Lack of association of post-discharge prophylactic antibiotics with decreased risk of surgical site infection following spinal fusion

Abstract

Objectives

To determine the prevalence and factors associated with post-discharge prophylactic antibiotic use after spinal fusion and whether use was associated with decreased risk of surgical site infection (SSI).

Methods

Persons aged 10–64 years undergoing spinal fusion between 1 January 2010 and 30 June 2015 were identified in the MarketScan Commercial Database. Complicated patients and those coded for infection from 30 days before to 2 days after the surgical admission were excluded. Outpatient oral antibiotics were identified within 2 days of surgical discharge. SSI was defined using ICD-9-CM diagnosis codes within 90 days of surgery. Generalized linear models were used to determine factors associated with post-discharge prophylactic antibiotic use and with SSI.

Results

The cohort included 156 446 fusion procedures, with post-discharge prophylactic antibiotics used in 9223 (5.9%) surgeries. SSIs occurred after 2557 (1.6%) procedures. Factors significantly associated with post-discharge prophylactic antibiotics included history of lymphoma, diabetes, 3–7 versus 1–2 vertebral levels fused, and non-infectious postoperative complications. In multivariable analysis, post-discharge prophylactic antibiotic use was not associated with SSI risk after spinal fusion (relative risk 0.98; 95% CI 0.84–1.14).

Conclusions

Post-discharge prophylactic oral antibiotics after spinal fusion were used more commonly in patients with major medical comorbidities, more complex surgeries and those with postoperative complications during the surgical admission. After adjusting for surgical complexity and infection risk factors, post-discharge prophylactic antibiotic use was not associated with decreased SSI risk. These results suggest that prolonged prophylactic antibiotic use should be avoided after spine surgery, given the lack of benefit and potential for harm.

Introduction

Surgical site infections (SSIs) are among the most common healthcare-associated infections in the USA.¹ The most recent CDC guidelines for the prevention of SSIs recommend against continuation of prophylactic antibiotics in clean surgeries after the surgical incision is closed, even in the presence of surgical drains,² due to lack of data showing a benefit. While spine procedures were not included in the Surgical Care Improvement Project performance measures, the North American Spine Society also recommends discontinuing prophylactic antibiotics at the completion of surgery, although consideration is allowed for prolonged prophylaxis in complex procedures or complex patients.³ In the setting of appropriate peri-incisional antibiotic administration, a recent meta-analysis of studies involving spinal decompression or spinal fusion found no benefit for continuation of postoperative antibiotic prophylaxis.⁴ In practice, compliance with antibiotic discontinuation in hospital within 24 h of surgery varied from 58% to 100% in studies of the National Surgical Quality Improvement Program and U.S. Veteran’s Administration hospitals.^5–7 In a recent Australian national study, compliance with discontinuation of prophylactic antibiotics within 24 h was only 51% for orthopaedic and 36% for neurosurgical procedures.⁸

In contrast to in-hospital antibiotic prophylaxis after incision closure, few data are available on continuation of antibiotic prophylaxis after hospital discharge. In a randomized controlled trial by Hellbusch et al.⁹ comparing a single dose of cefazolin (with redosing for long procedures) versus extended prophylaxis that included oral cefalexin for 7 days, a non-significant decrease in the SSI rate after spinal fusion was found in the extended prophylaxis group (1.7% versus 4.3% in the single-dose group; P > 0.25). In a recent study involving three academic medical centres, oral antibiotics were prescribed in 3.3% at hospital discharge after non-traumatic spinal fusion. Similar to the study by Hellbusch et al., post-discharge antibiotic use was not associated with decreased 90 day SSI rate (1.7% with versus 0.9% without post-discharge antibiotics; P = 0.114).¹⁰

Exposure of patients to unnecessary antibiotics results in higher costs, adverse drug events and selection of antibiotic-resistant organisms,¹¹ and promotes Clostridioides difficile infection.^7,12–14 Outpatient antibiotic prescriptions constitute the majority of antibiotic use, and stewardship interventions in outpatient settings may have greater potential to decrease the prevalence of antibiotic-resistant organisms than inpatient antibiotic stewardship efforts.¹⁵

Given the heterogeneous and largely older literature regarding prophylactic antibiotic use in spine surgery, we sought to determine: (1) the prevalence and factors associated with post-discharge prophylactic antibiotic use; and (2) whether post-discharge prophylactic antibiotic use was associated with decreased risk of SSI after spinal fusion using a very large database of US commercially insured persons.

