Incidence and risk factors for delirium in cardiac surgery intensive care unit patients: A single-center study

Abstract

BACKGROUND

Delirium is a frequent and serious complication following cardiac surgery, particularly in patients admitted to the intensive care unit (ICU), where it can negatively impact recovery and outcomes.

AIM

To investigate the incidence, risk factors, and clinical consequences of delirium in post-cardiac surgery ICU patients.

METHODS

A total of 202 consecutive patients admitted to the cardiac surgery ICU at the Onassis Cardiac Surgery Center were evaluated for delirium using the Confusion Assessment Method-ICU scale.

RESULTS

Delirium was observed in 29.2% of the cohort and was independently associated with older age, longer total anesthesia/sedation duration, and higher EuroSCORE II. These variables showed poor-to-fair/modest discriminatory ability for delirium prediction. Patients who developed delirium experienced more frequent reintubation, higher rates of ICU-acquired weakness and hemodialysis, prolonged mechanical ventilation, and extended ICU stays.

CONCLUSION

These findings underscore the multifactorial nature of postoperative delirium and suggest that older age, higher EuroSCORE II, and longer anesthesia exposure are associated with increased delirium risk. However, given the exploratory design and modest discriminatory performance of the identified variables, these findings should not be interpreted as a validated risk prediction tool and require confirmation in larger multicenter studies.

Key Words: Delirium; Incidence; Cardiac surgery; Intensive care unit; Risk factors; EuroSCORE II

Core Tip: In this prospective single-center study of 202 patients admitted to a cardiac surgery intensive care unit (ICU), delirium occurred in 29.2% during ICU hospitalization. Older age, longer total anesthesia-sedation duration, and higher EuroSCORE II were independently associated with delirium. Patients with delirium also had higher rates of reintubation, ICU-acquired weakness, and hemodialysis, as well as longer mechanical ventilation and ICU stay. These findings support systematic postoperative delirium surveillance in high-risk cardiac surgical patients, while highlighting the need for larger multicenter studies to refine risk assessment and prevention strategies.

INTRODUCTION

Delirium is an acute neuropsychiatric syndrome characterized by fluctuating disturbances in attention, awareness, and cognition, often accompanied by altered consciousness and perceptual disturbances. It occurs in 20%-70% of hospitalized patients[1,2], reaching up to 87% in intensive care unit (ICU) patients[3-5]. Regardless of the classification, there are three subtypes of delirium categorized according to psychomotor behavior: (1) Hyperactive delirium; (2) Hypoactive delirium; and (3) Mixed delirium[6], with the mixed type (52.5%) being the most commonly observed in the ICU setting[7]. The Confusion Assessment Method-ICU (CAM-ICU) is a reliable tool that has been used for delirium diagnosis and for evaluating delirium over time[8].

In the context of cardiac surgery, delirium represents a common and serious postoperative complication, typically manifesting within the first few days after the procedure[9]. Patients undergoing cardiac surgery are at increased risk of postoperative delirium, which contributes to prolonged ICU and hospital stay.

We hypothesize that delirium occurs frequently following cardiac surgery, mainly due to perioperative risk factors and pre-existing disease, and is associated with poor outcomes. The purpose of this study is to evaluate the incidence, risk factors, and outcomes of delirium in patients undergoing cardiac surgery. Although several studies have identified risk factors for postoperative delirium after cardiac surgery, prospective data from specialized cardiac surgery ICUs and from different healthcare environments remain limited. The aim of the present study was not to propose a novel mechanistic explanation or predictive model, but to prospectively assess the incidence, perioperative correlates, and short-term clinical consequences of delirium in a consecutive cohort from a high-acuity cardiac surgery ICU using systematic CAM-ICU screening. By providing data from a real-world specialized center, this study adds context-specific evidence that may support risk stratification and prevention strategies in similar clinical settings.

MATERIALS AND METHODS

Study design

This is a prospective study that consecutively enrolled patients after cardiac surgery who were hospitalized in the Cardiac Surgery Intensive Care Unit of the Onassis Cardiac Surgery Center during the period from March 2019 to June 2019. This study was approved by the Institutional Review Board and Ethics Committee of the Onassis Cardiac Surgery Center, Athens, Greece, Protocol No. 641/14.03.2019 and was conducted in accordance with the Declaration of Helsinki. The manuscript was reported in accordance with the CONSORT 2010 guidelines for observational cohort studies.

