The bars above and below the dots indicate 95% CI.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Li X, Wilson M, Nylander W, Smith T, Lynn M, Gunnar W. Analysis of Morbidity and Mortality Outcomes in Postoperative Clostridium difficile Infection in the Veterans Health Administration. JAMA Surg. 2016;151(4):314–322. doi:https://doi.org/10.1001/jamasurg.2015.4263
This study analyzes and reports Clostridium difficile infection (CDI) rates, risk factors, and associations with postoperative outcomes in the Veterans Health Administration (VHA).
To report 30-day postoperative CDI rates and outcomes and identify associated risks by surgical procedures and preoperative patient demographics in a large integrated health care system.
Design, Setting, and Participants
In a retrospective observational study conducted from September 2014 to April 2015, the Veterans Affairs Surgical Quality Improvement Program database and the Decision Support System pharmacy database were linked to analyze the association of postoperative CDI with patients’ demographics, preoperative comorbidities, operative characteristics, and preoperative medications. The Veterans Affairs Surgical Quality Improvement Program assessments from October 1, 2009, to September 30, 2013, were investigated. The study was conducted at 134 VHA surgery programs, and the study population represents 12 surgical specialties: general, gynecological, neurosurgical, oral, orthopedics, otolaryngologic, plastic, podiatric, thoracic, transplant, urologic, and peripheral vascular.
Main Outcomes and Measures
Thirty-day postoperative CDI rates, risk factors of CDI, and association of CDI with postoperative morbidity and mortality.
Among 468 386 surgical procedures, the postoperative CDI rate was 0.4% per year and varied by the VHA Surgery Program (0.0% to 1.4%) and surgical specialty (0.0% to 2.4%). Thirty-day CDI rates were higher in emergency procedures, procedures with greater complexity and higher relative value units, and those with a contaminated/infected wound classification. Patients with postoperative CDI were significantly older, more frequently hospitalized after surgery (59.9% vs 15.4%), had longer preoperative hospital stays (9.1 days vs 1.9 days), and had received 3 or more classes of antibiotics (1.5% vs 0.3% for a single antibiotic class) (all P < .001). Patients with CDI had higher rates of other postoperative morbidity (86.0% vs 7.1%), 30-day mortality (5.3% vs 1.0%), and longer postoperative hospital stays (17.9 days vs 3.6 days). Independent risk factors for CDI included commonly identified patient factors (albumin, functional class, and weight loss), procedural characteristics (complexity, relative value units, emergency, and wound classification), surgical program complexity, the number of preoperative antibiotic classes, and length of preoperative hospital stay.
Conclusions and Relevance
The number and class of antibiotics administered after surgery, preoperative length of stay, procedural characteristics, surgical program complexity, and patient comorbidities are associated with postoperative CDI in the VHA.
Clostridium difficile is an anaerobic, spore-forming bacterium. Clostridium difficile infection (CDI) is associated with life-threating conditions. The clinical features of CDI are diverse, ranging from mild diarrhea to fever, abdominal pain, abdominal distention, leukocytosis, hemorrhage, and necrosis.1Clostridium difficile infection is a common cause of health care–associated infectious diarrhea. It has become a growing cause of nosocomial morbidity, long hospital stays, high hospital costs, and mortality in North America and across the world.2,3 In the United States, CDI is the most frequently reported nosocomial pathogen and the ninth leading gastrointestinal cause of death.4
Studies have shown that CDI is a significant complication for surgical patients. The reported 30-day complication rate of CDI ranges from 0.2% to 9.0%, depending on the surgical population.5-7 The health care–associated burden of CDI is increasing among surgical patients, and the burden varies among surgical procedures.8Clostridium difficile infection is associated with a greater length of mechanical ventilation, longer intensive care unit and hospital stay, and higher mortality risk.9
Several CDI risk factors have been identified, including advanced age, severe comorbidity, hospitalization, and antibiotic exposure.10,11 Several studies have examined the risk factors for colorectal cancer surgery and cardiac surgery.12-14 To our knowledge, no study has reported risk factors associated with CDI across a variety of surgical specialties in the Veterans Health Administration (VHA), a predominantly male and elderly surgical population.15
The 134 VHA Surgery Programs perform approximately 400 000 surgical procedures annually. In 2007, the Veterans Affairs (VA) Surgical Quality Improvement Program (VASQIP) started collecting 30-day postoperative CDI data in eligible noncardiac surgical procedures. The objectives of this study were to document the CDI incidence in the VHA over a 4-year period across different surgical procedures, identify the risk factors associated with CDI, and determine the impact of CDI on postoperative mortality, morbidity, and hospital length of stay.
