Does Timing Matter? The Effect of Intensive Care Unit Arrival Timing on Elective Cardiac Surgery

Abstract

BACKGROUND:

Due to staffing changes at scheduled intervals and decreases in essential staff in the evenings, late intensive care unit (ICU) arrivals may be at risk for suboptimal outcomes. Utilizing a regional collaborative, we sought to determine the effect of ICU arrival timing on outcomes in elective isolated coronary artery bypass (CABG).

METHODS:

Adults undergoing elective, isolated CABG from 17 hospitals between 2013–2023 were identified. Patients with missing predicted risk of mortality or missing ICU arrival time were excluded. Late ICU arrival time was defined as between 18:00–06:00. Hierarchical logistic regression with appropriate predicted risk scores was utilized for outcome risk adjustment.

RESULTS:

We identified 11,638 patients, with 972 (8.4%) experiencing late ICU arrival. Late ICU arrival patients had higher predicted risk of morbidity or mortality (8.2% [5.6%, 12.0% vs. 7.7% [5.5%, 11.5%], p=0.048) compared to early ICU arrival patients with longer median cardiopulmonary bypass times (96 minutes [78, 119] vs. 93 [73, 116], p<0.001). Late ICU arrival patients experienced more unadjusted complications including prolonged ventilation (7.7% vs. 4.2%, p<0.001) and operative mortality (2.0% vs. 1.1%, p=0.02), although no difference in failure-to-rescue (11.0% vs. 10.4%%, p=0.84). Logistic regression with risk adjustment demonstrated late ICU arrival as a predictor of prolonged ventilation (OR=1.49 [1.12–1.99], p=0.006).

CONCLUSIONS:

Following adjustment, late ICU arrivals experienced higher rates of prolonged ventilation, although this did not translate to failure-to-rescue.

Staffing and organizational structure of intensive care units (ICUs) influence the clinical outcomes of critically ill patients¹. Human factors such as fatigue, time on the job, circadian rhythms, scheduling, and staffing have been shown to have an impact on patient care across all medical specialties^2–4. Additionally, staffing changes in ICUs frequently occur at nighttime which specifically predisposes patients admitted during off hours to worse outcomes^1,5. In fact, nighttime medical care has previously been associated with worse outcomes in general medicine, cardiology, general surgery, and abdominal organ transplant surgery⁵. Possible mechanisms of this off-hour effect include lower levels of medical and allied health staff, inadequate access to specialized services, discontinuity of care, and reduced availability of certain procedures which leads to reduced accessibility and quality of medical services during off hours⁶.

Specifically in cardiac surgery, there is opposing data on clinical outcomes of surgeries occurring after hours^2,5,7–9. Some studies have suggested no differences in perioperative outcomes, operative mortality, length of stay, or total hospital cost for elective cardiac surgery cases started after hours^7,9. Other studies have concluded that starting non-emergent cardiac surgery cases later in the day is associated with significantly higher absolute and risk-adjusted mortality and that these results should be carefully considered when scheduling elective cardiac surgery cases^10–12.

Utilizing a regional collaborative, we sought to determine the effect of ICU arrival timing on outcomes in elective isolated coronary artery bypass (CABG) to address inconsistencies in the literature. Given concern for decreases in essential staff ratios, particularly in the late evenings, we hypothesized that patients arriving to the ICU later in the day would be at particular risk for worse outcomes including failure-to-rescue (FTR), a quality metric focused on hospital systems.