Patients

Patients aged > 18 years who were admitted to the ICU of the Onassis Cardiac Surgery Center post-cardiac surgery were included in the study. Individuals who refused written consent, patients with severe pre-existing mental disorders (e.g., severe dementia), patients with a critical clinical condition with a mortality risk > 80%, as well as those with an inability to communicate due to severe encephalopathy or deafness, were excluded from the study.

Immediately postoperatively, upon admission to the cardiac surgery ICU, patients underwent an assessment for delirium upon awakening. The assessment was repeated daily, twice a day, and continued for 7 days after the first assessment or until ICU discharge (if it occurred within less than 7 days). This 7-day assessment window was chosen based on clinical practice and evidence indicating that most cases of postoperative delirium occur within the first postoperative week. For instance, postoperative delirium is known to manifest from shortly after anaesthesia up to 7 days post-surgery[10]. While this timeframe likely captures the majority of cases, we acknowledge that it may lead to under-recognition of delirium in patients with prolonged ICU stays or late-onset presentations.

Data collection was performed using the interview method. Patients who met the study inclusion criteria were verbally informed about the purpose and methods of the research, if feasible. The researcher then requested oral and written consent from the patients or their relatives (first-degree kin) for participation in the study.

All patient data files used in the study maintained patient anonymity through special coding.

Medical history and clinical examination

All participants underwent a detailed personal and family history, which included information regarding age, gender, type of surgery [coronary artery bypass grafting (CABG)/valve repair/mixed/chronic aneurysm repair/acute ascending aortic dissection/peripheral "extracorporeal membrane oxygenation"/other], their personal medical history [hypertension, diabetes mellitus, dyslipidemia, smoking habits, coronary artery disease, chronic obstructive pulmonary disease (emphysema), interstitial lung disease, lung surgery, chronic heart failure, chronic renal failure, and other comorbidities], and their personal psychiatric history (anxiety/depression/psychotic disorder). Moreover, EuroSCORE II was calculated in all patients to assess the risk of in-hospital mortality after cardiac surgery. EuroSCORE II was calculated based on the original methodology[11], available at http://www.euroscore.org.

Parameters including extracorporeal circulatory support (duration of extracorporeal circulation, duration of aortic occlusion, duration of general anesthesia), hemodynamic characteristics (blood pressure, heart rate, mean arterial pressure, central venous pressure), biochemical indicators before and after surgery (hematocrit, hemoglobin, plasma urea, plasma creatinine, plasma lactate, etc.), and respiratory parameters were recorded in detail. Total anesthesia/sedation time was calculated by adding the duration of postoperative sedation during mechanical ventilation in the ICU to the duration of intraoperative general anesthesia. The Kidney Disease: Improving Global Outcomes (KDIGO) criteria were used to classify acute kidney injury (AKI). According to the KDIGO criteria, AKI is defined as an increase in serum creatinine of ≥ 0.3 mg/dL within 48 hours, or an increase in serum creatinine of ≥ 1.5 times its baseline value (or that considered to have been present during the previous 7 days), or urine volume ≤ 0.5 mL/kg/hour for 6 hours[12]. For AKI diagnosis, only serum creatinine values were used, as urine output data may be confounded by diuretic administration.

Finally, mechanical respiratory support [type of mechanical ventilation (pressure control/volume control/synchronized intermittent mandatory ventilation/pressure support/pressure regulated volume control), tidal volume (in mL), tidal volume relative to ideal weight (in mL/kg), respiratory rate, end-expiratory pressure, maximum airway pressure, and duration of stay on the ventilator], pharmaceutical support for sedation/analgesia (propofol, midazolam, morphine/fentanyl, myorelaxation, dexmedetomidine), hemodynamic pharmaceutical support with vasoconstrictors (noradrenaline/adrenaline/vasopressin), and inotropes (dobutamine), the duration of stay in the ICU, as well as the outcome upon discharge from the ICU/hospital, were also recorded.

