Gajdos C, Hawn MT, Kile D, Robinson TN, Henderson WG. Risk of major nonemergent inpatient general surgical procedures in patients on long-term dialysis.. Arch Surg.. Published online October 15, 2012. doi:10.1001/2013.jamasurg.347.
eTable. General surgery CPT codes used for procedure selection
Gajdos C, Hawn MT, Kile D, Robinson TN, Henderson WG. Risk of Major Nonemergent Inpatient General Surgical Procedures in Patients on Long-term Dialysis. JAMA Surg. 2013;148(2):137-143. doi:10.1001/2013.jamasurg.347
Author Affiliations: Division of GI, Tumor, and Endocrine Surgery, Department of Surgery (Drs Gajdos and Robinson) and Colorado Health Outcomes Program (Ms Kile and Dr Henderson), University of Colorado Anschutz Medical Campus, Aurora; and Section of Gastrointestinal Surgery, Department of Surgery, University of Alabama at Birmingham (Dr Hawn).
Hypothesis Patients on long-term dialysis undergoing major nonemergent general surgical procedures are thought to have high rates of postoperative complications and death.
Design Retrospective cohort study.
Setting Academic and private hospitals.
Patients The American College of Surgeons National Surgical Quality Improvement Program database was used to select dialysis and nondialysis patients who had undergone nonemergent major general surgical procedures between 2005 and 2008. Multivariable logistic regression analysis was used to examine the effect of dialysis on 30-day surgical outcomes adjusted for age, race, sex, work relative value units, American Society of Anesthesiologists class, and recent operations (within the past 30 days).
Main Outcome Measures Patient morbidity, mortality, and failure-to-rescue rates.
Results Dialysis patients undergoing major nonemergent general surgical procedures were significantly more likely to develop pneumonia, unplanned intubation, ventilator dependence, and need for a reoperation within 30 days from the index procedure. Dialysis patients also had a higher risk of vascular complications and postoperative death. Older dialysis patients (aged ≥65 years) had a significantly higher postoperative mortality rate compared with their younger counterparts. Dialysis patients were significantly more likely to die after any complication occurred, and mortality rates were especially high following stroke, myocardial infarction, and reintubation. Abnormalities in potentially modifiable preoperative variables (blood urea nitrogen level, albumin level, and hematocrit) did not increase the risk of postoperative complications or death in dialysis patients compared with nondialysis patients.
Conclusions Dialysis patients undergoing nonemergent general surgery have significantly elevated risks of postoperative complications and death, particularly if they are aged 65 years or older.
More than half a million individuals in the United States undergo dialysis,1 a number that has been forecasted to increase by 7% annually.2 End-stage renal disease requiring dialysis is associated with poor health care outcomes, including a 10-fold increase in risk of hospitalization3 and an expected lifespan between one-fourth and one-sixth of that of the general population.4 Despite the well-described outcomes of the effect of end-stage renal disease on community-dwelling individuals, little is known about the contribution of long-term dialysis to the risks of postoperative complications and death.
Most of the currently existing surgical literature examining the postoperative outcomes of dialysis patients is limited to small retrospective reviews of single-institution experiences.5- 9 While complication and mortality rates of dialysis patients have been a subject of several articles in cardiac surgery,10,11 there is a relative paucity of data on the complication and mortality rates of dialysis patients undergoing general surgery. As a result, the optimal perioperative medical and surgical management of these patients is poorly understood.
The purposes of our study were to better characterize the risk of major general surgery in long-term dialysis patients and to identify potentially modifiable preoperative risk factors that could be moderated to improve surgical outcomes. Using the large American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) database, we examined the postoperative morbidity and mortality in patients undergoing major nonemergent general surgery with and without dialysis.