PATIENTS AND METHODS

Patients

Utilizing the Virginia Cardiac Services Quality Initiative (VCSQI), Society of Thoracic Surgeons (STS) Adult Cardiac Surgery Database (ACSD) data from all patients undergoing elective, isolated CABG from January 2013 until June 2023 were identified in the state of Virginia. The collaborative consists of 17 total hospitals, including 4 academic institutions. Late ICU arrival time was defined as arrival between 18:00 and 06:00 during the entire week, including weekends, and patients were stratified based on either late ICU arrival or early ICU arrival. Our primary outcome of interest was FTR as defined by patients experiencing mortality following one of four prespecified complications: prolonged ventilation, renal failure, permanent stroke, or reoperation. Analyses of our FTR outcome only included patients that experienced one of the four prespecified complications and excluded all other patients that did not experience any of these complications. FTR was chosen as the primary outcome of interest because it is a measure of system level outcomes which we believe would be most impacted by late ICU arrival time. Secondary outcomes of interest included other morbidity measures such components of FTR including prolonged ventilation, postoperative stroke, renal failure, and reoperation. Lastly, cardiac arrest and operative mortality were also compared. Prolonged ventilation was defined as requiring ventilator support after surgery for more than 24 hours. This study was approved by the University of Virginia’s Institutional Review Board (#23305).

Statistical Analysis

Continuous variables were analyzed via Wilcoxon rank sum test with expression as either mean and standard deviation or median with interquartile ranges (IQR). Categorical variables were expressed as absolute numbers with percentages and analyzed via chi-square or Fisher’s Exact Testing for expected outcomes of less than 5 events. A hierarchical multivariable logistic regression model was utilized for risk adjustment to account for total operative time in hours, number of total intraoperative blood products as a continuous variable, whether surgery was performed at an academic center, mechanical circulatory support, center volume of CABG defined by quartiles, year of surgery, and lastly ICU arrival time as a dichotomous variable of late arrival vs. early arrival. CABG volume was defined as the total number of isolated CABG procedures performed by each hospital during the entire study period and was divided into the respective quartiles; <506 cases, 506 to 864 cases, 865 to 1644 cases, and >1644 cases. This volume variable was used categorically in reference to the 1^st quartile for logistic regression. Yearly case volume for each center was also analyzed via linear regression over the study period to rule out significant case volume variability over the study period. For each outcome of interest, the respective STS risk score was utilized for risk adjustment where applicable such as predicted risk of prolonged ventilation for modeling prolonged ventilation and predicted risk of mortality (PROM) for modeling operative mortality. If no relevant risk score was available for an outcome, STS predicted risk of morbidity or mortality (PROMM) was utilized. Hospital was utilized as a random effect to account for potential clustering of patients. An interaction term was created between center CABG volume and late ICU arrival to determine if perhaps lower volume centers with less resources may have contributed to changes in outcomes, particularly prolonged ventilation. A plot of the marginal effect of increasing hour of ICU arrival on predicted prolonged ventilation was created utilizing our logistic regression to control for the same risk factors specified above. All statistical analyses were carried out using SAS Version 9.4 (SAS Institutive, Cary, NC) and R Version 4.3.0 (Posit, PBC, Boston, MA) with a p-value less than 0.05 determining significance.

RESULTS

Population Characteristics

From 2013 until 2023, a total of 11,638 patients underwent isolated CABG with 8.4% (972) experiencing late ICU arrival (Supplemental Figure 1); only 34 (0.3%) patients underwent elective surgery on the weekend. A histogram of patient ICU arrival time is demonstrated in Figure 1. Patient demographics are displayed in Table 1. Patients arriving late to the ICU were of similar ages (66 years vs. 66 years, p=0.95), proportion female (20.1% vs. 20.7%, p=0.63), and were more likely to undergo surgery at an academic center (62.1% vs. 38.1%, p<0.001). Late ICU arrival patients did have a higher rate of preoperative atrial fibrillation or flutter (8.3% vs. 6.6%, p=0.03) and higher median PROMM (8.2% [5.6%, 12.0%] vs. 7.7% [5.5%, 11.5%], p=0.048) but similar predicted risk of mortality (PROM) scores (0.8% [0.5%, 1.5%] vs. 0.8% [0.5%, 1.4%], p=0.06). Late arrival ICU patients were less likely to undergo off-pump CABG (2.8% vs. 5.7%, p<0.001) but had similar rates of reoperative cardiac surgery (1.5% vs. 2.0%, p=0.56) and atrial fibrillation ablation (3.7% vs. 2.7%, p=0.08).

Figure 1.

Histogram of patient intensive care unit arrival time.

Table 1.

Baseline characteristics and demographics.