CAM-ICU scale

Cardiac surgery ICU patients underwent a delirium assessment using the CAM-ICU scale[13]. All delirium assessments were conducted by trained ICU physicians or nurses who had completed standardized training in the use of the CAM-ICU tool. While inter-rater reliability was not formally assessed during the study, efforts were made to ensure consistency by having the same trained staff perform repeated evaluations in individual patients. The assessment of delirium is essentially part of the overall assessment of consciousness. Consciousness is determined by two things, the level of wakefulness and the adequacy of consciousness. The first step in assessing consciousness is to assess the level of consciousness. This is best achieved by using a weighted scale for assessing the level of alertness/sedation. Specifically, the Richmond Agitation-Sedation Scale (RASS) score was used. The next step is to assess the adequacy of consciousness. At levels of deep sedation (e.g. RASS: -4 or -5) it is difficult to confirm the adequacy of consciousness because the patient is unresponsive. These levels are referred to as stupor or coma, and in these cases CAM-ICU is not performed, and the patient is reported as “unassessable”. However, at higher levels of consciousness (e.g. RASS -3 and above), patients may demonstrate at least the initial signs of substantial responsiveness (e.g., response to verbal stimuli). At these levels, clarity of thought, and specifically delirium, can be assessed. According to the scale, patients are diagnosed with delirium when an acute onset of altered mental status or a fluctuation in course and inattention is accompanied by either disorganized thinking or a disturbance in the level of consciousness. All CAM-ICU assessments were conducted by trained intensive care physicians or nurses who had completed standardized training in the application of the scale. The same trained staff did repeated tests on each patient to make sure the results were the same, but the study did not formally check for inter-rater reliability.

Statistical analysis

Continuous variables with a normal distribution are presented as mean ± SD, and those with a non-normal distribution as median and interquartile range (25^th-75^th percentile). Categorical variables are shown as n (%). The Shapiro-Wilk test was used to assess normality.

Between-group comparisons were conducted using the Student’s t-test or the Mann-Whitney U test for continuous variables, and the χ² test for categorical variables. Variance equality was assessed using the Levene test.

To assess variables associated with delirium, we first performed univariate comparisons between patients with and without delirium. Variables showing statistical significance in univariate analysis, as well as variables considered clinically relevant based on prior literature, were screened for potential inclusion in multivariable logistic regression. However, statistically significant variables were not automatically entered into the model, because 59 patients developed delirium and the number of covariates had to be restricted to reduce the risk of overfitting. An initial candidate model included 6 variables: Age, total anesthesia/sedation duration, EuroSCORE II, preoperative plasma urea, postoperative plasma lactate, and vasopressor use. These variables were selected to represent clinically relevant and relatively non-overlapping domains, including patient vulnerability, global surgical risk, anesthesia/sedation exposure, baseline renal/metabolic status, postoperative metabolic stress, and hemodynamic support. Other variables that were significant in univariate analysis, including mean arterial pressure, postoperative creatinine, AKI, total bilirubin, high sensitivity C-reactive protein (hs-CRP), and base excess, were not included because of overlap with selected variables, missing data, concern for overfitting, or their potential role as postoperative organ dysfunction markers or mediating variables rather than independent baseline predictors. Collinearity among the variables entered into the candidate model was assessed using variance inflation factors, and no relevant multicollinearity was identified, with all VIF values below 1.3.

The initial 6-variable candidate model demonstrated poor calibration according to the Hosmer-Lemeshow goodness-of-fit test (χ² = 51.47, df = 8, P < 0.001). Therefore, this model was not retained as a final prediction model. To reduce overfitting and improve interpretability, a simplified final multivariable logistic regression model was constructed including the variables that remained independently associated with delirium: Age, total anesthesia/sedation duration, and EuroSCORE II. The final model was interpreted as an exploratory association model and not as a validated risk prediction tool.

Receiver operating characteristic (ROC) curves were generated to evaluate the discriminatory power of significant predictors, and area under the curve (AUC) values were reported with 95% confidence intervals (95%CIs). Statistical analyses were performed using SPSS v25.0 (SPSS Inc., IL, Chicago, United States), and a P value < 0.05 was considered statistically significant.

No formal a priori sample size calculation was performed based on expected delirium incidence, desired precision of the incidence estimate, or requirements for multivariable regression. Instead, all eligible consecutive adult patients admitted to the cardiac surgery ICU during the predefined study period were included. Therefore, the study should be interpreted as an exploratory prospective cohort analysis. Nevertheless, the number of variables entered into the multivariable logistic regression model was restricted in relation to the number of delirium events observed, in order to reduce the risk of model overfitting.

