Association of Operative Time of Day With Outcomes After Thoracic Organ Transplant FREE

Since the Institute of Medicine published a report suggesting that medical errors result in more than 98 000 deaths annually,¹ increasing emphasis is being placed on systems-based approaches to improve patient safety. Of the significant complications alleged to be caused by medical errors, surgical complications were specifically cited as the second most common cause of preventable morbidity and mortality.^1- 2

Although the causes of medical errors are likely multifactorial, many have suggested that medical staff fatigue associated with delivering medical care outside of daytime working hours is an important driving factor.^2- 5 Nighttime medical care has previously been associated with worse outcomes in general medicine,⁴ cardiology,⁶ general surgery,^5,7- 9 and abdominal organ transplant surgery.^2,10

To our knowledge, no study has assessed the relationship between nighttime surgery and outcomes following thoracic organ transplants. Therefore, we undertook this study to assess whether performing orthotopic heart transplants or lung transplants at night was associated with adverse outcomes for transplant recipients.

United Network for Organ Sharing (UNOS) data were extracted from the standard transplant analysis and research files. These files represent a prospectively collected open cohort of all US thoracic organ transplants since 1987. These data are collected and compiled by trained data-entry personnel. The Johns Hopkins Medicine institutional review board waived informed consent because this study used deidentified patient data.

We conducted a retrospective cohort study of all adults older than 18 years who underwent heart or lung transplants from January 2000 through June 2010. Patients undergoing retransplant surgery were excluded. Patients were stratified by operative time of day as defined below. The primary end point was all-cause mortality. Secondary outcomes included commonly observed postoperative complications.

Patients were stratified by operative time of day. Daytime was defined as between 7 AM and 7 PM. Nighttime was defined as between 7 PM and 7 AM. Organ ischemic time was added to the donor cross-clamp time to estimate when the critical portion of the recipient transplant operation took place. This time point was used for stratification. Patients missing time data were excluded.

Recipient variables examined included demographic factors, comorbidities, and measures of patient acuity. Donor and operative variables examined included demographic factors and comorbidities. Transplant variables included cytomegalovirus mismatching, human leukocyte antigen mismatching, transplant center volume (high volume was defined as >15/y for heart transplants and >20/y for lung transplants),^11- 12 waitlist time, and allograft ischemic time. Recipient and donor race were determined by patient self-identification. The primary outcomes examined were 30- and 90-day and 1-year mortality. Secondary outcomes included common complications (need for reoperation, pacemaker placement, noncardiac surgery, drug-treated infection, cerebrovascular accident, and airway dehiscence) and total length of stay.

We compared baseline characteristics using the t test for continuous variables and χ² tests or Fisher exact probability test for categorical variables. The Wilcoxon rank-sum test was used for continuous nonparametric variables.

The association between operative time of day and survival was evaluated using the Kaplan-Meier method. The log-rank test was used to compare these survival estimates by strata. Risk of 90-day mortality was assessed using a multivariable Cox proportional hazards regression model. Independent covariates supported by previously published literature, biological plausibility, or a probability value less than 0.2 on univariate analysis were considered variables of interest. Variables of interest were incorporated into the Cox model in a forward and backward stepwise fashion using the likelihood ratio test and Akaike's information criterion in a nested model approach to maximize the explanatory power of our model. Covariates for which more than 15% of data are missing include recipient hypertension, recipient educational level, donor inotropic status, and a history of prior lung surgery. Because models were constructed by case-wise deletion, these covariates were excluded. Because different variables are important for heart transplants and lung transplants, 2 separate multivariable models were constructed.

The final model for heart transplants included operative time of day, recipient factors (age, sex, race, creatinine level, preoperative dialysis status, cardiac output, and mean pulmonary artery pressure), measures of acuity (need for preoperative ventilator, extracorporeal membrane oxygenation, or ventricular assist device [VAD] support, preoperative hospitalization, preoperative intensive care unit [ICU] care), donor factors (age, sex, cigarette use), and transplant factors (days on the waiting list, cytomegalovirus mismatch, human leukocyte antigen mismatch, allograft ischemic time, and center volume).

The final model for lung transplant included operative time of day, recipient factors (age, sex, race, creatinine level, preoperative dialysis status, cardiac output, and mean pulmonary artery pressure), measures of acuity (need for preoperative ventilator, or extracorporeal membrane oxygenation support, preoperative hospitalization, preoperative ICU care), forced expiratory volume in 1 second, donor factors (age, sex, cigarette use), and transplant factors (allograft ischemic time and center volume).

