A Longitudinal Study of S. aureus Infection in a National Cohort of Surgical Patients

Abstract

Background

Postoperative infections are a common and often preventable complication of surgery. Staphylococcus aureus is a prevalent organism cultured in these infections and is associated with morbidity, mortality, and increased health care utilization. However, the long-term burden of S. aureus infection in surgical patients is not well studied. The purpose of this retrospective observational study was to assess the incidence, time trend, and burden of S. aureus infection up to 1 year after surgery.

Methods

We obtained manually reviewed data from the VA Surgical Quality Improvement Program (VASQIP) to identify surgeries in all major specialties. These were combined with laboratory microbiology and pharmacy data to identify pneumonia and infections of the urinary tract, surgical site, and blood.

Results

In the study population of 559 550 patients, S. aureus incidence decreased each year, from 2.1% in 2008 to 1.1% in 2015. Among these, incidence of methicillin-resistant S. aureus infection decreased from 0.7% to 0.4%, and methicillin-susceptible S. aureus decreased from 1.4% to 0.7%. S. aureus infection was associated with increased length of stay, emergency department utilization, inpatient admissions, and a 4-fold increase in mortality.

Conclusions

This is one of the largest studies describing the long-term incidence of S. aureus in the surgical population of a national integrated health care system. We conclude that the burden of S. aureus infection extends well beyond the conventional 30-day postoperative window, and late infection should be included in assessing the effects of interventions.

In this large retrospective study of Veterans undergoing major surgery, we found a nearly 50% reduction in 1-year S. aureus infections from 2008-2015, and only 1/4 of infections occurred within 30 days after surgery.

Postoperative infections are associated with increased mortality, morbidity, length of stay, and cost [1]. In the Veterans Health Administration (VA), methicillin-resistant Staphylococcus aureus (MRSA) colonization is present in nearly 13% of veterans admitted to the acute care setting [2]. These colonized patients are at significant risk of developing an active infection, especially when they lose mobility, undergo invasive procedures and surgery, have implants, or receive antibiotics that select for more resistant organisms [3]. The prevention of postsurgical infections is a high priority in surgical practice, and more information is needed about their immediate and long-term incidence and effects. S. aureus is of particular concern in comparison with most other pathogens due to its high prevalence and potential for resistance.

Numerous studies have described the incidence and harms of S. aureus infections in various populations and health care settings, with emphasis on those caused by MRSA, as well as the effects of interventions to prevent them. Our prior work has concluded that preoperative MRSA nasal colonization is associated with increased risk of postoperative MRSA clinical cultures and postoperative infections [4]. In 2007, the VA implemented the MRSA Prevention Initiative, a quality improvement program in which inpatients were screened for MRSA nasal colonization and placed on contact precautions, clinical staff were trained on hand hygiene, and cultural change in regard to infection control was emphasized. The initiative was generally regarded as successful [5]. Incidence of MRSA health care–acquired infections (HAIs) among patients in intensive care units declined by 62% in the first 4 years of the program, with the largest reduction (79%) being among MRSA bloodstream infections not related to a device [2]. During the first 2 years of the initiative, a single-center VA study found a decrease of MRSA HAIs and transmissions [6]. The MRSA HAI rates in 22 VA acute care spinal cord injury units decreased 81% during 2007–2011 [7]. In 127 acute care VA hospitals, the rates of MRSA HAIs from 2009 to 2015 decreased by 87% in intensive care units (ICUs) and 81% in non-ICUs [8]. Several non-VA studies have reached similar conclusions [9, 10]. However, much of this work is limited to the 30-day postoperative interval (or 1 year for implants). Less is known about the long-term incidence of S. aureus infections in the surgical patient population.

The goal of this study was to estimate the long-term incidence and time trend of S. aureus infections in veterans undergoing major surgery in 130 acute care hospitals and ambulatory surgical centers during 2008–2015. We hypothesized that the incidence of S. aureus infection would decrease over the 8-year study period and that patients undergoing certain types of procedures would face higher risk of infection, controlling for other patient characteristics and perioperative factors.