The VASQIP prospectively collects preoperative demographics, comorbidities, operative characteristics, and 30-day postoperative outcomes for eligible cases in VA facilities with an approved VHA Surgery Program.16-18 By policy, the VHA National Surgery Office analyzes VASQIP data on a quarterly basis, publishes quarterly and annual reports, and monitors quality improvement.19 This operational retrospective cohort study analyzed all noncardiac VASQIP-assessed cases from 134 VHA surgical programs from October 1, 2009, to September 30, 2013. For patients with multiple operations within 30 days, only the first case (index procedure) was included in the analysis. Postoperative CDI incidences within 30 days were captured at the VA facility and recorded by a VHA surgical quality nurse. Clostridium difficile infection cases were included regardless of whether diagnosis occurred during index hospitalization, readmission, or outpatient treatment. Clostridium difficile infection was defined with 2 components: a positive culture result of C difficile and/or a toxin assay and a clinical diagnosis of C difficile documented in the Computerized Patient Record System, the VHA’s electronic health record. The laboratory database in the VHA Centralized Data Warehouse was used for identifying all C difficile tests conducted in VHA facilities. The VA Decision Support System pharmacy database of outpatient prescription fills was linked with the VASQIP database to analyze the association of preoperative antibiotic prescription with the incidence of CDI. All surgical cases with preoperative vancomycin hydrochloride, metronidazole, and fidaxomicin administration were excluded from the analysis because these antibiotics are predominantly used to treat CDI. The antibiotic classes were defined as aminoglycoside, β-lactam, fluoroquinolone, glycopeptide, lincosamide, lipopeptide, macrocyclic, macrolide, nitroimidazole, oxazolidinone, penicillin, polymyxin, rifamycin, streptogramin, sulfonamide, tetracycline, and other. Per VHA policy (Handbook 1058.05), the information presented in this article represents VHA operations activity and therefore did not require informed consent or institutional review board approval.
All VASQIP preoperative laboratory values were the closest value to the date of surgery and were within 90 days of the operation. Missing data for preoperative demographics and clinical characteristics in VASQIP are rare, but any noted missing data were imputed by a multiple-imputation process. Some numeric variables were categorized based on clinical significance or standard practice. Univariate analyses were performed using χ2 tests for categorical variables and t tests or the Wilcoxon rank-sum test for continuous variables. A Bonferroni correction was applied to all of the multiple comparisons. The Hosmer-Lemeshow20 purposeful selection model building algorithm was used for initial predictor variable selection. The final multivariable logistic model included significant variables selected from the approach described here, as well as variables known to be significant risk factors for CDI in other studies. P ≤ .05 was the criterion established for statistical significance in the final model. The goodness of fit of the final model was assessed by the Hosmer-Lemeshow20 test. The cluster effect of patients within hospitals was assessed by the generalized linear mixed model. For sensitivity analysis, patients with CDI were matched with patients without CDI with a 1:2 ratio by the principal Current Procedural Terminology (CPT) code, surgery program, surgery year, and predicted 30-day mortality risk probability by VASQIP reporting models. All statistical analyses were performed using SAS version 9.3 (SAS Institute Inc).
A total of 468 386 procedures were included in the analysis after removal of cases with preoperative antibiotic use suggestive of preoperative CDI treatment (8465 [1.8%] of the total sample). This study only reported the 30-day postoperative CDI rate as recorded for VASQIP-assessed procedures. Demographic characteristics, preoperative comorbidities of patients, intraoperative characteristics, and 30-day postoperative outcomes are shown in Table 1. Patients with CDI were significantly older than patients without CDI (mean age, 67.4 vs 60.6; P < .001). Quiz Ref IDSeveral preoperative comorbidities were more common in patients with CDI than those without CDI, including congestive heart failure, history of chronic obstructive pulmonary disease, ascites, acute renal failure, impaired sensorium, hemiplegia, rest pain/gangrene, bleeding disorder, wound infection, and weight loss greater than 10% within 6 months before surgery. Patients with impaired functional status were more prevalent in the CDI group than in the non-CDI group. A higher American Society of Anesthesiology (ASA) classification was significantly associated with CDI, with a moribund patient having significantly higher risk of CDI compared with a healthy patient (1.7% vs 0.03%; P < .001; Table 1).
