You are here

  • Home
  • Archive
  • Volume 15, Issue 5
  • 1-year survival in critically ill elderly medical patients treated with a conservative or an invasive approach (OCTO-REVERSE study): a nationwide observational study

Article Text

Article menu

Download PDFPDF

Intensive care
Original research
1-year survival in critically ill elderly medical patients treated with a conservative or an invasive approach (OCTO-REVERSE study): a nationwide observational study
  1. Clement Leclaire1,
  2. Alexandre Georges1,
  3. Matthieu de Stampa2,3,
  4. Philippe Aegerter2
  1. 1Paris Public Hospital at Home (HAD AP-HP), Greater Paris University Hospitals, Assistance Publique—Hopitaux de Paris, Paris, France
  2. 2Inserm U1018, Center for Research in Epidemiology and Population Health (CESP), University of Versailles Saint-Quentin-en-Yvelines (UVSQ), CESP, Villejuif, France
  3. 3Departmental Gerontological Center, Marseille, France
  1. Correspondence to Dr Clement Leclaire; clement.leclaire{at}aphp.fr

Abstract

Objective To determine whether an invasive approach is associated with favourable long-term outcomes among elderly medical patients in the intensive care unit (ICU), compared with a conservative approach.

Design Nationwide observational study (OCTO-REVERSE study) using data prospectively collected in the National French Healthcare Database (covering 99% of the population, 66 million people).

Setting Comprehensive multicentre study through the linkage of large-scale national registries (including all public or private facilities) from 2013 to 2018 to avoid ambiguities related to the COVID-19 pandemic.

Participants All non-surgical patients aged 80 years or older admitted to an ICU in France during the period (n=107 014 patients at 822 hospitals).

Outcome measures The main outcome was the 1-year survival rate. The association of the two approaches with 1-year survival was estimated using a time-dependent Cox model and a propensity score (PS) adapted to time-to-event analysis, yielding the average treatment effect in the treated and extended weighted Kaplan–Meier curves.

Results 107 014 patients were categorised into two groups based on the type of care received: invasive (n=51 680 (48%) received invasive ventilation and/or vasopressor support) or conservative (n=55 334 (52%) received neither). 1-year survival rate was significantly lower in the invasive group than in the conservative group (27% vs 59% estimated with extended time-dependent Kaplan–Meier method). The risk of death in the invasive group remained significantly higher after time-dependent PS weighting (HR 1.64; 95% CI 1.60 to 1.69; p<0.001). The loss in restricted mean survival time was 67.7 days (95% CI 65.7 to 69.8) in this group and 31% of deaths occurred the day of initiation of the procedure, or the following day.

Conclusion Among the whole population of critically ill elderly medical patients in France, the invasive approach was unknowingly associated with end-of-life care in nearly three quarters of cases. Further research is needed to align intensive care with compassionate goals.

  • PALLIATIVE CARE
  • Triage
  • Person-Centered Care
  • Caregivers
  • INTENSIVE & CRITICAL CARE
  • Health Services for the Aged

Data availability statement

Data may be obtained from a third party and are not publicly available. The data used in this study are derived from the French National Health Data System, which is managed by the French National Health Insurance Fund. Access to these data is subject to regulatory approval and can be requested through the official data access platform under authorisation number TPS 962976. Further details on data access procedures can be obtained by contacting snds.cnam@assurance-maladie.fr.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

https://doi-org.ezproxy.u-pec.fr/10.1136/bmjopen-2024-089835

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    STRENGTHS AND LIMITATIONS OF THIS STUDY

    • This study uses the National French Healthcare Database, which records all medical procedures coded for reimbursement, ensuring exhaustiveness and real-world representativeness.

    • This study employs a robust time-dependent analysis using the extended Kaplan–Meier method, which accounts for variations in treatment exposure over time and improves survival estimates.

    • The simplified acute physiology score II, used in the propensity score model as the main severity adjustment factor, proved to be the primary determinant in balancing patient characteristics between groups.

    • The main limitation is the presence of potential unmeasured confounding factors, including key clinical parameters that could influence both treatment decisions and patient outcomes.

    • Information on whether care was withheld or withdrawn is not available, making it difficult to assess the impact of end-of-life decisions on mortality differences.

    Introduction

    The proportion of older patients in the intensive care unit (ICU) has been steadily increasing over the past two decades, although the overall benefit to patient outcomes remains unclear.1 Many studies have explored this topic, but variations in patient selection (age 65–90, surgical vs medical and invasive vs conservative care) and outcome measures (mortality vs functional status and early vs delayed assessment) have led to uncertain conclusions.2 Additional research is needed to address the care consequences of this growing population. However, a randomised trial has been deemed infeasible on ethical grounds.3 The ensuing uncertainty may have led to a decision-making process based solely on clinical judgement, while triage systems based on even limited evidence would have been preferable.4 In addition, overoptimistic expectations5 of outcomes could have favoured the expansion of time-limited trials,6 some of which turned out to be disproportionate. Previous studies on this issue7–9 have identified older age (≥80 years),10 non-operative condition11 and exposure to invasive procedures (primarily invasive mechanical ventilation and vasopressor therapy)12–14 as prognostic factors that should be considered especially as to long-term outcomes. By conducting a nationwide study that combines these three factors, we aimed to assess the characteristics, management and 1-year survival of the entire target population of elderly patients admitted to the ICU for medical reasons in France.