The proportional hazard assumption was tested graphically by evaluating scaled Schoenfeld residuals and complementary log-log plots for each covariate. Inspection of residuals plots and complementary log-log plots suggested the proportional hazard assumption held for our model. Because the study population bridged the introduction of the lung allocation score, this variable was excluded from the multivariable model, though many of the covariates that contribute to the score were included.

For all statistical measures including the t test, χ² test, Fisher exact test, and Wilcoxon rank-sum test, 2-tailed P < .05 was considered significant. Continuous variables are presented as mean values with standard deviations. Categorical variables are presented as whole numbers and percentages. Medians are presented with interquartile range (IQR). Hazard ratios (HRs) are shown with their 95% confidence intervals (CIs). Statistical analysis was performed with Stata 9.2 SE software(StataCorp, College Station, Texas).

A total of 36 394 thoracic organ transplants were performed during the study period. After excluding 1364 repeat organ transplants (3.75%), 3575 pediatric transplants (9.82%), and 249 combined heart-lung transplants (0.68%) as well as 3396 with inadequate timing data (9.33%), 27 118 patients comprised the final study population. This included 16 573 heart transplants (59.59%): 8346 (50.36%) during the day and 8227 (49.64%) during the night and 10 545 lung transplants (39.92%): 5179 (49.11%) during the day and 5366 (50.89%) during the night. A total of 8061 patients (28.99%) died during a median follow-up of 32.2 months (IQR, 11.2-61.1 months).

Of those undergoing heart transplants, 400 had a concomitant organ transplant (54 liver, 340 kidney, and 6 liver-kidney transplants). For patients undergoing lung transplants, 30 had a concomitant organ transplant (22 liver, 6 kidney, 2 pancreas transplants).

Patients undergoing heart transplant during the daytime had higher creatinine levels, lower mean pulmonary artery pressures, lower pulmonary capillary wedge pressures, more preoperative dependence on inotropes, and less preoperative VAD implantation (Table 1). Additionally, their donors were younger, had less hypertension, and more commonly needed inotropic support before organ recovery.

In patients undergoing lung transplants, daytime transplants occurred in fewer elderly recipients (Table 2). Recipients were more likely to be white, less likely to have a diagnosis of idiopathic pulmonary fibrosis, had a higher forced vital capacity, were less likely to be hospitalized or in the ICU preoperatively, more likely to have donors with a smoking history, and tended to have a longer time on the waitlist.

Although several differences in baseline characteristics were statistically significant between patients undergoing either procedure, the absolute differences were small and unlikely clinically relevant.

In unadjusted analysis, operative time of day (daytime vs nighttime) was not associated with 30-day survival (95.02%; 95% CI, 94.53%-95.47% vs 95.22%; 95% CI, 94.73%-95.66%; P = .78), 90-day survival (92.55%; 95% CI, 91.96%-93.10% vs 92.71%; 95% CI, 92.12%-93.26%; P = .93), or 1-year survival (87.98%; 95% CI, 87.24%-88.67% vs 87.65%; 95% CI, 86.89%-88.37%; P = .46; Figure) for all heart transplants. Additionally, in the subset bridged to transplant with a VAD, 1225 (46.77%) underwent heart transplantation during the day and 1398 (53.23%) during the night. Operative time of day was not associated with 30-day survival (93.51%; 95% CI, 91.95%-94.77% vs 93.97%; 95% CI, 92.55%-95.12%; P = .87), 90-day survival (90.67%; 95% CI, 88.85%-92.20% vs 90.79%; 95% CI, 89.09%-92.24%; P = .63), or 1-year survival (84.57%; 95% CI, 82.28%-86.59% vs 85.16%; 95% CI, 83.03%-87.05%; P = .98). For lung transplants, while operative time of day was not associated with 30-day survival (96.01%; 95% CI, 95.44%-96.52% vs 95.52%; 95% CI, 94.93%-96.05; P = .13), nighttime lung transplants were associated with a lower unadjusted 90-day survival than daytime lung transplants (91.66%; 95% CI, 90.87%-92.38% vs 92.73%; 95% CI, 91.87%-93.41%; P = .03). However, this difference is clinically small (absolute difference, 1.07%, number needed to treat, 94). For lung transplants, operative time of day was not associated with survival at 1 year (83.83%; 95% CI, 82.76%-84.85% vs 82.58%; 95% CI, 81.49%-83.62%; P = .08).

Additionally, there were no differences in the incidence of complications for heart transplants (Table 3). Although the incidence of most postoperative complications was similar for lung transplants (Table 3), there was an increased incidence of airway dehiscence for the 57 of 5022 patients (1.14%) whose lung transplant was performed during the night compared with 87 of 5224 (1.67%) of those performed during the day (P = .02). Again, this difference is small (absolute difference, 0.53%; number needed to treat, 189).