A total of 1833 30-day postoperative CDIs were diagnosed. Of these, 1239 cases (67.6%) were diagnosed as having a CDI during index hospitalization, and others were diagnosed as having a CDI on readmission or as outpatients. Table 2 shows the distribution of case volumes and VASQIP 30-day postoperative CDI rates across the surgical specialties. There were 12 noncardiac main surgical specialties represented in the cohort. In this VASQIP surgery population, general surgery and orthopedic surgery were the specialties with the most prevalent case volumes (32.1% and 28.4%, respectively), and transplant and oral surgery were the specialties with the lowest case volumes (0.2% and 0.4%, respectively). The overall 30-day postoperative CDI rate was 0.4% in this cohort. The CDI rate was significantly different among the surgery specialties. Quiz Ref IDTransplant surgery had the highest CDI rate with 2.37%, and there was no CDI incidence among oral surgery procedures during the 4-year period. The median time to postoperative CDI was 9 days, but it was noted to vary among surgery specialties. The CDI rate also varied among the 134 VHA Surgery Programs, ranging from 0% to 1.4% during the 4-year period. Surgery programs with high CDI rates were more likely to stay high, and programs with low CDI rates stayed low in the 4-year study. Of the 1833 CDI events, 1078 events (58.8%) were the first or only complication, and 755 (41.2%) were associated with other VASQIP-recorded complications.
The VHA Surgery Programs are assigned surgical complexity levels in accordance with VHA policy,16-18 standardizing the facility infrastructure and surgical procedure assignments to ensure high quality and safe delivery of surgical services. Most VASQIP-assessed surgery procedures were performed in inpatient complex and inpatient intermediate surgical programs (74.4% and 20.3%, respectively; Table 2). The mean CDI rate of inpatient complex surgical programs was 0.44%, with a range from 0.0% to 1.4%. The mean CDI rates of ambulatory surgery centers and inpatient standard surgical programs were lower compared with those of the inpatient intermediate and inpatient complex surgical programs (Table 2). The CDI rates for complex inpatient, inpatient intermediate, inpatient standard, and ambulatory surgery center surgical programs were 0.44%, 0.31%, 0.06%, and 0.04%, respectively (P < .001) (Table 2). The data did not show a significant temporal trend at the national level. The CDI rates were 0.4%, 0.4%, 0.4%, and 0.4%, respectively, from fiscal year 2010 to 2013 (Figure).
The VHA policy assigns a surgical procedure complexity designation to each surgical procedure identified by the CPT code.17,18 In the 4-year period, most of the VASQIP-eligible noncardiac surgical procedures were defined as standard (57.1%) or intermediate (39.4%) based on the CPT code, while 2.4% were complex surgical procedures (Table 2). Quiz Ref IDSurgical complexity designation of the surgical procedure was significantly associated with CDI. The CDI rates for complex, intermediate, and standard surgical procedures were 1.16%, 0.59%, and 0.23%, respectively (P < .001). The CDI rate was significantly higher in emergency procedures (1.43%) compared with nonemergency procedures (0.34%; P < .001) (Table 2).
There were significant differences in intraoperative work relative value unit assignments and wound classification between CDI cases and non-CDI cases (Table 1). The mean (SD) intraoperative work relative value unit was 20.6 (10.7) for those with postoperative CDI and 14.5 (8.5) for those without postoperative CDI (P < .001). Intraoperative wound classification was significantly associated with increased risk of CDI. The CDI rates for clean, clean/contaminated, contaminated, and infected wound classifications were 0.2%, 0.59%, 1.28%, and 1.4%, respectively (P < .001) (Table 2).