    Methods

    Study design and data sources

    We performed a nationwide retrospective analysis (OCTO-REVERSE study) using data prospectively collected in the Système National des Données de Santé (SNDS), the national French healthcare database that covers 99% of the population regardless of socioeconomic status (66 million). The SNDS was created for hospital payment and government purposes, and regular audits are conducted to ensure reliability and completeness. The SNDS also facilitates epidemiological research, with a specific interest in ageing by an extended follow-up. Formal methods have been developed to assess the strengths and limitations of databases, and population-level data have enhanced our understanding of the use and outcomes of mechanical ventilation,12 particularly for elderly patients hospitalised in ICUs in France.15–17 The SNDS links the national hospital discharge database of diagnosis-related groups (DRGs) ‘PMSI (Programme de Médicalisation des Systèmes d’Information (French national hospital discharge database))’ with the national death registry ‘CépiDC’, enabling the tracking of individual patient trajectories from any hospital stay through to death. This allows for the identification of the date and place of death, whether it occurs in hospital or in the community. It contains comprehensive medical and administrative data for all patients in the country. This linkage of large-scale French national registry populations contributes to guiding public decisions. Each patient is assigned a unique identification number with pseudonymised information, allowing individuals to be followed over time by linking inpatient and outpatient data from each hospitalisation in all public or private facilities in France. Procedures, including surgery or life-sustaining treatments, are time stamped and coded using a national classification system called CCAM (Classification Commune des Actes Médicaux (Common Classification of Medical Procedures)) (online supplemental appendix 1).

    The study was submitted to and approved by the independent national data protection authority (CNIL (Commission Nationale de l'Informatique et des Libertés (French Data Protection Authority)) Registration: 920181). All data were deidentified for research purposes, and French law waives the need for informed consent. An independent institutional review board approved the study protocol (Institutional Review Board (IRB) Registration: #00011928).

    Study population

    All patients aged 80 years or older with complete data, admitted to hospitals in France from January 2013 to December 2018 (to avoid any ambiguities related to COVID-19), were eligible. We then selected ICU stays where the simplified acute physiology score (SAPS) II is comprehensively collected. To select the non-operative condition, we excluded stays that involved a surgery or whose DRG was surgical (or severe burn injury), either during the ICU stay or within 30 days prior to the ICU admission. To avoid counting multiple hospital outcomes for a single patient, only the first hospital stay, including a medical admission to an ICU during the study period, was considered for each patient, and readmission stays were excluded. Finally, we considered the invasive condition for patients who underwent invasive mechanical ventilation (CCAM codes GLLD004, GLLD008 and GLLD015) or vasopressor support (CCAM codes EQLF001 and EQLF003) (online supplemental table 2, appendix 1).

    Variables and exposure

    The exposure variable was defined as the first occurrence of one of the following invasive procedures during the ICU stay: vasopressor support or invasive mechanical ventilation. We focused on these two main modifiable factors of interest because clinicians often question their benefit for the oldest patients. This choice was also driven by their coding reliability, as there is a clear-cut bedside decision without delay, marking a turning point in the patient’s trajectory. We did not include non-invasive ventilation (NIV) because its use as a palliative treatment for dyspnoea has become increasingly common. We also did not include renal replacement therapy (RRT) due to the large grey area in coding between acute and chronic situations, and because its initiation can often be delayed without threatening the prognosis.18 We addressed potential confounding resulting from variation in the case mix by controlling for age, severity of illness, pre-existing medical conditions expressed by the Charlson comorbidity index, primary diagnosis at admission, preadmission location and academic status of the ICU. These confounders were specified a priori to develop a risk adjustment model for mortality. Patients were followed until death or the end of the year 2019, whichever occurred sooner. The invasive and non-invasive cohorts were mutually exclusive.

    Outcomes

    The primary objective of our analysis was to describe the 1-year mortality of two populations of elderly medical ICU patients based on the treatment intensity (invasive vs conservative care). Additionally, we analysed intermediate mortality rates and lengths of stay, both in ICU and in hospital.

    Statistical analysis

    Characteristics of patients according to the use of invasive procedures were described using mean and SD for age and median and IQR for other continuous variables. Categorical variables were presented as counts and percentages. Comparisons between the two groups were performed using the Wilcoxon test for continuous variables and the χ2 test for categorical variables. Crude mortality rates were calculated at ICU and hospital discharge, along with 1-year readmission rates and the duration of procedures and hospital stays, for both groups and for each invasive procedure (online supplemental appendix 2, pp. 11–12).

    Survival was measured from the date of ICU admission to the date of death or censored at the last follow-up. To limit immortal time bias, the occurrence of any invasive procedure was considered as a time-dependent covariate. Time precision was day, and the time of death was shifted by 0.1 day to avoid ties with the time of the procedure. Multivariate analysis of survival was performed by a Cox proportional hazards (PHs) model producing HRs. The handling of ties used the Efron approximation. The PH assumption was verified by scaled Schoenfeld residuals. The time-varying effect of a procedure was modelled by adding an interaction term between the corresponding covariate and some transformation of time. The restricted mean survival time (RMST) method was selected because it is not dependent on the number of events and on the assumption of PHs, as is the case in time-to-event analyses. RMST reflects the life expectancy of patients up to a specified time. Details regarding the statistical methods are provided in online supplemental appendix 1. As patients receiving invasive procedures may differ from the conservative ones, a propensity score (PS) was used to minimise the effects of confounding. Assumed confounders included age, sex, pre-existing comorbidities (Charlson comorbidity index), SAPS II score, year, reason for admission, availability of a palliative or geriatric care team in the hospital, preadmission location, ICU level and hospital characteristics (online supplemental figure S2 and table S1, appendix 1). However, the population at risk changes over time due to attrition (death or discharge). Therefore, we used an inverse probability of treatment weighting (IPTW) method adapted to the time-dependent context.19 IPTW usually estimates the average effect of treatment when the entire sample is moved from control to treated. But this seemed unrealistic for invasive support in the elderly population, so here IPTW targeted the average treatment effect in the treated (ATT). We divided the time span into strata within which a distinct PS could be calculated. These strata were each day until day 14, then periods 14–20, 20–30, every 15 days until 120 and then every 30 days until 365 (case base). This time-stratified PS was built using a Cox model that predicted the use of invasive procedures by variables available at admission. The influence of continuous variables (age and SAPS II) was modelled by fractional polynomials. Variable balance was assessed by examining the standardised mean differences before and after weighting. Weights were trimmed at the 1st and 99th percentiles. To remove residual confounding as much as possible, we combined weighting at the design stage with regression adjustment on the same confounders (as listed in online supplemental figure S2 and table S1, appendix 1) at the analysis stage when estimating treatment effects.20 A robust sandwich-type estimator was used to calculate SEs.

