingness to participate. Thus, 100 patients were included in the analysis. Table 1 reports their main features. Only a fifth of patients had no clinically significant preexisting condition, and the median Charlson Comorbidity Index was 5.

Figure 1.

Flowchart of the study. LST-L = life-supporting treatment limitation.

TABLE 1.

Main Features of the Study Patients

Patient Characteristics	Overall Cohort (n = 100)	Full-Care Group (n = 38)	Life-Supporting Treatment Limitation Group (n = 62)	p
Demographics and comorbidities
Males, n (%)	67 (67)	20 (52.6)	47 (75.8)	0.03
Age, yr, median (IQR)	68 (59–74)	63 (59–69.8)	72 (64–77.8)	0.02
Body mass index, kg/m², median (IQR)	26.8 (23.5–31.1)	27.6 (25.6–31.2)	26.3 (22.3–29.7)	0.10
Comorbidities
Cardiovascular disease, n (%)a	27 (27)	4 (10.5)	23 (37.1)	< 0.01
Chronic obstructive pulmonary disease, n (%)	14 (14)	2 (5.3)	12 (19.4)	0.10
Mild cirrhosis, n (%)	2 (2)	0 (0)	2 (3.2)	0.70
Cirrhosis with portal hypertension, n (%)	10 (10)	6 (15.8)	4 (6.5)	0.24
Diabetes mellitus, n (%)	15 (15)	3 (7.9)	12 (19.4)	0.20
Chronic kidney disease with glomerular filtration rate < 50 mL/min, n (%)	9 (9)	1 (2.6)	8 (12.9)	0.17
Immunosuppression, n (%)b	37 (37)	12 (31.6)	25 (40.3)	0.51
Alcohol abuse, n (%)	19 (19)	8 (21.1)	11 (17.7)	0.88
Current smoking, n (%)	16 (16)	3 (7.9)	13 (21.0)	0.15
Charlson Comorbidity Index, median (IQR)	5 (3–7)	4 (2–5)	5.5 (4–8)	< 0.01
ECOG performance status score, median (IQR)	1 (0–2)	1 (0–2)	2 (1–2)	0.06
ECOG performance status score ≥ 2, n (%)	15 (15.3)	4 (10.8)	11 (17.7)	0.50
Comorbidities known at initial request for ICU admission, n (%)	87 (87)	32 (84.2)	55 (88.7)	0.73
Self-sufficiency
Clinical frailty score, median (IQR)	3 (2–4)	2 (1–4)	3 (2–4)	0.10
Clinical frailty score ≥ 4, n (%)	42 (42)	12 (31.6)	30 (48.4)	0.16
Self-sufficiency assessed at the initial request for ICU admission, n (%)	77 (77)	28 (73.7)	49 (79.0)	0.60
Ability to walk				0.26
Need for material or human assistance, n (%)	24 (24)	5 (13.2)	19 (30.6)
Unable to walk, n (%)	7 (7)	1 (2.6)	6 (9.7)
Lives independently at home, n (%)	32 (32)	13 (34.2)	19 (30.6)	0.80
At least one home-support provider, n (%)	27 (27)	7 (18.4)	20 (32.2)	0.17

ECOG = Eastern Cooperative Oncology Group, IQR = interquartile range.

^aCongestive heart failure, ischemic heart disease, and peripheral arterial disease.

^bHematological malignancy, active solid cancer, and immunosuppressive therapy.

Admissions occurred during on-call periods in 66 patients and for medical reasons in 96 patients. Table 2 lists the main reasons for admission. Overall, the severity of the critical illness was high, with a median SAPS II score of 83, and multiple organ dysfunction syndrome in 26 patients.

TABLE 2.

