Association between the confusion assessment method for the intensive care unit (CAM-ICU) and prolonged mechanical ventilation following major cardiac surgery

Abstract

Background

Delirium is common after cardiac surgery and has been associated with adverse outcomes in critically ill patients. The Confusion Assessment Method for the Intensive Care Unit (CAM-ICU) is a validated tool for detecting delirium; however, few studies have explored its association with prolonged mechanical ventilation (PMV) following major cardiac surgery. This study aimed to investigate whether a positive CAM-ICU assessment is associated with PMV and other postoperative outcomes in this population.

Methods

We conducted a retrospective cohort study of 1,437 adult patients who underwent cardiac surgery in 2022. Delirium was assessed daily using the CAM-ICU during the early postoperative period. PMV was defined as ventilation lasting more than 24 h after surgery. Clinical outcomes were analyzed using logistic regression and time-dependent Cox regression, with adjustments made using inverse probability of treatment weighting (IPTW).

Results

Among 1437 patients, 281 (19.6%) were assessed as CAM-ICU positive in the early postoperative period. After IPTW adjustment, a positive CAM-ICU assessment was significantly associated with increased risk of PMV (odds ratio [OR] 3.60; 95% CI, 2.53–5.10; p < 0.001), reintubation (OR 3.37; 95% CI, 1.52–7.31; p = 0.002), intensive care unit (ICU) readmission (OR 2.53; 95% CI, 1.22–5.03; p = 0.010), and prolonged ICU stay after extubation (OR 2.95; 95% CI, 2.19–3.96; p < 0.001). Neither early death nor all-cause mortality differed significantly between groups after IPTW adjustment. Age-related analyses indicated that predicted ventilator duration was generally longer with a positive CAM-ICU assessment across most age groups, with attenuation of this difference in the very elderly (≥ 80 years).

Conclusion

An early postoperative positive CAM-ICU assessment was associated with higher risks of PMV, reintubation, ICU readmission, and prolonged ICU stay after major cardiac surgery. Routine CAM-ICU screening may help identify patients at risk of prolonged ventilation and enable timely, targeted interventions to improve postoperative respiratory recovery.

Introduction

Delirium is a common and serious complication among critically ill patients, including those undergoing cardiac surgery. It is associated with adverse outcomes in hospitalized patients, such as prolonged length of stay, an increased likelihood of discharge to institutional care, and elevated mortality rates [1–3].

The Confusion Assessment Method for the Intensive Care Unit (CAM-ICU) is a bedside diagnostic algorithm that evaluates four features—(1) acute onset or fluctuating course, (2) inattention, (3) altered level of consciousness, and (4) disorganized thinking—to detect delirium in ICU patients [4, 5]. A patient is considered CAM-ICU positive when features 1 and 2 are present with either feature 3 or 4. CAM-ICU has been validated in both mechanically ventilated and non-ventilated ICU populations and demonstrates high inter-rater reliability and diagnostic accuracy. Owing to its brevity, ease of use, and routine adoption in ICU screening, CAM-ICU is well suited for early postoperative surveillance in cardiac surgical patients [3, 6].

Prolonged mechanical ventilation (PMV) following major cardiac surgery is a well-recognized risk factor for postoperative complications and mortality [7, 8]. Specifically, patients who require mechanical ventilation for 24 h or more after cardiac surgery exhibit higher mortality rates, and this threshold is widely accepted as a clinically meaningful definition of PMV [9, 10]. Identifying patients at risk for PMV in the early postoperative period is important for implementing timely and individualized management strategies that may improve outcomes. Although previous studies have demonstrated associations between delirium and delayed ventilator weaning [11, 12], evidence regarding the predictive value of CAM-ICU assessments in the cardiac surgical population remains limited.

This study aimed to investigate whether a positive CAM-ICU assessment in the early postoperative period is associated with the risk of PMV and other adverse outcomes in patients undergoing major cardiac surgery.

Methods

Study cohort

From the Institutional Cardiac Surgery Database of Asan Medical Center, we identified 2,000 consecutive adult patients (aged ≥ 18 years) who underwent major cardiac surgery between January and December 2022. Patients were excluded if they (1) did not have a documented CAM-ICU assessment; (2) died within 24 h after surgery; or (3) were not admitted to the intensive care unit (ICU) postoperatively. Consequently, 1,437 patients were included in the final analysis (Fig. 1).