Quiz Ref IDPostoperative 30-day mortality was significantly higher for patients with postoperative CDI than those without CDI (5.29% vs 1.02%; P < .001) (Table 1). Thirty-day postoperative morbidity events, including cardiac complications, central nervous system events, pulmonary complications, renal impairment, and surgical site infection, occurred more frequently in the CDI group than in the non-CDI group. In the CDI group, 20.95% had 30-day pulmonary complications vs only 2.45% in the non-CDI group (P < .001). The renal impairment complication rate was higher in the CDI group than the non-CDI group (16.2% vs 2.0%; P < .001; Table 1). The mean postoperative hospital length of stay was significantly longer for patients with CDI than those without CDI (17.9 days vs 3.6 days; P < .001). The number of different antibiotic classes administered within 60 days before surgery was significantly associated with CDI. The CDI rate for only 1 antibiotic class administration was 0.3%, while it was 1.5% for cases with 3 or more antibiotics used preoperatively (Table 1).
Sensitivity analysis using the matched cases validated the association of postoperative CDI with postoperative mortality, morbidity, and hospital stay (Table 3). Patients with CDI were associated with 3.9 increased odds of 90-day mortality (95% CI, 2.9-5.4; P < .001). Patients with CDI were 3.0 times more likely to die within 180 days compared with patients without CDI (95% CI, 2.4-3.9; P < .001). Quiz Ref IDPatients with CDI also had longer postoperative hospital stays compared with matched control individuals (15.6 days vs 8.1 days).
Independent predictors of CDI were identified by multivariable regression analysis, as outlined in Table 4. Advanced age, albumin less than 3.5 g/dL (to convert to grams per liter, multiply by 10), hematocrit 38.0% or less (to convert to proportion of 1.0, multiply by 0.01), and white blood cell count greater than 11.0 × 1000/μL3 (to convert to ×109 per liter, multiply by 0.001) were significantly associated with increased risk for CDI. The following preoperative comorbidities and intraoperative characteristics were also independent risk factors of CDI: a higher ASA classification, bleeding disorder, complexity of surgical procedure, higher intraoperative work relative value unit, emergent case status, nonindependent functional status, hemiplegia, history of chronic obstructive pulmonary disease, rest pain/gangrene, wound classification, weight loss within 6 months greater than 10.0%, hospital complexity, number of antibiotic classes prescribed before surgery, and length of preoperative hospital stay.
The overall postoperative incidence of CDI of 0.4% for this VHA study population is low compared with previously published rates of postoperative CDI.5-7 To our knowledge, this is the first study to report that the rate of postoperative CDI is significantly associated with the number of antibiotic classes administrated within 60 days before surgery. To our knowledge, this is also the first study to report that the rate of postoperative CDI is significantly associated with the complexity of the surgical procedure and the complexity of the surgical program in which the procedure is performed. The diversity of the surgical population and surgical procedures may contribute to differences in reported CDI incidence. Selection bias may also play a role in the variability of reported CDI rates because most published studies of CDI are nonrandomized trials. To minimize the selection bias, this study included all the VASQIP-assessed surgical cases performed by the VHA for a 4-year period for 12 specialties.
The study team did not find a temporal trend for CDI rates in the VHA surgical patient population for the 4-year period studied. This finding appears contradictory to reported rising CDI rates among hospitalized patients in the United States.21,22 Some surgical procedures have been reported as a risk factor for CDI.23,24 In this study, transplant patients had the highest 30-day postoperative CDI rate (2.37%) while gynecology and oral surgery had the lowest CDI rates (0.06% and 0.00%, respectively). The incidence of CDI for solid organ transplant from one institutional study was 12.4%.25 In this study, the CDI rate was 2.6% and 3.1% for kidney and lung transplant cases, respectively. Our findings indicate that patient demographics and preoperative clinical characteristics vary significantly among surgical procedures. These differences likely contribute to the CDI rate variability across surgical specialties.
Independent risk factors for CDI were found to be advanced age, albumin less than 3.5 g/dL, higher ASA classification, bleeding disorder, dialysis, impaired functional status, hematocrit less than 38%, hemiplegia, history of chronic obstructive pulmonary disease, rest pain/gangrene, open wound infection, weight loss greater than 10% within 6 months before surgery, duration of hospitalization before surgery, and preoperative cumulative antibiotic use. Moreover, the number of antibiotic classes administrated within 60 days before surgery was significantly associated with an increased risk of CDI. The CDI rate of cases with 3 or more different antibiotic classes prescribed was almost 6 times higher than that for cases with no or only 1 antibiotic class prescribed within 60 days before surgery.