Methods

Source of data and inclusion/exclusion criteria

We established a retrospective cohort of children and adults aged 10–64 years in the USA who underwent spinal fusion from 1 January 2010 to 30 June 2015 in the IBM^® MarketScan^® Commercial Database. The Commercial Database includes medical and outpatient pharmacy claims and enrolment information for over 100 million persons from 2010 to 2015. Individuals in the database include employees, dependants and persons with COBRA continuation covered by employer-sponsored and other commercial health plans. The MarketScan Database is a deidentified limited dataset; for this reason the study was considered exempt from oversight by the Washington University Human Research Protection Office.

Persons undergoing spinal fusion were identified from the inpatient and outpatient medical claims files based on a Current Procedural Terminology (CPT) code or International Classification of Diseases, Ninth edition, Clinical Modification (ICD-9-CM) procedure code for spinal fusion (Table S1, available as Supplementary data at JAC Online). Spinal fusion procedures were restricted to those performed in an inpatient hospital, outpatient hospital or ambulatory surgical centre. We implemented additional measures to verify that a spinal fusion procedure was performed and to identify the date of the procedure, as previously described (Table S1).¹⁶

We applied additional exclusions for complicated admissions and procedures in which post-discharge antibiotics were not possible or could have been for therapeutic indications (Figure 1, Table S2). Exclusions during the index surgical admission included coding for motor vehicle accident or gunshot wound, in-hospital death and additional surgery(ies) other than spine using the National Healthcare Safety Network (NHSN) procedure list.¹⁷ To exclude patients who may have received antibiotics related to a recent or current infection, fusion procedures were excluded in persons with an ICD-9-CM diagnosis code for a systemic or serious infection from 30 days prior to the fusion hospitalization through to 2 days after surgical discharge (Table S2). Similarly, spinal fusion procedures in which more minor infections (e.g. urinary tract infection, Table S2) were coded within 14 days prior through to 2 days after surgical discharge were also excluded.

Figure 1.

Consort flow diagram of spinal fusion procedures included in analysis, 1 January 2010 to 30 June 2015.

The population was restricted to those persons with continuous medical and prescription drug insurance enrolment from 365 days before through to 90 days after the spinal fusion surgery to assess comorbidities, complications and post-discharge antibiotic utilization.

Identification of exposures, outcomes and covariates

The primary exposure of interest was post-discharge prophylactic antibiotic use (full list in Table S3), defined as a paid prescription filled between the surgical discharge date through to 2 days post-discharge. If a patient had a prescription for the same antibiotic in the 30 days prior to surgery, it was considered prior use and not prophylactic.

Demographic variables of interest included sex, age, rural residence (defined by absence of a metropolitan statistical area) and US census region of residence. Comorbidities were identified in the year before surgery, primarily based on the Elixhauser classification (Table S1).^18,19 Prescription drug claims were also used for diabetes and smoking to increase the sensitivity of identification.

Operative factors evaluated included the number of spine levels, site of fusion (cervical, thoracic and/or lumbar) and approach (anterior, posterior or both), identified using CPT and ICD-9-CM procedure codes. Other potential factors associated with post-discharge antibiotics and/or SSI captured during the spinal fusion admission included haematoma, haemorrhage, dehiscence, dural tear, fall or fracture prior to hospital admission, instrumentation and non-local autograft harvest (from a separate incision) and spine cancer. Details of exposure and covariate definitions are included in Table S1.

