Association between In-hospital Mortality and Low Cardiac Output Syndrome with Morning versus Afternoon Cardiac Surgery: A Retrospective Cohort Study

This retrospective, observational, single-center cohort study was approved by the Cleveland Clinic Institutional Review Board with waived informed consent. We obtained data from the Cleveland Clinic Perioperative Health Documentation System, the Automated Record Keeper System (electronic intraoperative anesthesia record), and our institutional Society of Thoracic Surgeons database. Patient data were linked across databases through medical record numbers and dates of surgery. Variable definitions and data sources are reported in Supplemental Digital Content, table 1 (https://links.lww.com/ALN/C559). This report follows the Reporting of studies Conducted using Observational Routinely collected health Data (RECORD) and STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) guidelines.

We initially limited our investigation to patients having aortic valve replacement; however, postoperative mortality and complications after aortic valve surgery are rare at our institution. We therefore broadened our inclusion criteria and considered all adults who had elective aortic valve replacement/repair, mitral valve replacement/repair, and/or coronary artery bypass grafting with cardiopulmonary bypass between 2010 and 2018 at the Cleveland Clinic main campus with aortic cross-clamp start times between 8 am and 5 pm.

We excluded patients who had end-stage renal disease requiring renal replacement therapy, elevated baseline troponin T, myocardial infarction within 7 days preoperatively, aortic cross-clamp start time before 8 am or after 5 pm, emergent surgery, heart transplantation, lung transplantation, preoperative cardiogenic shock, and patients who were scheduled for insertion of right or left ventricular assist device.

Post hoc, we realized that surgical approach was not recorded for patients who had surgery in 2010 and was only sometimes recorded for patients who had surgery from January to June 2011, most likely because of a change in data collection protocol. Because surgical approach was an important confounder, associated with both exposure and outcome, we restricted our analysis to patients who had surgery from January 1, 2011, to December 31, 2018. Patients who were missing surgical approach from the first half of year 2011 were excluded under a complete case analysis, assuming missing at random for all missing variables.

Patients were divided into morning and afternoon groups based on their aortic cross-clamp start times. Patients with aortic cross-clamp start times between 8 and 11 am were defined as having morning surgery, and those with aortic cross-clamp start times between 2 and 5 pm were defined as having afternoon surgery. Patients with aortic cross-clamp application between 11 am and 2 pm were not included in the primary analysis to provide clear separation between the groups. All patients were followed until hospital discharge.

Standard American Society of Anesthesiologists (Schaumburg, Illinois) monitors, brachial arterial catheters, central venous or pulmonary artery catheters, and transesophageal echocardiography were used for all patients. Anesthesia was induced and maintained with midazolam, fentanyl, propofol, etomidate, ketamine, methadone, depolarizing and/or nondepolarizing muscle relaxants, and volatile anesthetics. Standard procedures were used for initiation and separation from cardiopulmonary bypass. Intermittent retrograde and antegrade Buckberg or del Nido cardioplegia were given for cardioplegic arrest. Low cardiac index during separation from cardiopulmonary bypass was treated with epinephrine, milrinone, dobutamine, or dopamine infusion to achieve a cardiac index of more than 2.0 l ⋅ min⁻¹ ⋅ m⁻². The indications for use of two or more inotropic agents include continued hemodynamic instability caused by myocardial failure/insufficiency assessed by pulmonary artery catheters with a cardiac index of less than 2.0 l ⋅ min⁻¹ ⋅ m⁻² or left and/or right ventricular dysfunction assessed by echocardiography after optimizing preload and afterload and correction of all electrolyte and blood gas abnormalities. Hypotension with adequate intravascular volume status was treated with norepinephrine or vasopressin infusion for a target mean arterial pressure of more than 70 mmHg.

The primary endpoint was a composite outcome of in-hospital mortality and/or low cardiac output syndrome. Low cardiac output syndrome was defined as one or both of the following criteria:

• (1) The need for two or more inotropic agents (epinephrine, milrinone, dobutamine, or high-dose dopamine [more than 5 μg · kg⁻¹ · min⁻¹]) at 24 to 48 h postoperatively with one or more of the following: (i) a documented low cardiac output (less than 2.0 l · min⁻¹ · m⁻²) postcardiopulmonary bypass and/or immediately after surgery in the intensive care unit; (ii) left and/or right ventricular dysfunction on intraoperative postcardiopulmonary bypass transesophageal echocardiography; and (iii) chart documentation of postoperative cardiac insufficiency/cardiogenic shock. The need for vasopressor support (norepinephrine, vasopressin, phenylephrine) without the use of two or more inotropic agents (epinephrine, milrinone, dobutamine, or high-dose dopamine [more than 5 μg · kg⁻¹ · min⁻¹]) does not fulfill the criteria for low cardiac output syndrome.