Data for this study were obtained from the ACS-NSQIP database, which assesses preoperative risk factors, operative data, and 30-day postoperative outcomes for sampled patients undergoing major surgery at participating hospitals. A trained surgical clinical nurse collects the data from the patient's medical records. All data contained within the data set are deidentified. On the 30th postoperative day, the nurse obtains outcome information through record review, reports from morbidity and mortality conferences, and communication with each patient or the patient's family by letter or telephone.12
The ACS-NSQIP database was used to select patients who had undergone major nonemergent general surgical operations between 2005 and 2008. Because emergent surgical operations are known to have further elevated risks of complications and death, all cases coded as emergent in the data set were excluded. Other exclusions included operations aimed at creating or revising hemodialysis access. Current Procedural Terminology codes were used to select patients whose procedure warranted more than overnight stay as we were interested in complication rates following major general surgical procedures (mean postoperative length of stay for all patients, 5.9 days). Additionally, patients with missing information for sex, race, wound classification, age, work relative value units, and whether the patient had undergone a prior operation within the past 30 days were excluded. Patients with missing values for albumin level (n = 51 461 [31.1%]), blood urea nitrogen (BUN) level (n = 18 280 [11.0%]), and hematocrit (n = 10 061 [6.1%]) were retained and missing values were imputed. The eTable lists all Current Procedural Terminology codes used to select patients for this study.
The “currently on dialysis” variable was used to categorize patients into dialysis and nondialysis groups. This variable was defined as yes if the patient had renal failure requiring treatment with peritoneal dialysis, hemodialysis, hemofiltration, hemodiafiltration, or ultrafiltration within 2 weeks prior to surgery according to the ACS-NSQIP data user guide. As all operations collected were nonemergent, the assumption is that few of the patients in the analysis had acute renal failure. Finally, patients were also grouped according to age, 65 years or older vs younger than 65 years (where <65 years is the reference group) as this is a threshold age at which reported mean survival is less than 5 years according to the US Renal Data Service.13
Postoperative outcomes of interest were complications occurring within 30 days of the index operation, return to the operating room within 30 days, postsurgical length of stay, and 30-day mortality. Postoperative complications included the following: surgical site infection (SSI) (including superficial and deep wound infections as well as wound disruptions); cardiovascular, pulmonary, urinary tract, and central nervous system complications; return to the operating room; and postoperative hospital length of stay. In addition, we also grouped complications to form a composite pulmonary outcome (pneumonia, failure to wean from ventilator >48 hours, or reintubation for cardiorespiratory failure) and a composite vascular outcome (stroke/cerebrovascular accident or myocardial infarction [MI]). An overall composite outcome was created by combining SSI, pulmonary complications, and vascular complications.
Baseline patient characteristics were compared among the dialysis and nondialysis groups using χ2 test of association for categorical variables and unpaired t test for continuous variables. Unadjusted postoperative outcome rates were compared by dialysis status using Pearson χ2 test of association.
For each of the dichotomous postoperative outcomes, the adjusted odds ratio (OR) was determined for the dialysis group vs the nondialysis group using multivariable logistic regression analysis. Analyses were adjusted for age, sex, race/ethnicity, work relative value units,14 American Society of Anesthesiologists classification, and prior operation within the past 30 days. The SSI outcome was additionally adjusted for wound classification. A second model of multivariable logistic regression was also done with additional adjustment performed for 6 different procedure types, using groupings outlined in Table 1.
While several preoperative comorbid conditions were more frequently present in dialysis patients, multivariable models were not additionally adjusted for all of these comorbidities individually as their presence resulted in an upgrade in the American Society of Anesthesiologists classification already.
Failure-to-rescue (FTR) rates were computed for the dialysis and nondialysis patients by computing the 30-day mortality rates for those patients who developed each type of postoperative complication. For example, the FTR rate for SSI was calculated as the number of patients who had a postoperative SSI and died within 30 days following the operation divided by the total number of patients who developed a postoperative SSI. These FTR rates were compared between the dialysis and nondialysis patients using relative risk and a 95% CI computed by the Cochran-Mantel-Haenszel method. The 95% CIs greater than 1 indicate FTR rates higher in the dialysis patients compared with the nondialysis patients.
Additional multivariable logistic regression analyses were performed to examine the relative importance of potentially modifiable preoperative risk factors (serum albumin level, BUN level, and hematocrit) in predicting 30-day postoperative morbidity and mortality between the dialysis and nondialysis patients. These models also included age, sex, race, diabetes, functional status, American Society of Anesthesiologists class, and prior operation as adjustment variables. A random normal imputation method was used to populate missing values for albumin level, BUN level, and hematocrit. These random variables were generated to approximate a range of values typically seen in healthy individuals because it was assumed that these values tend to be missing more often for healthy patients than unhealthy patients. All analyses were performed using SAS version 9.3 statistical software (SAS Institute, Inc).