Fig. 1.

Flowchart of study patients

ICU management and ventilator weaning protocol following cardiac surgery

Patients were admitted directly to the ICU following surgery. Hemodynamic status was continuously monitored using arterial catheters, and fluid therapy, inotropes, and vasopressors were administered based on the type of surgery and the patient's condition.

Sedation was administered according to the attending physician’s judgment based on the patient’s clinical condition. Indications for sedation included postoperative bleeding, the presence of cardiac assist devices such as an intra-aortic balloon pump or extracorporeal membrane oxygenation, ventilator weaning failure, or hemodynamic instability. Routine sedation was targeted to achieve light sedation, typically using dexmedetomidine, remifentanil, or a combination of both agents. Deep sedation was applied in hemodynamically unstable or complex cases, most commonly using remifentanil with propofol, or a combination of remifentanil, propofol, and dexmedetomidine, with or without neuromuscular blockade. Propofol was administered as repeated boluses of 10–20 mg when rapid sedation was required and was discontinued as soon as possible to minimize hypotension and the risk of propofol infusion syndrome, with subsequent transition to dexmedetomidine when appropriate. Rocuronium bromide was used for neuromuscular blockade and discontinued within 3 days. For short procedures or patient transport, midazolam (2–3 mg intravenous bolus) was administered as needed. In emergency situations such as intubation or re-sternotomy in hemodynamically unstable patients, etomidate (10 mg intravenous bolus, repeatable once) was used, and fentanyl was administered additionally for analgesia during the procedure.

Mechanical ventilation was initiated in pressure-controlled mode, with an initial tidal volume set at 8–10 mL/kg of predicted body weight, a positive end-expiratory pressure (PEEP) of 5 cmH₂O, a fraction of inspired oxygen (FiO₂) of 0.6, a respiratory rate of 12 breaths/min, and an inspiratory-to-expiratory ratio of 1:2. Ventilator weaning was initiated when patients met all of the following criteria: alert and oriented mental status; adequate spontaneous respiratory effort with a respiratory rate (RR) ≥ the ventilator setting plus 2 breaths/min; tidal volume ≥ 6 mL/kg of predicted body weight; FiO₂ of 0.40; PEEP of 5 cmH₂O; pressure support (PS) level < 15 cmH₂O; hemodynamic stability, defined as systolic blood pressure > 90 mm Hg and heart rate < 120 beats/min; and total chest tube drainage < 100 mL/h.

Extubation was performed after patients successfully tolerated PS mode with an FiO₂ of 0.40, PEEP of 5 cmH₂O, and a PS level of ≤ 10 cmH₂O for at least 1 h. The arterial blood gas analysis criteria for extubation included a pH of ≥ 7.3, a PaO₂ of ≥ 70 mm Hg, an SpO₂ of > 95%, and an RR of < 30 breaths/min. All patients were managed in accordance with our center’s standardized ventilator weaning and extubation protocol. Daily delirium screening was performed using a Korean-translated version of the CAM-ICU at least every 8 h during the postoperative ICU stay.

Data collection and clinical outcomes

The primary outcome of interest was PMV, defined as the need for ventilator support lasting more than 24 h following surgery. Secondary outcomes included early complications that occurred within 30 days of surgery or during the same hospitalization, such as early death, reintubation, ICU readmission, and an ICU stay exceeding 48 h after extubation. Both ICU readmission and reintubation were defined as any occurrence for any reason after the initial discharge from the ICU or extubation.

Clinical follow-up was conducted until December 31, 2024. Survival status and date of death were verified using data from the Korean National Health Insurance Service. Additional clinical data were collected through a comprehensive review of institutional electronic medical records and telephone interviews. Patients who did not experience adverse clinical events were censored at the end of the follow-up period.

Statistical analysis

Statistical analyses were performed using R version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria). Categorical variables, presented as frequencies (%), were compared using the χ² test or Fisher’s exact test. Continuous variables, presented as mean ± standard deviation for normally distributed data or as median (interquartile range) for skewed data, were compared using Student’s unpaired t-test or the Mann–Whitney U test, as appropriate.

