Operative Start Times and Complications After Liver Transplantation


Corresponding author: Jorge A. Ortiz, OrtizJor@einstein.edu


The recent national focus on patient safety has led to a re-examination of the risks and benefits of nighttime surgery. In liver transplantation, the hypothetical risks of nighttime operation must be weighed against either the well-established risks of prolonging cold ischemia or the potential risks of strategies to manipulate operative start times. A retrospective review was conducted of 578 liver transplants performed at a single institution between 1995 and 2008 to determine whether the incidence of postoperative complications correlated with operative start times. We hypothesized that no correlation would be observed between complication rates and operative start times. No consistent trends in relative risk of postoperative wound, vascular, biliary, or other complications were observed when eight 3-h time strata were compared. When two 12-h time strata (night, 3 p.m.–3 a.m., and day, 3 a.m.–3 p.m.) were compared, complications were not significantly different, but nighttime operations were longer in duration, and were associated a twofold greater risk of early death compared to daytime operations (adjusted OR 2.9, 95% CI 1.16–7.00, p = 0.023), though long-term survival did not differ significantly between the subgroups. This observation warrants further evaluation and underscores the need to explore and identify institution-specific practices that ensure safe operations regardless of time of day.


In 1999, the Institute of Medicine (IOM) summarized the first phase of its Health Care Quality Initiative in a report that, overnight, attracted the attention of medical professionals, administrators, politicians, and the general public alike, and drastically changed the current climate of medical practice (1–3). In this report, the staggering estimates that medical errors had cost the health care system up to $29 billion and had resulted in as many as 98 000 patient deaths triggered a nationwide, multilevel focus on patient safety and a subsequent widespread and ongoing effort to identify systems-based strategies to minimize medical errors. Of the IOM's findings, two had particular relevance to surgeons and surgical trainees. The first was the identification of surgical complications as the second most common cause of preventable morbidity and mortality (second only to medicine administration errors). The second was the suggestion that physician fatigue and excessive duty hours were associated with a higher incidence of medical errors. Attempts to scientifically evaluate the extent to which sleep deprivation and fatigue impacts performance have been reported in the literature, but the results of these studies have been conflicting at best (4–9). Nonetheless the suggestion that sleep deprivation and fatigue may be associated with an increased incidence of surgical complications has called into question the safety of operating overnight, and has led, at least within some surgical subspecialties, to a re-examination of what truly constitutes a need for emergency surgery and what can be safely managed by delaying operative intervention until the following day (10–12).

The field of transplantation is unique among the surgical subspecialties in that the timing of an operation is largely driven by the time of donor death, and historically, transplant operations are begun when donor organs become available, with little regard to time of day. For deceased-donor kidney operations, the fairly broad window of cold ischemia time (CIT) that can be accrued without measurable detriment to postoperative graft function might potentially allow for the possibility of postponing a nighttime operation until the following morning. At least some centers have adopted this strategy in light of single center data suggesting worse outcomes are associated with nighttime recipient operations (13,14). For liver transplantation, however, the adverse impact of prolonging cold ischemia is well established, and the window of acceptable CIT is decidedly narrower, to the extent that delaying a recipient operation in favor of a daytime operation could render some grafts unusable (15). Alternative means by which to manipulate the start time of the recipient operation, such as delaying the donor operation, can be envisioned but the potential risks of this are difficult to quantify and, furthermore, this strategy is likely to impact the appropriate recovery of other organs also sensitive to time issues.

Thus the suggestion to delay a donor operation solely for purposes of orchestrating a daytime recipient operation would mandate the support of clear evidence that nighttime surgery carries a higher risk for adverse events or poorer outcomes. We hypothesized that no difference in outcomes would be observed between nighttime and daytime liver transplant operations. In order to determine whether recipients of nighttime liver transplant operations have worse outcomes compared to those who undergo daytime operation, we have performed a retrospective, single center analysis of complications after orthotopic liver transplantation as they relate to operative start times.


