Prediction of survival time after terminal extubation: the balance between critical care unit utilization and hospice medicine in the COVID-19 pandemic era

The data for this study were obtained from interdisciplinary palliative care team at Chang Gung Memorial Hospital in Keelung and Lovers’ Lake Branch, before and after palliative extubation from 2015 to 2020. All participants were terminally ill, defined as having end-of-life status and no chance of returning to a meaningful life based on the judgment of at least two specialist physicians. Eight patients and/or their families who refused to forgo life-sustaining therapies or withdraw mechanical ventilation were excluded. The terminal extubation process, consistent with the Hospice Palliative Care Act (Natural Death Act) Amendment, was initiated by the families or intensivists after consensus with the family and other medical staff. This retrospective 6-year study was approved by the Institutional Review Board of Chang Gung Medical Foundation Institution, and the requirement for participants’ informed consent was waived (IRB file No. 202101929B0).

Different clinical and demographic characteristics were summarized according to extubation status. All possible physiologic and respiratory parameters associated with survival time were calculated. Continuous data are expressed as mean ± standard deviation (SD) or median and interquartile ranges (IQR), and categorical variables are expressed as proportions. Vital signs and oxygen saturation were recorded at the discontinuation of mechanical ventilation. The APACHE II score was reassessed according to the latest laboratory data and physiological variables at extubation. Any unknown or out-of-date variables were scored as 0 points when calculating the APACHE II score.

Continuous variables were analyzed using the Mann–Whitney U test, with comparisons of medians for variables with skewed distribution. Categorical variables were compared using Pearson’s Chi-square test or Fisher’s exact test at survival times of 1 h and 24 h. We conducted Cox proportional hazards regression analysis to compute the hazard ratios (HRs) and 95% confidence intervals (CIs) to identify the independent predictors associated with death within 1 h and survival beyond 1 day. Two multiple linear regression models with backward stepwise elimination were conducted on all factors with a p-value ≤ 0.2 in univariate analysis. A two-sided p-value < 0.05 was considered to indicate statistical significance. Data analyses were performed using SPSS version 26.0 (IBM Corp., Armonk, NY).

The mean age of the study population was 67 years (range 19–94 years), and there were more males (62.1%) than females. Eighty-nine (63.6%) patients were treated with surgical services. Cerebrovascular accident was the main comorbidity (55.0%), followed by congestive heart failure (54.3%), renal failure (45.7%), chronic respiratory disease (42.1%), diabetes mellitus (40.7%), advanced malignancy (37.9%), and chronic liver disease (30.7%). The mean APACHE II score assessed at ICU admission was 25.2 (range 6–42), which increased to 31.3 (range 14–66) at extubation. The clinical characteristics are listed in Table 1.

The Kaplan–Meier curve for time to death in the 140 extubation patients is shown in Fig, 1B. The time to death after the extubation ranged from 0.02 to 401.72 h (median 0.79 h). Seventy-six patients (54.3%) died within 1 h, and 35 patients (25%) survived beyond 24 h. After extubation, most patients died in the ICU (72.1%), while others died in the ward, hospice and home according to individual circumstances. The mean ICU stay was 18.6 days (range 1–65 days). Of the eight patients (5.7%) whose family members were willing to donate organs, two (1.4%) eventually completed organ transplantation (Additional file 1).

The results of univariate analysis of continuous variables are listed in Table 2. The significant factors (p-value < 0.05) associated with death at 1 h of extubation were total Glasgow Coma Scale (GCS) score, diastolic blood pressure (DBP), mean arterial pressure (MAP), pulse rate, respiratory rate, fraction of inspired oxygen (FiO2) from the ventilator, positive end-expiratory pressure (PEEP), static pressure, minute ventilation (MV), and APACHE II score at extubation. The significant factors associated with death at 1 day were GCS score, systolic blood pressure (SBP), DBP, MAP, peripheral arterial oxygen saturation (SpO2) from pulse oximetry, FiO2 from the ventilator, PEEP, static pressure, and MV.

The results of univariate analysis of categorical variables are listed in Table 3. The significant factors associated with death at 1 h of extubation were comorbid cerebrovascular disease, chronic respiratory disease, advanced malignancy, SpO2 ≥ 96%, MV ≥ 12 L/min, spontaneous breathing trial (SBT) within the past 24 h, use of inotropic agents in the past 12 h, and APACHE II score at extubation ≥ 25 and ≥ 30. The significant factors associated with death at 1 day were SpO2 ≥ 96%, MV ≥ 12 L/min, SBT within the past 24 h, and the use of inotropic agents in the past 12 h.

The results of Cox proportional hazards regression analysis for survival at 1 h and 1 day are shown in Table 4. In the 1-h model, intubation duration, SpO2 from pulse oximetry, total dose of opioids within 24 h, total dose of BZDs within 24 h, education level, comorbid cerebrovascular accident had p-values ≤ 0.2 in the univariate analysis and were entered into the multivariate analysis. Items that overlapped including GCS, DBP, MAP, pulse rate, respiratory rate, FiO2 from the ventilator in APACHE II score were excluded. To increase clinical relevance, we used MV 12 L/min in regression multivariate analysis according to a previous study [13]. In the 1-day model, intubation duration, APACHE II score at extubation, and DNR orders signed before family meeting had p-values ≤ 0.2 in the univariate analysis and were included in the multivariate analysis. Items correlated with APACHE II score were removed. To enhance application, SpO2 < 96% was used in the analysis.