To mitigate potential confounders and selection biases inherent in observational comparisons, differences between patients who tested positive and negative for CAM-ICU were adjusted using inverse probability of treatment weighting (IPTW) based on propensity scores. Propensity scores were estimated using logistic regression that incorporated baseline demographics, comorbidities, and surgical characteristics (Table 1). Stabilized IPTW with trimming of extreme weights was used to reduce variance and limit the influence of outliers. Covariate balance between exposure groups was assessed using standardized mean differences (SMD), with SMD < 0.20 indicating acceptable balance [13].

Table 1.

Baseline characteristics and operative profiles according to the CAM-ICU status in the unadjusted and IPTW-adjusted cohorts

	Original				IPTW-adjusted
Variables	CAM-ICU (−) (N = 1156)	CAM-ICU (+) (N = 281)	p-value	SMD	CAM-ICU (−) (N = 1156)	CAM-ICU (+) (N = 281)	p-value	SMD (%)
Age (years)	59.6 ± 12.8	66.3 ± 11.2	< 0.001	55.5%	61.1 ± 12.8	60.6 ± 12.8	0.718	3.8
Female	429 (37.1)	121 (43.1)	0.076	12.2%	439 (38.0)	104 (37.0)	0.820	2.0
BMI (kg/m2)	24.2 ± 3.5	24.0 ± 4.3	0.494	4.8%	24.1 ± 3.5	24.2 ± 4.4	0.785	2.6
Hypertension	627 (54.2)	180 (64.1)	0.004	20.1%	645 (55.8)	163 (57.9)	0.630	4.3
Diabetes mellitus	282 (24.4)	90 (32.0)	0.011	17.0%	302 (26.2)	77 (27.3)	0.755	2.5
Cerebrovascular disease	109 (9.4)	39 (13.9)	0.036	13.9%	124 (10.8)	31 (11.1)	0.881	1.2
COPD	52 (4.5)	30 (10.7)	< 0.001	43.9%	70 (6.0)	15 (5.3)	0.614	3.3
eGFR (mL/min/1.73 ${\text{m}}^2$)			< 0.001	43.9%			0.883	3.1
< 30	47 (4.1)	26 (9.3)			59 (5.1)	13 (4.7)
30–60	137 (11.9)	71 (25.3)			167 (14.4)	38 (13.6)
> 60	972 (84.1)	184 (65.5)			930 (80.5)	230 (81.7)
History of cardiac surgery	164 (14.2)	67 (23.8)	< 0.001	24.8%	192 (16.6)	46 (16.3)	0.910	0.9
Hemoglobin (g/dL)	12.8 ± 1.9	11.9 ± 2.2	< 0.001	44.7%	12.6 ± 2.0	12.7 ± 2.1	0.487	6.0
PMV	8 (0.7)	9 (3.2)	0.001	18.2%	17 (1.5)	4 (1.3)	0.815	1.4
History of ICU admission	55 (4.8)	42 (14.9)	< 0.001	34.7%	78 (6.8)	18 (6.3)	0.749	2.0
Preoperative ECMO support	8 (0.7)	11 (3.9)	< 0.001	21.6%	17 (1.5)	4 (1.4)	0.890	0.8
Operative profile
Operation type			0.002	28.0%			0.979	7.7
Valve	624 (54.0)	129 (45.9)			602 (52.1)	143 (50.8)
CABG	184 (15.9)	48 (17.1)			186 (16.1)	53 (18.8)
Aorta	95 (8.2)	40 (14.2)			117 (10.1)	29 (10.3)
Valve + CABG	39 (3.4)	16 (5.7)			43 (3.7)	10 (3.6)
Valve + Aorta	88 (7.6)	23 (8.2)			92 (8.0)	21 (7.4)
CABG + Aorta	3 (0.3)	3 (1.1)			4 (0.3)	1 (0.4)
Others	123 (10.6)	22 (7.8)			113 (9.7)	25 (8.8)
Operative time, hours	3.8 ± 1.4	4.7 ± 1.8	< 0.001	59.6%	4.0 ± 1.7	4.0 ± 1.5	0.562	4.9
MICS	598 (51.7)	96 (34.2)	< 0.001	36.1%	551 (47.7)	132 (46.9)	0.862	1.5
Emergent operation	110 (9.5)	49 (17.4)	< 0.001	23.4%	134 (11.6)	28 (9.9)	0.459	5.3
Sedative use, n (%)	239 (20.7)	198 (70.5)	< 0.001	115.4%	261 (22.5)	185 (65.7)	< 0.001	96.5
ICU stay (days)	1 (1–2)	3 (2–8)	< 0.001	54.6%	1 (1–2)	2 (1–4)	< 0.001	27.5
Hospital stay (days)	8 (7–12)	14 (9–36)	0.023	9.5%	9 (7–13)	10 (8–21)	0.302	9.6