Critical Illness Features and Treatments

Circumstances and Management in the ICU	Overall Cohort (n = 100)	Full-Care Group (n = 38)	Life-Supporting Treatment Limitation Group (n = 62)	p
ICU admission
Sequential Organ Failure Assessment score, median (IQR)	12 (9–15)	13.5 (11–15.8)	11 (9–12.3)	< 0.01
Simplified Acute Physiology Score II, median (IQR)	83 (69–94)	84 (70–98.3)	80.5 (66–93)	0.21
Admission mode, n (%)				0.28
Medical	96 (96)	38 (100)	58 (93.5)
Scheduled surgery	2 (2)	0 (0)	2 (3.2)
Emergent surgery	2 (2)	0 (0)	2 (3.2)
Admission during on-call hours, n (%)	66 (66)	24 (63.2)	42 (67.7)	0.80
Reasons for admission, n (%)
Respiratory failure	40 (40)	12 (31.6)	28 (45.2)	0.26
Neurologic failure	36 (36)	16 (42.1)	20 (32.3)	0.44
Kidney failure	15 (15)	7 (18.4)	8 (12.9)	0.64
Septic shock	34 (34)	14 (36.8)	20 (32.3)	0.80
Heart failure	6 (6)	1 (2.6)	5 (8.1)	0.50
Hemorrhagic shock	17 (17)	8 (21.1)	9 (14.5)	0.57
Cardiogenic shock	5 (5)	2 (5.3)	3 (4.8)	1
Cardiac arrest	4 (4)	0 (0)	4 (6.5)	0.28
Trauma	10 (10)	7 (18.4)	3 (4.8)	0.06
ICU management, n (%)
Invasive mechanical ventilation	72 (72)	33 (86.8)	39 (62.9)	< 0.01
Noninvasive ventilation	9 (9)	1 (2.6)	8 (12.9)	0.18
Renal replacement therapy	11 (11)	8 (21.1)	3 (4.8)	0.03
Norepinephrine	72 (72)	33 (86.8)	39 (62.9)	< 0.01
Surgery	2 (2)	2 (5.3)	0 (0)	0.27
Death
Time from admission to death, hr, median (IQR)	13.2 (5.8–29.8)	16.7 (7.0–27.5)	12.9 (5.2–30.2)	0.41
Cause of death, n (%)
Multiple organ failure	56 (56)	25 (65.8)	31 (50)	0.14
Refractory hypoxemia	10 (10)	1 (2.6)	9 (14.5)	0.12
Cardiac arrest or terminal extubation	40 (40)	12 (31.6)	28 (45.2)	0.30

IQR = interquartile range.

Primary Endpoint

Of the 100 patients, 62 received LST-L decisions (LST-L group) and 38 did not (full-care group). LST-L decisions were made at admission in 28 of 62 patients (45.2%), and later during the stay in 34 of 62 patients (54.8%) (Fig. 1). Of the 28 patients with LST-L decisions at admission, six had changes in those decisions during the ICU stay. Thus, this represented a total of 68 LST-L decisions.

Considerable variability in the proportion of patients with LST-L decisions was noted across centers (eTable 1, https://links.lww.com/CCX/B537).

Factors Associated With Life-Supporting Treatment Limitation Decisions

Patients in the LST-L group were older, had a higher median Charlson Comorbidity Index, and more often had cardiovascular comorbidities (Table 1). At ICU admission, their SOFA score was lower, although their SAPS II score was not significantly different compared with the full-care group. None of the variables reflecting self-sufficiency differed significantly between the two groups. After adjusting for age, Charlson Comorbidity Index, SAPS II score, CFS, and whether the decision was made during an on-call period, men had 65% lower odds of receiving an LST-L decision compared with women (95% CI, 0.13–0.93; p = 0.03) (eTable 2, https://links.lww.com/CCX/B537).

During the ICU stay, invasive mechanical ventilation, renal replacement therapy, and norepinephrine therapy were used significantly less often in the LST-L group than in the full-care group (Table 2).

Life-Supporting Treatment Limitations: Decision-Making Process

Data on the main criteria for LST-L decisions were available for 65 of 68 decisions. At admission, a single patient had established advance directives in accordance with French law; 91 had no advance directives, and for eight no information was available on this point. Throughout their clinical course and before the LST-L decision-making process, two of 100 patients had expressed a preference for LST-L, 78 of 100 had not indicated any preference, and 20 of 100 had not communicated their wishes at all.

Data on self-sufficiency and medical history were available for 20 of 28 (71.4%) and 30 of 37 patients (81.1%) with LST-L decisions taken at ICU admission and later during the stay, respectively.

Twenty of 28 (71.4%) and 21 of 37 (56.8%) decisions at admission and later on, respectively, were made during on-call hours.