A total of 165 600 patients who underwent elective major general surgery between 2005 and 2008 were included from the ACS-NSQIP data set. The subgroup with dialysis included 1506 patients (0.9%). Patient demographic characteristics for dialysis and nondialysis patients are shown in Table 1. More than 98% of the surgical procedures were done by general surgeons as a surgical specialty.
The incidence of postoperative adverse events is presented in Table 2. Dialysis patients compared with nondialysis patients had a markedly greater rate of 30-day overall complications (composite outcome, 28.6% vs 10.7%, respectively; P < .001), death (12.7% vs 1.5%, respectively; P < .001), and return to the operating room (18.5% vs 4.9%, respectively; P < .001). Most of the increase in complications was attributed to increased rates of pulmonary complications. Additionally, the average length of postoperative surgical stay was more than twice as long for dialysis patients compared with nondialysis patients (13.4 vs 5.8 days, respectively; P < .001).
There were no reported complications for 84.3% of nondialysis patients compared with 62.1% of dialysis patients. Among nondialysis patients, 10.9% had 1 complication and 2.9% had 2 complications; among dialysis patients, 19.6% had 1 complication and 9.8% had 2 complications. There were 3 or more complications in 2.1% of nondialysis patients compared with 8.7% of dialysis patients (P < .001).
The ORs of postoperative events adjusted for race, sex, age, work relative value units, American Society of Anesthesiologists class, and prior operation within 30 days are presented in Table 3. Dialysis patients undergoing elective surgery were significantly more likely to develop pneumonia (OR = 1.28; 95% CI, 1.04-1.57), unplanned intubation (OR = 1.82; 95% CI, 1.48-2.23), ventilator dependence (OR = 1.94; 95% CI, 1.65-2.29), and return to the operating room within 30 days (OR = 1.94; 95% CI, 1.68-2.25) than nondialysis patients. In addition, dialysis patients were more likely to develop a composite pulmonary outcome (OR = 1.89; 95% CI, 1.64-2.18), vascular outcome (OR = 1.69; 95% CI, 1.04-2.75), composite outcome (OR = 1.55; 95% CI, 1.37-1.75), and death (OR = 2.57; 95% CI, 2.15-3.08) within 30 days from surgery. This model was further adjusted for procedure type using 6 different Current Procedural Terminology procedure groups outlined in Table 1. Further adjustment for procedure type resulted in no significant change in the results presented in Table 3 (data not shown).
We examined whether certain procedure types carried a higher risk of postoperative complications. Secondary to sample size, we ran analysis only for the first 4 groups listed under the Current Procedural Terminology groupings in Table 1 (group 1, gastric, small bowel; group 2, hepatobiliary; group 3, colorectal; and group 4, spleen, hernia). Dialysis patients in groups 1 and 3 had much higher frequencies of SSI (13.0% and 13.5%), composite pulmonary outcome (30.7% and 21.4%), all complications combined (39.3% and 31.6%), and mortality (19.6% and 16.2%) compared with groups 2 and 4 (SSI, 5.1% and 7.5%; composite pulmonary outcome, 10.7% and 15.0%; all complications, 15.4% and 15.0%; and mortality, 6.3% and 5.7%).
Further stratification of dialysis patients based on age (<65 years, n = 918 [61.0%]; ≥65 years, n = 588 [39.0%]) was performed. Older dialysis patients undergoing elective general surgery were significantly more likely to die postoperatively compared with younger dialysis patients (OR = 2.65; 95% CI, 1.88-3.74).
