1 Introduction

Delirium, known as an acute confusion state, is a prevalent complication during invasive mechanical ventilation (IMV) [1]. Several factors contribute to the incidence of delirium, including advanced age, exposure to infection, and the use of certain drugs, so that all patients have the risk for delirium. Patients in intensive care units (ICUs), who are typically among the most critically ill in the hospital, might be at an even higher risk for delirium. Sedation is necessary during IMV, and guidelines recommend the use of nonbenzodiazepine sedatives, such as propofol, for critically ill patients undergoing IMV [2]. Propofol, which acts on gamma-aminobutyric acid receptors, is widely used for sedation and is employed in the induction and maintenance of general anesthesia, as well as in the treatment of delirium due to its notable efficacy [3]. However, propofol was reported to have a potential association with the incidence of delirium [4-6]. Despite this, the mode by which propofol may induce delirium remains unclear. Therefore, it is important to investigate the relationship between delirium and propofol.

Delirium, characterized by an acute change in mental status, poses significant diagnostic challenges. Many researches have demonstrated that in-hospital delirium is associated with poor prognosis and reduced postdischarge life quality [7]. Thus, identifying high-risk patients is crucial for the early detection and management of delirium.

Previous studies have primarily concentrated on comparing propofol with other sedatives, such as midazolam or dexmedetomidine, with limited attention devoted to the correlation between the rate of propofol infusion and the incidence of delirium. The study was launched to identify the high delirium-risk patients and explore potential strategies to reduce the incidence of delirium.

2 Methods and Materials

3 Results

3.1 Baseline Characteristics

The flowchart of the inclusion is depicted in Figure 1. A comprehensive sample of 17,015 patients (delirium: n = 5811; control: n = 11,204) was enrolled. Baseline characteristics, including gender, ethnicity, last care unit, first-day SOFA, first-day GCS, SIRS, LODS, OASIS, and SAPS II, and clinical information are summarized in Table 1. To minimize the potential bias of the study, PSM was employed to adjust for disparities in baseline characteristics. A refined sample of 16,956 patients was ultimately included after PSM (delirium: n = 5805; control: n = 11,151). Group demographics and cohort characteristics are listed in Table S1. The assessment of PSM performance is presented in Figure S1.

TABLE 1. Group demographics and characteristics.

	Delirium (n = 5811)	Control (n = 11,204)	p
Age, year
Median, IQR	66.32 (54.89, 75.99)	65.91 (55.29, 75.61)	0.442
Gender
Female	2380	4165	< 0.001
Male	3431	7039
Ethnicity
White	3533	7664	< 0.001
Black	584	796
Other	1694	2744
Last care unit
TSICU	912	1548	< 0.001
MICU/SICU	2899	3923
NICU	459	147
CVICU	1151	5137
CCU	390	449
First-day SOFA
Median, IQR	7 (5, 11)	5 (3, 8)	< 0.001
First-day GCS
Median, IQR	12 (8, 14)	14 (10, 15)	< 0.001
SIRS
Median, IQR	3 (2, 3)	3 (2, 3)	0.807
LODS
Median, IQR	7 (5, 9)	5 (3, 8)	< 0.001
OASIS
Median, IQR	39 (34, 45)	36 (31, 41)	< 0.001
SAPS II
Median, IQR	41 (32, 51)	37 (31, 47)	< 0.001
Average propofol infusion rate, μg/(kg*h)
1 h
Median, IQR	37.65 (22.71, 48.92)	40.76 (29.35, 51.78)	< 0.001
2 h
Median, IQR	36.50 (21.41, 48.62)	40.16 (29.09, 50.98)	< 0.001
3 h
Median, IQR	34.81 (20.70, 47.80)	39.04 (25.55, 50.04)	< 0.001
4 h
Median, IQR	32.94 (20.24, 46.52)	35.53 (22.69, 48.09)	< 0.001
5 h
Median, IQR	31.57 (19.98, 45.24)	32.38 (20.30, 45.59)	0.017
6 h
Median, IQR	30.49 (19.35, 43.93)	30.08 (18.79, 43.24)	0.220
7 h
Median, IQR	29.96 (18.44, 42.85)	28.30 (17.01, 41.48)	< 0.001
8 h
Median, IQR	29.38 (17.51, 42.22)	26.32 (15.45, 40.33)	< 0.001
9 h
Median, IQR	28.69 (16.89, 41.47)	24.75 (14.14, 39.46)	< 0.001
10 h
Median, IQR	28.07 (16.15, 40.78)	23.25 (13.02. 38.28)	< 0.001
11 h
Median, IQR	27.41 (15.45, 40.32)	21.88 (12.00, 37.11)	< 0.001
12 h
Median, IQR	26.92 (14.91, 39.99)	20.79 (11.13, 36.06)	< 0.001
18 h
Median, IQR	23.74 (11.99, 36.75)	15.71 (7.79, 30.08)	< 0.001
24 h
Median, IQR	21.08 (10.11, 33.76)	12.32 (5.90, 25.33)	< 0.001
30 h
Median, IQR	19.08 (8.52, 31.78)	10.12 (4.77, 21.53)	< 0.001
36 h
Median, IQR	17.17 (7.42, 29.78)	8.53 (4.00, 18.82)	< 0.001
42 h
Median, IQR	15.55 (6.52, 27.81)	7.38 (3.44, 16.63)	< 0.001
48 h
Median, IQR	14.05 (5.83, 26.33)	6.50 (3.02, 14.77)	< 0.001
In-hospital mortality
Yes	953	1452	< 0.001
No	4858	9752

