Share this article on:

Interaction Effects of Acute Kidney Injury, Acute Respiratory Failure, and Sepsis on 30-Day Postoperative Mortality in Patients Undergoing High-Risk Intraabdominal General Surgical Procedures

Kim, Minjae MD, MS*; Brady, Joanne E. PhD; Li, Guohua MD, DrPH*†

doi: 10.1213/ANE.0000000000000915
Critical Care, Trauma, and Resuscitation: Research Report

BACKGROUND: Acute kidney injury (AKI), acute respiratory failure, and sepsis are distinct but related pathophysiologic processes. We hypothesized that these 3 processes may interact to synergistically increase the risk of short-term perioperative mortality in patients undergoing high-risk intraabdominal general surgery procedures.

METHODS: We performed a retrospective, observational cohort study of data (2005–2011) from the American College of Surgeons-National Surgical Quality Improvement Program, a high-quality surgical outcomes data set. High-risk procedures were those with a risk of AKI, acute respiratory failure, or sepsis greater than the average risk in all intraabdominal general surgery procedures. The effects of AKI, acute respiratory failure, and sepsis on 30-day mortality were assessed using a Cox proportional hazards model. Additive interactions were assessed with the relative excess risk due to interaction.

RESULTS: Of 217,994 patients, AKI, acute respiratory failure, and sepsis developed in 1.3%, 3.7%, and 6.8%, respectively. The 30-day mortality risk with sepsis, acute respiratory failure, and AKI were 11.4%, 24.1%, and 25.1%, respectively, compared with 0.85% without these complications. The adjusted hazard ratios and 95% confidence intervals for a single complication (versus no complication) on mortality were 7.24 (6.46–8.11), 10.8 (8.56–13.6), and 14.2 (12.8–15.7) for sepsis, AKI, and acute respiratory failure, respectively. For 2 complications, the adjusted hazard ratios were 30.8 (28.0–33.9), 42.6 (34.3–52.9), and 65.2 (53.9–78.8) for acute respiratory failure/sepsis, AKI/sepsis, and acute respiratory failure/AKI, respectively. Finally, the adjusted hazard ratio for all 3 complications was 105 (92.8–118). Positive additive interactions, indicating synergism, were found for each combination of 2 complications. The relative excess risk due to interaction for all 3 complications was not statistically significant.

CONCLUSIONS: In high-risk general surgery patients, the development of AKI, acute respiratory failure, or sepsis is independently associated with an increase in 30-day mortality. In addition, the development of 2 complications shows significant positive additive interactions to further increase the risk of mortality. Our findings suggest that interactions between these 3 perioperative complications increase the risk of mortality more than would be expected by the independent effects of each complication alone.

Published ahead of print August 13, 2015

From the *Department of Anesthesiology, Columbia University Medical Center, New York, New York; and Department of Epidemiology, Columbia Mailman School of Public Health, New York, New York.

Accepted for publication May 30, 2015.

Published ahead of print August 13, 2015

Funding: Supported, in part, by a grant from the International Anesthesia Research Society.

The authors declare no conflicts of interest.

Presented as a poster at the American Society of Anesthesiologists Annual Meeting, New Orleans, LA, October 2014.

Reprints will not be available from the authors.

Address correspondence to Minjae Kim, MD, MS, Department of Anesthesiology, Columbia University Medical Center, 622 West 168th St., PH 5, Suite 505C, New York, NY 10032. Address e-mail to mk2767@cumc.columbia.edu.

Acute kidney injury (AKI), acute respiratory failure, and sepsis are recognized as distinct but interrelated pathophysiologic processes with similar mortality rates among critically ill patients.1 Patients who develop more than one of these conditions have a markedly increased risk of mortality. For example, AKI and acute respiratory failure are associated with mortality rates between 43% to 66% and 30% to 45%, respectively, but those with both processes have a >80% chance of dying in the hospital.1,2 In addition, a given process increases the risk of developing the other 2 processes. Among mechanically ventilated patients in the intensive care unit (ICU), AKI is associated with more severe disease, including higher rates of sepsis and a longer duration of mechanical ventilation.3 Sepsis is both a cause of and complication of AKI in the ICU, with a high correlation between the severity of sepsis and the incidence of AKI.4

The effects of AKI,5 acute respiratory failure,6,7 or sepsis8 in surgical patients have been studied independently and associations with other complications have been described.9 However, no study to date has evaluated any type of interaction between these 3 complications to assess their combined effects on perioperative mortality. We hypothesize synergism among AKI, acute respiratory failure, and sepsis increases the risk of mortality in intraabdominal general surgery patients. Although other complications can occur in the perioperative period, we focused on these 3 because of their high incidence and risk of mortality in critically ill patients.1 We used data from the American College of Surgeons-National Surgical Quality Improvement Program (ACS-NSQIP), a large, multicenter database of surgical outcomes from hospitals throughout North America. Understanding the epidemiologic links between AKI, acute respiratory failure, and sepsis in the perioperative period may help clinicians optimize perioperative outcomes and provide insight into the role of interactions between these acute processes.