All decisions were made in consultation with a senior intensivist and at least one other physician. During the decision-making process, patients’ wishes received a score greater than five of ten for only three of 28 (10.7%) and seven of 37 (18.9%) decisions at admission and later, respectively. Wishes as reported by relatives received scores greater than 5 for four of 28 (14.3%) and nine of 37 (24.3%) decisions at admission and later on, respectively; corresponding values for the nursing staff were eight of 28 (28.6%) and 12 of 37 (32.4%), respectively, and for external consultant seven of 28 (25%) and 13 of 37 (35.1%), respectively. The most decisive criteria for LST-L decisions made at admission were clinical severity, frailty, and comorbidity burden. A terminal illness with no reasonable therapeutic options was the primary criterion for decisions made during the stay (Fig. 2).

Figure 2.

Main criteria for guiding life-supporting treatment limitation (LST-L) decisions, as assessed by the intensivists at ICU admission and during the ICU stay. Each factor was assessed on a scale from 0 to 10. The proportion of decisions assigned each score is on the y-axis.

Life-Supporting Treatment Limitations: Types of Decisions

Figure 3 details the types of decisions. Data on the types of LST-L decisions were available for all decisions but two (66/68). Decisions could include several components, such as withholding treatments and plans to withdraw treatments. Withholding decisions were more common at admission and withdrawing decisions during the stay. Withholding decisions included not starting cardiopulmonary resuscitation in 19 of 28 (67.9%) and 15 of 38 (39.5%) decisions at admission and during the stay, respectively. Corresponding values were 15 of 28 (53.6%) and 13 of 38 (34.2%), respectively, for not starting extracorporeal membrane oxygenation; 12 of 28 (42.9%) and 11 of 38 (28.9%), respectively, for not starting renal replacement therapy; and 11 of 28 (39.3%) and seven of 38 (18.4%), respectively, for not starting invasive mechanical ventilation.

Figure 3.

Types of life-supporting limitation decisions (n = 66): 28 decisions at ICU admission and 38 decisions later during the stay; some decisions had several components. CPR = cardiopulmonary resuscitation, ECMO = extracorporeal membrane oxygenation, RRT = renal replacement therapy, NIV = noninvasive ventilation.

Withdrawal was planned in 18 of 66 (27.3%) of all decisions, including one of 28 (3.6%) at admission and 17 of 38 (44.7%) during the stay. A time-limited trial (TLT) was decided for only four of 66 (6.1%).

Staff Perceptions

At least one questionnaire was completed for each patient participating in the study, including 192 questionnaires for 62 patients with LST-L (70 senior intensivists, 32 residents, 64 nurses, and 26 nursing assistants) and 92 questionnaires about 38 patients in full-care group (39 senior intensivists, 11 residents, 37 nurses, and five nursing assistants).

Physicians felt that a consensus was reached regarding ICU admission in 84.2% of cases overall (79.4% for the LST-L group and 94% for the full-care group) and regarding the LST-L decision in 95.1% of cases overall. However, 26 of 152 (17.1%) physicians felt the management had not been optimal, due to delayed diagnosis (n = 10), delayed ICU admission (n = 9), and/or perceived non beneficial care (n = 7).

Among senior intensivists, 29 of 109 (26.6%) felt with hindsight that ICU admission of 28 patients had not been the best option. Of these 28 patients, 25 were in the LST-L group. Reported reasons for this perception were excessive frailty or comorbidity burden (20 patients), extreme severity of the acute condition that led to ICU admission (15 patients), and palliative care unit as a better option (n = 23).

Among nursing staff, 69 of 132 (52.3%) felt no consensus was achieved about ICU admission; the proportions were 51% for the LST-L group and 54.8% for the full-care group. A higher proportion of nursing staff (66/90 [73.3%]) felt there had been a consensus about the LST-L decision. Only 15 of 132 (11.4%) nursing staff reported perceiving suboptimal care, the most common reason (n = 8) being delayed ICU admission.

No agreement was observed between physicians’ and nurses’ perceptions of consensus regarding the ICU admission decision for the same patient (both overall and for the LST-L and full-care groups). In contrast, when two physicians had to assess whether ICU admission represented the best option for the patient, agreement was moderate (Table 3).

TABLE 3.