The mean length of postoperative hospital stay for functionally independent patients was 3 days compared with 19 days for totally functional dependent patients, and the mean length of postoperative hospital stay for ASA classification I patients was 0.6 days compared with 19 days for ASA classification V patients. Preoperative hospital stay can be associated with a higher exposure to pathogens and antibiotics, and these are known significant risk factors for CDI.26 Several intraoperative characteristics (surgical complexity, emergent case status, and complexity of the surgical program) were identified as statistically significant risk factors for CDI after surgery. These factors reflect the illness of patients, duration of operation, and hospital setting; each is an established risk factor for CDI.26
The CDI rate significantly varied among the 134 VHA Surgery Programs (0% to 1.35%). Programs with high CDI rates remained high, and programs with low CDI rates remained low in the study. Based on this study, this VA facility variance is likely owing to the case mix of patients and the complexity of the surgical procedures performed. However, medical practice may also play a significant role in CDI. In this report, surgical procedures and patients at risk for CDI have been identified. Surgical administrators and clinical teams may consider the results of this study to target interventions for specific patients undergoing high-risk procedures. Such interventions include selective antibiotic administration, early testing of at-risk patients, hand hygiene with nonalcohol agents, early contact precautions, and specific environmental cleaning protocols. The results of this study can help inform best practice and provide actionable data to VHA leadership for the prevention of future increases in CDI rates.
The VASQIP has collected data that enable risk-adjusted postoperative 30-day mortality and morbidity as measures of surgical quality of care. Because the VASQIP does not collect postoperative outcomes beyond 30 days, the program cannot evaluate later postoperative outcomes. With improvements to VHA Centralized Data Warehouse data, we will investigate the possibility of postoperative 90-day CDI rates in future analyses, as this is a more established time frame for defining health care–associated infections.
This study had limitations, the first of which was the retrospective analysis. Second, most patients were middle-aged to elderly men. The generalizability of the results to women and to the non-VA population may be limited. Nevertheless, this large study evaluated CDI rates and identified risk factors in the VHA surgery population.
An overall postoperative CDI rate of 0.4% was identified in the VHA and remained unchanged during the 4-year period. The postoperative CDI rates varied significantly across surgical specialties and VHA Surgery Programs. Patients with postoperative CDI were significantly older, more frequently hospitalized following surgery, had longer preoperative hospital stays, and had received 3 or more classes of antibiotics in the 60-day preoperative period. Patients with CDI had significantly higher rates of other postoperative morbidity, 30-day mortality, and longer postoperative hospital stays. Independent risk factors for CDI included commonly identified patient factors (albumin, functional class, and weight loss), procedural characteristics (complexity, relative value unit, emergency, and wound classification), surgical program complexity, the number of preoperative antibiotic classes, and length of preoperative hospital stay. Mitigation of these CDI risk factors has the potential to decrease the overall incidence of CDI.
Corresponding Author: Xinli Li, PhD, National Surgery Office, Veterans Health Administration, 4100 E Mississippi Ave, Ste 310, Glendale, CO 80246 (email@example.com).
Correction: In the Original Investigaton titled “Analysis of Morbidity and Mortality Outcomes in Postoperative Clostridium difficile Infection in the Veterans Health Administration,” there was a typographical error in Table 1. The number (%) of patients in the age group of younger than 50 years should be 79 694 (17.0), maintaining a total number of patients in all age categories of 468 386. This article was corrected online on March 23, 2016.
Accepted for Publication: August 8, 2015.
Published Online: November 25, 2015. doi:10.1001/jamasurg.2015.4263.
Author Contributions: Dr Li had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Li, Wilson, Gunnar.
Acquisition, analysis, or interpretation of data: Li, Nylander, Smith, Lynn, Gunnar.
Drafting of the manuscript: Li, Nylander, Gunnar.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Li, Nylander, Smith.
Administrative, technical, or material support: Li, Nylander, Lynn, Gunnar.
Study supervision: Nylander, Smith, Gunnar.
Conflict of Interest Disclosures: None reported.
Previous Presentation: This paper was presented at the 39th Annual Meeting of the Association of VA Surgeons; May 3, 2015; Miami Beach, Florida.
Create a personal account or sign in to: