|Home | About | Journals | Submit | Contact Us | Français|
Each year, as many as two million operations are complicated by surgical site infections in the United States, and surgical patients account for 30% of patients with sepsis. The purpose of this study was to determine recent trends in sepsis incidence, severity, and mortality rate after surgical procedures and to evaluate changes in the pattern of septicemia pathogens over time.
Analysis of the 1990–2006 hospital discharge data from the Healthcare Cost and Utilization Project (HCUP) State Inpatient Databases (SID) for New Jersey. Patients ≥18 years who developed sepsis after surgery were identified using International Classification of Diseases, Ninth Revision, Clinical Modification codes as defined by the Patient Safety Indicator “Postoperative Sepsis” developed by the Agency for Healthcare Research and Quality (AHRQ). Severe sepsis was defined as sepsis complicated by organ dysfunction.
A total of 1,276,451 surgery discharges (537,843 elective [42.1%] and 738,608 non-elective [57.9%] procedures) were identified. After elective surgery, 5,865 patients (1.09%) developed postoperative sepsis, of whom 2,778 (0.52%) had severe sepsis. The incidence of postoperative sepsis after elective surgery increased from 0.67% to 1.74% (p<0.0001) and severe sepsis after elective surgery from 0.22% to 1.12% (p<0.0001). The sepsis mortality rate for elective procedures showed no significant change over time. The proportion of severe sepsis after elective cases increased from 32.9% to 64.6% (p<0.0002). The rates of postoperative sepsis (4.24%) and severe sepsis (2.28%) were significantly greater for non-elective than for elective procedures (p<0.0002). Non-elective surgical procedures had a significant increase in the rates of postoperative sepsis (3.74% to 4.51%) and severe sepsis (1.79% to 3.15%) over time (p<0.0001) with the proportion of severe sepsis increasing from 47.7% to 69.9% (p<0.0002). The in-hospital mortality rate after non-elective surgery decreased from 37.9% to 29.8% (p<0.0001).
Sepsis and death were more likely after non-elective than elective surgery. Sepsis and severe sepsis has increased significantly after elective and non-elective procedures over the last 17 years. The hospital mortality rate was reduced significantly after non-elective surgery, but no improvements were found for elective surgery patients who developed sepsis. Disparities in age, sex, and ethnicity and the development of postoperative surgical sepsis were found. Population-based studies may assist in defining temporal trends, disparities, and outcomes in sepsis not elucidated in smaller studies.
More than 40 million major surgical operations are performed annually in the United States of which 800,000 to two million are complicated by surgical site infections [1–3]. Sepsis is an extreme manifestation of the infectious process that is associated with increased resource utilization and poor outcome. Surgery patients can be defined as a high-risk group for developing sepsis, as procedures evoke substantial metabolic, hematologic, and immunologic responses . The immune function after surgery may be transformed substantially by perioperative bacterial infection, hemorrhage, blood transfusion, or anesthesia up to severe suppression that promotes sepsis [4–8].
In the United States, the incidence of sepsis during the last decades has increased considerably and has been accompanied by a significant increase in the disease severity [9–11]. Despite the reduction in the disease case fatality rate over time, sepsis remains one of the leading causes of death in the United States . Surgical patients account for approximately 30% of all sepsis patients , and epidemiologic investigation of sepsis incidence and death after surgical procedures may help to define patient populations at greatest risk and characterize temporal changes in sepsis and outcomes. Most of the current data on post-surgical sepsis are based on analysis from single institutions and small geographic areas, which makes their generalizability difficult [14–20]. We evaluated trends in incidence, severity, mortality rate, and associated pathogens of sepsis after surgical procedures on a population level.
Data for this study were collected from the New Jersey State Inpatient Databases (SID) from 1990 to 2006. The SID is a publicly available database, developed as a part of the Healthcare Cost and Utilization Project (HCUP) and sponsored by the Agency for Healthcare Research and Quality (AHRQ) . The New Jersey SID includes inpatient discharge abstracts from the acute care community hospitals in the state that are collected annually by the state Department of Health and Senior Services. They contain clinical and non-clinical information on hospitalized patients with all types of insurance and the uninsured, and cover annually more than one million hospitalizations with several hundred characteristics.