The FTR rates, or mortality rates following a complication, were significantly elevated for dialysis patients (Table 4). The FTR rate was significantly higher for all complications studied, with stroke (66.7%), MI (50.0%), and unplanned intubation (38.0%) being the most lethal complications among patients on dialysis (Table 4). The top 3 complications leading to mortality were the same in nondialysis patients, but the order was different: unplanned intubation (25.2%), MI (25.1%), and stroke (23.2%). The likelihood of surviving a complication was the highest following SSI in both groups (98.0% in nondialysis patients vs 87.3% in dialysis patients). The overall risk of dying after any major complication was also significantly higher in the dialysis patients than in the nondialysis patients (21.7% vs 6.4%, respectively; relative risk = 3.40; 95% CI, 2.89-4.00).
We examined the effect of potentially modifiable preoperative variables (preoperative nutrition for albumin level; preoperative dialysis for BUN level; preoperative transfusion for hematocrit) on the risk of developing any postoperative complication or death within 30 days in multivariable analysis (data not shown). All variables were studied as continuous variables. There was clear evidence of more liberal use of intraoperative blood transfusion in dialysis patients. One or more units of blood was received by 14.2% of dialysis patients compared with 6.1% of nondialysis patients. As a limitation of the data, potential interventions to improve low preoperative albumin levels or high BUN levels could not be studied. We did not find that the effect of these variables (low albumin level, high BUN level, low hematocrit) in predicting mortality or morbidity was higher in dialysis patients vs nondialysis patients.
This study is one of the largest on the complication rate and short-term outcomes of nonemergent general surgery in dialysis patients. Dialysis patients had significantly higher rates of postoperative morbidity and mortality in both unadjusted and adjusted analyses compared with their nondialysis counterparts. The elevated number of pulmonary complications (pneumonia, unplanned intubation, ventilator dependence) and reoperations accounted for most of the differences in morbidity. As a result, dialysis patients also had a significantly prolonged postsurgical length of stay and a markedly increased risk of postoperative death. Mortality rates in dialysis patients aged 65 years and older were significantly elevated compared with those of their younger counterparts. Once a complication occurred, dialysis patients were less likely to survive compared with nondialysis patients, with stroke, MI, and unplanned intubation being the most lethal complications in dialysis patients. Potentially modifiable preoperative variables (albumin level, BUN level, hematocrit) did not differentially affect complication rates or the risk of postoperative mortality in dialysis patients compared with nondialysis patients.
Despite reasonable outcomes from small single-center studies arguing for the relative safety of operating on dialysis patients,5,8 larger population-based studies had significantly different findings.15 Patients on long-term dialysis were found to have an increased risk of postoperative adverse events, longer length of stay, and higher rate of death following colorectal surgery.15 While the overall in-hospital mortality for all patients was 3.7% (including emergent cases), elective admissions for colorectal surgery were still associated with a 10.3% mortality in the setting of dialysis. After multivariable adjustment, dialysis patients had a 6-fold risk of mortality compared with their nondialysis counterparts. More than half of the dialysis patients had at least 1 complication, with infectious complications being the most common.
The lower morbidity and mortality rates from single-institution series can be attributed to mixing minor procedures with major interventions, combining a relatively small number of laparotomies with endoscopic procedures, anal fissure operations, dialysis access operations, and parathyroidectomies. The complication and death rates following the latter operations are obviously much lower. The postoperative mortality rate reported in small single-institution series of 1% to 6%5,6,8 is in sharp contrast with findings from larger database studies such as that by Drolet at al15 (10.3%) and our study (12.7%). Some of the small single-institution series reported about a 1% postoperative death rate for dialysis patients,5,8 which is lower than the death rate for nondialysis patients in our study (1.5%). A likely explanation for this discrepancy is that less complex surgical procedures are reported in the single-institution series.
The rate of pulmonary complications in dialysis patients was high in our study (21.6%) and comparable to that reported by Drolet et al.15 The exact reason for the high rate of pulmonary complications in dialysis patients remains unknown. Some studies speculate that dialysis treatment induces hypoxemia and carbon dioxide diffuses in the dialysate, leading to hypocapnia and reflex hypoventilation.16 These events combined may lead to a higher risk of developing atelectasis and pneumonia in the postsurgical setting.