	Late ICU Arrival (n = 972)	Ealy ICU Time Arrival (n = 10,666)	P value
Age (years)	66 ± 10	66 ± 10	0.95
Female	195 (20.1%)	2209 (20.7%)	0.63
BMI (kg/m2)	30.4 ± 5.8	30.3 ± 5.8	0.73
Albumin (gm/dL)	4.1 [3.8, 4.3	4.0 [3.7, 4.3]	<0.001
Preoperative Ejection Fraction (%)	57 [50, 60]	57 [50, 60]	0.86
Hypertension	893 (91.9%)	9732 (91.2%)	0.51
Diabetes	513 (52.8%)	5351 (50.2%)	0.12
Sleep Apnea	219 (22.5%)	2424 (22.7%)	0.89
Chronic Lung Disease	234 (24.1%)	2667 (25.0%)	0.52
Atrial Fibrillation or Flutter	81 (8.3%)	699 (6.6%)	0.03
Prior cardiac surgery	15 (1.5%)	192 (2.0%)	0.56
Surgery at academic center	604 (62.1%)	4063 (38.1%)	<0.001
Median Predicted Risk of Mortality	0.8% [0.5%, 1.5%]	0.8% [0.5%, 1.4%]	0.06
Median Predicted Risk of Morbidity or Mortality	8.2% [5.6%, 12.0%]	7.7% [5.5%, 11.5%]	0.048

Intraoperative Outcomes

Intraoperative findings are displayed in Table 2. Both cardiopulmonary bypass (96 minutes [78, 119] vs. 93 minutes [73, 116], p<0.001) and cross clamp times (74 minutes [58, 91] vs. 70 minutes [53, 88], p<0.001) were longer in the late ICU arrival group. Late ICU arrival patients received more intraoperative blood products including cryoprecipitate (8.2% vs. 2.2%, p<0.001), fresh frozen plasma (6.0% vs. 2.3%, p<0.001), and red blood cells (9.9% vs. 6.8%, p<0.001). They also required extracorporeal membrane oxygenation (ECMO) (0.6% vs. 0.2%, p=0.01) and intra-aortic balloon pumps (IABP) (2.4% vs. 1.4%, p=0.02) more frequently in the operating room and postoperatively.

Table 2.

Intraoperative and postoperative patient outcomes.

	Late ICU Arrival (n = 972)	Ealy ICU Time Arrival (n = 10,666)	P value
Atrial Ablation	36 (3.7%)	291 (2.7%)	0.08
Off-pump Approach	27 (2.8%)	611 (5.7%)	<0.001
Cardiopulmonary bypass time (minutes)	96 [78, 119]	93 [73, 116]	<0.001
Cross Clamp time (minutes)	74 [58, 91]	70 [53, 88]	<0.001
Total Time in Operating Room (hours)	5.4 [4.8, 6.2]	5.2 [4.5, 6.0]	<0.001
Intraoperative Cryoprecipitate transfusion	80 (8.2%)	232 (2.2%)	<0.001
Intraoperative Fresh Frozen Plasma Transfusion	58 (6.0%)	243 (2.3%)	<0.001
Postoperative Cryoprecipitate Transfusion	54 (5.6%)	325 (3.1%)	<0.001
Postoperative Fresh Frozen Plasma Transfusion	50 (5.1%)	345 (3.2%)	0.002
Postoperative Platelet Transfusion	24 (2.5%)	165 (1.6%)	0.03
Postoperative Red Blood Cell Transfusion	220 (22.6%)	1926 (18.1%)	<0.001
Extubated in Operating Room	7 (0.7%)	149 (1.4%)	0.08
ECMO Use Intraoperative or Postoperative	6 (0.6%)	22 (0.2%)	0.01
Intra-aortic balloon pump (IABP) Intraoperative or postoperative	23 (2.4%)	150 (1.4%)	0.02
Intraoperative or Postoperative ECMO and/or IABP Use	26 (2.7%)	164 (1.5%)	0.007
Postoperative Permanent Stroke	12 (1.2%)	94 (0.9%)	0.27
Postoperative Cardiac Arrest	16 (1.7%)	99 (0.9%)	0.03
Reoperation for Bleeding	20 (2.1%)	149 (1.4%)	0.10
Reoperation Any Reason	24 (2.5%)	177 (1.7%)	0.06
Postoperative Renal Failure	22 (2.3%)	177 (1.7%)	0.16
Postoperative Dialysis	17 (1.8%)	121 (1.1%)	0.09
Postoperative Pneumonia	21 (2.2%)	182 (1.7%)	0.30
Postoperative Atrial Fibrillation	221 (22.7%)	2509 (23.5%)	0.58
Prolonged Ventilation	75 (7.7%)	448 (4.2%)	<0.001
Total Ventilator Hours	5 [4, 9]	4 [3, 6]	<0.001
Initial ICU Hours	45 [23, 85]	48 [26, 76]	0.006
Reintubation	22 (2.3%)	173 (1.6%)	0.14
Failure to Rescue (FTR) (n=815)	11 (11.0%%)	74 (10.4%)	0.84
Length of Stay (days)	6 [5, 8]	5 [4, 7]	<0.001
Operative Mortality	19 (2.0%)	116 (1.1%)	0.02
Readmission	83 (8.5%)	817 (7.7%)	0.33

