Early deep-to-light sedation versus continuous light sedation for critically ill patients with mechanical ventilation: a cohort study

doi:10.21203/rs.3.rs-3212526/v1

Download PDF

Research Article

Early deep-to-light sedation versus continuous light sedation for critically ill patients with mechanical ventilation: a cohort study

https://doi.org/10.21203/rs.3.rs-3212526/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

The sedation strategies have not been well established for patients with invasive mechanical ventilation (MV). This study compared alternative sedation strategies – including early deep-to-light sedation (DTLS), continuous deep sedation (CDS) and continuous light sedation (CLS, the currently recommended strategy) – on ventilator, ICU or hospital mortality.

Methods

A cohort study was conducted using two large validated ICU databases, including the Registry of Healthcare-associated Infections in ICUs in China (ICU-HAI) and the Medical Information Mart for Intensive Care (MIMIC). Patients who received MV for more than 3 days with one of three sedation strategies were included. Multi-variable survival analyses with inverse probability-weighted competing risk models were conducted separately for ICU-HAI and MIMIC cohorts. The adjusted estimates were pooled using fixed-effects model.

Results

In total, 6700 patients (2627 ICU-HAI, 4073 MIMIC) were included in cohort study, of whom 2689 received CLS, 2099 CDS and 1932 DTLS. Compared to CLS, DTLS was associated with lower ICU mortality (9.3% vs. 11.0%; pooled adjusted HR 0.78, 95% CI 0.66–0.94) and hospital mortality (16.0% vs. 14.1%; 0.86, CI 0.74-1.00); and CDS was associated with higher ventilator mortality (32.8% vs. 7.0%; 5.38, 4.49–6.46), ICU mortality (40.6% vs. 11.0%; 3.39, 2.95–3.90) and hospital mortality (46.8% vs. 14.1%; 3.27, 2.89–3.71) than CLS. All HRs were qualitatively consistent in both cohorts.

Conclusion

Compared to the continuous light sedation, early deep-to-light sedation strategy may improve patient outcomes, and continuous deep sedation was confirmed with poorer patient outcomes.

Dynamic sedation protocol

Critical care

Mechanical ventilation

Mortality

Sedation is an important strategy for patients with invasive mechanical ventilation (MV) in the intensive care units (ICU) ^1–3. Although commonly used, the optimal sedation strategy has not been well established ^4–6. In the earlier practice guideline (i.e., PADIS 2018) ¹, light sedation was weakly recommended over deep sedation; however, the beneficial effects of light sedation on time to extubation and tracheostomy were only marginally observed, and its impact on mortality yielded conflicting results.

Up to now, the depth of sedation remains controversial ^7–12, since the effects of sedation depth may not be polarized, and the timing and length of deep sedation may underlie the effect. On one hand, prolonged deep sedation may excessively suppress spontaneous respiratory effort, which impedes ventilator weaning and delays time to extubation ⁸. On the other hand, plausible biological mechanisms suggested that a short time of early deep sedation may have potential clinical benefits. Previously published studies showed that, at the MV initiation, deep sedation could comfort patients by relieving anxiety, stress and pain at the onset ^13,14, facilitating lung-protective mechanical ventilation, reducing breath-stacking asynchrony, and preventing patient self-extubation ^15,16.

Given these characteristics, a dynamic sedation strategy – that is an early deep sedation (e.g., within 48 hours of MV initiation) followed by a switch to light sedation – is biologically plausible to improve patient outcomes. No clinical evidence, however, is available to support this hypothesis. Although previous studies have examined the effects of early deep sedation on clinical outcomes, these studies did not take into account the subsequent sedation level. Among these studies, most patients with early deep sedation continuing to receive deep sedation in subsequent ICU stays^5,9,17. In fact, previous studies evaluated the effects of continuous deep sedation.

Therefore, we conducted an exploratory observational study aimed to investigate the effects of deep-to-light dynamic sedation, deep sedation, or light sedation on patient-important outcomes, using two large validated ICU databases, one from China and another (MIMIC) from the United States.

This study was approved by the Ethical Committee of West China Hospital (WCH2018-409). It was reported according to the REporting of studies Conducted using Observational Routinely-collected Data (RECORD) Checklist¹⁸.

Data source

We conducted a retrospective cohort study based on two large-scale validated ICU databases: the Registry of Healthcare-associated Infections in ICUs in China (ICU-HAI) and the Medical Information Mart for Intensive Care (MIMIC) -IV database.