Back to Top | Article Outline

METHODS

Data

This study was not subject to review by the Columbia University Medical Center Institutional Review Board (New York, NY) because it did not require access to protected health information. The ACS-NSQIPa is a validated, prospectively collected national data set aimed at improving surgical quality and outcomes.10 The data collected include demographic characteristics, presurgical comorbidities, intraoperative variables, and 30-day postoperative morbidity and mortality data. All data are carefully reviewed by each site’s surgical clinical reviewer, and centers not meeting specific criteria for quality are removed from the data set. The systematic sampling process and criteria for maintaining the high quality of the data set have been described previously.11

Back to Top | Article Outline

Patient Selection

We selected patients undergoing high-risk intraabdominal general surgery using data from the 2005 to 2011 ACS-NSQIP participant use files. The Clinical Classifications Software for Services and Procedures (Agency for Healthcare Research and Quality, Rockville, MD)b were used to classify procedures based on the primary Current Procedural Terminology (American Medical Association, Chicago, IL) code. Fourteen intraabdominal general surgery categories were identified for consideration (Appendix 1), resulting in an initial sample of 795,154 records. Patients classified as outpatient were excluded (n = 209,936) because they have a low risk of adverse perioperative outcomes. Patients were also excluded if they had evidence of preoperative renal injury (n = 10,627), lung injury (n = 11,607), sepsis (n = 91,787), or had missing data on these variables (n = 60,896). The following conditions, as defined by the data set,c were excluded: (1) preoperative acute renal failure, (2) dialysis, (3) ventilator dependence, (4) current pneumonia, and (5) systemic sepsis (systemic inflammatory response syndrome, sepsis, or septic shock; Appendix 2). Finally, we excluded 4 procedure categories (other hernia repair [n = 48,538], cholecystectomy [n = 52,651], gastric bypass [n = 70,345], and appendectomy [n = 45,066]) that were considered low risk because their likelihood of postoperative AKI, sepsis, and acute respiratory failure were all below the average risks for the entire cohort. The final sample for analysis consisted of 217,994 records.

Back to Top | Article Outline

Baseline Demographic and Operative Variables

Demographic, comorbidity, and operative variables were collected directly from the ACS-NSQIP data set. Patients who are 90 years or older were reported as being 90 years old. Race/ethnicity was categorized as Caucasian versus non-Caucasian. Body mass index was calculated from height and weight data. The estimated glomerular filtration rate (mL/min/1.73 m2) was calculated using the Chronic Kidney Disease Epidemiology Collaboration formula incorporating creatinine, sex, age, and race12 and categorized into estimated glomerular filtration rate groups corresponding to the stages of chronic kidney disease13: <30, 30 to 60, 60 to 90, >90, or missing. Not all patients require a full laboratory workup before surgery, and a missing value may be an important prognostic indicator for perioperative morbidity and mortality. The patient was identified as having a history of cancer if at least one of the following criteria was present: (1) a history of disseminated cancer, (2) chemotherapy for malignancy within 30 days before procedure, (3) radiotherapy of malignancy within 90 days before procedure, or (4) a final International Classification of Diseases, Ninth Revision (ICD-9, Centers for Disease Control and Prevention, Hyattsville, MD) diagnosis code for neoplasm (ICD-9 range 140–239). Data on other comorbidities and patient characteristics were collected directly from the data set.

We obtained data on total operative time and perioperative blood transfusions. The ACS-NSQIP made changes to the reporting of blood transfusions in 2010. For patients in the 2005 to 2009 data set, any intraoperative transfusion or the transfusion of >4 units in the first 72 hours after surgery are recorded as a perioperative transfusion. For 2010 to 2011, any intra- or postoperative transfusion (up to 72 hours after surgery) was recorded as a perioperative transfusion. As a consequence, a patient from the 2005 to 2009 data set who did not receive an intraoperative blood transfusion but received 1 to4 units in the first 72 hours after surgery will be misclassified as having received no transfusion when they actually were transfused.

Back to Top | Article Outline

Clinical End Points

The follow-up period for patients participating in the ACS-NSQIP is 30 days after their operation. Our outcomes of interest were the development of AKI, sepsis, or acute respiratory failure (Appendix 2). AKI was defined as having (1) progressive renal insufficiency, defined as a rise in the creatinine level >2 mg/dL above the preoperative value, and/or (2) the need for dialysis in a patient who did not require dialysis before the operation. Sepsis was defined as the development of sepsis or septic shock. Acute respiratory failure was defined as (1) mechanical ventilation for a period of >48 hours after the procedure, and/or (2) unplanned tracheal intubation. The postoperative day in which each outcome was first diagnosed was also determined. Data on other clinical outcomes, such as postoperative myocardial infarction (MI), cardiac arrest, pulmonary embolus (PE), pneumonia, and 30-day mortality, were also collected from the data set.