Nighttime heart transplants were not associated with an increased risk of 30-day, 90-day, or 1-year mortality on multivariable analysis (Table 4). In the heart transplant model for 90-day mortality (eTable 1), the factors associated with an increased risk of mortality included increased recipient age, creatinine level, mean pulmonary artery pressure, preoperative need for ventilation, ICU care, extracorporeal membrane oxygenation, VAD support, waitlist time, donor age, and ischemic time. Male donors, Hispanic race, cytomegalovirus mismatch, human leukocyte antigen mismatch, and high- center volume were associated with reduced mortality risk.

Nighttime lung transplants were not associated with an increased risk of 30-day mortality on multivariable analysis (Table 4). However, nighttime lung transplants were associated with an increased risk of 90-day mortality on multivariable analysis (eTable 2). This association was not demonstrated in the multivariable model for 1-year mortality. Other factors associated with an increased risk of 90-day mortality in this model of lung transplants included recipient age, increased creatinine level, higher mean pulmonary artery pressures, need for preoperative hospitalization, ICU care, extracorporeal membrane oxygenation support, donor cigarette use, and longer ischemic time. Male donors and transplants performed at a high-volume center were associated with reduced mortality risk.

To our knowledge, this analysis represents the only study to date of the association of operative time of day with outcomes after thoracic organ transplantation. We found no difference in short-term survival between daytime and nighttime heart transplants regardless of prior implantation with a VAD. Additionally, we found no difference in short-term mortality for lung transplants at 30 days or at 1 year. There was a small difference in absolute mortality at 90 days in both unadjusted and adjusted analysis. Although this difference warrants consideration and further investigation, we speculate this difference is detectable due to the high power associated with large registry data sets but is not clinically meaningful. The fact that 30-day and 1-year mortality are not different further suggests that this difference while statistically real probably is not clinically significant. Therefore, we conclude that operative time of day is not associated with short-term survival after thoracic organ transplant procedures.

Although survival is ultimately the most important outcome measure, a detailed examination of secondary outcomes is crucial to understanding the potential adverse effects associated with nighttime transplantation. Diligent postoperative care may ameliorate the potential detrimental effects of nighttime surgery on survival; however, postoperative complications are measures not so readily offset by meticulous postoperative care. With only one exception, there was no difference in complications after either heart or lung transplants. We did find a slightly higher incidence of airway dehiscence in patients undergoing lung transplants at night. Airway dehiscence is a rare complication; however, it may be influenced by surgeon fatigue because airway anastomosis requires meticulous surgical technique and occurs near the end of the operation. Additionally, although vascular anastomotic problems are often immediately evident, airway problems may not become apparent for days after the transplant. However, the absolute difference we observed is small and likely not clinically significant but requires further evaluation.

Postoperative bleeding requiring reoperation is not an uncommon complication after complex cardiac surgery. Theoretically, surgical fatigue could lead to careless technique, resulting in additional bleeding requiring reoperation. Moreover, operative exhaustion may increase the tendency to overlook surgical bleeding. However, for both heart and lung transplants, the rate of cardiac reoperation was the same regardless of operative time of day suggesting that the fatigue that has been associated with nighttime operations is not associated with poor surgical technique.

Moreover, we found total hospital length of stay to be similar regardless of operative time of day. Length of stay is a broad marker of multiple adverse outcomes. Postoperative patients transferred to the ICU late at night could encounter critical care clinicians who may be fatigued, potentially resulting in inattention to detail in the crucial early postoperative period. Such inattention, theoretically, could lead to less aggressive weaning of ventilatory and inotropic support, resulting in a greater duration of ventilation and need for central venous access, possibly increasing infections and overall length of stay. However, we found that length of stay is similar regardless of operative time of day.

Previous studies examining nighttime medical care have suggested it is associated with poorer outcomes. In a large study of almost 60 000 patients, Peberdy et al⁴ found that nighttime cardiac arrests were associated with a lower survival to discharge, lower 24-hour survival, and unfavorable neurological outcomes. In another study, Glaser et al⁶ found that patients who underwent percutaneous coronary intervention for acute myocardial infarction (MI) had a higher incidence of in-hospital death and MI as well as a higher rate of procedural complications.

In the surgical literature, Ricci et al⁸ reported that urgent orthopedic surgery at night was associated with a higher rate of unplanned reoperation and the need to remove painful hardware. Similarly, Komen et al⁵ reported that nighttime surgery was an independent risk factor for anastomotic leakage after colorectal surgery.