The ICU-HAI registry database included all patients admitted to six ICUs (general, surgical, respiratory, neurological, thoracic, and pediatric) in the West China Hospital (WCH) system. The WCH system is a national critical care centre. There were over 8300 ICU admissions per year, and approximately 93% of them had undergone MV. The ICU-HAI registry was established by integrating data from three systems (electronic medical record (EMR) system, ICU-HAI system, and ICU system) using the patient’s unique ID. This registry contained over 1000 variables with detailed information on demographics, laboratory and microbiology results, diagnosis, treatment, outcome, hospital costs, etc. Among these, many variables were recorded hourly by well-trained specialist nurses, such as the Richmond Agitation-Sedation Scale (RASS), vital signs, fluid intake and output volume, and life support (oxygen, mechanical ventilation, positive end-expiratory pressure, and fraction of inspired oxygen). The ICU-HAI registry has been validated and proven to be of high quality. It has supported various ICU-related studies ^19–22. A detailed database profile of the ICU-HAI registry was described in our previously published study ²³. In this study, we used data from April 2015 to December 2018, which contained information on 28,848 ICU patients.

MIMIC-IV is sourced from two data systems (an EHR system and an ICU system) of the Beth Israel Deaconess Medical Center in Boston, Massachusetts. The data profile of MIMIC-IV is available elsewhere²⁴. In this study, MIMIC-IV v2.0 was used, containing 76,943 ICU admissions of 69,639 patients from 2008 to 2019, with abundant clinical and intensive care information.

Study population and sedation strategies

We identified ICU patients who consecutively received MV for at least three days from the ICU-HAI registry and MIMIC-IV. We excluded patients who were missing basic information such as age, sex and diagnosis at discharge, or who were under 18 years of age. To reduce indication bias, we also excluded patients who underwent MV for less than three days, stayed in the ICU for more than 30 days, or were in a coma (RASS = -5) before MV treatment.

We defined three sedation strategies according to the daily RASS within 7 days after the initiation of MV, and the strategies included continuous light sedation (CLS), light to deep sedation (LTDS) and continuous deep sedation (CDS). CLS referred to light sedation within 2 days after initiation of MV followed by nondeep sedation; LTDS referred to deep sedation within 2 days after initiation of MV followed by nondeep sedation; and CDS was defined as continuous deep sedation after MV initiation. We categorised the depth of sedation according to the mean daily value of the RASS: deep sedation (RASS≦-3), light sedation (-3 < RASS≦--1), calm (-1 < RASS < 1) and restless/agitated (RASS ≧ 1) ^1,25.

Study outcomes

Our outcomes included ventilator mortality, ICU mortality and hospital mortality. Patients who died one day after extubation were defined as ventilator mortality, indicating extubation failure. Similarly, patients who died one day after ICU discharge were defined as ICU mortality.

Covariates

Potential confounding covariates included demographic information (age and sex), ICU type, Acute Physiology and Chronic Health Evaluation (APACHE) II score at ICU admission, comorbidities (diabetes, ischaemic heart disease (IHD), heart failure, kidney failure, renal failure, pulmonary vascular disease, hypertension), acute conditions (acute respiratory distress syndrome (ARDS), gastrointestinal bleeding, shock, pneumonia, sepsis, trauma), and treatments. Treatments included prescriptions (sedatives, neuroleptics, opioids, antithrombotics, neuromuscular blockers, acid inhibitors, antibiotics, intestinal probiotics, immunosuppressants, vasopressors), surgery (cranial or cardiac surgery) and other interventions (fibreoptic bronchoscopy, tracheotomy, enteral nutrition, gastrointestinal decompression and head-of-bed elevation).

For the MIMIC dataset, ethnicity (white, Asian, black, Latin American and other) was also taken into account. The MIMIC dataset did not collect the APACHE II score, instead, we used two other clinical scores, the Oxford Acute Severity of Illness Score (OASIS) and Acute Physiology Score III (APSIII), to assess the health status of patients on ICU admission.

Demographic information, type of ICU, clinical scores on ICU admission and comorbidity/acute conditions were measured as time-fixed variables. Treatments were measured as time-dependent variables, which were recorded daily from admission to discharge.

Statistical analyses

All analyses were conducted separately in the ICU-HAI and MIMIC cohorts, and the adjusted hazard ratios (HRs) of three sedation strategies were pooled through meta-analysis.

We described the sedation pattern of patients who received consecutive MV for at least 3 days by plotting the individual trajectory of the sedation depth within 30 days of MV initiation. The patients were classified into 3 categories according to the clinical outcome within 30 days: dead, successfully extubated and still requiring ventilatory support.

We compared the characteristics and clinical outcomes among patients who underwent three sedation strategies of interest using descriptive analyses and univariate tests. We depicted numeric variables as the means (standard deviation) and categorical variables as frequencies (percentages). The Student’s t test was applied to assess differences in the means. The \({\chi }^{2}\) test or Fisher’s exact test was used to assess differences in percentages.