RESULTS

Delirium incidence and patient characteristics

The total study population consisted of 202 consecutive patients. Based on CAM-ICU assessment, the incidence of delirium during ICU hospitalization after cardiac surgery was 29.2% (95%CI: 23.2%-35.8%).

Demographic and clinical/Laboratory characteristics of patients with and without delirium during ICU hospitalization following cardiac surgery are presented in Table 1.

Table 1 Demographics, comorbidities, type of surgery, and other parameters for all study participants are presented according to delirium occurrence.

Parameters	All (n = 202)	No delirium (n = 143)	Delirium (n = 59)	P value
Males	143 (70.8)	103 (72.0)	40 (67.8)	0.481
Age (years)	66 ± 13	65 ± 13	71 ± 11	0.002
Body mass index (kg/m2)	28.5 ± 5	28.6 ± 5	28.2 ± 5	0.644
Comorbidities
Hypertension	158 (78.2)	111 (77.6)	47 (79.7)	0.811
Diabetes mellitus	65 (32.2)	45 (31.5)	22 (37.3)	0.656
Dyslipidemia	97 (48.0)	68 (47.6)	29 (49.2)	0.963
Smoking	47 (23.3)	33 (23.1)	14 (23.7)	0.39
Chronic obstructive pulmonary disease	17 (8.4)	10 (7.0)	7 (11.9)	0.235
Chronic kidney disease	13 (6.4)	6 (4.2)	7 (11.9)	0.04
Coronary arterial disease	67 (33.2)	49 (34.3)	18 (30.5)	0.683
Stroke	3 (1.5)	1 (0.7)	2 (3.4)	0.199
Thyroid disease	35 (17.3)	27 (18.9)	8 (13.6)	0.4
Pacemaker	5 (2.5)	3 (2.1)	2 (3.4)	0.572
Peripheral arterial disease	10 (5.0)	8 (5.6)	2 (3.4)	0.727
Anemia	11 (5.4)	7 (4.9)	4 (6.8)	0.516
EuroSCORE II	1.67 (1.1-2.8)	1.38 (0.9-2.2)	2.32 (1.6-4.6)	< 0.001
Type of surgery				0.18
CABG	80 (39.6)	58 (40.6)	22 (37.3)
CABG and valve replacement or valvuloplasty	34 (16.8)	21 (14.7)	13 (22.0)
Valvuloplasty	5 (2.5)	4 (2.8)	1 (1.7)
Valve replacement	50 (24.8)	32 (22.4)	18 (30.5)
Aortic dissection	3 (1.5)	1 (0.7)	2 (3.4)
Aneurysm repair and other cardiac surgery	22 (10.9)	19 (13.3)	3 (5.1)
Duration of extracorporeal circulatory support (minute)	121 ± 52	117 ± 49	131 ± 59	0.076
Duration of aortic closure (minute)	80 (59-107)	76 (57-104)	89 (67-109)	0.052
Duration of general anesthesia (minute)	240 (190-300)	231 (183-300)	244 (208-324)	0.07

Continuous variables with a normal distribution are presented as mean ± SD, and those with a non-normal distribution as median and interquartile range (25^th-75^th percentile). Categorical variables are shown as n (%). CABG: Coronary artery bypass grafting.

Variables associated with delirium

Patients with delirium showed a trend toward a longer duration of both extracorporeal circulatory support and general anesthesia, although these differences did not reach statistical significance (P = 0.076 and P = 0.070, respectively).

Clinical and laboratory parameters for the total sample and subgroups immediately after cardiac surgery are presented in Supplementary Table 1 in the appendix. Patients with delirium had significantly lower mean arterial pressure, higher preoperative urea, and elevated postoperative creatinine, bilirubin, and C-reactive protein values, as well as greater base excess deficits and higher lactate levels compared to patients without delirium.

Details on pharmaceutical support for all study participants, stratified by delirium occurrence, upon admission to the cardiac surgery ICU and during ICU stay, are provided in Supplementary Table 2 in the appendix.

Upon ICU admission, a higher proportion of patients with delirium received dexmedetomidine, noradrenaline, and dopamine, while during the ICU stay, they more frequently received vasopressin, furosemide, dopamine, dobutamine, and norepinephrine compared to patients without delirium.