Unadjusted Postoperative Outcomes

Postoperative outcomes are also displayed in Table 2. Late ICU arrival patients experienced more complications including cardiac arrest (1.7% vs. 0.9%, p=0.03), and prolonged ventilation (7.7% vs. 4.2%, p<0.001). Figure 2 demonstrates the unadjusted risk of prolonged ventilation based on arrival time in the ICU. There was no difference in the rate of reoperation for any reason between groups (2.5% vs. 1.7%, p=0.06) nor differences in renal failure (2.3% vs. 1.7%, p=0.16) or stroke (1.2% vs. 0.9%, p=0.27). Composite FTR complications were higher in the late ICU arrival group (10.3% vs. 6.7%, p<0.001) driven by higher rates of prolonged ventilation. There was no difference in FTR among patients that experienced these pre-specified complications (11.0% vs. 10.4%, p=0.84). Unadjusted operative mortality was higher in the late ICU arrival group (2.0% vs. 1.1%, p=0.02) as was hospital length of stay (6 days [5, 8] vs. 5 days [4, 7], p<0.001). There was no difference in rate of readmission (8.5% vs. 7.7%, p=0.33).

Figure 2.

Unadjusted incidence of prolonged ventilation based on intensive care unit arrival time with 18:00 to 01:00 signifying late arrival.

Adjusted Postoperative Outcomes

A hierarchical multivariable logistic regression model with hospital as a random effect was utilized to determine if late ICU arrival was independently associated with any significant postoperative outcomes. There were 11,632 patients included in all of the adjusted analyses as 6 patients were excluded for missing hospital center data, all of which came from the early ICU arrival group. Modeling of our FTR outcome included only the 815 patients that experienced one of the four prespecified complications. Of all patients, 75 (7.7%) late ICU arrival patients and 448 (4.2%) early ICU arrival patients experienced prolonged ventilation. Late ICU arrival was found to be a significant predictor of prolonged ventilation (OR=1.49 [1.12–1.99], p=0.006) (Figure 3, Supplemental Table 1). However, late ICU arrival time was not found to predict FTR (OR=0.87 [0.39–1.95], p=0.74) (Table 3), cardiac arrest (OR=1.63 [0.90–2.96], p=0.11), or operative mortality (OR=1.51 [0.86–2.67], p=0.15). Inclusion of an interaction term between center volume of CABG and late ICU arrival in the prolonged ventilation model was not found to be significant (p=0.77).

Figure 3.

Forest plot of multivariable logistic regression predicting outcome variable of prolonged ventilation. Red signifies significant relationship.

Table 3.

Predictors of failure to rescue utilizing late arrival to the ICU as a categorical variable.