Back to Top | Article Outline

Statistical Analysis

The differences in preoperative patient characteristics and comorbidities between the complication groups were compared with a t test for continuous variables and the χ2 test for categorical variables. Poisson regression modeling with robust variance14 was performed to assess the relative risk of mortality among the different clinical outcomes groups using risk ratios (RRs). Cox proportional hazard modeling15 was used to calculate hazard ratios (HRs) as this technique allows for the modeling of postoperative complications as time-dependent variables, which account for the postoperative day in which each complication occurred.16 Continuous variables were assessed to determine how they would optimally be entered into the Cox regression model, with a separate analysis for each. The variables were grouped into deciles and plotted against the log (HR) to visualize the relationships. We determined that age, body mass index, hematocrit, and log (operative time) would be entered as continuous variables in the Cox model. For multivariable analysis, we separately evaluated each potential variable in a Cox model that included AKI, acute respiratory failure, and sepsis. Because of the large sample size, only variables with P < 0.0001 were retained in the multivariable model. Finally, variables with P > 0.01 in the multivariable model were also excluded.

Additive interactions between our main complications of interest (AKI, sepsis, and acute respiratory failure) were assessed using the relative excess risk due to interaction (RERI).17 Methods for assessing additive interaction between 2 factors are well described. In addition, we developed a formula for assessing the RERI among 3 factorsd:

CV

CV

Confidence intervals (CIs) for the 3-way interactions were determined using simulation methods.18 Repeated estimates of RERIabc were made with each element in the formula (RRabc, RRab, etc.) sampled assuming a normal distribution based on the RRs and variances estimated from the Cox regression. The resulting sample distributions were used to determine the 95% CI and P value.19 The CIs and P values were stable when using 1000, 10,000, 100,000, and 1,000,000 estimates.

The effects of missing data were assessed with multiple imputation using a sequential regression strategy,20 with IVEware software (University of Michigan, Ann Arbor, MI). Complete case analyses were also performed. Statistical analyses were performed using SAS Software version 9.4 (SAS Institute, Cary, NC) and R software version 3.1.0 (The R Foundation for Statistical Computing, Vienna, Austria). In all analyses, statistical significance was determined with a P value <0.05.

Back to Top | Article Outline

RESULTS

Baseline Characteristics of Patients by Postoperative Complication

Table 1

Table 1

Among the 217,994 patients undergoing high-risk intraabdominal general surgery in our analysis, AKI, acute respiratory failure, and sepsis occurred in 2751 (1.3%), 7956 (3.7%), and 14,723 (6.8%) patients, respectively. Ninety-one percent (198,533 patients) did not develop any of these complications. Those who developed AKI, sepsis, or acute respiratory failure had higher rates of comorbidities and risk factors than those not developing these complications, including older age, higher ASA physical status classification, higher rates of emergency procedures, and decreased preoperative renal function (Table 1).

Back to Top | Article Outline

AKI, Acute Respiratory Failure, Sepsis, and Mortality in the Postoperative Period

The number of patients diagnosed per day with AKI and acute respiratory failure increased from postoperative day 0 to day 2 and steadily declined on subsequent days (Fig. 1, A and B). However, the number of patients diagnosed per day with sepsis remained steadily high in the first 7 postoperative days before slowly declining on subsequent days (Fig. 1C). The number of patients who died per day in the first 30 days after surgery was high in the first 15 postoperative days, with a slight decline in the next 15 postoperative days (Fig. 1D).

Figure 1

Figure 1

The overall risk (95% CI) of 30-day mortality in our sample was 1.93% (1.88%–1.99%). The risks of 30-day mortality among patients developing sepsis, acute respiratory failure, and AKI were 11.4% (10.8%–11.9%), 24.1% (23.2%–25.1%), and 25.1% (23.5%, 26.8%), respectively. In contrast, the 30-day mortality risk among those not developing any of these complications was 0.85% (0.81%–0.90%). The 30-day mortality risk increased with the number of complications developing in a given patient. Among those developing only one complication, the 30-day mortality risk ranged from 4.2% (3.8%–4.6%) with sepsis to 18.8% (17.5%–20.2%) with acute respiratory failure (Table 2). Among those with 2 complications, the 30-day mortality risk ranged from 21.2% (17.5%–25.3%) with AKI and sepsis to 46.1% (40.2%–52.1%) with AKI and acute respiratory failure. The 30-day mortality risk of those developing all 3 complications was 42.0% (38.8%–45.4%).

Table 2

Table 2

On a relative basis, the crude RRs for 30-day mortality with one complication compared with no complications ranged from 4.91 (4.42–5.46) with sepsis to 22.0 (20.2–24.0) with acute respiratory failure (Table 2). Among those with 2 complications, the RRs ranged from 24.8 (20.6–29.8) with AKI and sepsis to 54.0 (47.2–61.8) with AKI and acute respiratory failure. The RR with all 3 complications was 49.2 (45.0–53.9).