Study design

Five hundred and seventy-eight patients underwent orthotopic liver transplantation at a single institution between June 21, 1995 and October 14, 2008. Donor and recipient demographic data, operative start times, total operative times, transfusion requirements, incidence of postoperative sepsis, primary graft nonfunction (PNF), and postoperative complications were retrospectively collected by chart and electronic medical record review. Approval to conduct this study was obtained from the Albert Einstein Healthcare Network Institutional Review Board (IRB), and exemption to analyze de-identified data was obtained from the Johns Hopkins Medicine IRB.

For purposes of data analysis, postoperative complications were grouped into four major categories. The following were designated wound complications: wound infection, hematoma, dehiscence, evisceration, and incisional hernias. The following were designated vascular complications: hepatic artery stenosis, hepatic artery thrombosis, hepatic artery blowout, hepatic artery pseudoaneurysm, inferior vena cava thrombosis, portal vein thrombosis, hepatic vein thrombosis, axillary vein hematoma, and arterio-venous fistula. The following were designated biliary complications: bile leak, biloma, bile duct stenosis, biliary sludge, and biliary cast syndrome. The following were designated other complications: peritonitis, gastrointestinal bleed from roux-en-Y, enteric fistula, intra-abdominal fluid collection, intra-abdominal hematoma, intrahepatic hematoma, and lymphocele. Complication data was collected retrospectively, over a median follow-up period of 54 months (IQR 18–100 months). Except for incisional hernias, all complications evaluated in the analyses performed here were limited to those that occurred within the first 30 days after transplantation. Incisional hernias were assumed to be related to the transplant operation regardless of the timing of diagnosis, and thus incisional hernias were considered for the entire follow-up period.

Time strata comprised of 3-h blocks, defined by time of skin incision (designated operative start time), were considered for analysis of complications. The incidence of vascular, biliary, wound, and other complications, as well as the incidence of sepsis, PNF, total operative time (from skin incision to skin closure, measured in hours and minutes), intraoperative blood product (measured in units of red blood cells and fresh frozen plasma) usage, and early death (in the first seven postoperative days) were compared among the different time strata using multivariate models. Complications and early deaths were binary and examined in logistic regression models; operative times were found to be pseudonormally distributed and were examined using linear regression; number of units of blood products were found to be overdispersed count data (skewed integers) and examined using negative binomial regression. Multivariate analyses were adjusted for recipient age, race, gender, body mass index (BMI), model for end-stage liver disease (MELD) score, indication for transplantation, diabetes, transplantation as Status 1, donor age, gender, BMI, donation after cardiac death (DCD) status, and CIT. For patients who were transplanted prior to the capture of MELD score, MELD was retrospectively calculated for this analysis from pretransplant laboratory data.

For purposes of comparing a nighttime operation subgroup to a daytime operation subgroup, the study population was divided into two 12-h start time strata. The daytime subgroup was comprised of the four 3-h strata for which the majority of operative time occurred during daytime hours (between 7 a.m. and 7 p.m.; Figure 1). The nighttime subgroup was comprised of the four 3-h strata for which the majority of operative time occurred during nighttime hours (between 7 p.m. and 7 a.m.). Proportion of day versus night operative time for each of the time strata was calculated using continuous integrals with a uniform distribution assumption, and was based on the mean total operative time of 9 h for the study population. For cases that started between 3 a.m. and 3 p.m., the majority of the operative time occurred between 7 a.m. and 7 p.m., and for cases that started between 3 p.m. and 3 a.m., the majority of the operative time occurred between 7 p.m. and 7 a.m. For this reason, the ‘daytime’ subgroup was defined as those cases that started between 3 a.m. and 3 p.m., and the ‘nighttime’ subgroup was defined as those cases that started between 3 p.m. and 3 a.m.

Figure 1.