We assessed the effects of the three sedation strategies on ventilator mortality, ICU mortality and hospital mortality. To control the time-dependent confounding affected by time-dependent sedation status, we conducted a weighted survival analysis ²⁶. This analysis involved three steps. First, we calculated the daily probability of patients being exposed to different sedation depths by estimating the propensity score (PS) for multivalued treatments. As the ICU data were highly complex with nonlinearity, interactions, etc., the XGBoost algorithm was used to generate the PS. Second, we constructed the inverse probability of treatment weights based on the cumulative probability of the sedation trajectory during MV ²⁷. Finally, a Fine-Gray subdistribution hazard competing risk model combined with inverse probability weighting (IPTW) was adopted to analyse the competing risk of extubation alive versus ventilator mortality, ICU discharge alive versus ICU mortality, and hospital discharge alive versus hospital mortality²⁸. Based on the Fine-Gray model weighted by IPTW, marginal structural models (MSMs) were used to assess the risk of sedation strategies on ventilator mortality, and doubly robust models were applied to analyse ICU mortality and hospital mortality with adjustment for confounders both during MV and after extubation ²⁹. The sedation effects for each mortality outcome estimated in the ICU-HAI and MIMIC cohorts were pooled using fixed-effect meta-analysis. Missing data were imputed by the last observation carried forwards.

Sensitivity analyses

We conducted six sensitivity analyses to test the robustness of our results: 1) we combined the estimated effects for each outcomes through random effect meta analysis; 2)we adopted multiple imputation to tackle missing data; 3) we assessed the association using complete cases without imputation; 4) we used an alternative popular boosting algorithm, namely the gradient boosting machine (gbm), for weighting; 5) we employed a time-dependent Cox model regardless of the competing risk; and 6) we used an alternative definition of deep to light sedation strategy, that is deep sedation within 1 day of MV initiation and then switching to nondeep sedation. In the last five sensitivity analyses, the estimated effects of ICU-HAI cohort and MIMIC cohort were combined by fix-effect meta analysis same as the main analysis.

We used R packages twang, RISCA and survival, version 4.2.2. A two-sided P < 0.05 was considered to be statistically significant. Data analysis was conducted from August 2022 to April 2023.

A total of 13,351 ICU patients receiving MV for at least 3 days were identified from the two databases (7058 ICU-HAI registry and 6293 MIMIC-IV), who were included in the analyses of sedation patterns. By further checking against eligibility criteria, 6700 patients (2627 ICU-HAI registry and 4073 MIMIC-IV) were finally included for investigating the effects of three sedation strategies (Fig. 1).

Sedation patterns among ICU patients receiving MV

The sedation patterns during MV were shown in Fig. 2. In the ICU-HAI registry, 21.0% experienced deep sedation within 2 days of MV initiation, and 77.3% of them continued to receive deep sedation after 2 days of MV; 82.1% achieved unassisted breathing, 11.9% died within 30 days after MV initiation, and the rest (5.6%) remained intubated.

In the MIMIC-IV database, 48.4% received deep sedation within 2 days of MV initiation, and 53.4% of them continued to receive deep sedation after 2 days of MV; 69.7% achieved unassisted breathing, 30.0% died within 30 days after MV initiation, and only 0.3% remained intubated. The details about sedation pattern were presented in Table S1.

Clinical characteristics of eligible patients in the two cohorts

Table 1 presented the clinical characteristics of eligible patients in the two cohorts. In the ICU-HAI cohort, 2627 patients were included (1507 undergoing CLS, 570 DTLS, and 550 CDS), with a mean age of 56 and a male proportion of 61.7%. The mean APACHE II score was 19.6 (8.6), which was higher in the CDS (23.6) than CLS (18.4) and DTLS groups (17.9). The incidence of acute conditions was similar among the three sedation groups, but the proportion of patients with heart failure, chronic lung disease, pulmonary vascular disease, liver failure and hypertension significantly differed.