We selected patients 18 years and older who were admitted to the hospital for surgical procedures. To identify surgical patients, we used all International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) operating room procedure codes and surgical discharge Diagnosis-Related Groups (DRGs).
To identify surgical patients developing sepsis, the Patient Safety Indicator “Postoperative Sepsis” (PSI 13) was utilized . Patient Safety Indicator 13 was designed to identify sepsis as a complication. The AHRQ and the Stanford-University of California-San Francisco Evidence-based Practice Center have developed Patient Safety Indicators (PSIs). These indicators were developed after a comprehensive literature review, analysis of ICD-9-CM codes, review by a clinician panel, implementation of risk adjustment, and empirical analyses. The PSIs focus on potentially preventable instances of complications and other iatrogenic events during hospitalization. They are tools to identify potential adverse events and are screening tools for highlighting areas in which quality should be investigated and for case finding and tracking and monitoring patient safety events. Among 20 hospital-level indicators, eight are related to surgical discharges. One of these indicators, PSI 13 was utilized to identify sepsis as a complication. Patient Safety Indicators are particularly applicable to surgical patients, as they are more homogeneous than medical patients, making it easier to account for case mix.
Observations with postoperative sepsis had the following ICD-9-CM diagnosis codes in any secondary position (suggesting a complication from the surgical procedure) for the diagnosis codes: Streptococcal septicemia (038.0), staphylococcal septicemia unspecified (038.10), Staphylococcus aureus septicemia (038.11), Other staphylococcal septicemia (038.19), pneumococcal septicemia (038.2), septicemia due to anaerobes (038.3), gram-negative organism unspecified (038.40), Haemophilus influenzae (038.41), Escherichia coli (038.42), Pseudomonas (038.43), Serratia (038.44), septicemia due to other gram-negative organisms (038.49), other specified septicemias (038.8), unspecified septicemia (038.9), septic shock (785.52), other shock without mention of trauma (785.59), systemic inflammatory response syndrome due to infectious process without organ dysfunction (995.91), or systemic inflammatory response syndrome due to infectious process with organ dysfunction (995.92).
To identify major organ dysfunction, the following ICD-9-CM diagnosis codes in any secondary position for the diagnosis code were used: cardiovascular failure (785.50, 785.51, 785.59, 458.0, 458.8, 458.9, 796.3, 427.5), respiratory failure (518.81, 518.82, 786.09, 799.1), acute renal failure (584.5, 584.6, 584.7, 584.8, 584.9), acute hepatic failure (570, 572.2, 573.4), coagulation failure (286.6, 286.9, 287.4, 287.5), and central nervous system failure (293.0, 348.1, 348.3, 780.01).
In accordance with the PSI 13 requirements, we excluded from the analysis all patients with preexisting sepsis or infection, with any ICD-9-CM code for immunocompromised state or cancer, with the Major Diagnostic Categories (MDC) code 14 (Pregnancy and Childbirth), and with hospital length of stay of less than four days. These restrictions increase the chance that a person developed sepsis as a result of the surgical procedure. In addition, among all septic patients, we selected a subgroup of those with severe sepsis defined as sepsis complicated by organ dysfunction . Classification of an operation as elective or non-elective was based on the HCUP variable for admission type.