Definition of daytime and nighttime subgroups. For each of the eight start time groups (each comprised of a 3-h time block), relative daytime operative time (7 a.m. to 7 p.m., time period between dashed vertical lines) is represented by orange portion of the parallelogram and relative nighttime (7 p.m. to 7 a.m., time periods outside dashed vertical lines) operative time is represented by blue portion of the parallelogram. %night and %day represent percentages of operative time within nighttime and daytime hours as calculated by continuous integration across a 9-h (average operative time) time frame for each start time group. Sun icon depicts the subgroup defined as ‘day.’ These cases had start times between 3 a.m. and 3 p.m., but the majority of operative time took place between 7 a.m. and 7 p.m. Moon icon depicts the subgroup defined as ‘night.’ These cases had start times between 3 p.m. and 3 a.m., but the majority of operative time took place between 7 p.m. and 7 a.m.

For purposes of addressing whether operative business, as defined by clustering of cases within a narrow time frame was associated with differences in complication rates, we defined and divided the cohort into ‘busy’ and ‘nonbusy’ subgroups. The ‘busy’ variable was applied to cases either if a transplant had been performed in the preceding day, or if two or more transplants had been performed in the preceding 4 days. Complication rates were compared between ‘busy’ and ‘nonbusy’ cases.

Statistical analysis

Long-term patient survival was estimated by Kaplan–Meier methodology and compared using log-rank tests. All statistical tests were two-sided with statistical significance set at α= 0.05. All statistical analyses were performed using Stata 11.0/MP for Linux (StataCorp, College Station, TX).


Donor, recipient, and graft characteristics

The cohort of recipients had a mean age of 51.2 years (SD 9.5), BMI of 27.3 (SD 5.1), and MELD score of 20.3 (SD 8.3); 32.8% of recipients were female, 11.6% were black, 4.8% had been the recipient of a previous transplant, and 8.5% were transplanted as Status 1 (Table 1). The indication for transplantation was hepatitis C cirrhosis in 45.3% and alcoholic cirrhosis in 19%. Indications observed less frequently included: hepatocellular carcinoma (9.6%), cryptogenic cirrhosis (8.4%), autoimmune hepatitis (3.0%), primary biliary cirrhosis (3.0%), primary sclerosing cholangitis (2.7%), fulminant hepatitis B (1.9%), and drug induced fulminant hepatic failure (1.1%). Other indications comprised the remaining 3.2% of the cohort. Donors had a mean age of 45.1 years (SD 18.4) and BMI of 25.3 (SD 4.7). 6.1% of grafts were DCD, and mean CIT was 9.2 h (SD 2.3). Warm ischemia time was on average 60 min (SD 20.7). Intensive care unit (ICU) length of stay (LOS) averaged 5.7 days (SD 11.0), and total hospital LOS averaged 18.6 days (SD 24.1). Mean operative time was 9.3 h (SD 3.6), mean units of packed red blood cells (PRBC) transfused was 8.7 (SD 10.1), and mean units of fresh frozen plasma (FFP) transfused was 10.8 (SD 10.1). A comparison of the nighttime and daytime subgroups revealed similar demographics (Table 1).

Table 1.  Recipient and donor demographics, graft and operation characteristics
 AllDaytime startNighttime startp-Value
  1. BMI = body mass index; MELD = model for end-stage liver disease; tx = transplant; ICU = intensive care unit; DCD = donation after cardiac death; CIT = cold ischemia time; EtOH = alcoholic cirrhosis; LOS = length of stay; DRI = donor risk index; PRBC = packed red blood cells; FFP = fresh frozen plasma; WIT = warm ischemia time. Age, MELD, BMI, CIT, total time, PRBC and FFP expressed as means ± standard deviation; all others are represented as proportions.