    We performed sensitivity analyses to assess the robustness of the overall HR associated with invasive procedures: initially by using ATT weights truncated at the 5th and 95th percentiles, then by using time strata defined by every day, an invasive procedure or death occurred.

    All tests were two sided, and a p value<0.05 was considered statistically significant. The analysis used the SAS Enterprise Guide V.7.15 (SAS Institute Inc., Cary, NC, USA) provided by the SNDS infrastructure.

    Patient and public involvement

    None.

    Results

    The final analysis included 107 014 first stays of patients aged 80 years or older who were admitted to an ICU for medical reasons at 822 hospitals, mostly publicly funded (65%), in France between 2013 and 2018. Among those patients, 51 680 (48%) received invasive ventilation or vasopressors, and 55 334 (52%) received neither of the two. A flowchart of patients and stays in the study is shown in figure 1.

    Figure 1
    Figure 1

    Flowchart of patients and stays in the study. DRG errors correspond to stays that could not be assigned to a DRG, mainly due to coding inconsistencies or inaccuracies in the classification of hospital stays, such as misclassification of diagnoses, procedures or administrative coding discrepancies. DRG, diagnosis-related group; ICU, intensive care unit.

    The demographic characteristics and the outcomes since ICU admission are shown in table 1. The mean age was 84.6 years. More than half (52%) were male in the invasive group, and 52% were female in the conservative group. Overall, 89% were admitted from home, with most of them having a stop at the emergency department (59%). Patients in the invasive group had a significantly higher severity of illness at admission as reflected by a higher SAPS II but fewer general coexisting conditions as reflected by a lower Charlson comorbidity index. Over the years, the flow of patients was increasingly directed towards the invasive group. Among 51 680 (48%) patients in the invasive group, 70% were dead at 1 year, while among 55 334 (52%) patients in the conservative group, 41% were dead at 1 year (χ2 test, p<0.001).

    View this table:
    Table 1

    Patient characteristics*

    Outcomes of patients in the invasive group since ICU admission are detailed in table 2. Among 38 427 patients who underwent vasopressors, 72% were dead at 1 year. Among 40 495 patients who received invasive ventilation, 72% were dead at 1 year. Patients who received a first invasive procedure were likely to receive the other one in 53% of cases, and 1 year mortality was then 77%, whereas it was 60% and 63%, respectively, for vasopressors only or ventilation only. Among 8761 patients who underwent RRT (regardless of the group), 74% were dead at 1 year. The median number of days in the ICU was 4 (IQR 2–8). The median number of days in the hospital was 8 (IQR 2–16). The median duration of ventilation was 3 days (IQR 2–7). There were 2582 patients who underwent prone positioning. In their case, the median number of sessions was 2 (IQR 1–3) per patient. The median duration of vasopressor support was 2 days (IQR 1–4). The time of the initial invasive procedure was the day of ICU admission for 79% of patients in the invasive group and the following day for an additional 9% (online supplemental figure S1, appendix 1). More than half of the patients in the invasive group had both procedures, generally on the same day (86%). If vasopressors predated ventilation (6%) or ventilation predated vasopressors (9%), the median interval of initiation was 1 day (IQR 1–3). 24% of deaths in the invasive group occurred on the day of the initiation of any procedure, and 7% the following day.

    View this table:
    Table 2

    Outcomes in the invasive group (n=51 680)*

    After the calculation of the time-stratified PS, the balance obtained on ATT weighted samples was good, with absolute standardised differences less than 0.1 (online supplemental figure S2, appendix 1). The test of Schoenfeld residuals for the PHs assumption was significant, as the HR was decreasing over time (online supplemental figure S4, appendix 1). Kaplan–Meier survival curves obtained before and after the time-stratified ATT weighting are presented in figure 2. ATT weighting induced a decrease in 1-year survival rate in the conservative group from 59% to 43% (95% CI 0.421 to 0.441) versus 27% in the invasive group (95% CI 0.268 to 0.276, remained unchanged after ATT weighting, as expected). The mortality in the invasive group remained significantly higher after time-dependent IPTW (HR 1.64; 95% CI 1.60 to 1.69; p<0.001). Life expectancy limited to 1 year was 117.5 days in the invasive group versus 185.2 days in the conservative group, resulting in a significant loss of 67.7 days (RMST, 95% CI 65.7 to 69.8).

    Figure 2
    Figure 2

    Probability of survival from ICU admission to 1 year, according to conservative or invasive approach, before and after time-dependent PS weighting (ATT). Kaplan–Meier method adapted to time-dependent covariate (with 95% CI). Dashed lines represent the time-dependent Kaplan–Meier plot of the probability of survival from ICU admission to 1 year among the original unweighted population. Solid lines represent the time-dependent Kaplan–Meier plot of the probability of survival on IPTW-ATT weighted sample using weights trimmed at the 1st and 99th percentile. Invasive curve remained unchanged after weighting, as expected. HR denotes the HR obtained from the ATT99 weighted Cox model and diff-RMST the difference in (1 year) RMST (days). ATT, average treatment effect in the treated; ICU, intensive care unit; IPTW, inverse probability of treatment weighting; PS, propensity score; RMST, restricted mean survival time.

    In the sensitivity analysis, neither the use of ATT weights truncated at the 5th and 95th percentiles nor the use of a different time-stratification pattern changed substantially the HR associated with invasive procedures. Finally, the combination of weighting and further adjustment on baseline covariates (year, age, sex, coexisting conditions, admission source, severity and academic status) produced a similar HR (HR 1.67; 95% CI 1.63 to 1.73; p<0.001) (online supplemental appendix 1).