Values are presented as n (%), mean ± standard deviation, or median (interquartile range: Q1–Q3), unless otherwise specified. IPTW, Inverse-probability-of-treatment weighting; SMD, Standardized mean difference; BMI, Body mass index; COPD, Chronic obstructive pulmonary disease; eGFR, Estimated glomerular filtration rate; PMV, Preoperative mechanical ventilation; ICU, Intensive care unit; ECMO, Extracorporeal membrane oxygenation; CABG, Coronary artery bypass graft surgery; MICS, Minimally invasive cardiac surgery

After adjusting for baseline differences using the IPTW method, logistic regression models were employed to evaluate early postoperative outcomes, with results expressed as odds ratios and 95% confidence intervals. All-cause mortality was analyzed using Cox proportional hazards models with IPTW-adjusted weights. For time-to-event outcomes involving competing risks (e.g., PMV vs. early death), Fine–Gray subdistribution hazard models were applied, treating death as a competing event. Long-term survival was assessed using the Kaplan–Meier method and compared through IPTW-adjusted Cox regression.

Subgroup analyses were conducted based on prespecified clinical variables (e.g., age group, sex, renal function, surgical type) to evaluate effect modification. Interaction terms between CAM-ICU status and the subgroups were formally tested within Fine–Gray models. To visualize age-related differences, we modeled ventilator duration by age and CAM-ICU group and plotted adjusted marginal effects (Fig. 4). Time-to-extubation curves by CAM-ICU group are displayed in Fig. 5. All reported p-values were two-tailed, and outcomes with p-values < 0.05 were deemed statistically significant.

Fig. 4.

Relationship between age and ventilator duration stratified by CAM-ICU assessment

Fig. 5.

Time-to-event curves for ventilator weaning stratified by CAM-ICU assessment

Results

Baseline characteristics

The mean age of the entire cohort was 60.9 ± 12.8 years, and 550 patients (38.3%) were men. Among the 1437 patients included in the study, 281 (19.6%) were assessed as CAM-ICU positive during the early postoperative period. Baseline demographic and clinical characteristics are summarized in Table 1.

Patients with a positive CAM-ICU assessment were significantly older than those with a negative CAM-ICU assessment (66.3 ± 11.2 vs. 59.6 ± 12.8 years; p < 0.001). Additionally, the CAM-ICU–positive group had a higher prevalence of comorbidities, including hypertension (64.1% vs. 54.2%; p = 0.004), diabetes mellitus (32.0% vs. 24.4%; p = 0.011), cerebrovascular disease (13.9% vs. 9.4%; p = 0.036), and chronic obstructive pulmonary disease (10.7% vs. 4.5%; p < 0.001).

Impaired renal function (estimated glomerular filtration rate [eGFR] < 60 mL/min/1.73 m²) was significantly more prevalent in the CAM-ICU–positive group (34.6% vs. 16.0%; p < 0.001). In addition, the CAM-ICU–positive group had a higher proportion of patients who received preoperative mechanical ventilation (3.2% vs. 0.7%; p = 0.001) and ECMO (3.9% vs. 0.7%; p < 0.001), as well as those who were admitted to the ICU prior to surgery (14.9% vs. 4.8%; p < 0.001).

Operative profiles

Operative characteristics are presented in Table 1. The CAM-ICU–positive group had a higher proportion of those who underwent emergent operations (17.4% vs. 9.5%; p < 0.001) and experienced longer operative times (4.7 ± 1.8 vs. 3.8 ± 1.4 h; p < 0.001). The distribution of surgical procedures also varied between the groups, with a higher proportion of aortic procedures observed in the CAM-ICU–positive group (23.8% vs. 16.1%; p < 0.001). In contrast, minimally invasive cardiac surgery was more prevalent among the CAM-ICU–negative group (51.7% vs. 34.2%; p < 0.001).