• (2) Use of mechanical circulatory support with extracorporeal membrane oxygenation, left ventricular assist device, or intraaortic balloon pump within the first five postoperative days to treat patients with refractory cardiac failure despite the use of multiple inotropic support. Patients who required extracorporeal membrane oxygenation for pulmonary insufficiency were excluded.

This clinically relevant outcome of low cardiac output syndrome was selected because it was associated with a 12-fold increase in risk for postoperative mortality after cardiac surgery in a large cross-sectional observational analysis that included more than 59,000 patients.²²  Importantly, those with low cardiac output syndrome comprised greater than 60% of postoperative deaths.

The Institutional Review Board–approved protocol included a detailed statistical analysis plan describing all variables required for the analysis, including the primary and secondary outcomes, exposure variables, and confounder variables, which were defined a priori before accession of the data. We initially planned to limit our analysis to patients having aortic valve replacement surgery, as in a previous analysis that demonstrated worse myocardial ischemic tolerance with morning surgery.¹⁸  However, the event rate is low in our aortic valve replacement surgery population, with an overall mortality of only ≈0.24% and low cardiac output syndrome of ≈1.2%. We therefore modified our statistical plan (before the data were accessed) by including more complex and combination procedures to increase the power and generalizability of our analysis.

Our a priori developed statistical plan included a primary composite outcome of mortality and/or low cardiac output syndrome and a secondary outcome of postoperative troponin concentrations. Patients with low cardiac output syndrome were identified by the need for postoperative hemodynamic support with two or more inotropic agents and/or need for mechanical circulatory support. Database and/or chart review were performed to confirm that patients who required two or more inotropic agents also received two or more inotropic agents at the end of surgery (to confirm that myocardial insufficiency began intraoperatively rather than following a postoperative complication) and also exhibited one or more characteristics consistent with low cardiac output syndrome, including a documented low cardiac index, left and/or right ventricular dysfunction, and/or postoperative cardiac insufficiency/cardiogenic shock as described above.

For troponin T concentrations, blood for troponin assessment in our institution is routinely sampled just once at approximately 2 am of the first postoperative morning. Post hoc, we realized that there were thus substantial systematic differences in the time elapsed between surgery and the troponin assay in the morning and afternoon groups, making the results uninterpretable. We therefore modified our a priori developed statistical plan and did not compare the prespecified secondary outcome of troponin concentrations.

Potential confounding factors including demographic characteristics, baseline comorbidities, preoperative medications, and surgery type were adjusted with stabilized propensity score weighting, in which the propensity score was calculated as the probability of cross-clamp occurring in the morning. The weight was then calculated for morning patients as the proportion of morning patients divided by the propensity score and for afternoon patients as the proportion of afternoon patients divided by (1 – propensity score). Balance in confounding factors between morning and afternoon was assessed through absolute standardized difference, and any confounding factors with an absolute standardized difference of more than 0.10 were treated as imbalanced and adjusted for in the main analysis model.

The association between aortic cross-clamp start time (morning vs. afternoon) and the incidence of the collapsed outcome were assessed through a logistic regression model. The interaction between aortic cross-clamp start time and surgery type was assessed at significance level of 0.15, and subgroup analyses were performed for each type of surgery whether or not interactions were significant. All tests were performed two-sided.

In an a priori planned sensitivity analysis, we treated aortic cross-clamp start time as a continuous exposure including data from all patients who had application of aortic cross-clamp between 8 am and 5 pm and assessed the association between cross-clamp start time and the incidence of the composite outcome through a logistic regression model, with confounding factors adjusted for directly in the model. Linearity of the association was visually checked by incorporating a restricted cubic spline term with smoothing parameter of 0.10 into the model. Upon finding a nonlinear association, we instead estimated the difference between each hour.

After the statistically nonsignificant association between cross-clamp hour and the outcome, we further assessed the association between surgery start hour and the composite outcome. Logistic regression with direct adjustment for confounding variables was used to assess the association, with surgery start hour analyzed as a categorical variable.