    Discussion

    Several studies support conservative management over invasive approaches,18 21 22 and there is concern that invasive procedures may worsen the prognosis of older ICU patients.23 24 Our large nationwide study was designed to describe the association between invasive care and mortality in real-world clinical practice. Our findings suggest that invasive procedures are a negative determinant, but as an observational study, it cannot be concluded as evidence of harm or non-inferiority between the two groups.

    However, the high mortality rate in the invasive group is per se our main finding. With a 1-year survival rate of 27%, it seems reasonable that a large portion of elderly patients may not be best served by invasive care. This calls into question the upward trend of time-limited trials that could lead to ethical misunderstanding, especially as the survival curves separated early on with a rapid decrease in the invasive group. The high proportion of early deaths in the invasive group most likely reflects the natural progression of an irreversible disease rather than complications from the procedure. This interpretation is supported by the strong culture of safety and patient-family engagement in ICU care, which prioritises harm prevention and appropriate treatment escalation.25

    The need for invasive care is not merely a treatment choice but also a marker of increased severity. Thus, the poor prognosis associated with invasive care is expected and reflects underlying disease severity rather than the direct impact of invasiveness itself. However, if invasiveness reflects increased severity, then a strict dichotomy between invasive and non-invasive approaches may seem overly simplistic, given that the relationship between the degree of invasiveness and prognosis is likely more gradual. This is supported by the sigmoidal relationship between SAPS II and mortality,26 resembling a dose–response effect. Nevertheless, we deliberately adopted this classification to avoid borderline situations and to describe the two extremes of the therapeutic spectrum in intensive care.

    Admittedly, the non-invasive group likely comprises two distinct populations: patients who are not severe enough to require invasive intervention and those who are too critically ill to benefit from it. However, when considered as a whole, the non-invasive group remains fundamentally distinct from the invasive cohort, where a paradigm shift in care is always required. The two populations we studied differ not in the degree of care received but in the nature of care itself. This distinction is crucial, as the decision between invasive and non-invasive management is never incidental27; it is always an active choice that carries significant implications for the patient, their family and the medical team. For this reason, we did not include different modalities of oxygenation, such as high-flow nasal insufflation or NIV, in our analysis.

    Furthermore, we aimed for the invasive group to reflect the maximal therapeutic investment—at least at the moment of procedure initiation. In this regard, NIV is a treatment that has been shown to improve survival and, as such, is not purely a palliative intervention. However, its role in alleviating respiratory discomfort is increasingly well documented. NIV has been found to be more effective than oxygen in reducing dyspnoea and decreasing the need for morphine in palliative care patients.28 This modality is used both to enhance comfort at the end of life29 and for patients who have declined tracheal intubation.30 More broadly, it is now integrated into an approach that seeks to improve palliative care in the ICU.31

    Finally, this distinction between invasive and non-invasive care also implies the notion of immediacy, which is why we excluded chronic conditions requiring invasive procedures. Both NIV and RRT are commonly used in chronic settings—NIV for chronic respiratory failure (eg, amyotrophic lateral sclerosis) and RRT for end-stage renal disease. The coding system does not allow for differentiation between acute and chronic indications. Including these cases would have introduced additional confounding factors, whereas our focus was on invasive procedures initiated for acute conditions. Nevertheless, our findings suggest that renal support worsens prognosis regardless of the treatment group, as among the 8761 patients who underwent RRT, 74% died within 1 year.

    The strength of this study is the size of the cohort with few missing data and the national-level database needed to overcome sampling biases32 that often limit epidemiological studies of the critically ill elderly. With 107 014 patients under observation, we had the statistical power to robustly assess this issue in the target population of medical ICU patients adjusted for known prognostic factors specified a priori. The comprehensive multicentre design enhances the generalisability of our findings, particularly important given that older patients are frequently excluded from clinical trials33 but are commonly treated in the ICU. Our study further provides valuable information on the epidemiology of circulatory and respiratory failures in older patients.

    The main limitation of our study is the risk of confounding inherent to any observational study, which the PS cannot entirely eliminate. As a summary of measured covariates, the PS cannot account for unmeasured confounding.34 In contrast to randomised controlled trials, where randomisation is expected to balance both measured and unmeasured covariates across treatment arms, observational studies remain subject to residual confounding.35 Nonetheless, best practices for the use of PS methods with survival or time-to-event36 outcomes using IPTW can help mitigate the effects of confounding in observational studies.37 38

    Residual confounding may result from the lack of detailed clinical or contextual data involved in the decision to initiate an invasive procedure. These parameters not only influence the decision itself but also impact patient outcomes. The initiation of vasopressor therapy relies on critical information, such as arterial pressure, heart rate and urinary output, while the decision to proceed with invasive mechanical ventilation depends on the patient’s level of consciousness, signs of respiratory failure and oxygen saturation. Comparing patients without this information is challenging. Nevertheless, although our methods do not directly incorporate these physiological parameters individually, they each contribute to the SAPS II score, which serves as the primary severity adjustment factor in our PS model. Notably, the density plots indicate that the SAPS II distribution is not only the strongest determinant of risk in our model but also closely aligned between invasive and conservative groups after weighting (online supplemental figure S3). In addition, the reason for admission integrates some clinical data.

    We acknowledge that clinical parameters influencing invasive procedures are strong predictors of mortality, regardless of the level of invasiveness. Moreover, the need for invasive care serves as a marker of disease severity rather than a direct determinant of prognosis. In our study, the two populations inherently differ in acute severity, which is a key driver of treatment decisions. However, our aim was not to assess the appropriateness of invasive procedures but rather to describe patient outcomes based on treatment intensity, once a treatment pathway has been adopted in real-world practice.

    The predominant contribution of SAPS II to our PS substantially mitigates the risk of confounding, although it cannot be broken down into its circulatory and ventilatory components, which would have allowed for a more refined adjustment. Nevertheless, SAPS II integrates various parameters that are associated with both prognosis and the decision to initiate an invasive procedure. This duality reflects the inherent overlap between prognosis and therapeutic decision-making in intensive care, where addressing clinical imbalance is expected to neutralise risk.