Table 1

Clinical characteristics among patients with three sedation strategies of interest in the ICU-HAI and MIMIC cohorts
	Total	CLS	DTLS	CDS	P-value
Data from the ICU-HAI cohort
	(n = 2627)	(n = 1507)	(n = 570)	(n = 550)
Age, years: mean (s.d.)	56.0 (19.1)	55.3 (20.6)	56.0 (18.5)	57.2 (17.5)	0.204
Sex: n (%)
Male	2171 (64.7)	932 (61.8)	374 (65.6)	376 (68.4)	0.016
Female	1187 (35.3)	575 (38.2)	196 (34.4)	174 (31.6)
ICU ward: n (%)
General ICU	1066 (40.6)	531 (35.2)	200 (35.1)	335 (60.9)	< 0.001
Neurological ICU	195 (7.4)	171 (11.3)	16 (2.8)	8 (1.5)
Pediatric ICU	79 (3.0)	63 (4.2)	12 (2.1)	4 (0.7)
Respiratory ICU	572 (21.8)	362 (24.0)	96 (16.8)	114 (20.7)
Surgical ICU	508 (19.3)	330 (21.9)	125 (21.9)	53 (9.6)
Thoracic surgery ICU	207 (7.9)	50 (3.3)	121 (21.2)	36 (6.5)
APACHEII: mean (s.d.)	19.6 (8.6)	18.4 (8.4)	17.9 (8.4)	23.6 (8.1)	< 0.001
Acute conditions: n (%)
ADRS	50 (1.5)	20 (1.3)	6 (1.1)	14 (2.5)	0.186
Shock	259 (7.7)	124 (8.2)	37 (6.5)	43 (7.8)	0.614
Gastrointestinal bleeding	74 (2.2)	32 (2.1)	12 (2.1)	16 (2.9)	0.654
Pneumonia	420 (12.5)	179 (11.9)	71 (12.5)	77 (14.0)	0.638
Trauma	276 (10.5)	165 (10.9)	67 (11.8)	44 (8.0)	0.085
Sepsis	144 (5.5)	84 (5.6)	18 (3.2)	42 (7.6)	0.004
Comorbidity: n (%)
Diabetes	281 (8.4)	142 (9.4)	44 (7.7)	45 (8.2)	0.191
Ischemic heart disease	64 (1.9)	21 (1.4)	13 (2.3)	12 (2.2)	0.081
Heart failure	241 (7.2)	72 (4.7)	77 (13.5)	27 (4.9)	< 0.001
Chronic lung disease	406 (12.1)	221 (14.7)	60 (10.5)	53 (9.6)	< 0.001
Pulmonary vascular disease	301 (9.0)	148 (9.8)	66 (11.6)	31 (5.6)	0.002
Cancer	305 (9.1)	148 (9.8)	40 (7.0)	48 (8.7)	0.247
Liver failure	95 (2.8)	28 (1.9)	11 (1.9)	32 (5.8)	< 0.001
Hypertension	600 (17.9)	299 (19.8)	120 (21.1)	74 (13.5)	< 0.001
Kidney failure	255 (7.6)	100 (6.6)	39 (6.8)	55 (10.0)	0.056
Data from the MIMIC cohort
	(n = 4073)	(n = 1182)	(n = 1362)	(n = 1529)
Age, years: mean (s.d.)	61.7 (16.7)	63.1 (16.7)	62.0 (16.6)	60.2 (16.6)	< 0.001
Sex: n (%)
Male	2348 (57.6)	627 (53.0)	809 (59.4)	912 (59.6)	< 0.001
Female	1725 (42.4)	555 (47.0)	553 (40.6)	617 (40.4)
ICU ward: n (%)
Cardiac Vascular ICU	497 (12.2)	48 (4.1)	212 (15.6)	237 (15.5)	< 0.001
Coronary Care Unit	471 (11.6)	116 (9.8)	148 (10.9)	207 (13.5)
Medical ICU	1126 (27.6)	334 (28.3)	350 (25.7)	442 (28.9)
Medical/Surgical ICU	578 (14.2)	219 (18.5)	180 (13.2)	179 (11.7)
Neuro Surgical ICU/ Neuro Intermediate/ Neuro Stepdown	197 (4.8)	53 (4.5)	71 (5.2)	73 (4.8)
Surgical ICU	550 (13.5)	159 (13.5)	194 (14.2)	197 (12.9)
Others	654 (16.1)	253 (21.4)	207 (15.2)	194 (12.7)
APSIII: mean (s.d.)	70.4 (29.0)	61.3 (23.1)	63.1 (25.6)	84.0 (30.7)	< 0.001
OASIS: mean (s.d.)	41.6 (8.3)	39.8 (7.9)	40.4 (7.8)	44.1 (8.3)	< 0.001
Acute conditions: n (%)
ADRS	91 (2.2)	18 (1.5)	15 (1.1)	58 (3.8)	< 0.001
Shock	1735 (42.6)	454 (38.4)	532 (39.1)	749 (49.0)	< 0.001
Gastrointestinal bleeding	168 (4.1)	44 (3.7)	44 (3.2)	80 (5.2)	0.019
Pneumonia	891 (21.9)	240 (20.3)	240 (17.6)	411 (26.9)	< 0.001
Trauma	697 (17.1)	172 (14.6)	195 (14.3)	330 (21.6)	< 0.001
Sepsis	1324 (32.5)	387 (32.7)	392 (28.8)	545 (35.6)	< 0.001
Comorbidity: n (%)
Diabetes	780 (19.2)	245 (20.7)	278 (20.4)	257 (16.8)	0.013
Ischemic heart disease	218 (5.4)	58 (4.9)	77 (5.7)	83 (5.4)	0.696
Heart failure	1216 (27.9)	388 (32.8)	427 (31.4)	401 (26.2)	< 0.001
Pulmonary vascular disease	240 (5.9)	82 (6.9)	83 (6.1)	75 (4.9)	0.078
Cancer	206 (5.1)	70 (5.9)	80 (5.9)	56 (3.7)	0.007
Liver failure*	9 (0.2)	2 (0.1)	3 (0.2)	4 (0.3)	0.919
Hypertension	1573 (38.6)	440 (37.2)	510 (37.4)	623 (40.7)	0.096
Kidney failure	925 (22.7)	286 (24.2)	327 (24.0)	312 (20.4)	0.024
CLS: continuously light sedation; DTLS: deep to light sedation; CDS: continuously deep sedation;
ICU: intensive care unit; APACHE: Acute Physiology and Chronic Health Evaluation; APS: acute physiology score; OASIS: Oxford Acute Severity of Illness Score; ADRS: acute respiratory distress syndrome.
*: Fisher’s exact test was applied.