Recipient characteristics
 Number of recipients578388190 
 Age (years)51.2 ± 9.550.7 ± 9.652.3 ± 9.2 0.05
 BMI27.3 ± 5.127.4 ± 5.127.2 ± 5.10.6
 MELD20.6 ± 8.320.4 ± 8.421.0 ± 8.10.4
 Female (%)32.930.136.80.2
 Black (%)11.610.314.20.2
 Hospitalized at time of tx (%)27.529.423.70.2
 ICU at time of tx (%)12.611.215.30.2
 Life support at time of tx (%) 6.77 6.70.8
 Retransplant (%)4.8 6.4 1.6 0.01
 Status 1 (%)8.5 8.5 8.40.9
   Hepatitis C (%)45.345.142.10.5
   EtOH (%)
 LOS (days) 18.6 ± 24.1 18.2 ± 23.419.7 ±25.50.5
 ICU LOS (days)  5.7 ± 11.0 5.4 ± 9.5  6.4 ± 13.60.3
Donor and graft characteristics
 Age (years) 45.1 ± 18.4 45.7 ± 18.9 44.0 ± 17.20.3
 BMI25.3 ± 4.725.4 ± 4.825.0 ± 4.40.3
 DCD (%)
 CIT (hours) 9.2 ± 2.3 9.3 ± 2.3 9.1 ± 2.20.4
 DRI 1.6 ± 0.5 1.6 ± 0.5 1.7 ± 0.50.4
Operation characteristics
 Total time (hours) 9.3 ± 3.6 9.1 ± 3.39.61 ± 4.20.3
 PRBC (units)  8.7 ± 10.1 8.2 ± 9.3  9.8 ± 11.70.1
 FFP (units) 10.8 ± 10.110.4 ± 9.5 11.7 ± 11.30.2
 WIT (min) 60.0 ± 20.7 60.1 ± 20.4 59.8 ± 21.40.9

Postoperative complications

The overall incidence of any complication among the entire cohort was 36.3% (Table 2). Biliary complications were the most frequently observed complications (observed in 16.3%) followed by wound (observed in 12.6%), vascular (observed in 11.1%), and other (observed in 10.2%) complications. Variability in the incidence of complications was observed among start time strata, and a dominant pattern was not obvious. For purposes of comparing time strata to one another, the 6 a.m. to 9 a.m. subgroup was chosen as the reference group because this time block is comprised of cases for which 98.1% (Figure 1) of operating time fell between the ‘daytime’ hours of 7 a.m. and 7 p.m. Cases with start times between 3 p.m. and 6 p.m. had the greatest odds of any complication, twice those of the reference group with start time between 6 a.m. and 9 a.m. (adjusted odds ratio [aOR] 2.22, 95% CI 1.15–4.28, p = 0.018; Table 3). This appears to be driven by a twofold greater odds of vascular complications in the 3 p.m.–6 p.m. start time group when compared to the reference group (aOR 2.60, 95% CI 1.10–6.17, p = 0.03). The likelihood of wound complications was greatest among the group with start times between 6 p.m. and 9 p.m. (aOR 2.87, 95% CI 1.00–8.21, p = 0.049). Cases that began between 9 a.m. and 12 p.m. also had twice the odds of wound complications compared to the reference group (aOR 2.65, 95% CI 1.16–6.02, p = 0.02). For cases that began during the 9 p.m. to 12 a.m. and the 12 a.m. to 3 a.m. strata, a trend towards greater odds of wound complications was observed but was not statistically significant. No statistically significant differences in biliary complications were observed among the different start time strata. Complications categorized as ‘other’ occurred with greatest odds in the group with start times between 12 p.m. and 3 p.m. (aOR 2.63, 95% CI 1.03–6.72, p = 0.043).

Table 2.  Overall incidence of early death and postoperative complications stratified by operative start times
Time categoryNumber of txptsNumber of early deaths (%)Number of complications
Any (%)Wound (%)Vascular (%)Biliary (%)Other (%)PNF (%)Sepsis (%)
  1. PNF = primary graft nonfunction; Early death, death within 7 days of transplant; Any, patients with at least one complication. Deaths and complications are represented as overall number and percentage of total transplants (txpts) within that time category. Night = 3 p.m.–3 a.m. start times; Day = 3 a.m.–3 p.m. start times.