C Statistic = 0.83	Odds Ratio	Confidence Limits	P Value
Late Arrival to ICU	0.87	[0.39–1.95]	0.74
Predicted Risk of Mortality	1.12	[1.04–1.21]	<0.001
Total Time in Operating Room (hours)	0.93	[0.76–1.14]	0.48
Year of Surgery	1.03	[0.94–1.12]	0.54
Surgery at academic center	0.97	[0.41–2.29]	0.94
Intraoperative composite of blood products (per unit)	1.08	[0.97–1.20]	0.15
Intra or postoperative mechanical circulatory support	3.53	[1.91–6.53]	<0.001
CABG Volume Quartile 4 (vs. 1)	0.27	[0.10–0.77]	0.01
CABG Volume Quartile 3 (vs. 1)	0.66	[0.24–1.82]	0.42
CABG Volume Quartile 2 (vs. 1)	0.39	[0.16–0.94]	0.04

Analysis of Marginal Effects

Arrival time was utilized as a continuous variable to determine the hourly effect ICU arrival had on risk of prolonged ventilation (Supplemental Table 2). A plot of marginal effect of increasing hour for ICU arrival on predicted prolonged ventilation was created using the previous regression model. This plot demonstrated a significant increase in risk of prolonged ventilation with increasing hour of ICU arrival (Figure 4).

Figure 4.

Margins plot of predicted risk of prolonged ventilation by intensive care unit arrival time. Time of day corresponds to 24-hour clock.

COMMENT

Through analysis of a large multicenter regional STS dataset, we found that late arrival to the ICU did not increase risk of FTR although did result in higher rates of prolonged ventilation following risk adjustment. We hypothesize this may be due to hesitation to extubate patients arriving at later times of the evening from perhaps reduced staffing as might be the case with lack of evening or overnight intensivist physical presence. Other possible explanations are different levels of experience among night nurses, respiratory therapists, or physicians. This suggests a need to review evening extubation protocols to help combat risk of prolonged ventilation and its potential consequences such as ventilator-associated pneumonia, prolonged length of stay, and mortality¹³. However, this only addresses a small component of prolonged ventilation as patients extubated the following day would not qualify for prolonged ventilation if within 24 hours from ICU arrival. Thus, the downstream effects from hesitation to extubate overnight of lengthened pharmacologic sedation and diaphragm weakening from excessive ventilator support may also play into delays with extubation.

Studies on late ICU arrival are mixed in cardiac surgery and other specialties with some suggesting that after hours elective cases pose no risk of worse outcomes and others that suggest patients arriving to the ICU in the evenings experience worse mortality, even after risk-adjustment ^2,7–10,14. In the present study, late arrival to the ICU was associated with an increase in unadjusted postoperative complications such as cardiac arrest, prolonged ventilation, and mortality. Although after risk adjustment, majority of these complications did not persist. This suggests late ICU arrival may not affect major morbidity or mortality in patients undergoing elective, isolated CABG. A study by Axtell et al. looked at elective cardiac surgery cases starting after 15:00 and found no increased risk of morbidity or mortality for coronary and valve surgery⁹. Similarly, a study from Tan et al. looking at elective CABG timing did not demonstrate worse outcomes for cases based on time of day, day of the week, or month of the year². They suggest this consistency in outcomes stems from protocols in place to standardize patient care after surgery. However, a study from Yount et al. found worse outcomes for these patients with elective cases starting after 15:00 despite shorter operative times¹⁰. Our study differs from these in a few ways. First, we did not look at elective case start time and instead focused on the actual arrival time to the ICU which would be more inclusive of cases compared to other studies. As such, cases starting much earlier than 15:00 that may have taken longer than expected would have been included in our patient cohort. Our study also represents findings from a regional collaborative rather than a single institution which importantly would allow for a diverse set of ICU management strategies and different staffing models such as centers with and without 24-hour intensivist coverage. We do not know which centers possess differing staffing strategies but could hypothesize that a wide variety of models existed in the 17 hospitals included in our study.