Agreement Among Staff Regarding Perceptions About Patient Management

Question and Applicable Groups	Overall		Full-Care Group		LST-L Group
	κ (95% CI)	p a	κ (95% CI)	p a	κ (95% CI)	p a
Physicians vs. nursing teamb
Consensus about ICU admission	0.01 (–0.15 to 0.17)	0.90	–0.06 (–0.17 to 0.05)	0.30	0.05 (–0.18 to 0.27)	0.70
Consensus about LST-L					0.07 (–0.13 to 0.27)	0.50
ICU management optimal	0.18 (–0.07 to 0.43)	0.15	0.25 (–0.14 to 0.64)	0.20	0.15 (–0.14 to 0.44)	0.30
Two physicians involved in caring for the same patient
Appropriateness of ICU admission	0.56 (0.27–0.84)	< 0.01	0.42 (–0.17 to 1)	0.16	0.57 (0.25–0.90)	< 0.01

LST-L = life-supporting treatment limitation.

^aKappa test.

^bNurses and nursing assistants.

DISCUSSION

In this multicenter study conducted in 12-Fr ICUs in 2022–2023, 100 of 1615 (6.2%) patients died within 48 hours after ICU admission. Among these 100 patients, 62 had LST-L decisions. Patients with LST-L decisions were older and had a heavier comorbidity burden. The SOFA score was lower in the LST-L group than in the full-care group but the SAPS II reflecting acute illness severity was not different. These results are probably ascribable to a limited number of organ failures at admission, as patients with worse impairments may have been deemed too ill for ICU admission, and to the older age in the LST-L group since age is among the items in the SAPS II score.

That LST-Ls are associated with older age and a heavier comorbidity burden has been reported previously (9). We found no significant association with self-sufficiency, perhaps due to the small sample size. In another study, a modified Rankin score above 2 was associated with LST-L within 48 hours after admission (odds ratio, 2.5; 95% CI, 1.5–4.4) (18). By multivariable analysis, men were less likely to be involved in an LST-L decision, as reported previously (28). Speculative explanations include women being more likely to express their wishes or caregivers misperceiving their frailty.

Most physicians felt that a consensus had been achieved about ICU admission decisions and LST-Ls. In contrast, the absence of a consensus was reported by the nursing staff for nearly half the patients regarding the ICU admission decision and a quarter regarding LST-L. Senior intensivists felt retrospectively that ICU admission had not been the most suitable option for about a quarter of the patients. Consensus is crucial, notably for recently admitted patients, given the rapid shift from the initial curative intent to an end-of-life care intent, which may cause distress among staff, patients, and families.

The rising proportion of ICU deaths preceded by LST-L decisions may reflect increased admission of older, frailer patients (2, 29), and the growing use of TLTs of full critical care (30). The proportion of patients with LST-L decisions in our cohort aligns with studies that did not focus on death within 48 hours after ICU admission (1, 2, 9, 18). The wide variability in this proportion across ICUs should be interpreted in light of the small sample sizes. Considerable variability has been reported previously across and within countries and within ICU teams (9). Differences in physicians’ perceptions of acceptable quality of life may also contribute to variability (8).

Our finding that half of LST-L decisions were made at admission and over a quarter early in the ICU stay in patients who died within 48 hours raises concern about both unsuitable ICU admission and delayed ICU admission. The senior intensivists considered, with hindsight, that 28% of patients should probably not have been admitted, chiefly due to severe comorbidities and frailty. Nonetheless, information on prior self-sufficiency and comorbidities was available for most patients at ICU admission. Thus, ICU admission to ensure survival while waiting for this information was probably uncommon.

Providing LSTs followed by withdrawal if ineffective may prevent loss of chance (4). A broad ICU admission policy may be justified, even the result is that some patients die shortly after admission (24). TLTs of full care with reevaluation should be used more widely than in our study. However, such trials may have been performed but not recorded, in patients whose prognosis was initially uncertain. This situation may have been perceived with hindsight as unsuitable admission. A structured approach to LST-L decisions and goals of care, including made during on-call periods, must be developed and implemented to prevent justified admissions from being subsequently perceived as not having been the best option and to ensure that no unjustified admissions occur.

Importantly, the wishes of the patient or relatives were not usually considered when making LST-L decisions. French law requires that efforts be made to ascertain the patient’s expressed wishes (31). In France, advance directives were made into law in 2005 (32) but are rarely established (33). In a recent study, only 32.7% of hematologists regularly discussed advance directives with their patients (34). In our study, 91% of the patients had not written advance directives, in keeping with another French study (35). Obtaining information about the patient’s desired level of care before clinical deterioration may also help avoid nonbeneficial ICU admissions (36). If this is not possible, the family should be consulted to gather the patient’s values and preferences. That the wishes of patients and families were often not considered in our study may reflect both the urgency of the situations, which may not have left enough time to collect information on wishes, and prompt severe clinical deterioration making the need for LST-L decisions obvious.

