Introduction

Up to 23% of patients die in the intensive care unit (ICU) worldwide, with decisions to forgo (withhold or withdraw) life-sustaining treatment (DFLST) preceding the majority of these deaths [1, 2]. The decision to forgo life-sustaining treatment in ICUs is influenced by a complex interplay of ethical, cultural, and medical factors [2,3,4]. The Ethicus-2 study, a worldwide study conducted between 2015 and 2016, including 12,850 patients who died or had limitation of treatments, highlighted significant regional variabilities in decisions to forgo life-sustaining treatment. In face of the considerable worldwide variation in end-of-life practices, the authors concluded that future research should investigate the causes of this variation with the aim to improve ethical practices. This was also highlighted in studies conducted in other former worldwide cohorts of various ICU patients [2,3,4,5,6,7]. Interestingly, a study highlighted that one of the factors associated with more DFLST was septic shock and pneumonia, suggesting that these practices seem to vary according to specific clinical conditions beyond country-level factors [3].

While these studies provide valuable insights, they also reveal the complexity and variability of DFLST due to the heterogeneous nature of the patient populations. By using the EUROBACT-2 cohort consisting of patients all admitted for hospital-acquired bloodstream infections (HABSI), we hypothesized that this approach would allow for a clearer understanding of the specific factors, including country, center, patient, and infection characteristics, that are associated with these critical DFLST.

Methods

EUROBACT-2 study design

We conducted a secondary analysis of the data from the EUROBACT-2 study, a prospective observational multicontinental cohort study conducted between June 2019 and January 2021. During this period, centers prospectively recruited patients, with a minimum of 10 consecutive HABSI patients or for a 3-month period [8]. Details on the study process, data collection, and variables can be found elsewhere [8]. Adult patients (≥ 18 years old) with a first episode of HABSI treated in the ICU were enrolled. A HABSI was defined as a positive blood culture sampled 48 h after hospital admission. Treatment in the ICU was defined as either the blood culture having been sampled in the ICU or the patient having been transferred to the ICU (i.e. in 48 h) for the treatment of the HABSI. For this analysis, we focused exclusively on non-COVID-19 patients who died in the hospital or within 28 days after HABSI. We focused on non-COVID-19 patients to avoid the confounding effects of the pandemic, in whom it has been shown that withholding therapies frequently preceded death [9].

The primary outcome of the present study was the occurrence of DFLST preceding death, which was a variable collected prospectively for every death in the EUROBACT-2 cohort.

Statistical analysis

First, a descriptive analysis of centers, patients' and pathogen’s characteristics according to the decision to forgo life-sustaining treatments was performed. Continuous variables were presented as median with interquartile range (IQR) and categorical variables as number of patients (n) and percentage (%). Chi-square or Fisher's exact tests were used to detect differences in categorical variables, as appropriate, and the Wilcoxon rank-sum test was used for continuous variables.

To identify factors associated with DFLST, we built a two-tiered hierarchical logistic mixed model. The effects of centre-based variables were included as random intercepts. We considered the hierarchical structure of the data, which may manifest as intraclass correlations. We performed univariable mixed logistic regression models to assess the association of each variable with DFLST.

Ethics

This study was approved by the ethics Committee from the Royal Brisbane & Women's Hospital Human Research (LNR/2019/QRBW/48376). Each study site then obtained ethical and governance approvals according to national and/or local regulations.

Results

Among the 333 centers recruited in the EUROBACT-2 study, 68 centers were excluded as they either did not include non-COVID-19 patients or did not report any death at day 28. Of the 1589 patients, 519 (32.7%) died, of which 191 (36.8%) had a decision to forgo life-sustaining treatment (Supplementary Table 1, Supplementary Fig. 1).