METHODS

This is a retrospective cohort study of patients undergoing major surgery in the VA during 2008–2015. The study was approved by the Boston VA Institutional Review Board.

Data Sources

We identified inpatient and outpatient surgeries from 2 sources. First, potential index surgeries were obtained from the Veterans Affairs Surgical Quality Improvement Program (VASQIP) data sets. VASQIP is a national program that provides risk-adjusted adverse outcomes data to VA Medical Centers, based on validated, high-fidelity risk adjustment models [11]. Surgeries eligible for VASQIP review are identified by principal Current Procedural Terminology (CPT) codes. VASQIP defines major surgery as “procedures performed under general, spinal, and epidural anesthesia, and all carotid endarterectomies and inguinal herniorrhaphies, regardless of anesthesia type” [12]. Data are collected by manual medical record review performed by trained nurses and include items for preoperative risk factors, information about the operative procedures and anesthesia, and postoperative outcomes and complications. Second, the VA Corporate Data Warehouse (CDW) Surgical Package was used to identify additional surgeries not selected for review by VASQIP. The CDW also provided information on patient demographics and clinical characteristics.

Index Surgeries

We obtained all VASQIP-reviewed noncardiac surgeries, across all available specialties, during calendar years 2008 to 2015. The CPT-based procedures reviewed by VASQIP have been described previously. We did not consider cardiac surgeries because VASQIP tracks a different set of postoperative complications compared with the noncardiac cases, making comparisons difficult. The patient’s first chronological surgery was flagged as a potential index surgery. If the patient had any subsequent surgery within 1 year, both the potential index and those subsequent surgeries were excluded. The purpose of this exclusion criterion was to allow us to attribute infection outcomes to 1 and only 1 index surgery. A patient could have more than 1 index surgery in the final cohort, provided the second occurred more than 1 year after the first.

Primary Outcomes

The primary outcome of the study was occurrence of any S. aureus infection within 1 year after index surgery. We identified outcomes differently depending on whether they occurred within 30 days after surgery (using VASQIP and microbiology data) vs during postoperative days 31–365 (using microbiology and antibiotic data). Infections were classified as surgical site infection (SSI; within 30 postoperative days), skin and soft tissue infection (SSTI; after day 30), bloodstream infection (BSI), urinary tract infection (UTI), and pneumonia (PNA). The VASQIP database included the determination of 30-day SSI (either superficial, deep, or organ/space infection), UTI, and PNA. VASQIP does not assess 30-day BSI; therefore, we identified these through the presence of a blood culture with S. aureus growth. In cases where VASQIP did flag an infection, we attempted to identify in the CDW microbiology database any positive culture taken 30 days postoperatively from an appropriate anatomical site with reported growth for S. aureus. If no such culture was found, we did not count that VASQIP-determined outcome as an S. aureus infection.

Infections during postoperative days 31–365 were identified using an algorithm we previously developed and which has been used in other large database studies of S. aureus infection [13, 14]. An infection was defined as a laboratory culture positive for S. aureus, combined with prescribing of systemic antibiotics within 5 days after obtaining the specimen. The exception was BSI, which we counted regardless of antibiotic use, as per standard clinical practice. We used the anatomical site of the specimen to determine the type of infection; that is, a positive sputum culture was flagged as PNA, a positive blood culture as BSI, a positive urine culture as UTI, and a positive skin culture as SSTI (as opposed to SSI, as the skin infection might not be related to a previous surgical site). The S. aureus infections were then distinguished as MRSA or methicillin-susceptible (MSSA) based on the organism description field in the CDW microbiology data.