Unadjusted clinical outcomes

The CAM-ICU–positive group had significantly higher rates of adverse outcomes (Table 2), including the incidence of PMV (37.4% vs. 5.4%; odds ratio [OR] 10.52, 95% CI 7.43–15.04; p < 0.001). Other early complications, including early death (3.2% vs. 0.8%; p = 0.003), reintubation (8.2% vs. 1.2%; p < 0.001), ICU readmission (8.5% vs. 1.8%; p < 0.001), and prolonged ICU stay beyond 48 h after extubation (45.9% vs. 17.1%; p < 0.001), were also significantly more prevalent in the CAM-ICU positive group.

Table 2.

Early and overall clinical outcomes according to the CAM-ICU status in the unadjusted and IPTW-adjusted cohorts

Outcomes	Original				IPTW-adjusted
	No. of events (rate)				No. of events (rate)
	CAM-ICU (−) (N = 1156)	CAM-ICU (+) (N = 281)	OR or HR (95% CI)	p-value	CAM-ICU (−) (N = 1156)	CAM-ICU (+) (N = 281)	OR or HR (95% CI)	p-value
Early outcomes
PMV (> 24 h)	62 (5.4)	105 (37.4)	10.52 (7.43–15.04)	< 0.001	91 (7.8)	68 (24.1)	3.60 (2.53–5.10)	< 0.001
Early death	9 (0.8)	9 (3.2)	4.21 (1.67–10.60)	0.003	19 (1.6)	5 (1.8)	1.13 (0.37–2.86)	0.805
Reintubation	14 (1.2)	23 (8.2)	7.27 (3.73–14.66)	< 0.001	15 (1.3)	12 (4.4)	3.37 (1.52–7.31)	0.002
ICU readmission	21 (1.8)	24 (8.5)	5.05 (2.77–9.28)	< 0.001	22 (1.9)	13 (4.6)	2.53 (1.22–5.03)	0.010
Prolonged ICU stay after extubation (> 48 h)	198 (17.1)	129 (45.9)	4.10 (3.10–5.44)	< 0.001	173 (15.0)	96 (34.0)	2.95 (2.19–3.96)	< 0.001
Overall outcomes
All-cause death	12 (1.0)	20 (7.1)	4.11 (1.63–10.36)	0.003	22 (1.9)	10 (3.4)	1.77 (0.79–3.95)	0.164

Values are presented as n (%). Odds ratios and hazard ratios are presented for early outcomes and overall outcomes, respectively. OR, Odds ratio; HR, Hazard ratio; CI, Confidence interval; IPTW, Inverse-probability-of-treatment weighting; ICU, Intensive care unit

Follow-up for all-cause mortality was completed for all patients, with a median follow-up duration of 29.8 months. During this period, deaths occurred in 20 patients (7.1%) with positive CAM-ICU assessments and in 12 patients (1.0%) with negative assessments (p < 0.001). In the unadjusted Cox regression model, a positive CAM-ICU assessment was significantly associated with increased all-cause mortality (hazard ratio [HR], 4.11; 95% CI, 1.63–10.36; p = 0.003) (Fig. 2A).

Fig. 2.

Kaplan–Meier curves for all-cause mortality according to CAM-ICU assessment in unadjusted (Panel A) and IPTW-adjusted (Panel B) cohorts

Adjusted clinical outcomes

After IPTW adjustment, baseline characteristics—including demographics, comorbidities, and operative profiles—were well balanced, with standardized mean differences of less than 20% across most variables. In the adjusted logistic regression models, a positive CAM-ICU assessment was independently associated with increased risks of PMV (OR 3.60; 95% CI, 2.53–5.10; p < 0.001), reintubation (OR 3.37; 95% CI, 1.52–7.31; p = 0.002), ICU readmission (OR 2.53; 95% CI, 1.22–5.03; p = 0.010), and a prolonged ICU stay beyond 48 h after extubation (OR 2.95; 95% CI, 2.19–3.96; p < 0.001).

In contrast, early postoperative mortality did not significantly differ between the groups (OR 1.13; 95% CI, 0.37–2.86; p = 0.805). Similarly, a positive CAM-ICU assessment was not significantly associated with all-cause mortality in the Cox regression model (HR 1.77; 95% CI, 0.79–3.95; p = 0.164) (Fig. 2B).