FTR was our primary outcome of interest as it is an important quality metric of hospital systems and their ability to respond to complications. FTR consists of mortality following any of the following four prespecified complications: reoperation, prolonged ventilation, stroke, and renal failure¹⁵. This metric was chosen as it is more reflective of systems-based issues which we hypothesized may play a role in worse outcomes for patients admitted to the ICU late in the evening. Specifically, we anticipated patients arriving to the ICU after hours would be predisposed to reductions in hospital resources such as in-house staff, logistical issues with procedural services, and staff turnover with potential communication discrepancies during handoffs. A systematic review and meta-analysis by Cavallazzi et al. did not suggest worse outcomes in late ICU admissions for all patients, both surgical and non-surgical, but did suggest that weekend ICU admissions were associated with worse in-hospital mortality¹⁴. However, among studies reviewed in this meta-analysis evaluating late ICU admissions, there was heterogeneity suggesting that organizational structures play a role in these outcomes.

Despite a lack of difference in majority of complications following extensive risk-adjustment, prolonged ventilation remained significantly associated with late ICU arrival. This suggests that despite baseline operative risk or intraoperative difficulties, patient admission to the ICU after 18:00 was independently associated with a higher risk of prolonged ventilation. We postulated that this may be due to hesitancy to extubate patients at night due to either lack of appropriate staff coverage at late hours or standardized protocols set in place. This increased rate of prolonged ventilation did not, however, translate to an increase in risk of FTR, pneumonia, or reintubation.

Prolonged ventilation is defined as >24 hours on the ventilator after surgery and is an important quality metric after cardiac surgery¹⁶. Even if these patients were not extubated overnight after arrival, if they were extubated the following day by 18:00, they would not qualify as developing this complication. This may be the result of a few things. Providers taking care of these patients the following day may have believed that the patients remained intubated for an appropriate reason and therefore did not move expeditiously to extubate them during the daytime. Patients left on the ventilator overnight may have remained on significant amounts of sedation leading to difficulties with moving toward extubation the following day from altered mental status and delayed sedation washout. Lastly, patients mechanically ventilated overnight may already show signs of diaphragmatic dysfunction and have difficulties with passing spontaneous breathing trials leading to a longer time before planned extubation¹⁷. Several studies have suggested that overnight extubation in cardiac surgery patients does not increase the risk of morbidity or mortality but does decrease length of stay and ventilator hours^18–20. Therefore, attempts should be made in appropriate patients to extubate in a reasonable timeframe no matter their ICU arrival time to prevent prolonged ventilation.

There are limitations to the present study. First, this is a retrospective study and is subject to unmeasured confounding. We were not able to determine ICU staffing models for each center and what respective effect that may have on certain complications. Hospitals may also have different nurse, respiratory therapist, and physician staffing ratios at night. However, the inclusion of 17 centers for analysis across the state of Virginia may have helped to select for a diverse group of ICU staffing models. Additionally, we included presence of an academic center in our model as this may reflect a certain type of ICU practice distinct from the private setting. Case volume based on quartiles was utilized to control for potentially higher resourced or more experienced centers. However, volume is not necessarily static over time and may vary based on hospital resources. Despite the relative consistency of case numbers of each center over the study period, there were still some year-to-year variations that were not accounted for statistically that may have affected our results. Intraoperative risk for patients arriving to the ICU late were different despite the patients having similar preoperative risk suggesting that their intraoperative course may have contributed to their worse, unadjusted outcomes. We attempted to control for these factors in our multivariable model but may not have completely accounted for all confounding factors including individual surgeons and anesthesiologists who performed the cases. Lastly, these results may not be generalizable to other hospital systems although the Commonwealth of Virginia possesses a socioeconomically diverse patient population with a multitude of different types of hospitals including both academic and private, and in rural, urban, and suburban settings.

Late arrival to the ICU was not associated with failure to rescue or operative mortality but was associated with prolonged ventilation following elective, isolated CABG. These findings may be reflective of communication handoff discrepancies or a hesitation to extubate patients overnight.

GLOSSARY OF ABBREVIATIONS

ACSD

Adult Cardiac Surgery Database

ECMO

Extracorporeal membrane oxygenation

FTR

Failure to rescue

IABP

Intra-aortic balloon pump

ICU

Intensive Care Unit

PROM

Predicted risk of mortality

PROMM

Predicted risk of morbidity or mortality

STS

Society of Thoracic Surgeons