The primary outcome of interest was SSI from 2 to 90 days after spinal fusion surgery, identified using ICD-9-CM diagnosis codes assigned during inpatient and/or outpatient encounters (Table S4). Diagnostic and other specialized claims were not used to identify SSI, to minimize capture of diagnostic workup as an outcome.¹⁹ Censoring was implemented at the time of a subsequent surgical procedure within 90 days, using the 2015 NHSN SSI surveillance procedure codes.¹⁷ Exceptions to censoring included the following: if the subsequent procedure was a spine surgery coded for SSI on the date of procedure, or if the person was coded for a spine-specific SSI (e.g. epidural abscess) after a subsequent non-spine surgery, as described previously.¹⁶

Statistical methods

Bivariate comparisons were performed using chi-squared tests for binary variables and Student’s t- or Mann–Whitney U-tests for continuous variables. Utilization of post-discharge prophylactic antibiotics by year was assessed using the Cochran–Armitage trend test. Independent factors associated with utilization of post-discharge prophylactic antibiotics or with SSI were identified in a generalized linear model, with calculation of robust standard errors. Variables with P < 0.05 in bivariate analysis or with clinical/biological plausibility were included in the initial full models, with the exception of post-discharge prophylactic antibiotics (primary exposure in the SSI model). Variables were removed sequentially with P < 0.05 as the threshold for retention. Potential multicollinearity of independent variables was assessed using variance inflation factors. All statistical analyses were performed in SAS version 9.4 (Cary, NC, USA), with P < 0.05 considered statistically significant.

Results

A total of 329 290 spinal fusion inpatient or ambulatory surgery procedures were identified among patients 10–64 years old between 1 January 2010 and 30 June 2015. A total of 172 844 surgical encounters were excluded for lack of continuous medical and prescription drug coverage, insufficient evidence for fusion, hospitalization with discharge status of death, or coded for motor vehicle accident or gunshot wound, an exclusion infection, another non-spine surgery performed during the fusion hospitalization, or unknown US region of residence (Figure 1). In the final study cohort of 156 446 spinal fusion encounters, the median age was 52 years (IQR 44–58), 4375 (2.8%) encounters occurred in paediatric patients, 86 130 (55.1%) were female, 125 436 (80.2%) were inpatient surgeries and 29 413 (18.8%) were among patients residing in a rural area (Table 1).

Table 1.

Characteristics of the commercially insured spinal fusion population (N = 156 446), 1 January 2010 to 30 June 2015

Characteristic	n (%)a
Age, years, median (IQR)	52 (44–58)
Paediatric patients (10–17 years)	4375 (2.8)
Female	86 130 (55.1)
Patient residence, rural	29 413 (18.8)
Patient residence, US census regionb
Northeast	18 485 (11.8)
North Central	40 319 (25.8)
South	74 356 (47.5)
West	23 286 (14.9)
Inpatient surgery	125 436 (80.2)
Spinal region and approach
cervical anterior	74 819 (47.8)
cervical A/P	1307 (0.8)
cervical posterior	5055 (3.2)
thoracolumbar anterior	7266 (4.6)
thoracolumbar A/P	8132 (5.2)
thoracolumbar posterior	59 477 (38.0)
missing	390 (0.2)
Number of intervertebral levels
1–2	127 452 (81.5)
3–7	25 653 (16.4)
8+	3341 (2.1)
Post-discharge prophylactic antibiotics	9223 (5.9)
SSI within 90 days	2557 (1.6)

^aNone of the variables had missing data. Persons with missing sex, age or missing information on US region were excluded from the population. The remaining variables were defined by ICD-9-CM diagnosis or procedure variables, and thus have no missing values.

^bFor comparison, in 2015 the US population based on census region was Northeast (17.5%), North Central/Midwest (21.2%), South (37.7%) and West (23.6%).

Prevalence and factors associated with receipt of post-discharge prophylactic antibiotics

Prophylactic antibiotics were prescribed post-discharge after 9223 (5.9%) spinal fusion admissions. Utilization declined slightly over time, ranging from 6.3% in 2010 to 5.8% in 2015 (P < 0.001). Among those with post-discharge antibiotics, 44.1% were prescribed cefalexin, 20.9% received levofloxacin or ciprofloxacin, 10.1% received trimethoprim/sulfamethoxazole and 7.8% received doxycycline (complete list in Table S3). The most common class of antibiotics prescribed was first-generation cephalosporins.