Back to Top | Article Outline

Time-Dependent Effects of AKI, Sepsis, and Acute Respiratory Failure on 30-Day Mortality

The effects of AKI, sepsis, and acute respiratory failure on 30-day mortality were evaluated using a Cox model to account for the specific postoperative day on which mortality occurred.15 In addition, the Cox model allows for postoperative complications to be modeled as time-dependent covariates that account for the postoperative day that the complication occurs, since complications occur at varying time points for each individual.16 Compared with those who did not develop these complications, in unadjusted analyses, the HRs (95% CIs) with only sepsis, AKI, or acute respiratory failure were 8.82 (7.91–9.84), 16.8 (13.5–20.8), and 27.4 (25.0–30.0), respectively. With 2 complications, the HRs (95% CIs) with acute respiratory failure and sepsis, AKI and sepsis, and acute respiratory failure and AKI were 52.3 (47.9–57.1), 64.3 (52.1–79.2), and 107 (89.5–128), respectively. With all 3 complications, the HR (95% CI) was 158 (141–177).

After adjusting for comorbidities and intraoperative variables, the HRs with one complication, compared with no complications, ranged from 7.24 (6.46–8.11) with sepsis to 14.2 (12.8–15.7) with acute respiratory failure (Table 3, left column). With 2 complications, the HRs ranged from 30.8 (28.0–33.9) with acute respiratory failure and sepsis to 65.2 (53.9–78.8) with AKI and acute respiratory failure. With all 3 complications, the HR was 105 (92.8–118).

Table 3

Table 3

We also conducted analyses incorporating other important complications (MI, PE, cardiac arrest, stroke, and pneumonia) as time-dependent variables to account for the effects of these events (Table 3, right column). In this model, the HRs with one complication, compared with no complications, ranged from 6.73 (6.00–7.55) with sepsis to 9.23 (7.32–11.6) with AKI. With 2 complications, the HRs ranged from 17.2 (15.5–19.2) with acute respiratory failure and sepsis to 34.5 (27.8–42.9) with AKI and sepsis. With all 3 complications, the HR was 38.7 (33.6–44.6).

In multivariable analyses, 6.5% of patients had missing data in one or more variables in the models. We performed both multiple imputation of missing data and complete case analysis and determined that the missing data did not meaningfully change the results of our analysis.

Back to Top | Article Outline

Additive Interaction of AKI, Acute Respiratory Failure, and Sepsis on 30-Day Mortality

On the basis of the Cox model for 30-day mortality, we calculated 2-way and 3-way RERIs to assess the additive effects of AKI, sepsis, and acute respiratory failure. After adjusting for comorbidities and intraoperative variables, the 2-way RERIs were all significantly positive, ranging from 10.4 (7.66–13.1) with acute respiratory failure and sepsis to 41.2 (28.9–53.4) for AKI and acute respiratory failure (Table 4, left column). The 3-way RERI for acute respiratory failure, sepsis, and AKI was not statistically significant.

Table 4

Table 4

After further adjusting for other postoperative complications (MI, PE, cardiac arrest, stroke, and pneumonia), the 2-way RERIs remained significantly positive, ranging from 3.78 (2.11–5.45) with acute respiratory failure and sepsis, 9.88 (4.64–15.1) with AKI and acute respiratory failure, and 19.6 (11.9–27.2) with AKI and sepsis (Table 4, right column). The 3-way RERI was significantly negative with a value of −16.4 (−27.8 to −5.49). This indicates that although a third process increased the risk of mortality, this increase was lower than expected, considering the independent effects of each complication alone and the joint effects of 2 complications.

Back to Top | Article Outline

Additive Interactions for AKI, Acute Respiratory Failure, and Sepsis

Table 5

Table 5

We fit 3 separate adjusted Cox models for each inflammatory process as the outcome to determine the relationship between them. The development of AKI or acute respiratory failure was associated with a significantly increased risk of developing sepsis (Table 5, left column). In addition, AKI and acute respiratory failure had a positive additive interaction as demonstrated by a significantly positive RERI. Similarly, sepsis and acute respiratory failure had additive interactive effects on the risk of AKI (Table 5, middle column) while sepsis and AKI had additive interactive effects on the risk of acute respiratory failure (Table 5, right column).

Back to Top | Article Outline

DISCUSSION

We examined the epidemiologic relationships between 30-day mortality and AKI, acute respiratory failure, and sepsis in a high-risk cohort of patients undergoing intraabdominal general surgery procedures. Although the development of a single complication was associated with an increased risk of mortality, the addition of a second complication demonstrated significant positive additive effects, providing evidence for interactions between these complications. Interestingly, the addition of a third complication when 2 were already present did not demonstrate further positive additive effects. Thus, we show in high-risk general surgery patients that while AKI, acute respiratory failure, and sepsis are each independently associated with an increased risk of short-term perioperative mortality, the greatest marginal effects of these complications on mortality may occur when a second process develops in addition to an initial process.