3 a.m.–6 a.m. 631 (1.6)18 (28.6)6 (9.5)7 (11.1)7 (11.1)6 (9.5)1 (1.6)4 (6.3)
6 a.m.–9 a.m.1276 (4.7)41 (32.2)11 (8.7)13 (10.2)19 (15.0)10 (7.9)4 (3.1)8 (6.3)
9 a.m.–12 p.m.1301 (0.8)50 (38.5)22 (16.9)11 (8.5)26 (20.0)13 (10.0)1 (0.8)4 (3.1)
12 p.m.–3 p.m. 683 (4.4)24 (35.3)8 (11.8)6 (8.8)11 (16.2)12 (17.6)5 (7.4)5 (7.4)
3 p.m.–6 p.m. 574 (6.7)28 (49.1)6 (10.5)13 (22.8)12 (21.1)3 (5.3)2 (3.5)2 (3.5)
6 p.m.–9 p.m. 453 (6.7)19 (42.2)8 (17.8)4 (8.9)7 (15.6)6 (13.3)04 (8.9)
9 p.m.–12 a.m. 503 (6.0)19 (38.0)7 (14.0)6 (12.0)9 (18.0)6 (12.0)05 (10.0)
12 a.m.–3 a.m. 382 (5.3)11 (28.9)5 (13.2)4 (10.5)3 (7.9)3 (7.9)1 (2.6)4 (10.5)
Night19012 (6.3)77 (40.5)26 (13.7)27 (14.2)31 (16.3)18 (9.5)3 (1.6)15 (7.9)
Day38811 (2.8)133 (34.3)47 (12.1)37 (9.5)63 (16.2)41 (10.6)11 (2.8)21 (5.4)
All57823 (4.0)210 (36.3)73 (12.6)64 (11.1)94 (16.3)59 (10.2)14 (2.4)36 (6.2)
Busy1082 (1.9)44 (40.7)16 (14.8)13 (12.0)23 (21.3)13 (12.0)3 (2.8)7 (6.5)
Table 3.  Relative odds of postoperative complications stratified by operative start times
Time categoryNumber of txptsAnyp-ValueWoundp-ValueVascularp-ValueBiliaryp-ValueSepsisp-ValueOtherp-Value
  1. Data presented are adjusted odds ratios and 95% confidence intervals. Reference group for 3-h time block categories was the 6 a.m.–9 a.m. start time group. Reference group for Night vs. Day comparison was Day subgroup. Reference group for Busy vs. Nonbusy comparison was Nonbusy group.

3 a.m.–6 a.m. 630.89 (0.45–1.77)0.71.25 (0.42–3.76)0.71.24 (0.46–3.37)0.70.74 (0.29–1.89)0.51.16 (0.30–4.45)0.81.43 (0.47–4.33)0.5
6 a.m.–9 a.m.127REF REF REF REF REF REF 
9 a.m.–12 p.m.1301.42 (0.84–2.42)0.22.65 (1.16–6.02)0.020.86 (0.37–2.02)0.71.44 (0.74–2.80)0.30.59 (0.16–2.11)0.41.48 (0.61–3.62)0.4
12 p.m.–3 p.m.681.27 (0.67–2.42)0.51.91 (0.69–5.24)0.20.86 (0.31–2.43)0.81.21 (0.53–2.78)0.61.50 (0.43–5.20)0.52.63 (1.03–6.72)  0.043
3 p.m.–6 p.m. 572.22 (1.15–4.28)  0.0181.56 (0.51–4.75)0.42.60 (1.10–6.17) 0.031.65 (0.72–3.75)0.20.51 (0.10–2.69)0.40.68 (0.18–2.66)0.6
6 p.m.–9 p.m. 451.71 (0.83–3.54)0.12.87 (1.00–8.21)  0.0490.92 (0.28–3.07)0.91.17 (0.44–3.10)0.71.79 (0.45–7.04)0.42.13 (0.68–6.68)0.2
9 p.m.–12 a.m. 501.35 (0.67–2.74)0.42.05 (0.69–6.10)0.21.22 (0.42–3.50)0.71.30 (0.53–3.19)0.62.03 (0.56–7.33)0.31.39 (0.45–4.28)0.6
12 a.m.–3 a.m. 380.95 (0.42–2.15) 0.032.19 (0.66–7.24)0.21.07 (0.31–3.62)0.90.48 (0.13–1.77)0.32.51 (0.64–9.86)0.21.17 (0.29–4.76)0.8
Night vs. day1901.34 (0.93–1.94)0.11.24 (0.72–2.15)0.41.56 (0.90–2.68)0.11.03 (0.64–1.67)0.91.55 (0.74–3.25)0.30.84 (0.46–1.54)0.6
Busy1081.20 (0.77–1.86)0.41.12 (0.59–2.11)0.71.10 (0.56–2.15)0.81.48 (0.86–2.54)0.20.93 (0.38–2.31)0.91.23 (0.62–2.44)0.6