Clinical outcomes and multivariable analysis

Patients with delirium demonstrated higher rates of hemodialysis and ICU-acquired weakness (ICUAW), longer sedation time, and longer duration of mechanical ventilation, with a higher rate of reintubation and prolonged ICU stay compared to those without delirium (Table 2).

Table 2 Delirium status and clinical outcome.

Parameters	All (n = 202)	No delirium (n = 143)	Delirium (n = 59)	P value
RASS scale				< 0.001
-5 to -4	0 (0)	0 (0)	0 (0)
-3 to -1	13 (6.4)	7 (4.9)	6 (10.2)
0	145 (71.8)	125 (87.4)	20 (33.9)
> 1	44 (21.8)	11 (7.7)	33 (55.9)
Day of onset of delirium (days)	0 (0-1)	0 (0-0)	2 (1-2)	< 0.001
Duration of delirium (days)	0 (0-1)	0 (0-0)	1 (1-2)	< 0.001
Hemodialysis in the ICU	5 (2.5)	1 (0.7)	4 (6.8)	0.023
Duration of anesthesia from ICU to awakening (minute)	360 (270-540)	300 (240-450)	540 (300-900)	< 0.001
Duration of total anesthesia/sedation (minute)	600 (484-842)	544 (480-713)	960 (562-1141)	< 0.001
Duration of mechanical ventilation (minute)	779 (586-1154)	717 (480-931)	1230 (774-1696)	< 0.001
Reintubation	7 (3.5)	0 (0)	7 (11.9)	< 0.001
ICU acquired weakness	15 (7.4)	3 (2.1)	12 (20.3)	< 0.001
Length of ICU (days)	2 (2-4)	2 (2-3)	4 (3-6)	< 0.001
ICU outcome				0.272
Improvement	190 (94.1)	138 (96.5)	52 (88.1)
Death	1 (0.5)	0 (0)	1 (1.7)

Continuous variables with a non-normal distribution as median and interquartile range (25^th-75^th percentile). Categorical variables are shown as n (%). RASS: Richmond Agitation-Sedation Scale; ICU: Intensive care unit.

ROC analysis was performed only as an exploratory assessment of the discriminatory ability of selected variables associated with delirium occurrence; the corresponding AUC values are presented in Table 3 and Figure 1. The observed AUC values indicated poor-to-fair/modest discrimination and were considered insufficient for individual-level clinical prediction. No threshold-based sensitivity/specificity analysis or cut-off derivation was performed.

Figure 1 Exploratory receiver operating characteristic curves of variables associated with delirium occurrence. The observed area under the curve values indicate limited discriminatory ability and should not be interpreted as sufficient for individual-level clinical prediction. ROC: Receiver operating characteristic.

Table 3 Exploratory receiver operating characteristic analysis of variables associated with delirium occurrence in the cardiac surgery intensive care unit.

Variable	AUC	Standard error	P value	95%CI
Duration of total anesthesia	0.702	0.042	< 0.001	0.619-0.785
Age	0.635	0.044	0.030	0.547-0.722
EuroSCORE II	0.730	0.038	< 0.001	0.655-0.804
Urea before surgery	0.690	0.041	< 0.001	0.609-0.771
Plasma lactate	0.622	0.043	0.008	0.537-0.707
Vasopressors	0.598	0.046	0.032	0.508-0.688

AUC: Area under the curve; 95%CI: 95% confidence interval.

In the initial 6-variable candidate model, calibration was poor according to the Hosmer-Lemeshow goodness-of-fit test; therefore, this model was not retained as a final predictive model. A simplified final multivariable logistic regression model was subsequently used. In this parsimonious model, older age, longer total anesthesia/sedation duration, and higher EuroSCORE II were independently associated with delirium occurrence (Table 4). These findings should be interpreted as exploratory associations rather than as a validated risk prediction model.

Table 4 Simplified multivariable logistic regression model with delirium as the dependent variable.

Variable	Odds ratio	95%CI	P value
Age	1.047	1.010-1.086	0.013
Duration of total anesthesia	1.001	1.000-1.002	0.006
EuroSCORE II	1.260	1.090-1.457	0.002

The initially considered 6-variable candidate model demonstrated poor calibration according to the Hosmer-Lemeshow goodness-of-fit test (χ² = 51.47, df = 8, P < 0.001) and was therefore not retained as a final prediction model. Odds ratios and 95% confidence intervals were derived from the simplified multivariable logistic regression model. The model included age, total anesthesia/sedation duration, and EuroSCORE II. The model was interpreted as an exploratory association model and not as a validated risk prediction tool. 95%CI: 95% confidence interval.

