Placement of cardiovascular implantable electronic devices (CIEDs), such as pacemakers and implantable cardioverter-defibrillators (ICDs) is rising as the population ages.Reference Bradshaw, Stobie, Knuiman, Briffa and Hobbs 1 Devices are increasingly implanted into patients older than 75 years of age and with multiple comorbidities, who have a higher risk of procedure and medication-related adverse events.Reference Greenspon, Patel and Lau 2 Infections complicate an estimated 1%–2% of CIED procedures, and incidence rates have nearly doubled over the past decade.Reference Greenspon, Patel and Lau 3 , Reference Voigt, Shalaby and Saba 4 Procedure-related CIED infections cause considerable harm; the absolute 6-month mortality following a deep cardiac device infection is 18%.Reference Baman, Gupta, Valle and Yamada 5 Thus, preventing these infections is a major clinical priority.
The American Heart Association (AHA) and the Infectious Diseases Society of America (IDSA) endorse preincisional antimicrobial prophylaxis and recommend single-dose cefazolin or cefuroxime for cardiac device proceduresReference Baddour, Epstein and Erickson 6 – Reference Bratzler, Dellinger and Olsen 9 ; continuing prophylaxis beyond 24 hours following skin closure is not recommended. The Centers for Disease Control guidelines for the prevention of surgical site infections (SSIs) recommend against post-closure antibiotics for clean and clean-contaminated procedures, including cardiac procedures.Reference Berrios-Torres, Umscheid and Bratzler 10 , Reference McDonald, Grabsch, Marshall and Forbes 11 The Surgical Care Improvement Project (SCIP) limited periprocedural antibiotic use to 24 hours (48 hours for cardiac surgery) as part of a larger surveillance and reporting program.Reference Rosenberger, Politano and Sawyer 12 The implementation of these measures are complimentary to recent studies that demonstrate expanded and prolonged antimicrobial prophylaxis following invasive surgical procedures is associated with increases in acute kidney injury (AKI) and Clostridium difficile infection (CDI).Reference Branch-Elliman, Ripollone, Strymish, Itani and Gupta 13 , Reference Branch-Elliman, Ripollone and O’Brien 14
The SCIP measures for outpatient procedures, including CIED procedures, did not include an assessment of postprocedural antibiotic use. Also, 2 recent studies have demonstrated that prolonged duration of antimicrobial prophylaxis is common following cardiac device implantation procedures; however, data are limited regarding the potential harms of this approach.Reference Branch-Elliman, Stanislawski and Strymish 15 , Reference Mehrotra, Gupta and Strymish 16 Thus, we sought to measure the association between prolonged postprocedural antimicrobial prophylaxis and incidence of adverse events in a large, multicenter cohort of Veterans Affairs (VA) patients.
Methods
Databases
The VA Clinical Assessment Reporting and Tracking (CART) program is a national quality initiative integrated into the VA electronic medical record; CART reporting is mandatory for all cardiac catheterization procedures and optional for electrophysiology procedures, including device implantations and revisions.Reference Branch-Elliman, Stanislawski and Strymish 15 , Reference Maddox, Plomondon and Petrich 17 Data collected prospectively as part of the CART program include procedure type, date of procedure, patient demographics (age, sex), vital status and comorbidities (eg, diabetes, renal disease, and heart disease). The CART data have been combined with other data from the VA Corporate Data Warehouse (CDW), including pharmacy and administrative data, to create a single national data repository.
Cohort development
Cardiac device procedures, including implantations and revisions of permanent pacemakers, ICDs, biventricular pacemaker ICDs, and biventricular pacemakers entered into the CART-EP program during the period from October 2008 to September 2016 were included. For patients with multiple procedures entered in CART, only the first was included. Cases that could not be matched to administrative data in the CDW were excluded. To ensure that postprocedural antimicrobials were administered for prevention and not for other clinical indications, a sample of cases from all participating facilities underwent manual review to delineate the reason for the antimicrobial order.
Outcomes and exposures
The primary outcome measure was incidence of laboratory-defined CDI during the 90-day period following the procedure (any positive C. difficile toxin assay, PCR or culture), based on established windows of C. difficile risk following antimicrobial exposure.Reference Hensgens, Goorhuis, Dekkers and Kuijper 18 All case patients with a CDI diagnosis underwent planned manual review by a trained clinician, including review of scanned-in paper records from outside facilities to identify all antimicrobial orders in the 6-month period prior to the procedure, the reason for the order, and confounding factors (e., recent hospitalization, comorbidities, proton pump inhibitor use, and previous CDI diagnosisReference Kuntz, Smith and Petrik 19 ) that may have impacted the relationship between prolonged prophylaxis and CDI diagnosis.
The secondary outcome was the 7-day incidence of postprocedural AKI, as defined by the Acute Kidney Injury Network (AKIN) recommendations; AKIN definitions are calculated based on a change from baseline renal function.Reference Mehta, Kellum and Shah 20
The primary exposure was postprocedural antimicrobial administration. Prolonged exposure was defined as inpatient or outpatient antimicrobial therapy lasting for >24 hours after the device implantation or revision procedure.Reference Branch-Elliman, Stanislawski and Strymish 15 , Reference Bratzler 21 , Reference Bratzler and Houck 22 Antimicrobial prescriptions filled within 7 days prior to the procedure were included if they continued for at least 24 hours after the procedure. Duration was determined based on the start date of the earliest antimicrobial prescription and calculated based on the number of doses dispensed. The impact of duration of antimicrobials was also evaluated for both outcomes. The risk of combination regimens was estimated for the AKI outcome. Combination regimens included multiple antibiotics prescribed at the same time or prescribed sequentially for the purposes of postprocedural prophylaxis.