Operative time, blood products, and early death

Cases that began between 12 a.m. and 3 a.m. were significantly longer than those that began within the reference time frame of 6 a.m. to 9 a.m. (145 min longer, 95% CI 67.7–223.3, p <0.001; Table 4). Trends towards greater blood product (PRBC and FFP) requirements were observed in cases with later start times but no differences among the start time strata achieved statistical significance. A trend toward greater risk of patient death within 7 days of transplantation was observed among the individual time strata between 3 a.m. and 3 p.m., but this did not achieve statistical significance.

Table 4.  Relative operative times, rates of blood product utilization, risk of primary graft nonfunction (PNF), and risk of early death stratified by operative start times
Time categoryNumber of txptsOperative timep-ValueTotal unitsp-ValuePNFp-ValueEarly deathp-Value
  1. Data presented are adjusted relative time in minutes (Operative time), adjusted rate ratios (Total units), adjusted odds ratios (PNF and early death), and 95% confidence intervals. Reference group for 3-h time block categories was the 6 a.m.–9 a.m. start time group. Reference group for Night vs. Day comparison was Day subgroup. Reference group for Busy vs. Nonbusy comparison was Nonbusy group.

  2. 1No cases of PNF observed in these subgroups.

3 a.m.–6 a.m.6332.66 (−31.70–97.03)0.30.87 (0.66–1.17)0.40.31 (0.02–4.71)0.40.31 (0.03–2.80)0.3
6 a.m.–9 a.m.127 REF REF REF REF 
9 a.m.–12 p.m.130 −3.29 (−55.04–48.45)0.90.87 (0.68–1.09)0.20.08 (0.00–1.64)0.10.16 (0.02–1.40)  0.097
12 p.m.–3 p.m.6819.96 (−42.65–82.57)0.50.92 (0.69–1.22)0.55.37 (0.87–33.23)  0.0710.95 (0.21–4.26)1
3 p.m.–6 p.m.5742.12 (−24.02–108.25)0.21.08 (0.80–1.45)0.63.13 (0.36–27.67)0.31.65 (0.40–6.74)0.5
6 p.m.–9 p.m.459.92 (−62.83–82.68)0.81.16 (0.83–1.61)0.41 1.49 (0.31–7.16)0.6
9 p.m.–12 a.m.5021.57 (−48.22–91.36)0.51.07 (0.78–1.47)0.71 1.87 (0.40–8.63)0.4
12 a.m.–3 a.m.38145.47 (67.67–223.28) <0.0010.83 (0.57–1.21)0.34.67 (0.31–70.44)0.31.52 (0.25–9.34)0.7
Night vs. day190 42.19 (4.98–79.39) 0.0261.14 (0.96–1.35)0.11.50 (0.33–6.86)0.62.85 (1.16–7.00)  0.023
Busy108 −2.44 (−47.41–42.52)0.90.99 (0.81–1.22)0.91.25 (0.28–5.45)0.80.46 (0.10–2.07)0.3