Subgroup analysis

Subgroup analyses were conducted to evaluate whether the association between a positive CAM-ICU assessment and the risk of PMV varied across predefined clinical subgroups (Fig. 3). The increased risk of PMV in patients with a positive CAM-ICU assessment was consistently observed across the majority of subgroups.

Fig. 3.

Subdistribution hazard ratios for prolonged mechanical ventilation associated with a positive CAM-ICU assessment in prespecified subgroups. COPD, chronic obstructive pulmonary disease; eGFR, estimated glomerular filtration rate; ECMO, extracorporeal membrane oxygenation

Although the overall association remained consistent, formal tests for interaction revealed statistically significant effect modification in specific subgroups, including patients aged ≥ 70 years, those requiring preoperative mechanical ventilation or ECMO support, and those with a history of prior cardiac surgery. In these groups, the relative effect size of a positive CAM-ICU assessment on PMV was diminished compared to their counterparts.

Age-related pattern in ventilator duration

To further investigate the modifying effect of age, we analyzed the relationship between age and ventilator duration based on the CAM-ICU assessment (Fig. 4). In most age groups, patients with a positive CAM-ICU assessment demonstrated longer ventilator durations compared to those with a negative assessment. However, this difference gradually diminished after the age of 70, and beyond 80 years, patients with a negative CAM-ICU assessment experienced longer ventilator durations.

Predictors of prolonged mechanical ventilation

In the univariate logistic regression analysis, several preoperative and surgical variables were significantly associated with PMV following major cardiac surgery (Table 3). Factors such as older age, chronic obstructive pulmonary disease, eGFR < 60 mL/min/1.73 m², lower hemoglobin levels, preoperative mechanical ventilation, a history of ICU admission, and preoperative ECMO support were significantly associated with an increased risk of PMV. Additionally, aortic surgery, a history of prior cardiac surgery, and emergent surgery were also significantly associated with PMV. In contrast, minimally invasive cardiac surgery was associated with a decreased likelihood of PMV.

Table 3.

Univariable logistic regression for predictors of prolonged mechanical ventilation after major cardiac surgery

Variables	OR (95% CI)	p-value
Age	1.03 (1.01–1.04)	< 0.001
Female sex	1.22 (0.88–1.69)	0.231
BMI	1.02 (0.97–1.06)	0.418
Hypertension	1.33 (0.96–1.86)	0.091
Diabetes mellitus	1.38 (0.97–1.96)	0.067
Cerebrovascular disease	1.29 (0.77–2.08)	0.305
COPD	2.28 (1.28–3.87)	0.003
eGFR
< 30 mL/min/1.73 ${\text{m}}^2$	5.29 (3.07–8.91)	< 0.001
30–60 mL/min/1.73 ${\text{m}}^2$	2.98 (2.01–4.38)	< 0.001
> 60 mL/min/1.73 ${\text{m}}^2$	Reference
History of cardiac surgery	4.10 (2.88–5.82)	< 0.001
Hemoglobin	0.73 (0.67–0.79)	< 0.001
Preoperative mechanical ventilation	19.59 (7.16–62.24)	< 0.001
History of ICU admission	8.21 (5.27–12.76)	< 0.001
Preoperative ECMO support	17.78 (6.92–51.24)	< 0.001
CAM-ICU	10.53 (7.43–15.04)	< 0.001
Operative profile
Operation type
Valve	Reference
CABG	0.70 (0.39–1.20)	0.215
Aorta	3.18 (2.00–5.01)	< 0.001
Valve + CABG	0.74 (0.22–1.88)	0.576
Valve + Aorta	2.21 (1.27–3.71)	0.004
CABG + Aorta	4.72 (0.65–24.67)	0.757
Others	1.34 (0.75–2.28)	0.296
Operative time	1.89 (1.70–2.10)	< 0.001
MICS	1.74 (0.89–3.39)	0.105
Emergent operation	5.93 (4.05–8.65)	< 0.001

OR, Odds ratio; CI, Confidence interval; BMI, Body mass index; COPD, Chronic obstructive pulmonary disease; eGFR, Estimated glomerular filtration rate; ICU, Intensive care unit; ECMO, Extracorporeal membrane oxygenation; CABG, Coronary artery bypass graft surgery; MICS, Minimally invasive cardiac surgery

In the multivariable analysis (Table 4), a positive CAM-ICU assessment remained as the strongest independent predictor of PMV (OR 7.35; 95% CI, 4.86–11.22; p < 0.001). Additional risk factors included a higher BMI, impaired renal function, history of cardiac surgery, preoperative mechanical ventilation, aortic surgery, longer operative time, and emergent surgery. Conversely, higher preoperative hemoglobin levels were associated with a reduced risk of PMV.