The results of bivariate analyses for factors associated with post-discharge prophylactic antibiotic receipt are shown in Table S5. In multivariable analysis, after adjusting for age, risk of post-discharge prophylactic antibiotics after spinal fusion was significantly higher in persons living in the South and West (compared with the North Central US census region), those residing in a rural area, and those with underlying comorbidities (i.e. diabetes, history of lymphoma or Staphylococcus aureus infection). Operative variables associated with increased utilization of post-discharge prophylactic antibiotics included 3–7 versus 1–2 levels fused, posterior and anterior/posterior (A/P) cervical, anterior, A/P and posterior thoracolumbar approaches compared with anterior cervical, and use of donor bone autograft. Documented dehiscence, haematoma and dural tear during the surgical admission were also independently associated with increased utilization of post-discharge prophylactic antibiotics (Table 2).

Table 2.

Patient and operative factors independently associated with receipt of post-discharge prophylactic oral antibiotics after spinal fusion in multivariable analysis

Variable	RR (95% CI)a
Demographics and comorbidities
patient residence, US census region
Northeast	0.97 (0.90–1.05)
North Central	Reference
South	1.32 (1.25–1.39)
West	1.22 (1.14–1.31)
patient residence, rural	1.23 (1.17–1.28)
diabetes	1.16 (1.10–1.23)
lymphoma	1.36 (1.01–1.81)
S. aureus infection in prior year	1.32 (1.01–1.74)
Operative factors
dehiscence	2.77 (1.78–4.30)
dural tear	1.19 (1.05–1.35)
haematoma	1.66 (1.28–2.15)
autograft harvested via a separate incision	1.21 (1.14–1.29)
spinal region and approach
cervical anterior	Reference
cervical A/P	1.24 (1.00–1.54)
cervical posterior	1.54 (1.39–1.71)
thoracolumbar anterior	1.35 (1.23–1.49)
thoracolumbar A/P	1.55 (1.43–1.68)
thoracolumbar posterior	1.44 (1.38–1.51)
number of intervertebral levels
1–2	Reference
3–7	1.09 (1.03–1.15)
8+	0.94 (0.79–1.12)

RR, relative risk.

^aAlso adjusted for age using a spline (5 knots) and missing level and approach. Variables entered into model, but not retained: staged fusion, obesity, serious lung/cardiac disease, sex, surgery year, chronic pulmonary disease, hypertension and psychoses.

Incidence of SSI and C. difficile infection

The 90 day incidence of SSI after spinal fusion was 1.6% (n = 2557). The incidence of SSI was highest after posterior and A/P cervical procedures (4.0% and 4.1%, respectively) and higher after A/P and posterior thoracolumbar procedures (2.8% and 2.6%, respectively) compared with anterior cervical procedures (0.6%; Figure S1, Table S6). The incidence of SSI did not vary by overall post-discharge prophylactic antibiotic use versus no use (1.8% versus 1.6%, respectively; P = 0.144), or by post-discharge prophylactic antibiotic use versus no use within individual categories of spinal region/approach and number of intervertebral levels (e.g. post-discharge antibiotics versus no use in posterior cervical procedures; Figure S1; all P > 0.3 except for 3–7 levels, P = 0.098). There was also no difference in the incidence of C. difficile infection within 90 days after surgery by post-discharge prophylactic antibiotic use (n = 13; 0.14%) versus no use (n = 136; 0.09%; P = 0.160).

Factors associated with SSI

The results of bivariate analyses for factors associated with SSI are shown in Table S7 and factors associated with SSI in multivariable analysis are in Table 3. Post-discharge prophylactic antibiotic use was not associated with SSI after spinal fusion in multivariable analysis (adjusted relative risk 0.98; 95% CI 0.84–1.14), after controlling for age, a variety of underlying comorbidities, inpatient admission in the prior month, and spinal operative region, approach and number of intervertebral levels (Table 3).

Table 3.