Other studies found correlation between mortality and nutritional parameters in the setting of long-term dialysis. A low serum albumin level was a predictor of mortality and an independent risk factor for death in dialysis patients.17,18 A small single-institution study found that a high BUN level, low hematocrit, and low albumin level increased the rate of postoperative complications in dialysis patients undergoing abdominal surgery.9 As all these variables are potentially modifiable preoperatively in a nonemergent setting (dialysis for a high BUN level; blood transfusion for a low hematocrit; preoperative nutrition for a low albumin level), we thought to further examine the effects of these factors on postoperative complication and death rates. We found no clear evidence that a high BUN level, a low hematocrit, and a low albumin level affect postoperative outcomes differentially in dialysis patients compared with nondialysis patients in multivariable analysis.
The combined systemic effects of chronic renal disease create an altered physiological state for dialysis patients in the perioperative period. This altered physiological state influences the way complications are handled by the body and decreases our ability to rescue the patient once a complication happens. This leads to a significantly higher mortality rate in dialysis patients for all complications studied once an adverse event happens. While MI and stroke with neurological deficit are rare postoperative events even in dialysis patients (0.6%-0.7%), the mortality rate for both of these complications is higher than 50%. Admittedly, the high mortality rate following MI and stroke has to be interpreted with caution given the low number of events in the dialysis group. The third most important FTR complication in both groups was reintubation (38.0% in the dialysis group vs 25.2% in the nondialysis group). In both groups, SSIs were the least likely events to cause death; however, the rate of mortality in the dialysis group compared with that in the nondialysis group was still substantially higher (12.7% vs 2.0%, respectively) and the relative risk of dying was 6.29 (95% CI, 4.05-9.77).
Using age 65 years as a cutoff, we found significantly elevated mortality rates in older dialysis patients compared with their younger counterparts. We chose age 65 years because this is the point at which the average mean life expectancy for dialysis patients decreases to less than 5 years according to the US Renal Data Service. Showing a higher number of serious complications and mortality in elderly patients should caution clinicians and would argue for a less liberal use of general surgery interventions in a patient population with an already significantly limited lifespan.
Our study has several limitations. The database that we used and the analyses that were conducted contain a large number of different types of general surgical operations. The sample size is not large enough to perform analyses for each type of operation. Despite a large number of patients overall, the dialysis subgroup is relatively small (n = 1506). Based on the codes used to define long-term dialysis, there is a chance that some patients with acute renal failure were included. However, the exclusion of emergent operations from the analysis makes this less likely. Furthermore, while all operations related to creation or revision of hemodialysis access as a primary procedure were excluded, the higher rate of 30-day reoperations in dialysis patients may be related to vascular access–related complications.
In summary, we found that long-term dialysis patients undergoing elective general surgery operations who were included in the ACS-NSQIP database have a significantly longer postoperative length of stay and elevated risks of postoperative complications and death compared with their nondialysis counterparts. The most common postoperative adverse events in dialysis patients were pulmonary complications, while the most lethal complications were vascular events (MI or stroke). Older dialysis patients (aged ≥65 years) were more likely to die postoperatively compared with younger patients. The significantly elevated risks of postoperative complications and death deserve an in-depth discussion with dialysis patients preoperatively.
Correspondence: Csaba Gajdos, MD, Department of Surgery, University of Colorado Anschutz Medical Campus, Mail Stop C313, 12631 E 17th Ave, Room 6001, Aurora, CO 80045 (firstname.lastname@example.org).
Accepted for Publication: July 30, 2012.
Published Online: October 15, 2012. doi:10.1001/2013.jamasurg.347
Author Contributions:Study concept and design: Gajdos, Robinson, and Henderson. Acquisition of data: Hawn and Henderson. Analysis and interpretation of data: Gajdos, Hawn, Kile, and Henderson. Drafting of the manuscript: Gajdos, Kile, and Robinson. Critical revision of the manuscript for important intellectual content: Gajdos, Hawn, Robinson, and Henderson. Statistical analysis: Kile and Henderson. Obtained funding: Gajdos. Administrative, technical, and material support: Gajdos and Henderson. Study supervision: Gajdos, Hawn, Robinson, and Henderson.
Financial Disclosure: None reported.
Funding/Support: Statistical analysis for this study was supported by the University of Colorado Department of Surgery and School of Medicine Funds.