Table 4.

Multivariable logistic regression for predictors of prolonged mechanical ventilation after major cardiac surgery

Variables	OR (95% CI)	P-value
BMI	1.07 (1.01–1.12)	0.015
eGFR
< 30 mL/min/1.73 ${\text{m}}^2$	2.63 (1.29–5.27)	0.007
30–60 mL/min/1.73 ${\text{m}}^2$	1.57 (0.94–2.58)	0.080
> 60 mL/min/1.73 ${\text{m}}^2$	Reference
History of cardiac surgery	2.27 (1.43–3.57)	< 0.001
Hemoglobin	0.87 (0.78–0.96)	0.007
Preoperative mechanical ventilation	6.70 (2.02–24.94)	0.003
CAM-ICU	7.35 (4.86–11.22)	< 0.001
Operative profile
Aortic surgery	1.85 (1.14–2.96)	0.011
Operative time	1.55 (1.37–1.75)	< 0.001
Emergent operation	3.40 (2.02–5.70)	< 0.001

Candidate covariates were initially screened with univariable analyses (Table 3). Clinically significant covariates with a p value lower than 0.10 in univariable models were used in the multivariable model. Covariates that remained in the multivariable model using the backward elimination technique were retained in the multivariable model

OR, Odds ratio; CI, Confidence interval; BMI, Body mass index; eGFR, Estimated glomerular filtration rate; CAM-ICU, Confusion assessment method for the intensive care unit

Ventilator weaning probability

Time-to-event analysis for ventilator weaning revealed a significantly lower probability of successful weaning in patients with a positive CAM-ICU assessment compared to those with a negative assessment (Fig. 5). In the CAM-ICU negative group, most patients were successfully weaned from mechanical ventilation within the first 24 h after surgery, showing a steep early increase in cumulative weaning probability. In contrast, the CAM-ICU positive group showed a delayed and more gradual trajectory of ventilator liberation, with the disparity persisting throughout the 96-h observation period (p < 0.001).

Discussion

In this contemporary cohort of patients undergoing major cardiac surgery, we investigated the clinical impact of a positive CAM-ICU assessment on PMV and other early postoperative outcomes. Patients with a positive CAM-ICU assessment experienced significantly higher rates of PMV, reintubation, ICU readmission, and extended ICU stays beyond 48 h following extubation. These associations remained robust after adjustment using IPTW, underscoring the independent prognostic value of a positive early postoperative CAM-ICU assessment.

Our findings are consistent with prior studies reporting that delirium is associated with adverse respiratory outcomes in critically ill populations [10, 14]. A positive CAM-ICU assessment was a significant predictor of PMV in our multivariable analysis. This may reflect the combined effects of impaired cognition, reduced patient cooperation, and the need for deeper or prolonged sedation—factors known to hinder successful ventilator weaning [15–17]. Additionally, the rates of reintubation and ICU readmission were higher among patients with a positive CAM-ICU assessment, further highlighting the detrimental impact of early postoperative delirium on recovery.

In contrast to previous studies that reported a significant association between delirium and mortality in ICU populations [18–20], we found no statistically significant association between a positive CAM-ICU assessment and either early or all-cause mortality after IPTW adjustment. While unadjusted comparisons showed higher mortality among patients with a positive CAM-ICU assessment, this effect disappeared after accounting for preoperative risk factors. Several explanations are plausible. First, our study population predominantly consisted of patients undergoing elective or semi-urgent procedures with relative hemodynamic stability, with only 11.1% of patients requiring emergent operations. This contrasts with general ICU cohorts experiencing sepsis or multiorgan failure, where delirium often reflects more severe systemic illness. Second, the relatively low mortality rate and limited follow-up duration in our cohort may have reduced the statistical power to detect significant long-term differences. Nevertheless, the lack of a mortality association does not diminish the clinical importance of delirium. Its impact may be more strongly expressed through postoperative morbidity—as shown in our study by increased risks of prolonged mechanical ventilation and prolonged ICU stay, both of which escalate resource utilization and healthcare costs. Future research should further quantify these dimensions to better capture the comprehensive morbidity burden of postoperative delirium in cardiac surgical patients.

