Late Operating Room Start Times Impact Mortality and Cost for Nonemergent Cardiac Surgery

Abstract

Background

There is growing concern over the effect of starting non-emergent cardiac surgery later in the day on clinical outcomes and resource utilization. Our objective was to determine the differences in patient outcomes for starting non-emergent cardiac surgery after 3pm.

Methods

All non-emergent cardiac operations performed at a single institution from July 2008 to 2013 were reviewed. Cases were stratified based on “early start” or “late start,” defined by incision time before or after 3pm. Rates of observed and risk-adjusted mortality, major complications, and costs were compared on a univariate basis for all patients and via multivariable linear and logistic regression for patients with a valid Society of Thoracic Surgeons (STS) Predicted Risk of Mortality (PROM).

Results

A total of 3,395 non-emergent cardiac operations were reviewed, including 368 late start cases. Compared to cases starting earlier, mortality was significantly higher for patients undergoing late operations (5.2% vs. 3.5%, p=0.046) despite similar preoperative risk (STS PROM 3.8% vs. 3.3%) and major complication rates (18.2% vs. 18.3%). Costs were 8% higher with late start cases ($51,576 vs. $47,641, p<0.001). After controlling for case type, surgeon, year, and risk, late cases resulted in higher mortality (odds ratio 2.04, p=0.041) despite shorter operative duration (16 min, p<0.001).

Conclusions

Starting non-emergent cardiac cases later in the day is associated with 2× higher absolute and risk-adjusted mortality. These data should be carefully considered—not only by surgeons and patients but also in the context of the operating room system—when scheduling non-emergent cardiac cases.

Introduction

Since the Institute of Medicine published a report suggesting that medical errors result in 100,000 deaths annually with an overall cost of $29 billion, an increased emphasis in discussions regarding health care reform has been placed on systems-based approaches to improve patient safety [1]. In response, the public and policymakers have focused their attention on fatigue and long work hours, resulting in significant cultural changes and increased work hour restrictions [2,3,4].

Beyond simply fatigue, there is also increased awareness that complex operations, like those in cardiac surgery, require the availability of a highly reliable team consisting of specialized surgeons, anesthesiologists, nurses, and perfusionists [5]. The entire team's familiarity with and constant repetition of complex steps—e.g., placing patients on cardiopulmonary bypass (CPB), performing key intraoperative studies, and executing specific operative steps—is critical to ensuring consistent and efficient performance. Consequently, some have advised delaying non-emergent operations until usual work hours given the potential decreased availability of requisite resources and experienced staff at night [6].

In the context of such controversy and scrutiny, there has been substantial impetus to study the associations among after-hours care and patient outcomes in multiple settings; however, results have been conflicting. Overnight care (7 p.m. to 7 a.m.) has been associated with worse outcomes in internal medicine [7], intensive care [8], obstetrics [9], interventional cardiology [10], laparoscopy [11,12], colorectal surgery [13], orthopedic surgery [14,15], and kidney transplantation [16]; but non-inferior outcomes when liver [17] or cardiothoracic transplantation [18] is performed at night. However, no studies have incorporated potentially more subtle—or even pernicious—timing differences, such as the effect of starting elective cases in the late afternoon or early evening at the end of a busy operative day. Our objective was to determine the differences in patient outcomes for starting non-emergent cardiac surgery at our institution after 3 p.m.

Patients and Methods

Patients

This study was approved by the University of Virginia Institutional Review Board, including a waiver for the need to obtain patient consent. All adult patients undergoing adult cardiac operations were retrospectively reviewed from July 1, 2008, to June 30, 2013, using our institutional operative log and our institutional Society of Thoracic Surgeons (STS) Adult Cardiac Surgery database. Only patients undergoing non-emergent cardiac surgery were included; thus, their STS-defined status was either elective (i.e., the patient's symptoms had been stable for days or weeks) or urgent (i.e., the patient's operation was required during the same hospitalization—but not necessarily immediately) [19].

Patients were stratified into 2 primary study cohorts based on whether the time of skin incision occurred after 3 p.m. Patient demographics, preoperative risk factors, operative surgeon, operative features, and postoperative outcomes were compared between the 2 study groups. Preoperative risk was assessed by the prevalence of comorbid disease as described by standard STS variable definitions and established calculated STS predictive indices, including Predicted Risk of Mortality (PROM) and Predicted Risk of Mortality or Morbidity (PROMM).