Data were analyzed with SAS 9.1 software (SAS Institute, Cary, NC). We used demographic characteristics of patients (age, sex, and race), principal and eight secondary diagnoses, principal and seven secondary procedures, admission source and type, disposition of patient at discharge, and calendar year. The rate of postoperative sepsis was defined as the number of postoperative sepsis cases divided by the number of patients with surgical procedures, expressed as a percentage. To evaluate the severity of the disease, we calculated the percentage of severe sepsis cases among all sepsis cases. The rate of hospital mortality for postoperative sepsis was calculated as the number of fatal surgical sepsis cases in the hospitals divided by the total number of hospitalized surgical patients with sepsis expressed as a percentage. To test the difference in estimates between various groups we used the Student t-test for continuous variables and chi-square analysis for categorical variables. The z ratio with p value for the difference between two independent proportions was utilized to compare two groups with results presented as percentages. The Cochran-Armitage trend test was employed initially to analyze changes in the rates of postoperative sepsis and its mortality rate from 1990–2006. Taking into account the changes in the age structure of population in the country over the study period, we adjusted the rates of sepsis incidence and death with the method of direct standardization and computed direct standardized rate with the standard error (SE) of this rate . The structure of the population of hospitalized surgical patients in the State of New Jersey in 2000 was used as a standard in computing standardized rates of sepsis incidence. Calculating the standardized rates of death, we used for the same purpose a structure of septic surgical patients hospitalized in the year 2000. Trends in age-adjusted rates were evaluated with linear regression analysis. All reported p values are two-sided; p<0.05 was considered significant.
A total of 1,276,451 surgery discharges that met the inclusion and exclusion criteria for our study were identified in the New Jersey acute care hospitals from 1990 to 2006. Among all surgical patients, 42.1% were admitted to the hospital electively and 57.9% non-electively. Demographic and clinical characteristics of patients undergoing surgical procedures of both categories, including those who developed sepsis, are displayed in Table 1. Overall, 2.9% of all surgical procedures were complicated by sepsis. Patients undergoing non-elective procedures developed sepsis more often than those having elective procedures (4.2% and 1.1%, respectively; p<0.0002).
From 1990 to 2006, the rates of postoperative sepsis in New Jersey increased significantly (Table 2; Fig. 1) for both non-elective (from 3.74% to 4.51%; p<0.0001) and elective (from 0.67% to 1.74%; p<0.0001) surgery. The proportion of cases of severe sepsis after non-elective procedures increased from 47.7% to 69.9% (p<0.0002) and almost doubled after elective procedures (from 32.9% to 64.6%; p<0.0002). As well, the rates of severe sepsis increased 1.8 times for non-elective admissions (from 1.79% to 3.15%; p<0.0001) and more than five times (from 0.22% to 1.12%; p<0.0001) for elective admissions (Table 2; Fig. 2).
Evaluating trends between 1990 and 2006, we found that the hospital mortality rate after non-elective surgery for patients developing sepsis was reduced significantly, from 37.9% to 29.8% (p<0.0001) (Table 2, Fig. 3). Evaluation of elective procedures did not show a significant decrease in the postoperative sepsis mortality rate over time (p=0.097). Trend analysis showed a decreasing hospital mortality rate for severe sepsis in both groups (Table 2; Fig. 4): From 55.4% to 38.0% (p<0.0001) after non-elective procedures and from 52.2% to 34.4% (p<0.0001) after elective procedures.
In the total cohort of patients with sepsis, 52.7% developed severe sepsis. The proportion of severe sepsis cases after non-elective procedures (53.7%) was significantly greater than after elective procedures (47.4%; p<0.0002). The systems most affected with organ dysfunction were the respiratory, cardiovascular, and renal. The overall sepsis mortality rate in the research cohort was 31.7%, and after non-elective surgical procedures (32.7%), it was significantly greater than after elective procedures (26.6%; p<0.0002). In the severe sepsis cases, the mortality rate was higher (45.7%; p<0.0002); it was still greater for non-elective procedures (46.2%) than for elective surgery (42.4%; p<0.0002).
There was a significant and steady increase in the rates of postoperative sepsis associated with age. Only 1.75% of patients younger than 50 years developed sepsis, whereas octogenarians had a rate of sepsis of 4.21% (p<0.0002). Interestingly, this trend predominated among patients having elective procedures (3.2-fold increase: From 0.61% to 1.97%; p<0.0002) whereas after non-elective surgical procedures, the increase was less extensive (1.7-fold: From 2.77% to 4.86%; p<0.0002) and in patients 65–79 years of age, the rates of postoperative sepsis did not differ from that in octogenarians.