Subgroup analyses revealed that the association between a positive CAM-ICU assessment and PMV was attenuated in certain high-risk groups, including patients aged ≥ 70 years, those requiring preoperative mechanical ventilation or ECMO support, and those with a history of cardiac surgery. These factors may serve as significant predictors of PMV [21, 22], thereby reducing the relative impact of a positive CAM-ICU assessment.

Age-stratified modeling indicated that ventilator duration was generally longer with a positive CAM-ICU assessment across most age groups, with attenuation after age 70 and a possible reversal among patients aged ≥ 80 years. Among the 18 early deaths in the cohort, 12 occurred in patients aged ≥ 70 years (6 in each CAM-ICU group), suggesting that early mortality may have limited the opportunity for prolonged ventilator support, particularly in elderly patients. Moreover, the apparent reversal in patients over 80 years may be partly attributable to statistical variability resulting from the small sample size (n = 55), which limits the reliability of subgroup estimates.

From a clinical perspective, early screening with the CAM-ICU tool may help identify patients at an elevated risk of respiratory complications following cardiac surgery. In our multivariable model, additional predictors of PMV included higher BMI, impaired renal function, history of cardiac surgery, low preoperative hemoglobin levels, preoperative ventilator use, aortic procedures, longer operative times, and emergent operations. Recognizing these risk factors could facilitate early intervention and optimize resource allocation.

Institution-level strategies to prevent and manage postoperative delirium—such as early mobilization, environmental optimization, and the minimization of deliriogenic medications—should be prioritized [23, 24]. Certain patients may also benefit from pharmacologic interventions, including dexmedetomidine [25]. Special attention is warranted for older adults with a positive CAM-ICU assessment, as they are at an increased risk for PMV and related complications.

This study has several limitations. First, as a retrospective, single-center analysis, it may be subject to selection biases that could not be fully addressed through statistical adjustments. Second, our cohort was confined to a single high-volume tertiary center, which may limit the generalizability of the findings to other practice settings. Third, CAM-ICU performance can be reduced in deeply sedated or hypoactive patients, potentially leading to under-detection; moreover, delirium severity, duration, and motor subtype were not captured and could modify risk estimates. Fourth, we did not assess patient-centered outcomes such as cognitive function, physical disability, or quality of life, which are increasingly emphasized in delirium research. Fifth, the overall mortality rate was low and the follow-up period may have been insufficient to detect modest long-term differences in survival. Finally, subgroup results—particularly for patients aged ≥ 80 years—should be interpreted with caution given small sample sizes, limited precision of age-interaction estimates, and the attendant risk of type II error.

Conclusion

A positive CAM-ICU assessment in the early postoperative period was associated with an increased risk of prolonged mechanical ventilation, reintubation, ICU readmission, and extended ICU stays among patients undergoing major cardiac surgery. Predicted ventilator duration was generally longer with a positive CAM-ICU assessment across most age groups, with attenuation in the very elderly. Routine CAM-ICU screening may help identify patients at risk of prolonged ventilation and enable timely, targeted interventions reduce the risk of delirium.

Abbreviations

CAM-ICU

Confusion Assessment Method for the Intensive Care Unit

PMV

Prolonged mechanical ventilation

ICU

Intensive care unit

PEEP

Positive end-expiratory pressure

FiO₂

Fraction of inspired oxygen

RR

Respiratory rate

PS

Pressure support

IPTW

Inverse-probability-of-treatment weighting

SMD

Standardized mean differences

eGFR

Estimated glomerular filtration rate

OR

Odds ratio

HR

Hazard ratio

Author contributions

JYK was responsible for the methodology, data curation, formal analysis, visualization, and drafting the original manuscript. YAK contributed to data curation and the drafting of the original manuscript. PJK supervised the study, managed the project, and critically revised the manuscript.

Funding

There are no funding sources to declare.

Availability of data and materials

The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the principles of the Declaration of Helsinki. The study protocol was approved by the Institutional Review Board of Asan Medical Center (Study no: 2025–0623; Date of approval: May 22, 2025). The requirement for informed consent was waived due to the observational nature of the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.