The median age was 65 (IQR 54–75), with 204 (39.3%) females. At the country level, significant geographical differences were noted (Fig. 1, Supplementary Table 1), with no DFLST reported in African countries and fewer in the Middle East compared to Western Europe, Australia and Asia. Patients with DFLST were more likely to be from countries with higher current health expenditure as a share of Gross Domestic Product (Supplementary Table 1). At the center level, patients with DFLST were less frequent in centers with open ICUs and higher patient-per-nurse ratios. At the patient level, patients with DFLST were older (median 67, IQR 58–76, vs. 64, IQR 51–75), but other factors such as Charlson comorbidity index, ICU admission reason, infection type, and mechanical ventilation use were not significantly different. No differences in pathogen groups were found between patients with DFLST and those without.

Fig. 1
figure 1

Proportion of hospital-acquired bloodstream infection mortality cases following a decision to forgo life-sustaining treatment. This map highlights the countries that reported Intensive Care Unit (ICU) deaths associated with Hospital-Acquired Bloodstream Infections (HABSI). The map displays the percentage of these deaths that occurred following a clinical decision to forgo life-sustaining treatment. Map created with Khαrtis (https://www.sciencespo.fr/cartographie/khartis/)

Full size image

Using univariable mixed logistic regression analysis (Table 1), only gross current health expenditure as a share of Gross Domestic Product (OR 1.79, 95% CI: 1.45–2.21, p < 0.01) and older age (OR 1.02, 95% CI: 1.002–1.05, p = 0.03) were associated with DFLST.

Table 1 Univariate analysis of center, patient, and pathogen factors associated with the decision to forgo life-sustaining treatment
Full size table

Discussion

These findings, in a worldwide population of critically ill ICU patients with HABSI, highlight the importance of economic and regional disparities on patient outcomes and ethical decision-making in ICUs. Once a center effect was considered, only age and health expenditure were significantly associated with DFLST, underscoring the importance of geographical differences. Interestingly, pathogen factors and source of infection were not associated with DFLST in HABSI patients.

The Ethicus-2 study [2], conducted between 2015 and 2016, highlighted significant regional variability in decisions to forgo life-sustaining treatment. This was also evident in studies conducted in other cohorts, where heterogeneous patient populations, while allowing for generalizing their findings to overall ICU patients, could present more challenges identifying specific factors associated with DFLST [2,3,4,5,6,7]. Conducted five years after Ethicus-2, our study benefits from including more centers and offers a better representation of African countries. Despite the smaller sample size, it provides a detailed assessment of the variables that matter in terms of DFLST.

Past studies have found that patients' severity upon ICU admission and comorbidities were associated with more DFLST, a finding not observed in our study[3]. Interestingly, our study also found that infection and pathogen factors are not significantly associated with DFLST, reinforcing that variability in these decisions is primarily attributed to center and country effects. This can be explained by several hypotheses, including cultural, religious, and hospital policy differences, emphasizing the need for context-specific guidelines [1,2,3].

Regarding the importance of age after accounting for a center effect, our results align with previous studies [2, 3]. It appears that even in older patient populations, age continues to matter as a previous study focused on DFLST in old critically ill patients in European countries found that in this specific population, age, frailty, admission SOFA score, and country were the most significant variables associated with decisions to forgo life-sustaining treatments [10].

Our study has several limitations. Due to the sample size, we could not assess for the presence of variabilities in DFLST within a single country, which has been previously highlighted as being important [3]. Additionally, while our data spans multiple countries and centers, due to the limited sample size and secondary analysis nature of the study, it may not fully represent each region's diversity, indicating that further studies with broader representation of centers and populations are warranted. We could not distinguish between withdrawal and withholding of life-sustaining treatment, which might have provided more nuanced insights. The initial code status of the patient upon ICU admission was not available, which could be a significant factor in DFLST when, for example, limited time contracts with patients are made. Lastly, while our study has the advantage of reducing the variability related to different diseases, the heterogeneity of other studies allows for findings to be more generalizable across a broader range of ICU patients and conditions.

In summary, our study underscores the critical impact of regional and economic disparities on ethical decision-making in ICUs worldwide. Future research and global policy development should consider these disparities to enhance equity in end-of-life care practices across different settings.