Outcomes

Primary outcomes of interest included differences in operative time, morbidity, and mortality between the 2 study groups. Operative time was defined as the time between skin incision and skin closure. A composite outcome of major complications was used as a proxy for morbidity. Standard STS definitions for postoperative events and complications were used, including, for example, prolonged mechanical ventilation (> 24 hours) and renal failure (increase in serum creatinine level > 2.0 or a doubling [2×] of the most recent preoperative creatinine level). Operative mortality was defined as all patient deaths occurring during hospitalization as well as those within 30 days of surgery regardless of discharge status. Observed-to-expected ratios were calculated only for operations for which there was a valid STS PROMM and STS PROM.

Costs Data

Cost data were abstracted from the University of Virginia Health System Clinical Data Repository. The methodology used in generating this database has been described elsewhere [20], but briefly, this database uses micro-costing algorithms to capture cost data in an actual utilization framework. Consequently, financial transactions are calculated not as third-party charges but as estimated costs based on such algorithms. Thus, costs in the present study are defined as the cost of care as estimated by the institution rather than charges relayed to patients and insurers or actual hospital reimbursement.

Statistical Analysis

All study outcomes and data comparisons were established a priori. Statistical analyses were designed to test the null hypothesis that no association exists between a late operation and patient outcomes. All categorical variables are expressed as a percentage of the group of origin; continuous variables are expressed as mean ± standard deviation. Univariate comparisons were performed using the Pearson χ² test for dichotomous variables and Student's t-test for continuous variables.

To verify the univariate comparisons between the study cohorts and control for confounding variables (e.g., effect of surgeon, operation type, etc.), multivariable regression was used to estimate the relationship between a late operation and mortality (via logistic regression) and operative time (via linear regression). STS PROM (or PROMM, as appropriate) was used as a validated measure to control for patient- and operation-related risk factors. Consequently, for the purposes of regression analysis only, patients without a cardiac STS PROM score were excluded from analysis. Additional variables for control included surgeon; operation type; academic year of operation; and in the case of regressions related to operative time, redo sternotomy. Variable coefficients and odds ratios are reported as the mean impact of each variable with a 95% confidence interval (CI).

All reported p values are 2-tailed, and a threshold of p ≤ 0.05 was used to test for significance. Data analysis was performed using Microsoft Excel (Redmond, WA) and open-source R statistical software (http://www.R-project.org).

Results

Univariate Comparisons

A total of 3,395 non-emergent cardiac operations, including 368 cases after 3 p.m., met criteria for inclusion in our analyses between 2008-2013. Few baseline differences existed between the two cohorts (Table 1). Patients undergoing late operations required more urgent operations (51.6% vs. 38.8%, p<0.001). However, STS PROMM [21.3% (CI 20.7%-21.9%) vs. 19.7% (CI 18.1%-21.6%), p=0.153) and STS PROM [3.8% (CI 3.6%-4.0%) vs. 3.3% (CI 2.7%-3.9%), p=0.346) were not significantly higher.

Table 1. Comparisons of patient characteristics between operating before versus after 3 p.m.

Variable	All Cases		P
	< 3pm	> 3pm
n	3027	368
Age (years)	65.0	64.0	0.110
BMI (kg/m2)	29.6	29.3	0.184
Creatinine (mg/mL)	1.3	1.3	0.811
Ejection Fraction (%)	51.1	47.8	0.814
Female (%)	33.0%	33.7%	0.780
Smoker (%)	18.9%	21.2%	0.290
Diabetes (%)	34.8%	33.4%	0.613
Hypertension (%)	78.8%	73.9%	0.031
Peripheral Art Disease (%)	16.2%	15.8%	0.821
Dialysis-Dependent (%)	3.3%	4.3%	0.315
Chronic Lung Disease (%)	21.6%	23.9%	0.319
Mild	10.3%	11.7%	0.427
Moderate	7.7%	9.5%	0.214
Severe	3.6%	2.7%	0.369
Prior PCI (%)	18.6%	17.7%	0.673
Prior Sternotomy (%)	38.1%	40.5%	0.372
Congestive Heart Failure (%)	39.3%	46.7%	0.006
NYHA II	6.4%	7.6%	0.379
NYHA III	17.8%	19.3%	0.493
NYHA IV	14.9%	19.8%	0.014
Status
Elective (%)	61.2%	48.4%	<0.001
Urgent (%)	38.8%	51.6%	<0.001
Surgeon
1	34.9%	17.9%	0.000
2	33.6%	42.4%	0.001
3	30.2%	37.2%	0.006
4	1.3%	2.4%	0.076
Consecutive Hours Surgeon Operated	1.58	5.67	<0.001
STS Risk Available? (%)	68.5%	66.0%	0.327
STS PROMM (%)	19.7%	21.3%	0.153
STS PROM (%)	3.3%	3.8%	0.346