AKI, acute respiratory failure, and sepsis are considered to be separate processes with similarly high mortality rates in critically ill patients.1 When evaluating the relationships between these processes on adverse outcomes, a typical approach is to select a sample of patients developing a primary process (e.g., sepsis) and then to describe the increased risk of adverse outcomes conferred by secondary processes (e.g., acute respiratory failure or AKI). For instance, in ICU patients with sepsis, more than half of mortality was associated with multiorgan dysfunction and respiratory failure.21 This approach is able to demonstrate associations between disease processes but does not specifically address whether there are interactions between them. Here, we examine the independent, as well as joint, effects of all 3 processes without restricting our analysis to an index population of patients with a specific disease process. We also examined the relationships between the complications on one another. Each individual process independently increased the risk of developing the other 2, and the joint effects of 2 complications demonstrated positive additive effects that increased the risk of developing the third complication.

A major strength of our analysis involves the use of time-dependent variables in the Cox model, which reflects the dynamic nature of perioperative complications and accounts for the timing of each complication by assessing the specific combination of complications present for each patient at each postoperative day. For instance, a patient developing 2 complications at different times will be analyzed in 3 groups: (1) the group with no complications until the first complication develops, (2) the group with only the first complication until the second one develops, and (3) the group with 2 complications until the patient either dies or is censored at the end of the follow-up period. Therefore, the risk of mortality for a given combination of complications will only be assessed at times when patients have actually experienced the complications. In a Cox model without time-dependent variables, the risk of mortality in those that developed AKI and acute respiratory failure was the same as the risk of mortality in those that developed all 3 complications of interest, similar to the relative risk regression. Only after incorporating time-dependent variables did we observe the more intuitive finding that, at any point in time, having 3 complications is significantly worse than having any combination of 2 complications.

One nuance that this approach does not address is the effect of the order in which complications develop. We determined that, on average, acute respiratory failure occurred the earliest, followed by sepsis, then AKI. Of those developing 2 or more complications, mortality risk was higher when 2 complications occurred on the same postoperative day (data not shown). In addition, AKI that followed acute respiratory failure or sepsis had a higher mortality than AKI that preceded these complications. There was no such relationship between acute respiratory failure and sepsis.

Another strength of our analysis was the use of the RERI22 to demonstrate additive interactions. This statistical tool measures additive interactions in models that are inherently multiplicative in nature,17 such as the Cox model. We found no evidence for multiplicative interactions in our model, and the additive interactions would not have been detected without the use of the RERI. From a statistical perspective, additive and multiplicative interactions differ in the scale in which they are measured. While there was some historical debate as to the proper scale for assessing interactions,23 in Rothman’s sufficient-cause model, the presence of additive interactions implies the presence of clinical interactions,22,24 and synergies are best interpreted for individual decision making on an additive scale.23 Hence, our statistical findings suggest that there may be synergisms among AKI, acute respiratory failure, and sepsis that increase the risk of perioperative mortality.

Our study, however, is epidemiologic in nature and does not provide direct evidence for the exact mechanisms underlying these synergisms. These interactions are a result of complex processes occurring in the perioperative period and likely result from a multitude of factors. Potential biological mechanisms may include crosstalk between the kidneys and the lungs,25,26 as well as bidirectional relationships between sepsis and both AKI and acute respiratory failure.4,27–29 These effects may be mediated by circulating cytokines and chemokines, such as tumor necrosis factor-α and interleukin 6, activation of leukocytes and the adaptive immune response, vascular effects, apoptosis, dysregulation of water clearance, and oxidative stress.4,25–29 Clinically, the treatment for injury to one organ system may lead to deleterious consequences in others, such as aggressive fluid management30 or mechanical ventilation,31 leading to intraabdominal hypertension and AKI. Social factors also likely play a role as a patient and/or family may be more likely to withdraw care as the clinical condition deteriorates. Our analysis captures the amalgam of these and other factors and demonstrates that, regardless of the underlying mechanisms, interactions exist among AKI, acute respiratory failure, and sepsis to synergistically increase the risk of mortality.

It might be expected that every additional complication would synergistically increase mortality risk, but we found that, while any 2 complications had significant additive interaction effects, there were no such interaction effects with all 3 complications. This finding does not imply that the mortality with 3 complications was lower than the risk with 2 complications. Instead, the findings suggest that when compared with a single complication, the increase in mortality associated with the development of 2 complications was so great that, by the time a third complication developed, much of the damage had already been done. Further studies may clarify the specific reasons for the lack of an interaction effect with 3 complications.

In the perioperative setting, the type of surgery affects the likelihood of perioperative sepsis and multiorgan dysfunction. Our study focuses on intraabdominal general surgery, as this group accounts for over 4 million procedures per year in the United States.32 Among these, esophageal, pancreatic, and gastric procedures have the highest risk of developing postoperative sepsis, although not the highest sepsis-related mortality.33 Perioperative AKI34 and acute respiratory failure35 have been studied mainly in high-risk populations, including cardiac and major vascular surgery, but we recently demonstrated that certain intraabdominal general surgery procedures also have a high risk of AKI.36 In addition, the effects of general anesthesia and mechanical ventilation in general surgery patients, especially with regard to tidal volumes and acute respiratory distress syndrome, are increasingly being studied.37