To adjust for a potential unmeasured surgical factor, we performed an analysis that included a variable for the attending surgeon in the propensity score. The surgeon was excluded if she/he usually operated at a specific time of day and performed either four or fewer surgeries in the morning, or four or fewer surgeries in the afternoon during the entire study time frame.

We also assessed the difference between morning versus afternoon patients on other major postoperative complications as defined by the Society of Thoracic Surgeons (Chicago, Illinois) including stroke, renal failure, prolonged ventilation, deep sternal wound infections, and cardiac reoperation.²³  All the outcomes were assessed through logistic regression, with confounders adjusted through stabilized inverse probability of treatment weighting, as described above. Because the analysis was exploratory, Bonferroni correction was not applied.

Only our statistical team had access to raw data. SAS 9.4 statistical software (USA) and R statistical software version 3.5.1 for 64-bit Windows operating system (R Foundation for Statistical Computing, Austria) were used for all data processing and analyses.

A low incidence of mortality (0.24%) and low cardiac output syndrome (1.18%) after aortic valve replacement at our center would require a sample size of 386,000 patients for 90% power to detect a difference in our composite outcome, which was not feasible. When mitral valve and coronary artery bypass grafting procedures were included, 9,734 patients and an incidence of 3% in the composite outcome provided 90% power to detect an odds ratio of 1.50 or higher comparing morning patients to afternoon patients.

We identified 15,349 cardiac surgeries from 15,224 patients that met inclusion and exclusion criteria from January 1, 2011, through December 31, 2018. The last index cardiac surgery for each patient was considered in the analysis. For this cohort, we assessed the missing pattern for the baseline variables, including surgical approach, and proceeded with a complete case analysis involving 13,344 patients (fig. 1).

For the main analysis, we included 9,734 patients who had aortic cross-clamping started between 8 and 11 am or 2 and 5 pm. In 6,859 (70.5%) patients, aortic cross-clamping occurred in the morning (8 to 11 am), and 2,875 (29.5%) had aortic cross-clamping in the afternoon (2 to 5 pm). Morning patients had less severe left ventricular dysfunction, fewer previous myocardial infarctions, and less preoperative aspirin use. The two groups were also imbalanced in year of surgery, surgical approach, and surgical procedure. All baseline confounding variables were balanced after propensity score weighting (table 1).

The raw incidence of mortality was 0.5% (49 of 9,734) overall, 0.4% (29 of 6,859) for morning surgery, and 0.7% (20 of 2,875) for afternoon surgery (morning vs. afternoon unadjusted odds ratio, 0.61; 95% CI, 0.34 to 1.07; P = 0.086) The raw incidence of low cardiac output syndrome was 2.7% (263 of 9,734) overall, 2.6% (175 of 6,859) for morning surgery, and 3.1% (88 of 2,875) for afternoon surgery (morning vs. afternoon unadjusted odds ratio, 0.83; 95% CI, 0.64 to 1.08; P = 0.158). The perioperative profiles of patients with low cardiac output are summarized in Supplemental Digital Content, table 2 (https://links.lww.com/ALN/C559). The incidence of the collapsed composite outcome of in-hospital mortality and/or low cardiac output syndrome was 3.0% (292 of 9,734) overall, 2.8% (195 of 6,859) for morning surgery, and 3.4% (97 of 2,875) for afternoon surgery (morning vs. afternoon unadjusted odds ratio, 0.84; 95% CI, 0.65 to 1.07; P = 0.162). The incidence of mortality and low cardiac output syndrome by surgery type is listed in table 2. The median (quartile 1, quartile 3) length of hospital stay was 7 (6, 10) days for morning patients and 8 (6, 10) days for afternoon patients.

Intraoperative use of inotropic and vasopressor medications, including epinephrine, norepinephrine, milrinone, dopamine, dobutamine, and vasopressin, was similar between groups. Intraoperative medications, blood product transfusion, and other descriptive variables are summarized in table 3.

The incidence of the composite outcome did not differ significantly when comparing morning and afternoon surgery, with an adjusted odds ratio estimated at 0.96 (95% CI, 0.75 to 1.24; P = 0.770). The interaction between morning versus afternoon and surgery type was also statistically nonsignificant (P = 0.965), and the time effect was nonsignificant in all of the surgery type subgroups (table 2).