Some patients had multiple positive cultures that were related to the same infection, for example, 2 positive cultures from the same infected wound a week apart. To avoid double-counting these as 2 distinct infection outcomes, we applied criteria to disregard the subsequent culture. In cases of SSI, a subsequent positive culture represented a new infection if it was taken from a different anatomical site than the first positive culture. In BSI, PNA, and UTI, a subsequent culture was counted if it occurred at least 14 days after the previous infection, regardless of anatomical site.

Medical Record Review

To validate that our criteria for identifying infections after postoperative day 30 (ie, requiring a positive culture combined with antibiotic use) correctly identified true infections, a medical record review was performed by a trained surgical resident and subsequently reviewed by a senior surgeon (K.I.). Based on a review of 400 patients with a 1-year S. aureus infection and 100 patients without a 1-year S. aureus infection, we further refined our algorithm to capture and classify infections. Cultures from wounds after 30 days were classified as skin and soft tissue infections rather than SSIs, consistent with CDC definitions for nonimplant surgeries, as these SSTI outcomes were not necessarily related to the surgical wound.

Secondary Outcomes

As surgical infections are associated with poor clinical outcomes, as well as increased utilization and cost, we describe selected secondary outcomes, stratified by patients who had a 1-year S. aureus infection, compared with those who did not. We identified 1-year all-cause admissions to VA hospitals, mortality, emergency department (ED) use, and lengths of stay for the index surgery.

Statistical Analysis

Descriptive statistics for patient demographics, clinical characteristics, and surgical factors were calculated and stratified by patients who did and did not have a 1-year S. aureus infection. Standardized mean differences were calculated for both continuous and categorical variables [15]. We used multivariate logistic regression to estimate the association between patient characteristics and perioperative factors with the outcome of any 1-year S. aureus infection. We chose covariates based on their significance in previous studies [16] and to be parsimonious and generalizable across surgical specialties. Construction of the analytic data set was done in SAS 9.4, and table/figure creation and statistical analysis were conducted in R 3.5.

RESULTS

Description of Cohort

The initial VASQIP database contained 966 877 surgeries during the 8-year period. After applying the exclusion criteria described above, the final cohort had 559 590 surgeries (Figure 1). The average age of patients (SD) was 59.3 (13.4) years, 74.6% were white, and 91.4% were male (Table 1). The most common surgical specialties were orthopedics (30.8%) and clean general surgery (21.8%), followed by general surgeries that were either clean/contaminated or dirty (13.5%). In 38.2% of index surgeries, the patient was obese (BMI ≥ 30); 19.2% had a history of diabetes, 11.4% had severe COPD, and 32.3% had a history of heavy smoking. Four point five percent of surgeries were emergent. The top 10 most frequent CPT-based procedures within each surgical specialty are presented in the Supplementary Data.

Figure 1.

Identification of index surgeries.

Table 1.

Characteristics of Veterans Undergoing Major Surgery During 2008–2015, Stratified by 1-Year Staphylococcus aureus Infection Outcome