Perceived consensus on ICU admission and LST-L decisions was common among physicians but less so among nurses. Intensivists, as experts, can decide on admissions without prior formal meetings, but briefings held once the patient is stabilized help to alleviate staff stress. In one study, nursing staff reported similar perceptions to those of physicians regarding admission decisions and nonbeneficial ICU stays (37). French legislation requires that LST-L decisions be taken collegially. Of the 12 study ICUs, 11 held dedicated staff meetings. In our study, two-thirds of decisions occurred during on-call hours but remained collaborative. A study in 57-Fr ICUs found that only 17% of decisions to implement deep end-of-life sedation were not taken collegially (38). The predominance of LST-L decisions made during on-call periods raises concerns about decision-making quality. The severity of acute illness often limits time for formal meetings and multidisciplinary input, potentially explaining differences in perception between physicians and nurses. Ensuring the participation of nurses and nursing assistants may be more difficult at night given the greater time constraints. However, a strong consensus on LST-L decision is essential. Nurses, as primary contacts for patients and families, offer valuable insights, and their greater involvement could enhance team cohesion, and the LST-L decision-making process.

Among the limitations of our study, we acknowledge the risk of a self-fulfilling prophecy, where perceiving a patient as unlikely to recover may increase the likelihood of LST-L decisions, thereby increasing the risk of death. Another key limitation is the small sample size, which may have hindered identification of factors associated with LST-L. We chose retrospective data collection to maximize statistical power, although this resulted in possible information bias (this was counterbalanced by a predefined protocol, continuous and standardized inclusions, and highly responsive data collection). To our knowledge, no other studies have assessed early LST-L decisions, and the present study was therefore conducted with exploratory intent. Additionally, key variables such as bed occupancy rate and nurse-patient ratio were not collected. The 2-month inclusion periods across 12 ICUs over 8 months covered both high- and low-occupancy periods, but we did not track denied ICU admission requests, which could have informed admission practices given the variability in early deaths across ICUs (1.2–12.9%). ICU admission was retrospectively deemed not the best option for 28% of patients, but alternative available options (e.g., palliative care unit admission) were not recorded. Furthermore, such admissions may at times be made for compassionate reasons, such as ensuring optimal end-of-life support (allowing time for family to gather or avoiding death alone), particularly during off-hours. In such contexts, early ICU death does not necessarily indicate inappropriate triage but may reflect a deliberate, humane approach to end-of-life care. We did not record the type of end-of-life care given to patients and families. Data on delayed ICU admission, possibly linked to early deaths, were also missing.

To our knowledge, this study provides the first information on LST-L for patients dying within 48 hours of ICU admission. The participation of 12 ICUs in university and community hospitals supports the external validity of the findings. However, that centers were in France may limit applicability to other countries. Another strength of our study is the evaluation of staff perceptions.

A key implication of our study is that patient-care preferences should be gathered as soon as their condition worsens and considered in LST-L decisions. TLTs, rarely applied in our cohort, should be used more widely to prevent loss of opportunity. The perception by many nursing staff that decisions lacked consensus highlights the need for improved communication within ICU teams. These scenarios, though infrequent, pose significant ethical challenges, as patient wishes are rarely known and alternatives to intensive care for vulnerable patients are often unclear within tight timeframes. Situations in which the intensivist decides on LST-L for a patient who dies shortly after admission are infrequent and not always foreseeable. Some of these admission decisions likely reflect unexpressed preferences of patients or their families, particularly concerning the desire for a dignified end-of-life process.

CONCLUSIONS

In this retrospective, observational, multicenter study, 6.2% of patients admitted to the ICU for reasons other than possible organ donation and end-of-life care died within 48 hours after admission. In 62% of these patients, the death was preceded by an LST-L decision. In hindsight, senior intensivist felt that ICU admission had not been the most suitable option for 28% of the patients. More work is needed to define optimal policies for ICU admission and LST-L decisions.

Supplementary Material

cc9-7-e1300-s001.pdf ^{(396.1KB, pdf)}