In the MIMIC cohort, 4073 patients were included (1182 CLS, 1362 DTLS and 1529 CDS), with a mean age of 61.7 years and a male proportion of 59.6%. The patients in the CDS group had higher incidence of acute conditions than the other two groups, and worse APSIII and OASIS. The proportion of patients with diabetes, heart failure, cancer and kidney failure significantly differed across the groups.

Effects of sedation strategies on clinical outcomes

In the ICU-HAI cohort, patients with DTLS had lowest ventilator mortality (DTLS 6.7%, CLS 7.4%, CDS 22.0%), ICU mortality (DTLS 8.2%, CLS 10.1%, CDS 24.0%), hospital mortality (DTLS 9.3%, CLS 11.4%, CDS 24.9%, Fig. 3). Compared with CLS (the guideline-recommended approach), DTLS was associated with lower ICU mortality (HR 0.74, 95% CI 0.54–0.99) and hospital mortality (HR 0.72, 95% CI 0.54–0.96), but not for ventilator mortality (HR 0.97, 95% CI 0.70–1.34). However, CDS was associated with higher risk across all outcomes (ventilator mortality 3.55, 95% CI 2.78–4.52; ICU mortality 2.24, 95% CI 1.72–2.92; hospital mortality 2.05, 95% CI 1.58–2.66).

In the MIMIC cohort, patients with DTLS had lowest ventilator mortality (DTLS 4.8%, CLS 6.5%, CDS 36.6%) and ICU mortality (DTLS 9.7%, CLS 12.2%, CDS 46.6%), patients with CLS had lowest hospital mortality (DTLS 18.8%, CLS 17.5%, CDS 54.7%, Fig. 3). Compared with CLS, DTLS was associated with lower ICU mortality (HR 0.81, 95% CI 0.65-1.00), but not for ventilator mortality (HR 0.81, 95% CI 0.60–1.11) and hospital mortality (HR 0.92, 95% CI 0.77–1.09). However, CDS was consistently associated with higher risks on all mortality outcomes (ventilator mortality 6.15, 95% CI 4.80–7.88; ICU mortality 3.99, 95% CI 3.38–4.71; hospital mortality 3.76, 95% CI 3.26–4.33).

By pooling the results of the two cohort studies, DTLS was associated with significantly lower risk for ICU mortality (HR 0.78, 95% CI 0.66–0.94) and hospital mortality (HR 0.86, 95% CI 0.74-1.00) than CLS, but the effect on ventilator mortality did not differ (HR 0.88, 95% 0.71–1.10). Compared with CLS, patients with CDS had higher ventilator mortality (HR 4.65, 95% CI 3.91–5.53), ICU mortality (HR 3.39, 95% CI 2.95–3.90) and hospital mortality (HR 3.27, 95% CI 2.89–3.71, Fig. 3).

Sensitivity analyses

HR estimates from the ICU-HAI and MIMIC cohorts across the sensitivity analyses were consistent with those of the main analysis. Pooled HRs in all sensitivity analyses indicated that DTLS was associated with lower ICU mortality compared with CLS, and CDS was associated with higher ventilator, ICU and hospital mortality, similar to the main analysis. The point estimates of pooled HRs for hospital mortality comparing DTLS versus CLS suggested a similar lower risk as in the main analysis, but the values of their upper confidence interval in sensitivity analyses 1 to 4 were slightly greater than 1 (Figure S1-S6).

Main findings and explanations

In a real-world ICU clinical setting, we identified significant heterogeneity in sedation patterns. Despite guidelines recommending the use of light sedation over deep sedation during MV ¹, deep sedation was widely used, particularly at the early stages. Among patients with MV for at least 3 days, 40.6% received deep sedation at some point during MV, of whom 83.0% received early deep sedation. In addition, among those with early deep sedation, 42.4% switched to light sedation after two days of MV initiation.