	Overall	No Infection	With Infection	P
No. of surgeries (%)	559 590	551 143 (98.5)	8447 (1.5)
Male, No. (%)	511 272 (91.4)	503 244 (91.3)	8028 (95.0)	<.001
Race, No. (column %)				<.001
White	417 444 (74.6)	410 963 (74.6)	6481 (76.7)
Black	95 386 (17.0)	94 018 (17.1)	1368 (16.2)
Other or not known	46 760 (8.4)	46 162 (8.4)	598 (7.1)
Age, mean (SD), y	59.29 (13.84)	59.23 (13.84)	63.28 (13.24)	<.001
Surgical specialty, No. (%)				<.001
General (clean)	122 017 (21.8)	120 787 (21.9)	1230 (14.6)
Orthopedics	172 162 (30.8)	170 270 (30.9)	1892 (22.4)
General (other than clean)	75 402 (13.5)	73 615 (13.4)	1787 (21.2)
Urology	55 190 (9.9)	54 258 (9.8)	932 (11.0)
Peripheral vascular	37 601 (6.7)	36 664 (6.7)	937 (11.1)
Other specialty	32 611 (5.8)	32 240 (5.8)	371 (4.4)
Neurosurgery	31 399 (5.6)	30 878 (5.6)	521 (6.2)
Thoracic	14 373 (2.6)	13 985 (2.5)	388 (4.6)
Plastic	12 902 (2.3)	12 656 (2.3)	246 (2.9)
Podiatry	5933 (1.1)	5790 (1.1)	143 (1.7)
ASA class 3–5, No. (%)	361 576 (64.6)	354 525 (64.3)	7051 (83.5)	<.001
Wound class, No. (column %)				<.001
Clean	397 034 (71.0)	392 498 (71.2)	4536 (53.7)
Clean/contaminated	135 696 (24.2)	132 986 (24.1)	2710 (32.1)
Contaminated	14 631 (2.6)	14 096 (2.6)	535 (6.3)
Dirty/infected	12 229 (2.2)	11 563 (2.1)	666 (7.9)
Obesity (BMI ≥ 30), No. (%)	213 908 (38.2)	210 779 (38.2)	3129 (37.0)	.024
History of diabetes, No. (%)	107 699 (19.2)	105 195 (19.1)	2504 (29.6)	<.001
History of severe COPD, No. (%)	63 624 (11.4)	61 949 (11.2)	1675 (19.8)	<.001
History of smoking, No. (%)	180 703 (32.3)	177 761 (32.3)	2942 (34.8)	<.001
RVU range, No. (column %)				<.001
0.0–9.9	230 111 (41.1)	227 606 (41.3)	2505 (29.7)
10–19.9	191 863 (34.3)	188 567 (34.2)	3296 (39.0)
20+	137 589 (24.6)	134 943 (24.5)	2646 (31.3)
Missing	27 (0.0)	27 (0.0)	0 (0.0)
Emergent surgery, No. (%)	25 449 (4.5)	24 538 (4.5)	911 (10.8)	<.001
Open or infected wound, No. (%)	864 (0.2)	613 (0.1)	251 (3.0)	<.001

One-year Staphylococcus aureus infection was defined as occurring during postoperative days 0–365, with culture positive for S. aureus.

Abbreviations: ASA, American Society of Anesthesiologists; BMI, body mass index; COPD, chronic obstructive pulmonary disease; RVU, relative value unit.

Incidence Rate of S. aureus Infection

The incidence rate of 1-year S. aureus infection was 1.5%, or 8447 initial infections in 559 590 index surgeries. The 1-year incidence rate of infection after surgeries performed in 2008 was 2.1%, which declined steadily to 1.1% in 2015 (Figure 2). MRSA incidence declined from 0.66% in 2008 to 0.44% in 2015, and MSSA incidence declined from 1.44% to 0.67%. We performed a chi-square test for yearly trend in the proportion of eligible surgeries that resulted in any 1-year S. aureus infection and rejected the null hypothesis of no trend (P < .001). Consistently throughout the study period, approximately one-quarter of all 1-year S. aureus infections occurred by day 30, another one-quarter between days 31 and 90, and the remaining half between days 91 and 365.

Figure 2.

Incidence of Staphylococcus aureus and methicillin-resistant S. aureus/methicillin-susceptible S. aureus infection by postoperative interval and infection type. Abbreviations: MRSA, methicillin-resistant S. aureus; MSSA, methicillin-susceptible S. aureus.

Infection Incidence Stratified by Surgical Specialty

We present the incidence rate of 1-year S. aureus infections stratified by index surgical specialty in Figure 3. SSI and SSTI had the highest incidence in podiatry surgeries (209.0 per 10 000 surgeries), followed by peripheral vascular (183.8) and thoracic surgeries (162.8). The lowest incidence was observed in orthopedics (80.2), clean general surgery (82.4), and other specialties (83.1). Other types of infection (BSI, UTI, and PNA) were less frequent in comparison. Among BSI and PNA, the specialty with the highest incidence was thoracic surgery (44.5 and 91.1, respectively), whereas urologic procedures resulted in the highest rate of S. aureus–related UTIs (58.0).