    For instance, in haemodynamics, the 65 mm Hg mean arterial pressure threshold was originally established based on the retrospective cohort studies39 40 that demonstrated a strong association between time spent below this threshold and mortality in patients with septic shock. However, vasopressors have since been shown to exert pleiotropic effects,41 making their overall impact more difficult to predict, or even potentially leading to increased mortality in older patients.42

    Thus, the traditional perspective in which invasive procedures were assumed to have an inherently positive—or at worst, neutral—impact by counteracting the excess risk associated with a high severity score has evolved. It is now recognised that their effect is far less predictable and, particularly in older patients, unlikely to be neutral.43

    Our aim was to describe the trajectories of two distinct ICU populations rather than to compare treatment efficacy. Thanks to the inclusion of all treated patients nationwide over a 6-year period, any biases in our data are more likely to reflect the overall clinical practice rather than individual decisions regarding the appropriateness of invasive or conservative treatment. This is arguably the main advantage of real-world evidence over randomised trials in informing public health decision-making.44

    Other limitations of our study include potential biases due to coding errors. Moreover, our study did not include the functional status nor the quality of life among survivors, which are more important than survival for many older persons.45–47 The study also lacks consideration for pre-existing frailty, despite its recognised importance as a determinant of mortality.48–50 However, while elderly patients with a higher functional baseline are more likely to survive, their chances of returning to their prior level are reduced compared with those with a lower functional baseline.51 52 In addition, whoever with a good functional status is resuscitated until it becomes poor is doomed to die with a poor functional status.

    The absence of consideration for triage process, treatment appropriateness or therapeutic limitations may introduce selection bias but reflects the real-world shared decision-making process among physicians, patients and families. Current studies suggest that the risk of overutilisation of invasive procedures outweighs the risk of underutilisation.22 24 53–55

    The Principles of Biomedical Ethics by Beauchamp and Childress, developed in 1979, outlines the four principles of medical care: autonomy, non-maleficence, beneficence and justice. Individuals have a significant preference for non-maleficence over the other principles, but it does not appear to be directly used in the decision-making process.56 Advance care planning is essential to support clinicians in targeting appropriate invasive, rehabilitative or palliative strategies, which are not exclusive. As stated in a recent review: all critically ill patients, by definition, have serious illness and, thus, have palliative care needs.57 The retrospective meaning of ‘end-of-life care’ seems problematic unless considering the oldest patients are living the last part of their lives. Oldest patients, similar to paediatric patients, have distinct healthcare trajectories and goals supported by physiological evidence for geriatric specificities.

    To clarify, our study does not suggest that elderly patients should be denied ICU admission. On the contrary, it highlights the potential benefits of an intensive integrative approach in the ICU rather than a purely technical approach that may only prolong the dying process.58 The conservative group exhibits a low mortality rate, including patients who were refused invasive treatments due to futility, and it is unlikely to be further reduced with a more invasive approach. Similarly, the high mortality rate observed in the invasive group is unlikely to be exacerbated by a less aggressive strategy.

    Our findings have strong ethical implications. Despite being highly selected for favourable outcomes in real-world practice, patients who received invasive procedures had a low survival rate. The incentive care policy for a standard of care regardless of age is questionable in the light of medical appropriateness22 24 53–55 and patients' wishes.24–26 While age should not be the sole criterion for ICU triage, the combination of age and the need for an invasive procedure must be considered attentively when deciding on treatment options. Growing evidence suggests a gap between patient preferences and the actual care provided59 and that focusing resources on patient preferences is possible.60 Age remains a potent trigger for clinicians to ensure that patients and families are well informed about the benefits, risks and harms associated with invasive care.61

    Conclusions

    In conclusion, in our nationwide study, an invasive approach was associated with a lower 1-year survival rate among elderly patients who were admitted to the ICU for medical reasons compared with a conservative approach. Early discussion, including the requirement for an invasive procedure, may reduce the incidence of avoidable aggressive end-of-life care and improve goal-concordant care achievement.

    Data availability statement

    Data may be obtained from a third party and are not publicly available. The data used in this study are derived from the French National Health Data System, which is managed by the French National Health Insurance Fund. Access to these data is subject to regulatory approval and can be requested through the official data access platform under authorisation number TPS 962976. Further details on data access procedures can be obtained by contacting snds.cnam@assurance-maladie.fr.

    Ethics statements

    Patient consent for publication

    Ethics approval

    The study was submitted to and approved by the independent national data protection authority (CNIL Registration: 920181). All data were deidentified for research purposes, and French law waives the need for informed consent. An independent institutional review board approved the study protocol (IRB Registration: #00011928).