In the comparison of CDS versus CLS, we found that CDS was strongly associated with a higher risk of mortality: HR 5.38 (4.49–6.46) for ventilator mortality, 3.39 (2.95–3.90) for ICU mortality and 3.27 (2.89–3.71) for hospital mortality. The HRs for mortality in MIMIC cohort were higher than those in ICU-HAI cohorts, which may be due to the higher incidence of acute conditions and higher prevalence of comorbidities among patients in the MIMIC cohort, particularly those with CDS. Nevertheless, the estimates from both cohorts consistently suggested higher risk across all outcomes, and sensitivity analyses further showed the robustness of the estimates. Given the large magnitude of effect estimates, consistency across outcomes and populations, we were confident that CDS increases the risk of adverse outcomes in ICU patients with MV, supporting the current recommendation against the use of CDS.

In the comparison of DTLS versus CLS, however, we observed a lower risk of ICU and hospital mortality associated with deep-to-light sedation. After fully controlling for potential confounders, the adjusted HRs showed moderate effects. Similar results were obtained from data of both Chinese and American populations. All the sensitivity analyses supported the robustness of the results, which improved our confidence in the effects.

The observed effect is also supported by plausible biological potential mechanisms. Critically ill patients often suffer from tremendous discomfort, pain and anxieties due to illness, examination and treatment procedures. Early deep sedation of MV initiation, a common invasive procedures, could relieve pain and anxiety, ease mental and physical distress ¹³, which could improve patient compliance and eliminate procedural delays^30,31. Early deep sedation also increases sleep efficiency in critically ill patients with MV, who typically experience severe sleep fragmentation and distorted sleep architecture, and this leads to improved clinical outcomes for patients ^32,33.

Our findings showed that the proposed new sedation protocol – that is early deep sedation in combination with subsequent light sedation - may improve the patient outcomes over currently-recommended strategy. The findings were based on a plausible clinical hypothesis that deep sedation – the strategy that facilitated clinicians to manage in routine care – could have a potentially beneficial effect if the timing and length is properly administrated. The finding has not been addressed by previous studies; although the newly proposed protocol would require a further test by randomized controlled trials, the current findings suggest a potential to guide future practice of sedation among patients receiving MV.

Comparison with other studies

Earlier studies have shown that the first 48 hours of MV initiation is a key timing ^5,9,17. Without considering the subsequent depth of sedation, however, those studies drew the conclusion that early deep sedation was independently associated with higher mortality. While a high proportion of patients receiving early deep sedation would continue to be sedated deeply, these studies failed to account for the difference in the duration of deep sedation between patients who switched to light sedation after the first 48 hours and those who received continuous deep sedation. Indeed, longer duration of deep sedation was strongly associated with a significant increase in negative clinical outcomes, consistent with our findings. A prospective multicentre cohort study also showed that compared to patients who spent 50% of their ICU stay in deep sedation, the risk of mortality was five times higher for patients who spent 94% ¹⁰.

Strengths and limitations

This study has several strengths. First, we conducted independent analyses using data from two large-scale ICU databases with 6700 patients contributing over 85000 person-days, to validate our results and demonstrate their robustness. Secondly, because the sedation protocol is dynamic, we applied MSM and a double robust method with IPTW to address its effect with adjustment for time-dependent confounding caused by changes in sedation depth. Finally, the results of the main analyses were assessed using extensive sensitivity analyses, and found to be reliable.

However, some limitations should be noted. This is an observational study based on routinely collected databases. Unmeasured residual confounding may exist, although we adjusted for an extensive array of confounders including many time-dependent variables and applied models to eliminate time-dependent confounding. This finding, based on observational study, may not provide sound evidence for direct clinical recommendations. However, this study provides a new insight for management of MV, which need randomised clinical trials to further test the hypothesis. Second, data for APACHE II score, fluid balance and RASS were missing. Despite this, the percentage of missing data was less than 20%, and the results were consistent when using different methods for dealing with missing data. Third, we defined three sedation strategies of interest depending on the sedation status of the first week after MV initiation to maintain a larger sample size. This is because, even though based on two large ICU databases, the number of patients with MV under a specific sedation trajectory decreased sharply with the extension of the sedation trajectory time. Finally, our HR estimates of DTLS versus CLS were marginally significant, further work needed to be conducted to improve the level of evidence.

Compared to the practice-recommended continuous light sedation, continuous deep sedation was confirmed with poorer patient outcomes, including higher risk of ventilator mortality, ICU mortality, and hospital mortality. On the contrast, early deep-to-light sedation strategy, that is deep sedation within 2 days after initiation of MV followed by light sedation, may decrease the risk of ICU mortality and hospital mortality. Although this finding requires a further test by randomized controlled trials, the newly proposed protocol may provide a potential to guide future practice of sedation among patients receiving MV.