Figure 3.

Incidence rate of 1-year Staphylococcus aureus infection by surgical specialty and infection type. Abbreviations: BSI, bloodstream infection; PNA, pneumonia; SSI, surgical site infection; SSTI, skin and soft tissue infection; UTI, urinary tract infection.

Adverse Outcomes Associated With Infection

Patients with S. aureus infections had higher rates of all the adverse secondary outcomes we tracked (Supplementary Data). Infection was associated with a nearly 3-fold increase in the index length of stay (17.1 vs 5.8 days) among those patients with inpatient index surgery. Any 1-year ED utilization was more frequent in patients with infection compared with those without (76.3% vs 37.7%), as was the mean number of ED visits (3.6 vs 1.7). Patients with infection were more likely to have an inpatient admission (58.6% vs 14.3%) and had a higher mean number of admissions (1.2 vs 0.2). Finally, those with infection were observed to have a nearly 4-fold higher rate of 1-year mortality (15.9% vs 3.6%). All comparisons had a P value <.001.

Association Between Baseline Characteristics and Infection

The results of multivariate logistic regression analysis are shown in Figure 4. Using clean general surgery as the reference specialty, most specialties were associated with increased risk of infection, including podiatry (odds ratio [OR], 2.68; 95% confidence interval [CI], 2.25–3.20), plastics (OR, 2.10; 95% CI, 1.83–2.42), and thoracic (OR, 1.66; 95% CI, 1.47–1.88). Orthopedics was the only specialty with no significant difference in risk of S. aureus infection compared with clean general surgery (OR, 1.02; 95% CI, 0.94–1.10). Male sex was associated with higher infection risk (OR, 1.36; 95% CI, 1.23–1.51), whereas race and age had estimated odds ratios not different from 1. An American Society of Anesthesiologists score of 3–5 was associated with higher risk of infection (OR, 1.91; 95% CI, 1.79–2.04).

Figure 4.

Multivariate logistic regression for 1-year Staphylococcus aureus infection. Reference surgical specialty is General Surgery (Clean). Abbreviations: ASA, American Society of Anesthesiologists; BMI, body mass index; COPD, chronic obstructive pulmonary disease; RVU, relative value unit.

Discussion

This is the first long-term study of S. aureus infection in surgical patients in a national integrated health system. We found an overall S. aureus infection rate of 1.5% throughout the 8-year study period, and these infections were associated with increased mortality, increased inpatient length of stay, readmissions, and more frequent ED use. Incidence rates declined throughout the study period, concurrent with a national VA initiative to reduce MRSA colonization and infection. Several surgical specialties, most notably podiatry, plastics, thoracic, and vascular surgery, were associated with higher risk of infection compared with clean general surgery. The novel contribution of this study is our description of incidence rates specific to the surgical population, which will be useful in designing interventions to prevent postoperative infections.

The identification and prevention of MRSA HAI became urgent in response to an increase in incidence rates from 2006 to 2008, though that trend may be partly due to more intensive surveillance [17]. Other studies have identified dramatic declines in infection rates in large national cohorts, including an 87.0% decrease in MRSA HAI among acute care ICU patients from 2007 to 2015 in the VA [8]. In roughly the same period but with a follow-up period of 1 year, our study found a 33.0% relative decline in MRSA infection rates, from 0.66% in 2008 to 0.44% in 2015. Most of the infections that occurred in our study were beyond the conventional 30-day measurement period; therefore, we assert that surgical patients remain at risk of S. aureus infection well beyond that period. Approximately three-quarters of S. aureus infections occurred after the 30-day surveillance period, potentially leading to underestimates of the burden of S. aureus infection after surgery.