Male patients were more likely to have postoperative sepsis than female patients (odds ratio [OR]=1.17; 95% confidence interval [CI] 1.15, 1.20), and this difference was greater after elective procedures (OR=1.37; 95% CI 1.30, 1.44) than after non-elective surgery (OR=1.10; 95% CI 1.08, 1.13). Racial disparities in the occurrence of postoperative sepsis were noted. We found the lowest rate of surgical sepsis in white patients (2.77%), whereas black patients had the highest rate (3.82%; p<0.0002). A similar distribution was found after both elective and non-elective procedures.
We compared patterns of the septicemia pathogens in surgical patients with elective and non-elective surgical procedures. Compared with the latter, in the whole elective subgroup, there were lower rates of septicemia attributable to Streptococcus (p=0.0443), Pneumococcus (p=0.0043), and E. coli (p<0.0002), whereas the rates of septicemia caused by Pseudomonas and other specified and unspecified septicemias were increased (p=0.0159 and p<0.0002, respectively). The rates of septicemia attributed to anaerobes and Staphylococcus in the two subgroups were similar (p=NS).
Trend analysis was conducted of the rates of septicemia from 1990 to 2006 according to the particular organism coded. In the subgroup of the elective surgical procedures, a significant increase in the rates of the streptococcal septicemia (p=0.0039) and staphylococcal septicemia (p=0.0041) was discovered. The rates of septicemia caused by E. coli, Pseudomonas, and anaerobes were unchanged over time. Among patients with surgical procedures who were admitted to the hospital non-electively, we found a significant negative trend in the rates of septicemias attributed to staphylococci (p<0.0001), anaerobes (p=0.0451), Pseudomonas (p=0.0298), and E. coli (p<0.0001).
More than 40 million surgical procedures are performed annually in the United States, and sepsis remains a major postoperative complication [1–3,25]. Surgical sepsis accounts for approximately 30% of all sepsis patients . The majority of published data on postoperative sepsis are derived from single institutions. Although these studies are important, they offer limited information about the demography of sepsis or temporal changes. As well, the generalizability of small series may be difficult. The use of population data for postoperative surgical sepsis is not well represented in the literature, although there have been multiple studies evaluating population data for the occurrence of medical sepsis [9–11,26,27].
These data have demonstrated that trends in the age-adjusted rates of postoperative sepsis and the proportion of severe sepsis among all sepsis cases has increased significantly for both elective and non-elective procedures over the 17-year study period. During the study period, the overall hospital mortality rate among surgical sepsis patients with non-elective admission was reduced. Of concern is that elective surgical cases failed to show a decrease in age-adjusted rates of mortality for postoperative sepsis. Previous population data evaluating all cases of sepsis (medical and surgical) have reported that the incidence of sepsis and the number of sepsis-related deaths are increasing, although the overall mortality rate among patients with sepsis is declining .
This population-level study demonstrates an increase in the rate of sepsis for elective surgical procedures over time with no significant improvements in the mortality rate. The proportion of cases of severe sepsis also was found to increase for elective surgery, from 32.9% to 64.6%. This significant increase suggests greater severity of sepsis as well as a higher incidence over time. The reasons for this increase in the rate and severity of sepsis after elective surgery remain unclear, but may reflect changes in the elective operative case mix over time. That is, more elective operations may be performed as outpatient procedures or with short lengths of stay, leaving sicker patients and those having more extensive procedures in the in-patient surgery cohort.
This study also demonstrates differences in the incidence of postoperative sepsis with age. Previous large population studies have looked at global sepsis rates (medical and surgical) and have demonstrated that the incidence of sepsis was disproportionately increased in elderly adults, and age was an independent predictor of death . The aged were more likely to develop sepsis and severe sepsis after surgery and had a higher mortality rate after developing sepsis. Possible reasons for this disparity may be more frequent co-morbidities, institutionalization, declining performance status, and age-associated immunosenescence with defects in immunologic function in the aged [28,29]. Further analysis is needed to address procedures associated with sepsis in the elderly, as the aged are the fastest-growing segment of the U.S. population .