3.2 Primary Outcomes

The disparities in the average propofol infusion rate between the delirium and control groups are graphically presented in Figure 2. Patients belonging to the delirium group exhibited a notably higher average propofol infusion rate in the initial 5 h of IMV, followed by a decrease to a lower rate subsequent to 7 h. Specifically, the median time to diagnose delirium was the 18th hour.

Following the application of RCS, 40 μg/(kg*h) within the first hour and 20 μg/(kg*h) within the initial 18 h were selected as the indicators to differentiate between high and low risks of delirium, respectively (Figure S2a and Figure 3). Additionally, stacked bar charts depicting these data are also shown in the figures (Figure S2b and Figure 3b). A higher proportion of delirium patients was observed in the group that received a higher average rate of propofol infusion.

In the multiple logistic regression model, a higher incidence of delirium was observed to be associated with an average propofol infusion rate exceeding 20 μg/(kg*h) during the initial 18 h. However, an average propofol infusion rate below 40 μg/(kg*h) within the first hour did not demonstrate any statistically significant difference in the incidence of delirium. Multicollinearity was also rigorously assessed, and the variables were appropriately adjusted accordingly (Tables S2 and S3). Notably, the results remained consistent in both 18 h model [OR = 1.84, 95% CI = (1.72, 1.98), p < 0.001] and 1 h model [OR = 0.95, 95% CI = (0.88, 1.02), p = 0.163] after adjustment (Table 2 and Table S4).

TABLE 2. Multiple logistic regression of delirium (18 h).

Delirium	OR	Standard error	z	p > |z|	95% CI
Age, years
(< 60)
≥ 60	1.318194	0.0506589	7.19	< 0.001	[1.222551, 1.421318]
Gender
(Female)
Male	0.9047045	0.0327836	−2.76	0.006	[0.8426785, 0.9712958]
Race
(Other)
White	0.8371806	0.0339406	−4.38	< 0.001	[0.7732326, 0.9064172]
Black	1.092213	0.0741012	1.30	0.194	[0.9562191, 1.247547]
Last care unit
(TSICU)
MICU/SICU	1.013774	0.0515744	0.27	0.788	[0.9175658, 1.120069]
NICU	6.082144	0.6461928	16.99	< 0.001	[4.9388, 7.490175]
CVICU	0.3574196	0.020043	−18.35	< 0.001	[0.3202179, 0.3989432]
CCU	1.122617	0.095265	1.36	0.173	[0.9506026, 1.325759]
First-day SOFA	1.126682	0.0051812	25.94	< 0.001	[1.116572, 1.136883]
High-risk (18 h), μg/(kg*h)
(< 20)
≥ 20	1.844288	0.0656746	17.19	< 0.001	[1.719957, 1.977606]
Constant	0.2103473	0.0142313	−23.04	< 0.001	[0.1842247, 0.240174]

3.3 Secondary Outcomes

Subgroup analysis of the logistic regression revealed that average propofol amount exceeding 20 μg/(kg*h) had close connection in specific characteristics of the population, including age (< 60) [OR = 2.11, 95% CI = (1.88, 2.36), p < 0.001], age (≥ 60) [OR = 1.97, 95% CI = (1.82, 2.14), p < 0.001], gender (female) [OR = 1.69, 95% CI = (1.53, 1.87), p < 0.001], gender (male) [OR = 2.24, 95% CI = (2.06, 2.43), p < 0.001], race (white) [OR = 2.10, 95% CI = (1.94, 2.28), p < 0.001], race (black) [OR = 1.64, 95% CI = (1.32, 2.03), p < 0.001], race (other) [OR = 1.83, 95% CI = (1.62, 2.07), p < 0.001], last care unit (TSICU) [OR = 1.16, 95% CI = (0.98, 1.37), p = 0.081], last care unit (MICU/SICU) [OR = 1.51, 95% CI = (1.37, 1.66), p < 0.001], last care unit (NICU) [OR = 1.86, 95% CI = (1.28, 2.71), p = 0.001], last care unit (CVICU) [OR = 2.96, 95% CI = (2.59, 3.37), p < 0.001], and last care unit (CCU) [OR = 2.27, 95% CI = (1.72, 3.00), p < 0.001]. The forest plot is shown in Figure 4.