We included 13,344 patients whose surgeries had aortic cross-clamp started between 8 am and 5 pm. The incidence of the composite outcome was similar across aortic cross-clamp start hours (overall P = 0.989; fig. 2), suggesting no significant variation by time. We also assessed the incidence of the composite outcome against surgery start hour. Because our usual surgery start time is 7 am on weekdays, a large number of surgery start times were clustered between 6:45 and 7:15. We thus partitioned the surgery start time at the half hour after every full hour (e.g., 7:30 am, 8:30 am, etc.). Overall, there was not a significant association (overall P = 0.291; Supplemental Digital Content, table 3, https://links.lww.com/ALN/C559).

We included and balanced for individual surgeons after propensity score matching. Consistent with the primary results, no difference between morning and afternoon surgeries was found, with estimated odds ratio of 1.00 (95% CI, 0.78 to 1.29; P = 0.987) comparing morning to afternoon.

We also compared the two groups on major postoperative morbidity as defined by the Society of Thoracic Surgeons. Afternoon patients had more prolonged ventilation (P = 0.036), but no difference was found for other postoperative complications, including stroke, renal failure, deep sternal wound infection, and cardiac reoperation. These results are summarized in Supplemental Digital Content, table 4 (https://links.lww.com/ALN/C559).

The composite outcome of mortality and low cardiac output syndrome did not differ in patients who had morning versus afternoon aortic valve, mitral valve, and/or coronary artery bypass grafting surgery. The results were consistent regardless of procedural type. The results were also consistent regardless of whether surgical timing was considered dichotomously or as a continuous variable. Comparing surgery start time rather than aortic cross-clamp start time also provided consistent results.

The absence of time-of-day variation in postoperative cardiac complications in our patients contrasts with the results of Montaigne et al.,¹⁷  who reported more major adverse cardiac events and higher postoperative troponin T concentrations in patients having aortic valve replacement (with or without coronary artery bypass grafting) in the morning. Although we evaluated various types of cardiac surgery, subgroup analyses identified no substantive morning versus afternoon outcome differences across the entire cohort (n = 9,338) or within individual procedures including aortic valve replacement (n = 2,547). Our results are consistent with another analysis of patients having cardiac surgery, which similarly did not identify daytime variation in outcomes, although sample sizes were smaller, and procedure type was not considered.⁵ 

Our primary endpoint was a composite of in-hospital mortality and low cardiac output syndrome, and our data were restricted to in-hospital outcomes. We did not include postoperative myocardial infarction as an outcome because this term is only defined for patients having coronary artery bypass grafting.²⁴  In the prospective study by Montaigne et al.,¹⁷  troponin T concentrations were lower following afternoon surgery. We could not conduct valid comparisons of troponin T concentrations because the biomarker was only evaluated on postoperative day 1 at a fixed time point (approximately 2:00 am) for all patients, resulting in systematic differences in the time to blood sampling in morning and afternoon patients.

Although the incidence of the composite outcome did not differ significantly between morning and afternoon patients, the 95% CI spanned from a 25% reduction to a 24% increase in the odds for the outcome after morning surgery. The CIs were wide largely because the incidence of our primary composite outcome was low, and we were only adequately powered to test an odds ratio of 1.50. Nonetheless, the odds ratio was near 1.0, suggesting that in-hospital mortality and low cardiac output syndrome do not substantively differ for morning and afternoon aortic valve, mitral valve, and/or coronary artery bypass grafting surgeries. At the same time, the CI of the odds ratio (0.75 to 1.24) included a clinically important boundary, 1.20, indicating that the real difference might in fact be clinically meaningful.

Low cardiac output syndrome was defined by the need for two or more inotropic agents after cardiopulmonary bypass, a cardiac index of less than 2.0 l · min⁻¹ · m⁻², documented left or right myocardial dysfunction, or documented cardiac insufficiency or cardiogenic shock postoperatively. Presumably, most patients experienced myocardial insufficiency/dysfunction consequent to cardioplegic arrest and ischemia–reperfusion injury, although we lacked elevated troponin concentrations or other data to prove this etiology.