Cardiac arrest occurred in 0.5% of our sample, a rate similar to that observed in all surgical patients.38 Although cardiac arrest was less common than AKI, acute respiratory failure, or sepsis, it had a much higher mortality of 74%. Perioperative MIs also occurred in 0.5% of patients and had a mortality of 21%, similar to that for AKI and acute respiratory failure. Thus, cardiac events overall were less frequent than the complications we studied but were associated with a higher mortality. When we accounted for other complications, including MI, PE, cardiac arrest, stroke, and pneumonia, the magnitude of the additive effects of AKI, acute respiratory failure, and sepsis on 30-day mortality was diminished (Table 3). Cardiac arrest is likely to be part of the pathway (i.e., mediator) through which a patient dies of AKI, acute respiratory failure, or sepsis. Indeed, among those who died within 30 days of surgery, cardiac arrest was involved in 11% of patients who did not develop AKI, acute respiratory failure, or sepsis versus 27% in those that developed at least one.

Our assessment of AKI, acute respiratory failure, and sepsis in the perioperative period is sensitive to how these complications are defined, but we are limited to the data provided by ACS-NSQIP. In addition, the data set does not identify the etiology of each complication. The currently used consensus criteria for sepsis and septic shock were last updated in 2001,39 and the ACS-NSQIP definitions for sepsis and septic shock are closely aligned with these criteria (Appendix 2). For AKI, the latest Kidney Disease: Improving Global Outcomes (KDIGO) guidelines require an increase in the serum creatinine of at least 0.3 mg/dL or 50%40 while ACS-NSQIP identifies a minimum increase in the serum creatinine of 2 mg/dL. Thus, the ACS-NSQIP data set identifies only severe AKI and underestimates the impact of mild-to-moderate AKI.41 Perioperative AKI can have many causes, including hypotension42 and nephrotoxic agents,43 but we are not able to determine the specific cause of AKI in each patient. Finally, acute respiratory failure in this study refers to the major pulmonary outcomes available in ACS-NSQIP: unplanned tracheal intubation and prolonged mechanical ventilation. However, current definitions of lung injury, now termed acute respiratory distress syndrome, refer to the degree of hypoxemia due to intrinsic lung disease,44 but not all patients requiring mechanical ventilation have intrinsic lung disease. Our study could not differentiate between patients tracheally intubated as a consequence of AKI or sepsis from those who developed an independent respiratory complication. Nevertheless, our findings—that the addition of acute respiratory failure to AKI or sepsis dramatically increased mortality—remain relevant to clinicians seeking prognostic or therapeutic guidance.

Other limitations of our analysis include the fact that the data set does not provide data on other organ systems that might have affected mortality, such as the gut45 or liver.46 In addition, there are certain limitations related to retrospective analysis of surgical data sets.36 Many important variables that affect the risk of perioperative AKI, acute respiratory failure, and sepsis are not collected by the data set, including intraoperative management and hemodynamics, whether the patient received postoperative ICU care, and whether the patient had a do-not-resuscitate order initiated following surgery. Finally, we are not able to determine whether complications we studied and mortality were a result of the surgical disease state or through other mechanisms such as a lack of adequate care.

In conclusion, we present epidemiologic evidence that perioperative AKI, acute respiratory failure, and sepsis each independently increase the risk of 30-day mortality in patients undergoing high-risk intraabdominal general surgery procedures. In addition, the development of 2 complications demonstrate positive additive interactions, implying that these complications interact synergistically to further increase the risk of mortality above and beyond what would be expected with each complication alone. Our observations underscore the importance of understanding the full clinical implications of altered organ system function and the recognition that these complications may have far-reaching effects when combined with insults to other systems. Further research will be required to determine the complex mechanisms behind these synergisms and to identify the best therapeutic targets for the prevention of further sequelae after these injuries develop.

Back to Top | Article Outline

Appendix 1. Clinical Classifications Software for Services and Procedures Categories for Intraabdominal General Surgery

Table

Table

Back to Top | Article Outline

Appendix 2. Definitions for Exclusions and Outcomes, American College of Surgeons-National Surgical Quality Improvement Program, 2005–2011

Table

Table

Back to Top | Article Outline

DISCLOSURES

Name: Minjae Kim, MD, MS.

Contribution: This author was involved in study design, conduct of study, data analysis, and manuscript preparation.

Attestation: Minjae Kim approved the final manuscript and attests to the integrity of the original data and analysis reported in this manuscript and is the archival author.

Name: Joanne E. Brady, PhD.

Contribution: This author was involved in study design, data analysis, and manuscript preparation.

Attestation: Joanne E. Brady approved the final manuscript and attests to the integrity of the original data and analysis reported in this manuscript.

Name: Guohua Li, MD, DrPH.

Contribution: This author was involved in study design, data analysis, and manuscript preparation.

Attestation: Guohua Li approved the final manuscript and attests to the integrity of the original data and analysis reported in this manuscript.

This manuscript was handled by: Avery Tung, MD.

Back to Top | Article Outline

ACKNOWLEDGMENTS

The authors gratefully acknowledge Tomas Andersson for his valuable assistance in the analysis of additive interaction.