Ethics approval and consent to participate

ICU-HAI was approved by the regional ethical committee (West China Hospital, WCH) in 2018 (WCH2018-409, Establishment and Application of ICU Hospital Infection Research Database). MIMIC-IV was approved by the Institutional Review Board at the Beth Israel Deaconess Medical Center, who granted a waiver of informed consent and approved the data sharing initiative. Both datasets were routinely collected data and were exempted from the requirement for informed consent.

Consent for publication

Not applicable.

Availability of data and materials

ICU-HAI dataset will not be shared. The access policy and procedures of MIMIC-IV dataset are available at https://physionet.org/content/mimiciv/0.4/.

Competing interests

The authors declare that they have no competing interests.

Funding

This study was supported by the National Natural Science Foundation of China (Grant No. 82225049, 72104155), and the Sichuan Provincial Central Government Guides Local Science and Technology Development Special Project (Grant No. 2022ZYD0127).

Authors' contributions

WW and XS are joint corresponding authors. they obtained funding. JX, W Wang, and XS designed the study. LL collected the data. QH, ZW and W Wu were involved in data cleaning and verification. JX and MW analyzed the data. JX drafted the manuscript. JX, W Wang and XS contributed to the interpretation of the results and critical revision of the manuscript for important intellectual content and approved the final version of the manuscript. All authors have read and approved the final manuscript. JX, W Wang and XS are the study guarantors.

Acknowledgements

We thank the participants for their time and assistance. We are grateful the funding from the National Natural Science Foundation of China (Grant No. 82225049, 72104155), and the Sichuan Provincial Central Government Guides Local Science and Technology Development Special Project (Grant No. 2022ZYD0127).