Demographics, including age and race, generally showed no significant association with infection outcome, except for sex. Prior work has shown that men are more susceptible than women to SSI after certain types of procedures [18], but our inclusion of many surgical specialties likely clouds any relationship. The surgical specialties with the highest odds ratios of infection risk were podiatry and peripheral vascular surgery. Diabetes has been shown to be associated with increased risk of 30-day SSI following foot and ankle surgery [19]; therefore, the strong association in our study is not surprising, given the high prevalence of diabetes among veterans. SSI risk following arterial interventions has been shown to be high relative to other types of clean surgery, and nasal colonization and diabetes are major risk factors [20].

Prevention of S. aureus infection in prosthetic joint operations is of particular importance, given the high morbidity associated with these infections [21]. We found that the odds ratio of long-term infection in orthopedic surgery patients was no different than that of clean general surgery patients. Our study’s 1-year incidence rate of SSI/SSTI in orthopedic surgery patients was 0.80%, which was low compared with previous studies. We do not have preoperative MRSA colonization data available to make a correlation with postoperative MSRA infections in our study, but clearly this has been shown in previous work. For example, a single-center screening and decolonization intervention in orthopedic surgery patients with a 1-year postoperative follow-up found an S. aureus SSI rate of 3.5% in control patients and 0 infections in the treatment group [22]. Another single-center nasal decolonization intervention of 7019 orthopedic surgery patients found a 30-day S. aureus infection rate of 0.45% before the intervention, followed by a postintervention rate of 0.19% [23]. A large, multicenter study of 93 907 Medicare beneficiaries undergoing orthopedic surgery found an S. aureus infection rate of 1.6% at postoperative day 60 and 2.5% at postoperative day 180. Finally, a prospective study of 2423 orthopedic surgery patients who were followed for at least 1 year found an MRSA SSI rate of 6.3% in MRSA nasal carriers vs 0.5% in noncarriers [26].

The major strength of this study is its large sample of reliable, validated, medical record reviewed data on perioperative factors and short-term outcomes. The combination of VASQIP and a rich national database of laboratory microbiology and drug data make the VA one of the few health care organizations capable of conducting this type of large-scale epidemiologic study and is an important data source for surgical outcomes research [27]. A limitation is that the long-term infection outcomes (after postoperative day 30) were derived using an electronic algorithm. It is possible that some infection outcomes were misclassified, for example, colonization without infection, data entry errors, or being flagged as the wrong type of infection. However, the manual review of a small subset of infections showed that the algorithm is generally reliable and allowed us to fine-tune it to prevent the small number of misclassifications that were found. A second limitation is that, like most VA population-based studies, more than 90% of participants are male, and our results may not generalize to other patient populations. Finally, we were unable to control for a patient’s history of S. aureus infection, which is likely a risk factor for future infection.

The typical measurement period for SSI is 30 days postoperatively. This study showed that the majority of S. aureus infections in our national cohort were presented after the 30-day window. We do not suggest that these late infections are attributable to the index surgery or should be included in quality measurement. However, they should be accounted for in description of the overall burden of S. aureus infection in the surgical patient population and in calculating the risks and benefits of perioperative prevention strategies.

Supplementary Data

Supplementary materials are available at Open Forum Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Acknowledgments

Financial support. This work was funded by a Cooperative Research and Development Agreement (CRADA) between Pfizer Inc. and the authors’ institution. Dr. Itani is a consultant to Eupraxia Pharmaceuticals and has received an institutional grant from Sanofi. Dr. Gupta is consultant to Iterum, Paratek, Ocean Spray, and Tetraphase and an author for UpToDate.

Potential conflicts of interest. All authors: no reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Prior presentation. Preliminary versions of this work were previously presented as a poster at the 2018 Surgical Infection Society Meeting and as an oral presentation at the 2018 Michael E. DeBakey International Surgical Society Meeting.