Risk of postoperative complications for nighttime compared to daytime operations

To determine whether the relatively small number of cases within each time stratum could explain the lack of statistical significance associated with trends toward greater odds of complications associated with nighttime operations, the cohort was divided into ‘night’ and ‘day’ subgroups. The night subgroup was comprised of cases that began between 3 p.m. and 3 a.m., since for these cases the majority of the operation took place during nighttime hours (between 7 p.m. and 7 a.m.; Figure 1). The day subgroup was comprised of cases that began between 3 a.m. and 3 p.m., since for these cases the majority of the operation took place during daytime hours (between 7 a.m. and 7 p.m.). No statistically significant difference in risk of wound, vascular, biliary, or other complications, the incidence of sepsis, or PNF, was observed when nighttime operations were compared to daytime operations (Tables 2 and 3). Operative times, however, were on average 42 min longer for cases starting at night (95% CI 5.0–79.4, p = 0.026; Table 4). There was a trend towards greater utilization of blood products for nighttime operations compared to daytime operations but this did not achieve statistical significance (adjusted rate ratio 1.14, 95% CI 0.96–1.35, p = 0.1). The overall rate of patient death within 7 days of transplant, however, was significantly greater among the cases that began at night (6.3% death in the night subgroup versus 2.8% death in the day subgroup; Table 2). Nighttime operation in this cohort was associated with twice the odds of death within 7 days compared to daytime operation, even after adjusting for other factors associated with early death (adjusted odds ratio [aOR] 2.85, 95% CI 1.16–7.00, p = 0.023; Table 4). A analysis of mortality within 30 days of transplantation demonstrated a trend toward a greater odds of death among the nighttime subgroup (aOR 1.86, 95% CI 0.93–3.70, p = 0.08). Long-term patient survival, however, was similar between recipients of daytime and nighttime operations (Figure 2).

Figure 2.

Longitudinal patient survival among recipients of daytime and nighttime operations. Overall patient survival of the daytime (solid line) and nighttime (dashed line) subgroups was estimated by Kaplan–Meier methodology and compared by log-rank test. Although recipients of nighttime operations had a greater risk of early death (see Table 4), no statistically significant difference in 5-year survival was observed between the two groups (p = 0.5).

The causes of death included graft failure (n = 2 among the nighttime subgroup and n = 3 among the daytime subgroup), pulmonary embolus (0 nighttime and 3 daytime), stroke (2 nighttime and 1 daytime), intraoperative hemorrhage (2 nighttime and 0 daytime), respiratory failure (2 nighttime and 0 daytime), cardiac arrest/failure (0 nighttime and 2 daytime), multiple organ system failure (2 nighttime and 1 daytime), and sepsis (1 nighttime and 2 daytime). Hepatic artery thrombosis occurred in one of deaths among the daytime subgroup, otherwise deaths were not clearly associated with any of the major vascular, wound, biliary, or other complications. The majority of the early deaths occured within the first 48 h postoperatively. For the recipients of daytime operations who died within 7 days, 6 deaths occurred on POD0, 2 on POD 1, 1 on POD 2, and 2 on POD 5. For the recipients of nighttime operations, 1 death occurred on POD0, 7 in POD 1, 1 on POD 2, 1 on POD 4, and 2 on POD7.


In this manuscript we report the results of a single center retrospective analysis of postoperative complications following liver transplantation as they relate to operative start times. When our cohort was divided into eight start time subgroups, comprised of the cases that began within 3-h time blocks, our analysis revealed variable incidence of complications among the different start time strata. The lack of any consistent trend was suggestive of a lack of impact of operative start time on the complications that were studied. Consistent with this inference, when the cohort was divided into two larger subgroups, a nighttime and daytime start group, again no statistically significant differences in wound, vascular, biliary, or other complications were observed. Nighttime operations were, however, slightly longer in duration and trended towards a greater utilization of blood products than daytime operations. Interestingly, the risk of early death (within 7 days of transplant) was twice as high in the nighttime start subgroup when compared to the daytime start subgroup. This death rate did not appear to be driven by complications, at least as they were analyzed under our categorization scheme.