Operative features and clinical outcomes between the two cohorts are presented in Table 2. Interestingly, operative times were shorter in the late operation cohort (231 min vs. 244 min, p=0.043). Similar trends were mirrored with shorter CPB time (104 min vs. 110 min, p=0.013) and cross-clamp time (71 min vs. 77 min, p=0.030). Markers of resource utilization, including length of intensive care and overall postoperative length of stay, were similar. Despite these minimal differences in surrogate markers of resource utilization, however, total costs of hospitalization were 8% higher in the late case cohort ($51,576 vs. $47,641, p<0.001) (Table 3).

Table 2. Comparisons of clinical outcomes between operating before versus after 3 p.m.

Variable	All Cases		P
	< 3pm	> 3pm
n	3027	368
Operative Duration (min)	244	231	0.043
Cardiopulmonary Bypass Time (min)	110	104	0.013
Cross-clamp time (min)	77	71	0.030
Intraoperative Blood Transfusion (%)	42.9%	44.6%	0.554
Postoperative Length of Stay (days)	8.3	9.5	0.953
Intensive Care Length of Stay (hours)	90	100	0.153
Intensive Care Readmission (%)	4.8%	3.5%	0.269
Atrial Fibrillation (%)	19.5%	19.6%	0.961
Major Complication (%)	18.3%	18.2%	0.952
Deep Sternal Wound Infection (%)	0.1%	0.4%	0.048
Prolonged Ventilation (%)	12.1%	10.6%	0.404
Pneumonia (%)	3.8%	3.0%	0.438
Sepsis (%)	0.6%	0.5%	0.904
Renal Failure (%)	6.5%	7.3%	0.545
Dialysis (%)	3.6%	3.3%	0.740
Reoperation (%)	3.4%	5.2%	0.094
Stroke (%)	2.7%	1.1%	0.062
Myocardial Infarction (%)	0.2%	0.3%	0.646
Cardiac Arrest (%)	2.3%	2.2%	0.867
Mortality (%)	3.5%	5.2%	0.046
Observed:Expected Ratio
Morbidity	0.71	0.64	-
Mortality	0.68	1.20	-

Table 3. Comparison of the total costs of hospitalization between operating before versus after 3 p.m.

Variable	All Cases		P
	< 3pm	> 3pm
n	3027	368
Total Costs ($)	$ 47,641	$ 51,576	0.009
Anesthesia ($)	$ 579	$ 587	0.583
Blood products ($)	$ 1,532	$ 1,798	0.066
Diagnostic ($)	$ 1,549	$ 1,988	0.010
Hemodialysis ($)	$ 221	$ 229	0.910
Intensive Care ($)	$ 12,996	$ 15,184	0.011
Operating Room ($)	$ 3,678	$ 3,724	0.632
Medications ($)	$ 1,824	$ 1,898	0.646
Imaging ($)	$ 954	$ 1,120	0.017
Rehabilitation ($)	$ 1,451	$ 1,368	0.724

Overall morbidity tended to be remarkably similar between the two cohorts (18.2% vs. 18.3%), and there were no discernable trends when morbidity was analyzed by specific complication. When analyzed by case type, more complex operations (e.g., bypass plus valve surgery) tended to demonstrate higher rates of morbidity in the late case cohort. However, when analyzed via the observed-to-expected (O:E) ratio, risk-adjusted morbidity across all operation types was relatively similar (0.64 vs. 0.71).

However, the most dramatic differences were observed in the rates of mortality. Despite shorter operative times and similar major complication rates, mortality was significantly higher (5.2% vs. 3.5%, p=0.046) in the late operation cohort. Although patients undergoing late operations had slightly higher STS PROM scores, the O:E ratio for mortality remained almost double (1.20 vs. 0.68). The trend was consistent among all operation types except isolated aortic valve replacement and concomitant coronary artery bypass grafting with aortic valve replacement.