DISCUSSION

In the present study, the incidence of delirium during ICU hospitalization following cardiac surgery was 29%. This finding aligns with previously reported data summarized in recent systematic reviews, which indicate an overall delirium incidence of 23%[14], with wide variations ranging from 3% to 64%[2,15-17], and occasionally reaching up to 82%[18]. Specifically, a recent meta-analysis by Igwe et al[19] involving 90 studies demonstrated a 32% delirium incidence among post-cardiac surgery ICU patients. In close agreement, Yokoyama et al[20] reported postoperative delirium in 31.9% (562 out of 1731) of patients undergoing cardiovascular surgery with cardiopulmonary bypass, while a recent cohort of 88 patients undergoing open-heart surgery, CABG, or valve procedures yielded a 25% incidence rate[21]. Conversely, certain investigators have reported substantially lower rates; Wang et al[22] observed a 12.1% overall incidence of postoperative delirium in a prospective study of 232 cardiac surgery patients, whereas Cheng et al[23] reported a rate of just 6.58% across two medical centers in Taiwan.

A key finding of our study was that advanced age emerged as an independent predictor of delirium, with affected patients being, on average, six years older than their non-delirious counterparts. This strong association is well-documented in prior literature[24,25]. For instance, Afonso et al[26] similarly identified both age and surgical duration as independent predictors of delirium in a post-cardiac surgery cohort with a mean age of 66 years.

From a mechanistic standpoint, the predisposition of older patients to delirium involves complex, multifactorial pathways, including metabolic and electrolyte derangements, neurotransmitter imbalances, systemic inflammatory responses, sympathetic overdrive, and genetic susceptibilities[27]. Crucially, advanced age is characterized by a diminished cholinergic reserve, a neurochemical shift widely believed to facilitate delirium onset. Furthermore, senescent cerebrovascular atherosclerosis, when compounded by postoperative systemic inflammation, can critically impair cerebral blood flow autoregulation[24]. Finally, the age-related accumulation of sensory and cognitive deficits may further lower the threshold for delirium in this vulnerable population.

It is well documented that sedation prolongs mechanical ventilation, initiating a vicious cycle that elevates the risk of both delirium and ICUAW[28]. Our analysis demonstrated that the total duration of anesthesia and sedation (including propofol) was an independent predictor of delirium. This aligns with a recent meta-analysis associating propofol with increased delirium risk in critically ill patients[29], as well as data suggesting that propofol accumulation correlates with higher delirium incidence[30]. Mechanistically, this may be driven by sedative-induced neurotoxicity; prolonged or high-dose exposure to specific agents has been implicated in altered neurotransmission, mitochondrial dysfunction, and neuroinflammatory signaling[31-34]. While the unit-based odds ratio for total anesthesia duration appeared modest (point estimate of 1.001 per minute), this effect size must be interpreted contextually based on the measurement scale: It corresponds to an approximate 6% increase in the odds of delirium for each additional hour of anesthesia, and a roughly 10% increase for every 100 minutes, underscoring its clinical relevance during prolonged exposures.

A high EuroSCORE II also emerged as an independent predictor of postoperative delirium. Similarly, Theologou et al[31] demonstrated that an elevated EuroSCORE II, prolonged endotracheal intubation, and extended ICU stay were independent risk factors for delirium in a cohort of 179 CABG patients. Interestingly, incorporating frailty assessments into the EuroSCORE II has been shown to improve its discriminatory ability for postoperative delirium[32].

Importantly, while age, anesthesia/sedation duration, and EuroSCORE II were independently associated with delirium, their joint discriminatory performance was limited. The observed AUC values fall within the poor-to-fair range, remaining below the threshold required for individual-level clinical risk prediction. Consequently, these variables should be interpreted as markers of heightened perioperative vulnerability rather than as reliable, stand-alone tools for clinical decision-making. Incremental increases in these modifiable and non-modifiable exposures elevate cumulative risk, highlighting the necessity of limiting potentially modifiable perioperative exposures whenever feasible.