It is important to recognize some limitations to this study, particularly those of potential misclassification bias, residual confounding, and generalizability. First, while every effort was made to ensure accuracy of the data set, we acknowledge that retrospective data collection is subject to errors including but not limited to data entry errors, inaccuracies or omissions in the original records, and inconsistencies in the interpretation of primary data by different reviewers. Given the relatively small size of this cohort, inaccuracies in the data set could impact the observed trends. Next, the time stratification scheme chosen here was one of many possible categorizations, and it is conceivable that other categorizations may have identified differences among start times not observed among strata we ultimately chose. Similarly, because of the small absolute numbers of individual complications, we grouped complications into four categories based on physiologic system: wound, vascular, biliary and other. It is possible that different groupings of individual complications may have identified differences or trends not observed with the grouping scheme we chose. In addition, the ability of our multivariate analysis to adjust for confounding factors when comparing the nighttime to the daytime start groups was limited by the parameters that we were able to capture in the data set. It is possible that demographic or other differences not adjusted for in the analysis could account for the difference in early death observed between the nighttime and the daytime subgroups. Finally, the relatively small sample size of this single center analysis limits the power to detect small but clinically meaningful differences, and the single center nature potentially limits generalizability. For example, our study had 80% power to detect an odds ratio of 2.15 between night and day cases for the outcome of ‘Vascular Complications;’ as such, it is impossible for us to infer from our findings whether the odds ratio of 1.56, which clearly would be clinically meaningful, would be statistically significant in the context of a larger study.

While our analysis revealed variable incidence of complications among the different start time strata, some strata were, in fact, associated with statistically greater odds of certain complications. Despite the lack of a clear trend towards a single start time block for which risks of complications were universally higher, it is nonetheless important to speculate on possible mechanisms that could contribute to or account for the differences observed. The 3 p.m. to 6 p.m. start time block, while not associated with greater risk of complications of all types, appeared to be associated with statistically significantly higher complication rates more frequently than other time strata. One possible explanation for this observation might be that at some point during the cases that began between 3 p.m. and 6 p.m., a shift change would have occurred in virtually all of the cases. Several scenarios can be envisioned in which change-of-shift creates at least a transient disharmony in the operating room that might lead to problems that ultimately contribute to complications. It is widely believed that many medical errors are attributed to lapses in communication during patient care hand-offs (16). In the operating room, one can envision either that miscommunication during hand-offs could lead to at least a temporary lack of preparedness, or that the actual act of signing out could serve as a distraction or result in a delayed response to an acute issue. The impact of hand-offs that occur in the operating room among nursing, anesthesia, and support staff is not known but probably warrants evaluation.

Certainly the observation that early postoperative death occurred more frequently among recipients of nighttime operations warrants an analysis of the extent to which surgeon or staff fatigue may have contributed to this outcome. The retrospective nature of this study precludes our ability to address this topic at present. A rigorous analysis of this issue, however, would be a worthy undertaking and would need to assess fatigue as it applies to surgeons, anesthesia staff, operating room nursing staff, as well as the intensive care unit staff who manage the patient in the critical, early postoperative period.

Our results must be interpreted in the context of staffing, both surgical and nonsurgical, at the hospital where this study was conducted. With regard to overnight staffing of transplant operations, nursing and anesthesia teams were comprised of the overnight in-house teams and were not transplant-specific. With regard to operating surgeons, the organ was recovered by the surgeon on call for recovery, with the assistance of a transplant surgery fellow; the attending surgeon for the recipient operation was the individual on call for liver transplants at the time the organ was offered, and was assisted by a transplant surgery fellow. In general, staffing policy at the level of the attending and recovering surgeon, nursing, and anesthesia staffing was consistent during the entire study period.

A multicenter or national database analysis would enable the evaluation of a larger patient population which would improve the statistical power of the study, and would enable a longitudinal exploration of whether any the risks of nighttime operation have changed with time. Notwithstanding, the ultimate merits of a multicenter analysis are not entirely clear-cut. The observation that nighttime operation is associated with poorer outcomes is most likely explained by factors that are center-specific. In fact, a comparison of multiple single center analyses of this sort might be especially informative, as this might help to identify which centers have adopted strategies that most effectively improve outcomes of nighttime operations. This evaluation of our single center cohort may not, at this point, support the universal recommendation to change current practice, but it underscores the need for continued assessment and for efforts to identify and implement institution-specific practices that ensure the safest possible operation at any time of day or night.


This work was supported by a grant from the Einstein Society (J.A.O.).