Multivariable Comparisons

Multivariable linear regression (Table 4) confirmed the shorter operative times observed in the late operation cohort. After controlling for year, risk, case type, surgeon, and redo sternotomy, operative time was 16.9 min shorter (CI 10.0-23.9 min, p<0.001); CPB time was 7.9 min shorter (CI 3.9-11.9 min, p<0.001); and cross-clamp time was 5.2 min shorter (CI 2.4-8.0 min, p<0.001). Thus, the time savings appear reasonably equally shared between CPB time and non-CPB time.

Table 4. Multivariable regressions.

The coefficient on each of the multivariable linear regression outcomes represents the mean impact (in minutes) of operating after 3 p.m. after controlling for surgeon, operation type, STS PROM, redo sternotomy, and year of operation. The coefficient on mortality in the multivariable logistic regression represents the odds ratio on mortality for starting a case after 3 p.m. after controlling for the same variables, except redo sternotomy.

Variable	Effect of Operating after 3pm (min)
Multivariable Linear Regressions on…	Coefficient	Lower 95%	Upper 95%	P
Operative Time (min)	-16.9	-23.9	-10.0	<0.001
Cardiopulmonary Bypass Time (min)	-7.9	-11.9	-3.9	<0.001
Cross-clamp Time (min)	-5.2	-8.0	-2.4	<0.001
Multivariable Logistic Regression on…	Odds Ratio	Lower 95%	Upper 95%	P
Mortality	2.04	1.03	4.17	0.041

Multivariable logistic regression (Table 4) confirmed the higher mortality observed in the late operation cohort. After controlling for year, risk, case type, and surgeon, patients undergoing later operations experienced statistically significant higher mortality with an odds ratio of 2.04 (CI 1.03-4.17, p=0.041).

Interestingly, using a variable indicating the number of consecutive hours the attending had operated (in place of operating after 3 p.m.) failed to meet the criteria for statistical significance. Similarly, using a variable for the numerical order of the case failed to show a relationship with increased mortality. Consequently, these effects appear to be more related to the timing of the operation than other potential confounding factors.

Comment

In summary, we found increased absolute and risk-adjusted mortality despite no difference in morbidity and shorter operative times for late start non-emergent cardiac operations. Furthermore, we found that late operations were associated with increased costs of care. To our knowledge, this analysis is unique in two aspects: (1) examining the association between operative time of day and outcomes in non-emergent cardiac surgery and (2) lowering the cut-off time to operations starting after 3 p.m.

Effect on Mortality

The increase in mortality is consistent with an emerging consensus in the literature in other specialties. Prior investigations have suggested that late operations may be affected by shift changes, changes in nighttime staffing patterns, fatigue, technical lapses, and deficits in nighttime postoperative care [18]. Consequently, our finding of increased mortality is not entirely surprising, given that cardiac surgery is especially vulnerable to such problems: the operations are complex, require extensive skill by the entire operative team and staff, and are resource intensive; patients are higher risk; and complications are high consequence. Furthermore, our finding of worse outcomes when lowering the threshold to operating after 3 p.m. remains in line with similar studies in non-surgical specialties. Gastroenterologists are 4.6% less likely to detect a colon polyp with each hour that passes on a given day of performing colonoscopies [21]. Similarly, anesthesia-related problems have been shown to be as low as 1% during surgeries that started at 9 a.m. but as high as 4.2% for those starting at 4 p.m. [22].

Effect on Major Complications and Resource Utilization

Although mortality is arguably the most important clinical outcome, a detailed examination of secondary outcomes remains critical. Deep sternal wound infections, the prevention of which requires meticulous attention by all staff to proper chest closure at the end of the case, were more likely to occur with later operations. Similarly, there is a trend towards increased need for reoperation, which could be due to a tendency to overlook surgical bleeding at the end of the case. While it is difficult to attribute these complications to specific relative deficits in either intraoperative or postoperative care, their existence implies a breakdown within what needs to be a highly reliable team for late cases.

As a broader marker of resource utilization and adverse outcomes, intensive care length of stay tended to be higher despite not meeting strict statistical significance. Costs, however, were significantly higher in the late operation cohort, with specific increases in intensive care, diagnostic testing, and imaging. Such findings are increasingly important as surgeons face increased oversight in health care delivery.

Similar composite morbidity—despite higher mortality—could be the result of a type I error or type II error. Alternatively, mortality may occur before the development of many postoperative complications, or our composite of major complications may not adequately capture more subversive adverse events that contribute to mortality. Such considerations also may explain why intensive care costs were higher in the late cohort despite similar length of stay.