Beyond these independent predictors, delirium was significantly associated with major ICU complications, including prolonged mechanical ventilation, increased reintubation rates, ICUAW, and extended ICU stay[33-35]. These findings chime with previous studies indicating that delirium signals a more complex, protracted, and potentially incomplete recovery trajectory, which inherently jeopardizes long-term cognitive and functional outcomes, readmission rates, and overall mortality[3,36,37]. Such complications contribute to severe physical deconditioning, cognitive impairment, and psychological distress—key determinants of post-discharge quality of life[38,39]. In cardiac surgery cohorts, which frequently comprise older individuals with multiple comorbidities, these adverse trajectories hinder rehabilitation, reduce functional independence, and increase the risk of long-term institutionalization[40,41]. Furthermore, longer durations of delirium secondary to hypoxia, sepsis, and sedation have been shown to predict worse cognitive function at 12 months[42].

Of note, several inflammatory and metabolic stress markers, including hs-CRP and lactate, as well as vasopressor use, were significantly higher in patients who developed delirium. Vasopressor use, particularly during hypotension or shock, may precipitate cerebral hypoperfusion—a core mechanism underlying acute brain dysfunction[39]. This underscores the importance of fastidious hemodynamic management and supports prior literature identifying intraoperative hypotension as a delirium predictor. However, these variables were not retained in our final multivariable model; hs-CRP was assessed only univariately to maintain a parsimonious covariate strategy, while lactate was excluded during model simplification. Therefore, these associations remain exploratory and hypothesis-generating.

Limitations & methodological considerations

The findings of this study should be interpreted primarily as confirmatory rather than exploratory. We do not claim that age, EuroSCORE II, or anesthesia duration are novel predictors; rather, the merit of this work lies in the prospective validation of these clinically relevant risk factors within a specialized, high-acuity cardiac surgery ICU using systematic, routine CAM-ICU assessments.

Nevertheless, several limitations must be acknowledged. First, this was a single-center study with a relatively small sample size and a limited proportion of female patients, which may restrict the generalizability of our findings. The 4-month enrollment period (March 2019 to June 2019) may not fully capture seasonal variations in surgical volume, case mix, staffing patterns, or perioperative practices. Second, the sample size precluded granular subgroup analyses regarding specific surgery types, extracorporeal circulation parameters, and detailed sedative or vasopressor exposure metrics. Third, long-term post-discharge follow-up was unavailable, delirium sub-phenotypes were not systematically tracked, and late-onset delirium occurring after the first seven ICU days may have been missed.

Methodologically, because mechanical ventilation duration, reintubation, ICUAW, and ICU length of stay were analyzed as clinical outcomes or correlates, no directionality or causality can be inferred. The link between delirium and prolonged ventilation may reflect a bidirectional relationship: Delirium can delay liberation from the ventilator, while prolonged mechanical ventilation and its attendant exposures may exacerbate brain dysfunction, or both may simply mirror greater baseline severity of illness. Additionally, no formal a priori sample size calculation was performed, meaning the study may be underpowered for certain secondary associations. Because the limited number of delirium events necessitated a strict restriction of covariates to avoid model overfitting, residual confounding from other univariately significant hemodynamic, renal, inflammatory, or metabolic variables cannot be entirely ruled out. Finally, our initial 6-variable model demonstrated poor calibration and was rejected; while the simplified model enhanced interpretability, no internal or external validation or cut-off analyses were performed.

Thus, our results do not suffice to guide perioperative management decisions independently, but they do substantiate the clinical relevance of routine delirium surveillance. Current guidelines recommend daily objective monitoring using validated instruments like the CAM-ICU or ICDSC, while NICE guidelines[43] advocate for non-pharmacological multicomponent interventions, including reorientation, family engagement, early mobilization, fluid balance optimization, and the correction of underlying physiological triggers.

CONCLUSION

In conclusion, delirium occurs frequently in patients admitted to the ICU following cardiac surgery and is associated with a substantial short-term clinical burden, including prolonged mechanical ventilation, high reintubation rates, ICUAW, and extended ICU stays. Age, total duration of anesthesia, and EuroSCORE II were independently associated with its onset. Although these risk factors are well-established, this study provides prospective, confirmatory evidence from a specialized cardiac surgery ICU setting. These variables should be viewed as markers of perioperative vulnerability rather than a validated risk-stratification tool. Future large-scale, multicenter cohorts are required to develop, calibrate, and externally validate robust prediction models and to evaluate targeted preventive strategies in this vulnerable population.