Sex differences in the occurrence of sepsis are suggested by these data. We found that male patients were more likely to have postoperative sepsis. Further focused studies assessing sex disparities will be needed; we were not focused on that subject in this project. Several studies have evaluated the effect of sex and hormone concentrations on sepsis [31,32], suggesting that sex may have a role in the development of sepsis. Others have suggested that sex hormones play a significant role in shaping the host response to trauma [33,34]. Further analyses at a population level may help to identify procedures, co-morbidities, and other factors that influence the likelihood of developing postoperative sepsis.
These data also suggest disparities among ethnicities in the incidence of postoperative sepsis. We have demonstrated that the lowest rates of sepsis were in whites and the highest rates were in blacks. These disparities were seen with similar distributions after elective and non-elective procedures. Although other studies have evaluated the effect of race on medical sepsis , there are few data evaluating ethnicity and its influence on postoperative sepsis. The reasons for these disparities remain unclear, but considerations may be more co-morbidities in blacks, different access to care, or physiological differences yet to be determined. Further, more detailed evaluations of race and postoperative sepsis are required.
With regard to the evaluation of pathogens associated with sepsis, elective surgical procedures demonstrated a significant increase in the rates of the streptococcal and staphylococcal septicemias. This finding is supported by other population studies, which have shown that the nosocomial blood stream infection rate in the hospital has nearly doubled in the past 10 years, largely secondary to an increase in primary staphylococcal bacteremia . As well, the epidemiology of severe surgical site infection (SSI) in community hospitals and the prevalence of methicillin-resistant S. aureus SSI has increased significantly over the past years . The inpatient S. aureus infection rate and the economic burden of S. aureus infections for all U.S. hospitals increased substantially from 1998 to 2003 .
This study has several limitations. The New Jersey State Inpatient Databases do not include patients in military hospitals or Veterans Affairs medical centers. Moreover, the administrative data originally were intended primarily for determination of reimbursement, although there are reports validating the use of administrative data for research purposes [9,39]. In addition, the potential for inclusion bias based on limited coding schemes for the many clinical entities cannot be entirely excluded. For this analysis, we selected codes for systemic infection that were effective and unchanged during the study period, and, therefore, the addition of new codes should not have affected our selection. It is possible that there has been more thorough capture of codes by institutions based on reimbursement over time and greater emphasis on capturing sepsis events. Although the code scheme remained constant throughout the study, there may be upcoding by the institutions, and this cannot be evaluated from the dataset. We assume this change to have been slow over time and unlikely to affect the findings for the severity of sepsis.
We realize that more elective operations may now be performed on an outpatient and 23-hour admission basis, leaving patients who are sicker and are having more complex procedures in the in-hospital surgery cohort. We also acknowledge that there is a trade-off in using administrative data on hundreds of thousands of patients compared with the use of smaller cohorts with more refined clinical information. Although both types of studies have drawbacks and strengths, we believe that administrative data provide valuable population-based information on trends and severity of sepsis.
In conclusion, these population data have shown a significant increase in age-adjusted rates of postoperative sepsis over time. Despite the higher incidence of sepsis, the overall mortality rate improved, perhaps secondary to progress in surgical critical care, greater utilization of surgical intensivists, or alterations in the antibiotics or strategies employed. Of significant concern is the lack of advances in the elective surgery cohort. Elective surgery had the greatest increases in sepsis rates as well as the greatest increases in the proportion of severe sepsis cases. Elective surgery also failed to show a decrease in age-adjusted rates of death for postoperative sepsis over time. Directions for future research using population data for postoperative sepsis may include the analysis of the specific procedures associated with sepsis and the focused evaluation on the various racial and sex disparities in the development of postoperative sepsis. These data may also serve to track the effectiveness of care and function as hypothesis generating to initiate future studies focused on improvement of surgical outcomes.
*Presented at the 28th Annual Meeting of the Surgical Infection Society, Hilton Head Island, South Carolina, May 8, 2008.
No competing financial interests exist.