Devlin JW, Skrobik Y, Gélinas C, et al. Clinical Practice Guidelines for the Prevention and Management of Pain, Agitation/Sedation, Delirium, Immobility, and Sleep Disruption in Adult Patients in the ICU. Critical care medicine. 2018;46(9):e825-e873.
Shehabi Y, Howe BD, Bellomo R, et al. Early Sedation with Dexmedetomidine in Critically Ill Patients. N Engl J Med. 2019;380(26):2506-2517.
Strom T, Martinussen T, Toft P. A protocol of no sedation for critically ill patients receiving mechanical ventilation: a randomised trial. Lancet. 2010;375(9713):475-480.
Chanques G, Constantin JM, Devlin JW, et al. Analgesia and sedation in patients with ARDS. Intensive Care Med. 2020;46(12):2342-2356.
Stephens RJ, Dettmer MR, Roberts BW, et al. Practice Patterns and Outcomes Associated With Early Sedation Depth in Mechanically Ventilated Patients: A Systematic Review and Meta-Analysis. Critical care medicine. 2018;46(3):471-479.
Shehabi Y, Bellomo R, Reade MC, et al. Early intensive care sedation predicts long-term mortality in ventilated critically ill patients. Am J Respir Crit Care Med. 2012;186(8):724-731.
Shehabi Y, Chan L, Kadiman S, et al. Sedation depth and long-term mortality in mechanically ventilated critically ill adults: a prospective longitudinal multicentre cohort study. Intensive Care Med. 2013;39(5):910-918.
Dzierba AL, Khalil AM, Derry KL, Madahar P, Beitler JR. Discordance Between Respiratory Drive and Sedation Depth in Critically Ill Patients Receiving Mechanical Ventilation. Critical care medicine. 2021;49(12):2090-2101.
Stephens RJ, Evans EM, Pajor MJ, et al. A dual-center cohort study on the association between early deep sedation and clinical outcomes in mechanically ventilated patients during the COVID-19 pandemic: The COVID-SED study. Critical care (London, England). 2022;26(1):179.
Aragon RE, Proano A, Mongilardi N, et al. Sedation practices and clinical outcomes in mechanically ventilated patients in a prospective multicenter cohort. Critical care (London, England). 2019;23(1):130.
Barends CRM, Driesens MK, van Amsterdam K, Struys M, Absalom AR. Moderate-to-Deep Sedation Using Target-Controlled Infusions of Propofol and Remifentanil: Adverse Events and Risk Factors: A Retrospective Cohort Study of 2937 Procedures. Anesth Analg. 2020;131(4):1173-1183.
Aitken LM, Kydonaki K, Blackwood B, et al. Inconsistent relationship between depth of sedation and intensive care outcome: systematic review and meta-analysis. Thorax. 2021;76(11):1089-1098.
Su L, Liu C, Chang F, et al. Selection strategy for sedation depth in critically ill patients on mechanical ventilation. BMC Med Inform Decis Mak. 2021;21(Suppl 2):79.
Grap MJ, Munro CL, Wetzel PA, et al. Sedation in adults receiving mechanical ventilation: physiological and comfort outcomes. Am J Crit Care. 2012;21(3):e53-63; quiz e64.
Chanques G, Kress JP, Pohlman A, et al. Impact of ventilator adjustment and sedation-analgesia practices on severe asynchrony in patients ventilated in assist-control mode. Critical care medicine. 2013;41(9):2177-2187.
Wisecup C, Kent J, Bintvihok M, Sun J, Kocak D, Mora A. SELF EXTUBATIONS IN THE ICU: A REVIEW OF SEDATION MANAGEMENT AS A POTENTIAL RISK. CHEST. 2021;160(4):A1106.
Balzer F, Weiß B, Kumpf O, et al. Early deep sedation is associated with decreased in-hospital and two-year follow-up survival. Critical Care. 2015;19(1):197.
Benchimol EI, Smeeth L, Guttmann A, et al. The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement. PLoS medicine. 2015;12(10):e1001885.
Wang W, Zhu S, He Q, et al. Fluid Balance and Ventilator-Associated Events Among Patients Admitted to ICUs in China: A Nested Case-Control Study. Critical care medicine. 2022;50(2):307-316.
He Q, Wang W, Zhu S, et al. The epidemiology and clinical outcomes of ventilator-associated events among 20,769 mechanically ventilated patients at intensive care units: an observational study. Critical care (London, England). 2021;25(1):44.
Zhu S, Kang Y, Wang W, Cai L, Sun X, Zong Z. The clinical impacts and risk factors for non-central line-associated bloodstream infection in 5046 intensive care unit patients: an observational study based on electronic medical records. Critical care (London, England). 2019;23(1):52.
Zhu S, Wang W, Kang Y, et al. Clinical outcomes and risk factors for mortality from ventilator-associated events: A registry-based cohort study among 30,830 intensive care unit patients. Infection control and hospital epidemiology. 2022;43(1):48-55.
Wang W, Zhu S, He Q, et al. Developing a Registry of Healthcare-Associated Infections at Intensive Care Units in West China: Study Rationale and Patient Characteristics. Clinical epidemiology. 2019;11:1035-1045.
Johnson AEW, Bulgarelli L, Shen L, et al. MIMIC-IV, a freely accessible electronic health record dataset. Scientific data. 2023;10(1):1.
Sessler CN, Gosnell MS, Grap MJ, et al. The Richmond Agitation-Sedation Scale: validity and reliability in adult intensive care unit patients. American journal of respiratory and critical care medicine. 2002;166(10):1338-1344.
Robins JM, Hernán MÁ, Brumback B. Marginal Structural Models and Causal Inference in Epidemiology. Epidemiology. 2000;11(5).
Chen T, Guestrin CJPotnASICoKD, Mining D. XGBoost: A Scalable Tree Boosting System. 2016.
Klompas M, Li L, Kleinman K, Szumita PM, Massaro AF. Associations Between Ventilator Bundle Components and Outcomes. JAMA internal medicine. 2016;176(9):1277-1283.
Clare PJ, Dobbins TA, Mattick RP. Causal models adjusting for time-varying confounding-a systematic review of the literature. Int J Epidemiol. 2019;48(1):254-265.
Choo MS, Lee DS. Does deep sedation with analgesia have positive effects on anxiety, pain and compliance in patients before and after prostate biopsy? International journal of clinical practice. 2020;74(8):e13517.
Hida S, Takemoto M, Masumoto A, et al. Clinical benefits of deep sedation with a supraglottic airway while monitoring the bispectral index during catheter ablation of atrial fibrillation. Journal of arrhythmia. 2017;33(4):283-288.
Hardin KA, Seyal M, Stewart T, Bonekat HW. Sleep in Critically Ill Chemically Paralyzed Patients Requiring Mechanical Ventilation. Chest. 2006;129(6):1468-1477.
Jean R, Shah P, Yudelevich E, et al. Effects of deep sedation on sleep in critically ill medical patients on mechanical ventilation. Journal of sleep research. 2020;29(3):e12894.

No competing interests reported.

Additionalfile1.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Early deep-to-light sedation versus continuous light sedation for critically ill patients with mechanical ventilation: a cohort study

Status:

Version 1

Abstract

Background

Methods

Results

Conclusion

Figures

Background

Material and methods

Data source

Study population and sedation strategies

Study outcomes

Covariates

Statistical analyses

Sensitivity analyses

Results

Sedation patterns among ICU patients receiving MV

Clinical characteristics of eligible patients in the two cohorts

Effects of sedation strategies on clinical outcomes

Sensitivity analyses

Discussion

Main findings and explanations

Comparison with other studies

Strengths and limitations

Conclusion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1