Back to Top | Article Outline

FOOTNOTES

a The American College of Surgeons-National Surgical Quality Improvement Program and the hospitals participating in the ACS-NSQIP are the source of the data used herein; they have not verified and are not responsible for the statistical validity of the data analysis or the conclusions derived by the authors.
Cited Here...

b Available at: http://www.hcup-us.ahrq.gov/toolssoftware/ccs_svcsproc/ccssvcproc.jsp. Accessed June 15, 2012.
Cited Here...

c ACS-NSQIP: User Guide for the 2011 Participant Use Data File, October 2012. Available at: http://site.acsnsqip.org/wp-content/uploads/2012/03/2011-User-Guide_Final.pdf. Accessed December 3, 2012.
Cited Here...

d Private communication with Tomas Andersson.
Cited Here...

Back to Top | Article Outline

REFERENCES

1. Vincent JL. Acute kidney injury, acute lung injury and septic shock: how does mortality compare? Contrib Nephrol. 2011;174:71–7
2. Ko GJ, Rabb H, Hassoun HT. Kidney-lung crosstalk in the critically ill patient. Blood Purif. 2009;28:75–83
3. Vieira JM Jr, Castro I, Curvello-Neto A, Demarzo S, Caruso P, Pastore L Jr, Imanishe MH, Abdulkader RC, Deheinzelin D. Effect of acute kidney injury on weaning from mechanical ventilation in critically ill patients. Crit Care Med. 2007;35:184–91
4. Matejovic M, Chvojka J, Radej J, Ledvinova L, Karvunidis T, Krouzecky A, Novak I. Sepsis and acute kidney injury are bidirectional. Contrib Nephrol. 2011;174:78–88
5. Kheterpal S, Tremper KK, Englesbe MJ, O’Reilly M, Shanks AM, Fetterman DM, Rosenberg AL, Swartz RD. Predictors of postoperative acute renal failure after noncardiac surgery in patients with previously normal renal function. Anesthesiology. 2007;107:892–902
6. Arozullah AM, Daley J, Henderson WG, Khuri SF. Multifactorial risk index for predicting postoperative respiratory failure in men after major noncardiac surgery. The National Veterans Administration Surgical Quality Improvement Program. Ann Surg. 2000;232:242–53
7. Johnson RG, Arozullah AM, Neumayer L, Henderson WG, Hosokawa P, Khuri SF. Multivariable predictors of postoperative respiratory failure after general and vascular surgery: results from the patient safety in surgery study. J Am Coll Surg. 2007;204:1188–98
8. Moore LJ, Moore FA, Todd SR, Jones SL, Turner KL, Bass BL. Sepsis in general surgery: the 2005-2007 national surgical quality improvement program perspective. Arch Surg. 2010;145:695–700
9. Wakeam E, Hyder JA, Jiang W, Lipsitz SA, Finlayson S. Risk and patterns of secondary complications in surgical inpatients. JAMA Surg. 2015;150:65–73
10. Fink AS, Campbell DA Jr, Mentzer RM Jr, Henderson WG, Daley J, Bannister J, Hur K, Khuri SF. The National Surgical Quality Improvement Program in non-veterans administration hospitals: initial demonstration of feasibility. Ann Surg. 2002;236:344–53
11. Hua M, Brady J, Li G. The epidemiology of upper airway injury in patients undergoing major surgical procedures. Anesth Analg. 2012;114:148–51
12. Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF III, Feldman HI, Kusek JW, Eggers P, Van Lente F, Greene T, Coresh JCKD-EPI (Chronic Kidney Disease Epidemiology Collaboration). . A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009;150:604–12
13. National Kidney Foundation. . K/DOQI clinical practice guidelines for chronic kidney disease: evaluation, classification, and stratification. Am J Kidney Dis. 2002;39:S1–266
14. Spiegelman D, Hertzmark E. Easy SAS calculations for risk or prevalence ratios and differences. Am J Epidemiol. 2005;162:199–200
15. Cox DR. Regression Models and Life-Tables. J Roy Statist Soc Ser B. 1972;34:187–220
16. Allison PD Survival Analysis Using the SAS System: A Practical Guide. 1995 Cary, NC SAS Institute
17. Andersson T, Alfredsson L, Källberg H, Zdravkovic S, Ahlbom A. Calculating measures of biological interaction. Eur J Epidemiol. 2005;20:575–9
18. Hallgren KA. Conducting simulation studies in the R programming environment. Tutor Quant Methods Psychol. 2013;9:43–60
19. North BV, Curtis D, Sham PC. A note on the calculation of empirical P values from Monte Carlo procedures. Am J Hum Genet. 2002;71:439–41
20. Raghunathan TE, Lepkowski JM, Van Hoewyk J, Solenberger P. A multivariate technique for multiply imputing missing values using a sequence of regression models. Survey Method. 2001;27:85–95