    References

    1. ↵
      1. Vallet H,
      2. Schwarz GL,
      3. Flaatten H, et al
      . Mortality of older patients admitted to an ICU: a systematic review. Crit Care Med 2021;49:324–34. doi:10.1097/CCM.0000000000004772
      OpenUrl
    2. ↵
      1. Baldwin MR
      . Measuring and predicting long-term outcomes in older survivors of critical illness. Minerva Anestesiol 2015;81:650–61.
      OpenUrl
    3. ↵
      1. Boumendil A,
      2. Latouche A,
      3. Guidet B, et al
      . On the benefit of intensive care for very old patients. Arch Intern Med 2011;171:1116–7. doi:10.1001/archinternmed.2011.102
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      1. Biddison LD,
      2. Berkowitz KA,
      3. Courtney B, et al
      . Ethical considerations: care of the critically ill and injured during pandemics and disasters: CHEST consensus statement. Chest 2014;146:e145S–55S. doi:10.1378/chest.14-0742
      OpenUrlCrossRefPubMed
    5. ↵
      1. Soliman IW,
      2. Cremer OL,
      3. de Lange DW, et al
      . The ability of intensive care unit physicians to estimate long-term prognosis in survivors of critical illness. J Crit Care 2018;43:148–55. doi:10.1016/j.jcrc.2017.09.007
      OpenUrl
    6. ↵
      1. Vink EE,
      2. Azoulay E,
      3. Caplan A, et al
      . Time-limited trial of intensive care treatment: an overview of current literature. Intensive Care Med 2018;44:1369–77. doi:10.1007/s00134-018-5339-x
      OpenUrlCrossRefPubMed
    7. ↵
      1. Guidet B,
      2. Vallet H,
      3. Boddaert J, et al
      . Caring for the critically ill patients over 80: a narrative review. Ann Intensive Care 2018;8:114. doi:10.1186/s13613-018-0458-7
    8. ↵
      1. Ferrante LE,
      2. Pisani MA,
      3. Murphy TE, et al
      . Functional trajectories among older persons before and after critical illness. JAMA Intern Med 2015;175:523–9. doi:10.1001/jamainternmed.2014.7889
      OpenUrl
    9. ↵
      1. Flaatten H,
      2. de Lange DW,
      3. Artigas A, et al
      . The status of intensive care medicine research and a future agenda for very old patients in the ICU. Intensive Care Med 2017;43:1319–28. doi:10.1007/s00134-017-4718-z
      OpenUrlCrossRef
    10. ↵
      1. Seethala RR,
      2. Blackney K,
      3. Hou P, et al
      . The association of age with short-term and long-term mortality in adults admitted to the intensive care unit. J Intensive Care Med 2017;32:554–8. doi:10.1177/0885066616658230
      OpenUrl
    11. ↵
      1. Nathanson BH,
      2. Higgins TL,
      3. Brennan MJ, et al
      . Do elderly patients fare well in the ICU? Chest 2011;139:825–31. doi:10.1378/chest.10-1233
      OpenUrlCrossRefPubMed
    12. ↵
      1. Wunsch H,
      2. Guerra C,
      3. Barnato AE, et al
      . Three-year outcomes for Medicare beneficiaries who survive intensive care. JAMA 2010;303:849–56. doi:10.1001/jama.2010.216
      OpenUrlCrossRefPubMedWeb of Science
    13. ↵
      1. Jubran A,
      2. Grant BJB,
      3. Duffner LA, et al
      . Long-term outcome after prolonged mechanical ventilation. A long-term acute-care hospital study. Am J Respir Crit Care Med 2019;199:1508–16. doi:10.1164/rccm.201806-1131OC
      OpenUrlCrossRefPubMed
    14. ↵
      1. Biston P,
      2. Aldecoa C,
      3. Devriendt J, et al
      . Outcome of elderly patients with circulatory failure. Intensive Care Med 2014;40:50–6. doi:10.1007/s00134-013-3121-7
      OpenUrlCrossRefPubMed
    15. ↵
      1. Fassier T,
      2. Duclos A,
      3. Abbas-Chorfa F, et al
      . Elderly patients hospitalized in the ICU in France: a population-based study using secondary data from the national hospital discharge database. J Eval Clin Pract 2016;22:378–86. doi:10.1111/jep.12497
      OpenUrl
    16. ↵
      1. Atramont A,
      2. Lindecker-Cournil V,
      3. Rudant J, et al
      . Association of age with short-term and long-term mortality among patients discharged from intensive care units in France. JAMA Netw Open 2019;2:e193215. doi:10.1001/jamanetworkopen.2019.3215
    17. ↵
      1. Laporte L,
      2. Hermetet C,
      3. Jouan Y, et al
      . Ten-year trends in intensive care admissions for respiratory infections in the elderly. Ann Intensive Care 2018;8:84. doi:10.1186/s13613-018-0430-6
    18. ↵
      1. Gaudry S,
      2. Hajage D,
      3. Schortgen F, et al
      . Initiation strategies for renal-replacement therapy in the intensive care unit. N Engl J Med 2016;375:122–33. doi:10.1056/NEJMoa1603017
      OpenUrlCrossRefPubMed
    19. ↵
      1. Hernán MA,
      2. Brumback BA,
      3. Robins JM
      . Estimating the causal effect of zidovudine on CD4 count with a marginal structural model for repeated measures. Stat Med 2002;21:1689–709. doi:10.1002/sim.1144
      OpenUrlCrossRefPubMedWeb of Science
    20. ↵
      1. Austin PC,
      2. Thomas N,
      3. Rubin DB
      . Covariate-adjusted survival analyses in propensity-score matched samples: Imputing potential time-to-event outcomes. Stat Methods Med Res 2020;29:728–51. doi:10.1177/0962280218817926
      OpenUrl
    21. ↵
      1. Auriemma CL,
      2. Van den Berghe G,
      3. Halpern SD
      . Less is more in critical care is supported by evidence-based medicine. Intensive Care Med 2019;45:1806–9. doi:10.1007/s00134-019-05771-2
      OpenUrlCrossRefPubMed
    22. ↵
      1. Chang DW,
      2. Shapiro MF
      . Association between intensive care unit utilization during hospitalization and costs, use of invasive procedures, and mortality. JAMA Intern Med 2016;176:1492–9. doi:10.1001/jamainternmed.2016.4298
      OpenUrl
    23. ↵
      1. Guidet B,
      2. Leblanc G,
      3. Simon T, et al
      . Effect of systematic intensive care unit triage on long-term mortality among critically Ill elderly patients in France: a randomized clinical trial. JAMA 2017;318:1450–9. doi:10.1001/jama.2017.13889