More specifically, in the transplant literature, Fechner et al¹⁰ reported that nighttime kidney transplantation was associated with a higher risk of graft failure and need for emergent reoperation. Although not all studies have found the same risk of nighttime renal transplants,¹³ the increased risk of graft failure associated with a valuable societal resource like a kidney is sobering. Lonze et al² reported a series of 578 consecutive liver transplants stratified into 12-hour strata and found that the nighttime operation group was associated with longer operations, a statistically nonsignificant increase in blood product utilization, and a startling 2-fold increase in early death.

Given the emerging consensus in the literature that nighttime operations are associated with an increase in complications, it is noteworthy that operative time of day is not associated with thoracic organ transplantation outcomes in a clinically meaningful way. Previous reports have suggested that nighttime operation may be affected by fatigue,² intraoperative shift changes,² changes in nighttime staffing patterns,⁴ biological and circadian factors,⁴ technical lapses,^7- 9 greater reliance on residents,⁵ and decreased situational awareness.⁵ Although each of these explanations has its merits, it is difficult to understand why thoracic transplant surgery teams would be uniquely immune to any of these problems.

It is likely that because urgent nighttime operations are common in all types of transplant surgery, health care personnel involved in the transplant have developed various systems to prevent errors and directly cope with the limitations associated with nighttime medical care. For example, at our institution, heart and lung transplant teams include surgeons, anesthesiologists, perfusionists, and operating room staff who work together routinely and can be mobilized rapidly for transplant operations. Such routine camaraderie and familiarity likely diminishes the novelty of the operation, lessening the burden and strain of the nighttime environment. Moreover, transplant surgeons spend years training to perform transplants, often at night, thus preparing them for nighttime transplants as attending surgeons. However, this does not entirely account for the disparate effect of nighttime surgery on thoracic and abdominal organ transplantation.

Previous studies have suggested that patients treated during nighttime have higher acuity and a greater urgency of treatment than those treated during the day. In thoracic transplantation, recipient acuity determines organ allocation independent of the time of day. Therefore, as the baseline characteristics of our cohort demonstrate, patients who undergo transplant surgery at night have similar acuity as those undergoing it during the day, and so the surgeries are performed with similar urgency. This difference may provide some insight into why thoracic organ transplant outcomes are similar regardless of time of day.

These equivalent results for thoracic organ transplants regardless of operative time of day should be taken in context. Although our study was retrospective, it does suggest that not all trends in patient safety can be applied empirically across all specialties and that not all purportedly beneficial safety interventions are necessary or will be effective. Further investigation is warranted to understand why thoracic organ transplants do not appear to be associated with deleterious outcomes previously reported to be associated with nighttime surgery. Insight into the systems-based interventions that mitigate the effects of fatigue and disrupted circadian rhythms may provide information to help ameliorate the adverse outcomes associated with nighttime medical care in other settings.

Several limitations are inherent in the use of registry data. Despite the large number of variables in the UNOS database, it is impossible to certify that all possible confounders have been considered in this retrospective study. The data are compiled from multiple centers allowing for potential bias associated with differences in data entry practices at different centers. We have assumed that both coding errors and missing data occur at random. If missing or miscoded data do not occur randomly, residual bias may still exist.

Additionally, systems-based practices and outcomes are difficult to define and measure, particularly retrospectively. For example, night shifts are often relegated to staff lacking the seniority to avoid them, resulting in less experienced staff to provide nighttime medical care. Information on staffing patterns and other systems-based practices are not currently available in the UNOS database. Moreover, while operative time of day may not directly affect morbidity and mortality, it may have a more subtle and less quantifiable influence on patients and clinicians.

Given the small differences in absolute mortality in each group, we cannot exclude the possibility of a type II error. For instance, we estimated that a study with 90% power to detect a similar difference for patients undergoing heart transplants, bridged to transplant, and lung transplants would require, respectively, more than 207 000, 77 000, and 18 000 in each group. However, achieving greater statistical power than our study would be technically difficult because we used the largest national registry available to examine a cohort of thoracic organ transplant patients.

In the literature, the most commonly used method to stratify operative time of day is by operative start time.^{2,4,6,8,10,13} For the UNOS data, the only time variables available were the time of donor cross clamp and the total ischemic time. We added these together to estimate the time of reperfusion, thus approximating the time of the critical portion of the operation. We acknowledge that these classifications are arbitrary. Different methods of determining when operations occurred as well as the construction of different time strata are all possible. We chose 7 AM to 7 PM and 7 PM to 7 AM as reasonable strata based on the occurrence of the nursing shift change at our institution and perhaps at other institutions, although acknowledge that other rational time points exist.

In this large study using national registry data to examine the relationship between operative time of day and thoracic organ transplantation, operative time of day was not associated with a clinically meaningful difference in clinical outcomes.