The final Cox regression model for death within 1 h showed that no SBT within the past 24 h, MV ≥ 12 L/min, and APACHE II score ≥ 25 were associated with higher mortality. Meanwhile, the model for survival beyond 1 day indicated that lower MV, SpO2 ≥ 96%, and SBT within the past 24 h were associated with longer survival.

Previous studies have reported that various factors are associated with time to death within 1 h after WLST, including age, FiO2, body temperature, MAP, blood pH, heart rate, respiratory rate, serum sodium, potassium, creatinine, white blood cell count, GCS and severe organ system insufficiency, which are also the main variables used to calculate the APACHE II score [3, 6, 12‐18]. The APACHE II scoring system is a simple and widely used reproducible ICU prognostic model, and our data showed that it could be used to predict survival time after compassionate extubation. Recalculation of the “Acute Physiology Score” part based on the updated status of the patient after extubation is relatively convenient for multidisciplinary teams. Although the reassessed APACHE II score did not reach significance in the 1-day model, an APACHE II score ≥ 25 closely predicted 1-h mortality (Fig. 2C). As an APACHE II score of 25 represents an approximately 50% mortality rate in clinical practice, a cutoff value of 25 has been well validated in predicting mortality in ventilator-associated pneumonia, emergency surgical patients, and patients with severe sepsis, carbon monoxide poisoning, and hematological cancer [7, 19‐21].

Our results are also consistent with discharge APACHE II score being superior to admission APACHE II score in predicting post-ICU mortality [22]. Given that more patient parameters and surgical status are taken into account in the APACHE II score, it can serve as a new prognostic tool for hospice care.

Respiratory variables are consistently associated with time to death [13, 15]. With regard to the items excluded from the APACHE II scoring system, MV was of greater significance in multivariate analysis compared with static pressure and PEEP. MV has also been shown to play an important role in predicting noninvasive ventilation failure in patients with early mild acute respiratory distress syndrome (ARDS) induced by pneumonia, successful extubation, and mortality caused by ARDS in patients with COVID-19 [23‐25].

SBT, an indicator for liberation from ventilation in different populations, is performed using T-piece ventilation and pressure support ventilation lasting between 0.5 and 2 h [26, 27]. Patients eligible for SBT are screened according to low FiO2 (< 0.5) and PEEP (< 5–8 cmH2O) requirements, stable hemodynamics, and the ability to initiate spontaneous breathing, all of which are also favorable predictors for longer survival after WLST [3, 18, 28]. In addition, the components of the SBT can be used to measure the burden of post-extubation symptoms and guide the anticipatory dose of medication after terminal extubation [29]. It is therefore reasonable that the subjects with lower MV and attempting SPT within 24 h had less dependency on mechanical ventilation and a longer survival (Fig. 2A, B).

Pulse oximetry is used to measure SpO2, and pulse oximeters are standard equipment in ICUs [30]. It is considered to be the “fifth vital sign” to monitor systemic oxygen delivery in a noninvasive and continuous fashion, especially in critically ill patients supported by extracorporeal membrane oxygenation and mechanical ventilation [31‐33]. A lower SpO2, compared with SpO2 ≥ 96%, has been associated with an increased risk of all-cause mortality in the general adult population [34]. In a systematic review conducted in 2018, oxygen therapy was associated with increased mortality in acutely ill adults with SpO2 > 96% [35]. The authors suggested that critically ill patients with SpO2 ≥ 96% have a lower oxygen demand and better compensation. Our study supports these findings, as SpO2 was the only factor significantly associated with 1-day survival (Fig. 2D). Moreover, the target of SpO₂ should be 92% to 96% in adults with COVID-19 who need supplemental oxygen according to treatment guidelines as home pulse oximetry has become increasingly popular during the COVID-19 era [36].

Our results showed that combining the APACHE II score with other respiratory parameters was effective in predicting 1-h mortality. Therefore, our model could be used to identify which terminal patients with irreversible illness should remain in the ICU without being transferred, regardless of comorbidities. Our 1-h model could help physicians to quickly detect suitable candidates for DCD. The 1-day model based on SpO2 could be used to identify patients with a likelihood of longer survival, as a significant minority are discharged alive after palliative extubation [5, 37]. Transition to a general care ward, hospice department ward or home where comfort-oriented care can be provided is suitable for patients who are predicted to survive for more than 1 day according to the consensus of family meetings [12] (Fig. 1B).

ICU facilities are important for patients with moderate and severe COVID-19 infection. When COVID-19 peaks occur, hospitals may run out of beds and other medical supplies. In this situation, available ICU beds are recruited by the government to avoid collapse of the health system. Hospital capacity is consequently reduced, and other medical practices including hospice care are also likely to be over-utilized. Under the “coexisting with the virus” and “zero severe cases” policy of the Ministry of Health and Welfare in Taiwan, subjects who are expected to survive for 1 h to 1 day can be transferred to palliative home care directly depending on religious needs and individual differences. Some studies have also suggested that ICU specialist opinion was closely associated with the time of death [16]. When staff are overwhelmed by the number of COVID patients, a simple global guide is needed to avoid overloading healthcare systems and the guilt that decision-making can create.

Furthermore, DNR orders, consideration of organ donation, and hospice medications for pain/symptom relief did not affect the time to natural death in the multivariate analysis. Adequate medications and supplies should be considered based on probable survival time to reduce distress and family anxiety during and after transfer from the ICU. Although not a perfect substitute, well-designed apps and online counseling are practical in home hospice practice [38].