Patient and operative factors independently associated with SSI after spinal fusion in multivariable analysis

Variable	RR (95% CI)a
Post-discharge prophylactic antibiotic use	0.98 (0.84–1.14)
Patient demographics and comorbidities
patient residence, US census region
Northeast	1.18 (1.03–1.34)
North Central	Reference
South	1.21 (1.10–1.33)
West	1.17 (1.04–1.33)
alcohol abuse	1.27 (1.00–1.60)
anaemia in prior 30 days	1.25 (1.13–1.39)
chronic pulmonary disease	1.23 (1.10–1.37)
depression	1.16 (1.04–1.29)
diabetes	1.42 (1.28–1.56)
hypertension	1.22 (1.12–1.33)
inpatient admission in prior 30 days	1.50 (1.22–1.86)
malnutrition	1.71 (1.23–2.39)
multiple myeloma/metastatic cancer	1.69 (1.26–2.26)
obesity	1.57 (1.44–1.72)
other neurological disorders	1.34 (1.13–1.58)
paralysis	1.48 (1.19–1.84)
psychoses	1.18 (1.03–1.35)
pulmonary circulation disease	1.54 (1.11–2.15)
rheumatoid arthritis/collagen vascular disease	1.33 (1.16–1.53)
S. aureus infection in prior year	3.27 (2.49–4.29)
ulcer colitis/Crohn’s disease	1.50 (1.11–2.03)
Operative factors
dural tear	1.67 (1.39–2.01)
spinal region and approach
cervical anterior	Reference
cervical A/P	5.36 (4.02–7.15)
cervical posterior	5.59 (4.70–6.65)
thoracolumbar anterior	2.65 (2.16–3.26)
thoracolumbar A/P	4.04 (3.43–4.76)
thoracolumbar posterior	4.07 (3.64–4.55)
number of intervertebral levels
1–2	Reference
3–7	1.33 (1.21–1.47)
8+	1.62 (1.30–2.00)

RR, relative risk.

^aAlso adjusted for age using a spline (3 knots) and missing level and approach. Variables entered into model, but not retained: staged fusion, haemorrhage, haematoma, instrumentation, serious lung/cardiac disease, sex, antibiotic prescribed in 30 days prior to index, congestive heart failure, chronic kidney disease, coagulopathy, drug abuse, hypothyroidism, liver disease, lymphoma, fluid and electrolyte disorders, peripheral vascular disease, weight loss and donor graft.

Discussion

In this retrospective claims database study of over 150 000 spinal fusion procedures from 2010 to 2015, we found that several comorbidities (S. aureus infection in the past year, longer fusions, thoracolumbar procedures, posterior approach and use of autograft from a separate incision) were associated with receipt of prophylactic antibiotics at hospital discharge. Utilization was higher in the South and West US regions, in persons living in rural areas, and in those coded for non-infectious wound complications or dural tear during the surgical admission. Use of post-discharge prophylactic antibiotics was, however, not associated with decreased risk of SSI. Most of the factors independently associated with receipt of post-discharge prophylactic antibiotics were also associated with increased risk of SSI.

Consistent with the literature, anterior cervical procedures had the lowest 90 day incidence of SSI (0.6% with and 0.5% without post-discharge antibiotics; Figure S1).^20–23 The incidence of SSI after posterior and A/P cervical procedures was 4.0% and 4.1%, respectively, while posterior and A/P thoracolumbar procedures had 90 day incidence of SSI of 2.6% and 2.8%, respectively.

In two recent meta-analyses, diabetes, obesity, longer duration of surgery and transfusion were associated with increased risk of SSI after spinal surgery.^24,25 Other reported operative risk factors for SSI included posterior approach,^24,26 number of vertebral levels^27,28 and local or metastatic cancer involving the spine.^29,30 Consistent with reports of risk factors for SSIs after orthopaedic procedures, we also found prior S. aureus infection or colonization,^31,32 malnutrition^30,33,34 and preoperative anaemia^35,36 to be associated with increased risk of SSI. Although our primary objective was not to investigate risk factors for SSI, the consistency between our findings and the published literature supports the generalizability of our results.