Low cardiac output syndrome is a clinically meaningful outcome, with a specific definition.^22,25–27  There are limitations, however, related to the selection of this outcome. Our definition included the need for two or more inotropic medications at 24 to 48 h, which aims to identify patients who experience sustained myocardial dysfunction/failure after cardiac surgery. It is possible that inotropic agents were given for other reasons, such as right ventricular distention or fluid overload, although neither practice is routine at the Cleveland Clinic (Cleveland, Ohio). Importantly, our definition excludes patients who require only vasopressor therapy or “renal-dose” dopamine.

It is also possible that surgical complications including postoperative graft failure, surgical traumatic myocardial injury, or even cardiac dysfunction from cardiopulmonary bypass-induced systemic inflammatory response contributed to the development of low cardiac output syndrome. Although all patients required two or more inotropic agents at 1 to 2 days after surgery, our data show that these patients also received two or more inotropic agents intraoperatively, consistent with myocardial insufficiency caused by an intraoperative event rather than a postoperative complication, suggesting that the cause of low cardiac output syndrome was not acute postoperative graft occlusion. Cardiac dysfunction from a pronounced inflammatory response is also less likely because this condition often resolves within the first 24 h. Intraoperative cardiac dysfunction from aortic clamping and myocardial ischemia–reperfusion thus remains a likely explanation.

Other reports compared outcomes by time of day to assess whether factors such as availability of hospital resources or provider fatigue affected outcomes.^1,5  Our analysis could not distinguish between logistic factors, daytime variation in tolerance to myocardial ischemia, and other causes of low cardiac output syndrome. Nonrandom scheduling, variation in hospital resources such as intensive care unit bed availability, provider fatigue, and follow-up restricted to in-hospital period might all contribute to time-of-surgery outcome differences, although none was observed. Overall, it seems likely that what subtle time-of-day variations might exist are overwhelmed by other perioperative factors.

Investigating risk for cardiac complications and mortality associated with time of day was the primary purpose of this study, although a post hoc analysis was performed to examine other major Society of Thoracic Surgeons complications, including stroke, renal failure, deep sternal wound infection, cardiac reoperation, and prolonged ventilation. No differences in these outcomes were found, with the exception of prolonged ventilation. We suspect that duration of ventilation was prolonged in afternoon surgery patients simply because those patients arrived late to the intensive care unit, and there was some reluctance to extubate at night.

Other limitations to our study include those inherent to its retrospective nature, most notably, the possibility of unmeasured confounders. All patients were cared for at a single quaternary care medical center with special expertise in cardiac care. The results might differ at smaller or less-experienced institutions, but it seems unlikely that institution-specific factors will make time of surgery an important factor. We were unable to examine the difference in troponin concentrations because of systematic differences in time to measurement. Pulmonary artery catheters are only used in about half of our patients, and thus cardiac index could not always be measured. Our analysis did not adjust for anesthesiologists, but the same anesthesia and surgical teams routinely managed both morning and afternoon cases. Furthermore, sensitivity analysis detected no difference by surgeon. Anesthesia medications were not included in the analysis, although the induction medications were similar in both groups. Atrial fibrillation was not included in the composite outcome because it is a less severe complication and occurs at a greater frequency than the other components of the composite and would heavily influence the result.²⁸  Finally, our analysis was restricted to in-hospital events because we did not have long-term outcomes postdischarge.

In summary, we found no evidence for time-of-day variation in mortality or low cardiac output syndrome in patients recovering from aortic valve surgery, mitral valve surgery, and/or coronary artery bypass grafting. Whereas there might be small time-of-day variations in outcome, they appear to be overwhelmed by other perioperative factors. Our results do not support selective morning or afternoon scheduling for cardiac surgery.

Supported by the Department of Cardiothoracic Anesthesiology, Outcomes Research, and the Anesthesiology Institute, Cleveland Clinic (Cleveland, Ohio).

Dr. Duncan received funding from Fresenius Kabi (Bad Homburg, Germany) for research that is not relevant to the current article. Dr. Sessler is a consultant for Edwards Lifesciences (Irvine, California), Mercury Medical (Clearwater, Florida), Pacira (Parsippany, New Jersey), the Health Data Analytics Institute (Dedham, Massachusetts), and B. Braun (Melsungen, Germany). He serves on advisory boards and has equity interests in Calorint (Philadelphia, Pennsylvania), TransQtronics (Wilmington, Delaware), the Health Data Analytics Institute, Medasense (Ramat Gan, Israel), Serenno (Yokeneam Illit, Israel), and Sensifree (Cupertino, California). None is relevant to the current article. The other authors declare no competing interests.