The effect on operative times is perplexing. Despite almost uniformly shorter operative, CPB, and cross-clamp times—even when controlling for the effect of surgeon—both absolute and risk-adjusted mortality were higher for late operations. Given that minimizing CPB time is traditionally thought to result in superior outcomes [23], shorter times would seem to imply that increased mortality is less likely the result of deficits in operative care than in postoperative care, but there are several other potential—albeit mostly speculative—explanations. At the end of a busy operative day, attending surgeons may be more motivated to help facilitate both critical and non-critical components of operations, as reflected by an 8 minute shorter CPB and non-CPB time. There is evidence suggesting that physicians are more motivated to accomplish tasks at the end of a day or week to avoid after-hours care; for example, unplanned Caesarian-section is most likely on Fridays between 3-9 p.m. [24]. Alternatively, fatigue may allow technical lapses to escape attention or motivate shortcuts, as suggested by increased deep sternal wound infections and re-exploration. Such potentially deleterious effects, while obviously concerning, remain difficult to measure and define—or even contemplate.

Limitations

Our retrospective study design is subject to the limitations of selection bias, and the reported results are limited to describing observed associations. Also, while it is difficult to capture the influence of surgeon experience or ability in performing late cases, neither our sub-analyses for each surgeon or our multivariable regressions found substantial differences. Additionally, our results are limited to short-term operative outcomes and thus do not comment on longer-term outcomes, the incidence of breaches in safety protocols, or “close calls.”

Our results are also influenced by a cut-off time of incision after 3 p.m. Although we acknowledge that other rational time cut-offs may exist based on technical staff, nursing, or resident shift change (various members of our staff turnover at 3 p.m., 5 p.m., and 7 p.m.), we selected 3 p.m. to adequately power the study and to attempt to discern potentially more subtle effects, such as starting non-emergent cases later in the day such that (1) operating room personnel turnover would likely occur intraoperatively and (2) the departure of the day-time intensivist and nursing team would likely occur after patient arrival to intensive care.

Implications

Undoubtedly, these results are likely to be controversial among surgeons who feel that the profession should be conditioned for late hours. Additionally, some academic leaders have expressed concern that delaying prompt treatment of seemingly elective problems may increase disease acuity [25]. It may also be possible that patients undergoing late operations have a higher acuity and urgency that is not fully captured by their STS risk score or status.

While we are not indifferent to such considerations, our data ultimately suggest that worse survival with late operations is a robust finding that persists even after controlling for confounding factors. Most importantly, they highlight significant areas for potential improvement within our own institution—which we suspect are mirrored almost universally—rather than mandate reactionary policy change.

Despite the inherent complexity of cardiac surgery, it is simply not a priority at every center to start non-emergent cardiac surgery in a timely manner. We have used our findings as motivation to improve on-time first case starts, hasten turnover, and facilitate parallel processing to prevent case delays. In fact, in reviewing the same dataset, we found that average first-start skin incision time was 8:27 a.m. (rooms open between 7-7:30 a.m.) and only 44% of first-start cases had an incision time by 8:15 a.m.; both findings demand improvement. Additionally, we have taken steps to increase the availability of attending surgeon support throughout the day, maintain the availability of a back-up surgeon for all cases, and mandate the availability of adequate on-call support staff coverage. Postoperatively, we have codified our care protocols (specifically fluid and hemodynamic management and notification triggers), increased closed-loop communication, extended night-time in-house attending intensivist staffing until 12 a.m., provided a back-up 24-hour in-house rescue intensivist, decreased resident cross-coverage, and hired additional night-time critical care nursing staff.

The timing of non-emergent operations should also be carefully considered in the context of the experience of the surgeon, the availability and opportunity cost of necessary staff and resources, and the needs of the patient. Simply an increased awareness of our findings can help foster a culture in which delaying non-emergent cases is seen as a more efficient use of specialized resources that should be focused on life-threatening emergencies when they are relatively scarce.

Conclusion

Starting non-emergent cardiac cases later in the day is associated with 2× higher absolute and risk-adjusted mortality. Operative factors alone do not appear to explain these differences. Consequently, systems must also focus to ensure on-time starts and strengthen care delivery for late arrivals to intensive care. These data should be carefully considered—not only by surgeons and patients but also in the context of the operating room system—when scheduling non-emergent cardiac cases.