4 Discussion

The results indicated that patients with delirium had a higher average rate of propofol infusion during the first 18 h, suggesting that an average propofol infusion rate of ≥ 20 μg/(kg*h) in the initial 18 h was a significant risk factor for delirium. The hypothesis was supported by the multiple logistic regression model. Conversely, although Figure 2 demonstrates differences both at the first hour and during the initial 18 h, the average propofol infusion rate under 40 μg/(kg*h) in the first hour was found not to be associated with the higher incidence of delirium in another multiple logistic regression model. The findings suggested that the accumulation of propofol, rather than a higher momentary rate, might be the actual contributing factor to the increased incidence of delirium. A similar result was observed in a previous study on anesthesia induction. No significant difference in delirium incidence was found between varying propofol injection rates [9]. Additionally, as depicted in Table 1, the incidence of delirium was connected to a higher in-hospital mortality (p < 0.001), indicating that reducing the accumulation of propofol might potentially decrease the in-hospital mortality.

The results indicated that, in addition to the average propofol infusion rate, age, gender, race, last care unit, and first-day SOFA were also associated with the incidence of delirium. Previous studies have generally identified the elderly as a risk factor for delirium [10, 11]. The SOFA score was designed to evaluate the degree of organ dysfunction or failure and the current research depicts that the SOFA score is a risk factor for both delirium and in-hospital mortality [12]. Besides, gender was also recognized as an important factor in the incidence of delirium in prior studies. An investigation in the relationship between delirium and gender found that delirium would more frequently affect female patients over 85 years old in the cardiac ICU, which might explain the difference in gender observed in the multiple logistic regression model [13]. Additionally, patients in the NICU exhibited a higher incidence of delirium. One potential explanation was that the neurological disease itself could result in delirium by inducing neurological inflammation and other mechanisms, making NICU residency a significant factor in the incidence of delirium [14, 15]. In contrast to the findings in cardiovascular surgery [16-18], although patients in CVICU showed similar outcomes as other ICUs during subgroup analysis, CVICU appeared to have a lower impact on delirium compared to other ICUs. This phenomenon might be attributed to the low proportion of patients who underwent extracorporeal circulation and open-heart operation.

To deal with the high incidence of delirium associated with propofol, several researches paid attention to the alternatives. Dexmedetomidine has also been widely utilized in patients undergoing IMV. Evidence-based studies revealed that dexmedetomidine might shorten the duration of IMV and reduce the incidence of delirium [4, 5]. However, dexmedetomidine might also increase the risk of cardiac events [19]. Another commonly used sedative was midazolam. Previous researches depicted that although midazolam exhibited similar sedative efficacy, it was associated with higher mortality and an increased incidence of delirium [20, 21]. Future studies should focus on alternative sedatives with both a lower incidence of delirium and safety.

In the treatment of delirium, pharmacological intervention was generally considered as a common practice in delirium treatment. However, recent researches demonstrated that drugs like haloperidol and other antipsychotics might not be as effective as expected. Prevention was considered to be the most optimal treatment. Nonpharmacologic interventions have been important approaches in the past few decades and were recommended by clinical guidelines [14, 22, 23]. Future treatment for delirium should concentrate on the combination of pharmacologic therapies, nonpharmacologic prevention, and monitoring.

A key strength of our study was that it was the first study to concentrate on the relationship between average propofol infusion rate and the incidence of delirium in patients undergoing IMV. Furthermore, the research included a relatively large sample size. The results of this study suggested that the role of the average propofol infusion rate during the initial 18 h made great sense to the incidence of delirium, which might contribute to the future management of ICU patients undergoing IMV.

Some limitations also existed in the study. Initially, the average rate of propofol infusion was calculated based on the weight, time, and total amount of propofol, which made it difficult to explore the relationship between delirium and temporary propofol infusion rate. Secondarily, the effects of other drugs were not comprehensively considered, though the influence of propofol was demonstrated by multiple logistic regression. Additionally, it was difficult to identify different subtypes of delirium within the MIMIC IV database, which should be taken into consideration in future studies.

5 Conclusion

An average propofol infusion rate ≥ 20 μg/(kg*h) during the initial 18 h was an independent risk factor for delirium, suggesting that the accumulation of propofol might be associated with an increased incidence of delirium. Future studies should focus on the different subtypes of delirium and comprehensive management of delirium.

Author Contributions

Qi-Yue Ge, Chao Zheng, and Yi Shen completed the conceptualization. The methodology was finished by Qi-Yue Ge, Chao Zheng, Zhuang-Zhuang Cong, Yan-Qing Wang, Bing-Wei Chen, and Yi Shen. The formal analysis was accomplished by Qi-Yue Ge and Chao Zheng. The investigation was contributed by Qi-Yue Ge, Chao Zheng, Jing Luo, Xiao-Bin Song, Hao-Tian Zheng, and Peng-Long Zhao. The data curation was done by Qi-Yue Ge and Chao Zheng. The visualization was achieved by Qi-Yue Ge and Chao Zheng. Zhuang-Zhuang Cong, Jing Luo, and Yi Shen made great contributions to the funding acquisition. Resources were completed by Qi-Yue Ge, Chao Zheng, and Yi Shen. Supervision was given by Yan-Qing Wang, Bing-Wei Chen, and Yi Shen. All authors wrote and/or reviewed the manuscript.

Ethics Statement

We have completed the National Institutes of Health's web-based course and passed the Protecting Human Research Participants exam (ID: 52932612) to access the MIMIC IV database; Due to the retrospective nature, the study was dispensed from obtaining individual inform consents.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.