There is little evidence for use of prolonged antibiotic prophylaxis after orthopaedic procedures to prevent SSI. Inabathula and colleagues³⁷ continued oral antibiotics for 7 days post-discharge in ‘high-risk’ patients undergoing hip and knee arthroplasty, although 60% and 70% of the patients in each group met the investigators’ definition of high risk. Hellbusch et al.⁹ reported a lower rate of infection after extended antibiotic prophylaxis in instrumented lumbar procedures, including 7 days of oral cefalexin, compared with a single dose of perioperative antibiotics, although only a small number of infections occurred and the difference between groups was not significant. Pivazyan et al.³⁸ reported an approximately 2-fold decrease in deep SSIs after spine surgery for degenerative conditions in patients given prolonged systemic antibiotics while drains were in place compared with patients in a later time period without prolonged antibiotics, although the difference was not significant. In our current study we did not find decreased risk of SSIs in multivariable analysis associated with receipt of post-discharge antibiotics.

Our results regarding the lack of association of post-discharge prophylactic antibiotic use with patient comorbidities (other than diabetes and lymphoma) is similar to the report on prolonged surgical prophylaxis by Branch-Elliman and colleagues at outpatient U.S. Veteran’s Administration facilities.³⁹ They found prolonged utilization of surgical prophylaxis in approximately 4% of patients after outpatient orthopaedic procedures, with peaks in prescription durations of 3, 5 and 7 days. In that study, patient comorbidity score was not associated with receipt of antibiotics; rather, utilization was higher at lower-complexity hospital outpatient surgical facilities than at higher-complexity facilities.³⁹

Advantages of our study include the very large cohort size, and use of a detailed methodological approach combining careful patient selection and follow-up with rigorous statistical analyses. Our study’s primary limitation was the use of retrospective health insurer claims data, which lack some demographic information (e.g. socioeconomic status, race), information on surgeon and hospital type, and clinical details, since the data were designed for reimbursement purposes. Consequently, there was a possibility of misclassification of therapeutic antibiotics as prophylactic if an infectious diagnosis (e.g. urinary tract infection) was not recorded perioperatively. However, we used strict exclusion criteria to mitigate the possibility of missed infectious diagnoses within the month prior to the surgery admission and within 2 days post-discharge. In addition, the very large population size may have resulted in significant associations with post-discharge antibiotic utilization that are not clinically relevant (e.g. US region). The SSI risk factors identified in our study have been broadly reported in the literature, supporting the external validity of our findings. Additional limitations include lack of generalizability to the elderly and Medicaid or uninsured populations and non-US populations, due to use of data from a commercially insured population. The requirement for 1 year of prior insurance enrolment to identify underlying conditions may have selected for an initially healthier population. In addition, the MarketScan database contains proportionately more enrollees from Southern USA than the general population, which may also limit generalizability.

Antibiotic stewardship programmes in the USA have focused largely on decreasing unnecessary utilization of antibiotics in hospitals. There is increasing recognition of the importance of stewardship in the outpatient setting, since the majority of antibiotic prescriptions are filled in the outpatient setting.⁴⁰ Although less frequently studied, continuation of surgical prophylaxis post-discharge contributes to unnecessary utilization of antibiotics in the outpatient setting. In this large study using a commercial claims database we found that the use of post-discharge prophylactic antibiotics after spinal fusion was not associated with patient benefit. While interventions to prevent SSI after spine surgery are needed, extended prophylactic antibiotics should not be considered as one of these strategies.

Funding

Funding for this project was provided by grant U54CK000482 from the Centers for Disease Control and Prevention (V.J.F.). The Center for Administrative Data Research is supported in part by the Washington University Institute of Clinical and Translational Sciences grant UL1 TR002345 from the National Center for Advancing Translational Sciences (NCATS) of the National Institutes of Health (NIH) and grant number R24 HS19455 through the Agency for Healthcare Research and Quality (AHRQ). J.K.G. was supported by funding from the AHRQ (grant number 1F32HS027075-01A1) and the Trasher research fund (award #15024).

Transparency declarations

M.A.O. reports consultant work with Pfizer and grant funding through Pfizer outside the submitted manuscript. V.J.F. reports that her spouse is the Chief Clinical Officer at Cigna Corporation. D.K.W. reports consultant work with Centene Corp., PDI Inc., Pursuit Vascular and Homburg & Partner, and is a sub-investigator for a Pfizer-sponsored study for work outside the submitted manuscript. No other authors report competing interests.

Supplementary data

Tables S1 to S7 and Figure S1 are available as Supplementary data at JAC Online.