21. Vincent JL, Nelson DR, Williams MD. Is worsening multiple organ failure the cause of death in patients with severe sepsis? Crit Care Med. 2011;39:1050–5
22. Rothman KJ, Greenland S, Lash TL Modern Epidemiology. 20083rd ed. Philadelphia, PA Wolters Kluwer Health/Lippincott Williams & Wilkins
23. Rothman KJ, Greenland S, Walker AM. Concepts of interaction. Am J Epidemiol. 1980;112:467–70
24. Ahlbom A, Alfredsson L. Interaction: a word with two meanings creates confusion. Eur J Epidemiol. 2005;20:563–4
25. Doi K, Ishizu T, Fujita T, Noiri E. Lung injury following acute kidney injury: kidney-lung crosstalk. Clin Exp Nephrol. 2011;15:464–70
26. Singbartl K. Renal-pulmonary crosstalk. Contrib Nephrol. 2011;174:65–70
27. Bhargava R, Altmann CJ, Andres-Hernando A, Webb RG, Okamura K, Yang Y, Falk S, Schmidt EP, Faubel S. Acute lung injury and acute kidney injury are established by four hours in experimental sepsis and are improved with pre, but not post, sepsis administration of TNF-α antibodies. PLoS One. 2013;8:e79037
28. Pettilä V, Bellomo R. Understanding acute kidney injury in sepsis. Intensive Care Med. 2014;40:1018–20
29. White LE, Chaudhary R, Moore LJ, Moore FA, Hassoun HT. Surgical sepsis and organ crosstalk: the role of the kidney. J Surg Res. 2011;167:306–15
30. Shear W, Rosner MH. Acute kidney dysfunction secondary to the abdominal compartment syndrome. J Nephrol. 2006;19:556–65
31. Matthew D, Oxman D, Djekidel K, Ahmed Z, Sherman M. Abdominal compartment syndrome and acute kidney injury due to excessive auto-positive end-expiratory pressure. Am J Kidney Dis. 2013;61:285–8
32. Rahbari NN, Knebel P, Diener MK, Seidlmayer C, Ridwelski K, Stöltzing H, Seiler CM. Current practice of abdominal wall closure in elective surgery—is there any consensus? BMC Surg. 2009;9:8
33. Vogel TR, Dombrovskiy VY, Carson JL, Graham AM, Lowry SF. Postoperative sepsis in the United States. Ann Surg. 2010;252:1065–71
34. Thakar CV. Perioperative acute kidney injury. Adv Chronic Kidney Dis. 2013;20:67–75
35. Kor DJ, Lingineni RK, Gajic O, Park PK, Blum JM, Hou PC, Hoth JJ, Anderson HL III, Bajwa EK, Bartz RR, Adesanya A, Festic E, Gong MN, Carter RE, Talmor DS. Predicting risk of postoperative lung injury in high-risk surgical patients: a multicenter cohort study. Anesthesiology. 2014;120:1168–81
36. Kim M, Brady JE, Li G. Variations in the risk of acute kidney injury across intraabdominal surgery procedures. Anesth Analg. 2014;119:1121–32
37. Futier E, Jaber S. Lung-protective ventilation in abdominal surgery. Curr Opin Crit Care. 2014;20:426–30
38. Kazaure HS, Roman SA, Rosenthal RA, Sosa JA. Cardiac arrest among surgical patients: an analysis of incidence, patient characteristics, and outcomes in ACS-NSQIP. JAMA Surg. 2013;148:14–21
39. Levy MM, Fink MP, Marshall JC, Abraham E, Angus D, Cook D, Cohen J, Opal SM, Vincent JL, Ramsay GSCCM/ESICM/ACCP/ATS/SIS. SCCM/ESICM/ACCP/ATS/SIS. . 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Crit Care Med. 2003;31:1250–6
40. National Kidney Foundation: Kidney Disease: Improving Global Outcomes (KDIGO) Acute Kidney Injury Work Group. . KDIGO clinical practice guideline for acute kidney injury. Kidney Int Suppl. 2012;2:19–36
41. Bihorac A, Brennan M, Ozrazgat-Baslanti T, Bozorgmehri S, Efron PA, Moore FA, Segal MS, Hobson CE. National surgical quality improvement program underestimates the risk associated with mild and moderate postoperative acute kidney injury. Crit Care Med. 2013;41:2570–83
42. Walsh M, Devereaux PJ, Garg AX, Kurz A, Turan A, Rodseth RN, Cywinski J, Thabane L, Sessler DI. Relationship between intraoperative mean arterial pressure and clinical outcomes after noncardiac surgery: toward an empirical definition of hypotension. Anesthesiology. 2013;119:507–15
43. Liangos O. Drugs and AKI. Minerva Urol Nefrol. 2012;64:51–62
44. Ranieri VM, Rubenfeld GD, Thompson BT, Ferguson ND, Caldwell E, Fan E, Camporota L, Slutsky AS. Acute respiratory distress syndrome: the Berlin definition. JAMA. 2012;307:2526–33
45. Clark JA, Coopersmith CM. Intestinal crosstalk: a new paradigm for understanding the gut as the “motor” of critical illness. Shock. 2007;28:384–93
46. Marshall JC. New translational research provides insights into liver dysfunction in sepsis. PLoS Med. 2012;9:e1001341
© 2015 International Anesthesia Research Society