      OpenUrlCrossRef
    24. ↵
      1. Sjoding MW,
      2. Prescott HC,
      3. Wunsch H, et al
      . Hospitals with the highest intensive care utilization provide lower quality pneumonia care to the elderly. Crit Care Med 2015;43:1178–86. doi:10.1097/CCM.0000000000000925
      OpenUrl
    25. ↵
      1. Thornton KC,
      2. Schwarz JJ,
      3. Gross AK, et al
      . Preventing harm in the ICU-building a culture of safety and engaging patients and families. Crit Care Med 2017;45:1531–7. doi:10.1097/CCM.0000000000002556
      OpenUrl
    26. ↵
      1. Aegerter P,
      2. Boumendil A,
      3. Retbi A, et al
      . SAPS II revisited. Intensive Care Med 2005;31:416–23. doi:10.1007/s00134-005-2557-9
      OpenUrlCrossRefPubMedWeb of Science
    27. ↵
      1. Burns KEA,
      2. Cook DJ,
      3. Xu K, et al
      . Differences in directives to limit treatment and discontinue mechanical ventilation between elderly and very elderly patients: a substudy of a multinational observational study. Intensive Care Med 2023;49:1181–90. doi:10.1007/s00134-023-07188-4
      OpenUrl
    28. ↵
      1. Nava S,
      2. Ferrer M,
      3. Esquinas A, et al
      . Palliative use of non-invasive ventilation in end-of-life patients with solid tumours: a randomised feasibility trial. Lancet Oncol 2013;14:219–27. doi:10.1016/S1470-2045(13)70009-3
      OpenUrlCrossRefPubMedWeb of Science
    29. ↵
      1. Azoulay E,
      2. Kouatchet A,
      3. Jaber S, et al
      . Non-invasive ventilation for end-of-life oncology patients. Lancet Oncol 2013;14:e200–1. doi:10.1016/S1470-2045(13)70141-4
      OpenUrlPubMed
    30. ↵
      1. Azoulay E,
      2. Kouatchet A,
      3. Jaber S, et al
      . Noninvasive mechanical ventilation in patients having declined tracheal intubation. Intensive Care Med 2013;39:292–301. doi:10.1007/s00134-012-2746-2
      OpenUrlCrossRefPubMedWeb of Science
    31. ↵
      1. Puntillo K,
      2. Nelson JE,
      3. Weissman D, et al
      . Palliative care in the ICU: relief of pain, dyspnea, and thirst--a report from the IPAL-ICU Advisory Board. Intensive Care Med 2014;40:235–48. doi:10.1007/s00134-013-3153-z
      OpenUrlCrossRefPubMed
    32. ↵
      1. Hua M,
      2. Gershengorn HB,
      3. Wunsch H
      . Assessing delivery of mechanical ventilation: risks and benefits of large databases. Intensive Care Med 2020;46:2297–300. doi:10.1007/s00134-020-06214-z
      OpenUrl
    33. ↵
      1. Ivie RMJ,
      2. Vail EA,
      3. Wunsch H, et al
      . Patient eligibility for randomized controlled trials in critical care medicine: an international two-center observational study. Crit Care Med 2017;45:216–24:216. doi:10.1097/CCM.0000000000002061
      OpenUrl
    34. ↵
      1. Yang JY,
      2. Webster-Clark M,
      3. Lund JL, et al
      . Propensity score methods to control for confounding in observational cohort studies: a statistical primer and application to endoscopy research. Gastrointest Endosc 2019;90:360–9. doi:10.1016/j.gie.2019.04.236
      OpenUrl
    35. ↵
      1. Austin PC
      . Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med 2009;28:3083–107. doi:10.1002/sim.3697
      OpenUrlCrossRefPubMedWeb of Science
    36. ↵
      1. Austin PC
      . The use of propensity score methods with survival or time-to-event outcomes: reporting measures of effect similar to those used in randomized experiments. Stat Med 2014;33:1242–58. doi:10.1002/sim.5984
      OpenUrlCrossRefPubMed
    37. ↵
      1. Austin PC
      . An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res 2011;46:399–424. doi:10.1080/00273171.2011.568786
      OpenUrlCrossRefPubMedWeb of Science
    38. ↵
      1. Austin PC,
      2. Stuart EA
      . Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Stat Med 2015;34:3661–79. doi:10.1002/sim.6607
      OpenUrlCrossRefPubMed
    39. ↵
      1. Dünser MW,
      2. Takala J,
      3. Ulmer H, et al
      . Arterial blood pressure during early sepsis and outcome. Intensive Care Med 2009;35:1225–33. doi:10.1007/s00134-009-1427-2
      OpenUrlCrossRefPubMed
    40. ↵
      1. Maheshwari K,
      2. Nathanson BH,
      3. Munson SH, et al
      . The relationship between ICU hypotension and in-hospital mortality and morbidity in septic patients. Intensive Care Med 2018;44:857–67. doi:10.1007/s00134-018-5218-5
      OpenUrlCrossRef
    41. ↵
      1. Andreis DT,
      2. Singer M
      . Catecholamines for inflammatory shock: a Jekyll-and-Hyde conundrum. Intensive Care Med 2016;42:1387–97. doi:10.1007/s00134-016-4249-z
      OpenUrl
    42. ↵
      1. Futier E,
      2. Lefrant J-Y,
      3. Guinot P-G, et al
      . Effect of individualized vs standard blood pressure management strategies on postoperative organ dysfunction among high-risk patients undergoing major surgery: a randomized clinical trial. JAMA 2017;318:1346–57. doi:10.1001/jama.2017.14172
      OpenUrlCrossRefPubMed
    43. ↵
      1. Lamontagne F,
      2. Richards-Belle A,
      3. Thomas K, et al
      . Effect of reduced exposure to vasopressors on 90-day mortality in older critically Ill patients with vasodilatory hypotension: a randomized clinical trial. JAMA 2020;323:938–49. doi:10.1001/jama.2020.0930
      OpenUrlPubMed
    44. ↵
      1. Sheldrick RC
      . Randomized trials vs real-world evidence: how can both inform decision-making? JAMA 2023;329:1352–3. doi:10.1001/jama.2023.4855
      OpenUrlCrossRefPubMed
    45. ↵
      1. Philippart F,
      2. Vesin A,
      3. Bruel C, et al
      . The ETHICA study (part I): elderly’s thoughts about intensive care unit admission for life-sustaining treatments. Intensive Care Med 2013;39:1565–73. doi:10.1007/s00134-013-2976-y
      OpenUrlCrossRefPubMedWeb of Science
    46. ↵
      1. Behforouz HL,
      2. Drain PK,
      3. Rhatigan JJ
      . Rethinking the social history. N Engl J Med 2014;371:1277–9. doi:10.1056/NEJMp1404846
      OpenUrlCrossRefPubMed
    47. ↵
      1. Le Guen J,
      2. Boumendil A,
      3. Guidet B, et al
      . Are elderly patients’ opinions sought before admission to an intensive care unit? Results of the ICE-CUB study. Age Ageing 2016;45:303–9. doi:10.1093/ageing/afv191
      OpenUrlCrossRefPubMed
    48. ↵
      1. Flaatten H,
      2. De Lange DW,
      3. Morandi A, et al
      . The impact of frailty on ICU and 30-day mortality and the level of care in very elderly patients (≥ 80 years). Intensive Care Med 2017;43:1820–8. doi:10.1007/s00134-017-4940-8
      OpenUrlCrossRefPubMed
    49. ↵
      1. Guidet B,
      2. de Lange DW,
      3. Boumendil A, et al
      . The contribution of frailty, cognition, activity of daily life and comorbidities on outcome in acutely admitted patients over 80 years in European ICUs: the VIP2 study. Intensive Care Med 2020;46:57–69. doi:10.1007/s00134-019-05853-1
      OpenUrlCrossRefPubMed
    50. ↵
      1. Muscedere J,
      2. Waters B,
      3. Varambally A, et al
      . The impact of frailty on intensive care unit outcomes: a systematic review and meta-analysis. Intensive Care Med 2017;43:1105–22. doi:10.1007/s00134-017-4867-0
      OpenUrlCrossRefPubMed
    51. ↵
      1. Heyland DK,
      2. Garland A,
      3. Bagshaw SM, et al
      . Recovery after critical illness in patients aged 80 years or older: a multi-center prospective observational cohort study. Intensive Care Med 2015;41:1911–20. doi:10.1007/s00134-015-4028-2
      OpenUrlCrossRefPubMed
    52. ↵
      1. Griffiths J,
      2. Hatch RA,
      3. Bishop J, et al
      . An exploration of social and economic outcome and associated health-related quality of life after critical illness in general intensive care unit survivors: a 12-month follow-up study. Crit Care 2013;17:R100. doi:10.1186/cc12745
    53. ↵
      1. Teno JM,
      2. Gozalo P,
      3. Khandelwal N, et al
      . Association of increasing use of mechanical ventilation among nursing home residents with advanced dementia and intensive care unit beds. JAMA Intern Med 2016;176:1809–16. doi:10.1001/jamainternmed.2016.5964
      OpenUrl
    54. ↵
      1. Sullivan DR,
      2. Kim H,
      3. Gozalo PL, et al
      . Trends in noninvasive and invasive mechanical ventilation among medicare beneficiaries at the end of life. JAMA Intern Med 2021;181:93–102. doi:10.1001/jamainternmed.2020.5640
      OpenUrl
    55. ↵
      1. Guidet B,
      2. Flaatten H,
      3. Boumendil A, et al
      . Withholding or withdrawing of life-sustaining therapy in older adults (≥ 80 years) admitted to the intensive care unit. Intensive Care Med 2018;44:1027–38. doi:10.1007/s00134-018-5196-7
      OpenUrlCrossRefPubMed
    56. ↵
      1. Page K
      . The four principles: can they be measured and do they predict ethical decision making? BMC Med Ethics 2012;13:10. doi:10.1186/1472-6939-13-10
    57. ↵
      1. Aslakson RA,
      2. Cox CE,
      3. Baggs JG, et al
      . Palliative and end-of-life care: prioritizing compassion within the ICU and beyond. Crit Care Med 2021;49:1626–37. doi:10.1097/CCM.0000000000005208
      OpenUrl
    58. ↵
      1. Cox CE,
      2. Curtis JR
      . Using technology to create a more humanistic approach to integrating palliative care into the intensive care unit. Am J Respir Crit Care Med 2016;193:242–50. doi:10.1164/rccm.201508-1628CP
      OpenUrl
    59. ↵
      1. Heyland DK,
      2. Dodek P,
      3. Mehta S, et al
      . Admission of the very elderly to the intensive care unit: family members’ perspectives on clinical decision-making from a multicenter cohort study. Palliat Med 2015;29:324–35. doi:10.1177/0269216314566060
      OpenUrlCrossRefPubMed
    60. ↵
      1. Deniau N,
      2. Shojaei T,
      3. Georges A, et al
      . Home emergency response team for the seriously ill palliative care patient: feasibility and effectiveness. BMJ Support Palliat Care 2024;14:187–90. doi:10.1136/spcare-2023-004385
      OpenUrlAbstract/FREE Full Text
    61. ↵
      1. Leblanc G,
      2. Boumendil A,
      3. Guidet B
      . Ten things to know about critically ill elderly patients. Intensive Care Med 2017;43:217–9. doi:10.1007/s00134-016-4477-2
      OpenUrl

    Footnotes

    • X @matthieudestampa

    • Contributors This analysis is an investigator-initiated study led by CL, he supervised the analyses, drafted the manuscript and is the guarantor. All coauthors contributed to the study design and made substantial contribution. AG and CL conceptualised the study and prepared the original study protocol, which was subsequently reviewed by MdS and PA. CL, AG and PA developed the statistical methods. CL and PA had full access to the data and take responsibility for the integrity of the data and accuracy of the analysis. All authors contributed to data interpretation, critically revised the drafted manuscript and approved the final submitted version. The guarantor accepts full responsibility for the conduct of the study, had access to the data and controlled the decision to publish. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

    • Funding This study was supported by Assistance Publique–Hôpitaux de Paris through an agreement between the Caisse Nationale d’Assurance Maladie and HAD-APHP (signed on 19 October 2020) for the OCTO-Reverse project (962976). The funding sources had no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; preparation, review or approval of the manuscript; or decision to submit the manuscript for publication.

    • Competing interests None declared.

    • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.