Statistical analysis
Demographic and clinical characteristics of CIED patients by use of prolonged postprocedural antibiotics were compared. We used χ2 tests to compare categorical variables and Mann-Whitney tests for continuous variables. Epidemic curves of the outcomes of CDI and AKI were constructed using the R package epitools. 23
The association between prolonged prophylaxis and the 90-day CDI outcome was estimated and tested using Cox proportional hazards models. To account for clustering by facility, a robust estimator of the covariance matrix was used. Due to the low number of CDI events, only 1 confounding variable was included in the regression models. Based on previous reports of CDI risk factors,Reference Kuntz, Smith and Petrik 19 days of hospitalization prior to procedure (ie, 0 days, 1–6 days, and ≥7 days) were chosen. To ensure results were robust, sensitivity analyses were completed using other known risk factors (ie, age, diabetes, sex, nursing home admittance within 60 days prior to the procedure, proton pump inhibitor use, and year of procedure). Similar models were used to evaluate the effect of the duration in days of postprocedural antimicrobial use.
Generalized estimating equations were used to evaluate AKI, controlling for age, diabetes, sex, nursing home admittance within 60 days prior to the procedure, and year of procedure. AKI was evaluated as a binary outcome with an assumed binomial distribution, logit link function, and an exchangeable structure, to allow for clustering by facility. Similar models were used to evaluate the effects of (1) duration of postprocedural antimicrobial use; (2) combination antimicrobial therapies (2 or more, 1 only, and none); and (3) combination therapy involving vancomycin and another antimicrobial (yes vs no). Combination regimens may have been coadministered (eg, cefazolin plus vancomycin) or serial (eg, cefazolin followed by trimethoprim and sulfamethoxazole).
All analyses were completed using SAS version 9.4 software (SAS Institute, Cary, NC) and R version 3.4.0 software. 24
Results
In total, 6,832 CIED procedures in 6,801 unique patients at 42 different VA medical centers were entered into the CART-EP database during the study period, and 329 follow-up device procedures after initial CART entry in 305 subjects were excluded. An additional 6 patients were excluded due to lack of administrative data. Thus, the final cohort included 6,497 procedures in unique patients. Based on previous analysis, this cohort represents ~30% of CIED procedures within the national VA healthcare system.Reference Branch-Elliman, Stanislawski and Strymish 15
Participants were predominantly male (98.1%) and the median age was 71.4 years (interquartile range [IQR], 64.8–80.8) (Table 1). Most procedures performed were permanent pacemaker implantations (56.4%) and ICD implantations (28.4%). Prolonged antimicrobial prophylaxis was administered following 3,331 CIED procedures (51.2%). In these cases, the median duration of postprocedural therapy was 5 days (IQR, 5–7 days). A small minority of these were from prescriptions initiated prior to the procedure and continued >24 hours postprocedure (N=67, 2%). Approximately 25% of the cohort underwent manual review to determine reason for the postprocedural antimicrobial; manual review validated that the vast majority of prescriptions were for prophylactic purposes. Characteristics of patients who did and did not receive prolonged postprocedural prophylaxis are presented in Table 1.
Table 1 Patient and Procedural Characteristics for Index Device Procedure
Note. IQR, interquartile range; BMI, body mass index; INR, international normalized ratio; GFR, glomerular filtration rate; CLC, community living center; PPI, proton pump inhibitor; ICD, implantable cardioverter-defibrillator.
a Unless otherwise specified.
Clostridium difficile infection
Prolonged prophylaxis was associated with a higher incidence of CDI when compared to patients who received SSI guideline-concordant prophylaxis (Table 2; unadjusted hazard ratio [HR], 2.72; 95% CI, 1.46–5.07). The association between prolonged prophylaxis and CDI outcome persisted after controlling for preprocedural hospitalization (HR, 2.90; 95% CI, 1.54–5.46). Sensitivity analyses of models including other potential confounding variables had similar results.
Table 2 Results of C. difficile and Acute Kidney Injury Outcomes in Patients Receiving Prolonged Postprocedural Antibiotics Following Cardiovascular Implantable Electronic Device Procedures
Note. HR, hazard ratio; CI, confidence interval.
Note. OR, odds ratio; CI, confidence interval.
a Controlling for days of hospitalization prior to procedure (0, 1–6, ≥7).
b Interval value: Each additional day of antibiotics.
c Controlling for year of procedure, age, sex, diabetes, hospitalization/healthcare exposure (CLC) in the previous 60 days.
Among the 27 patients who developed CDI, one case was misclassified as a permanent device procedure (patient received a temporary pacing wire rather than a permanent device intervention). After excluding this case, the remaining 26 cases were incident cases in unique patients. N=11 CDI patients (42%) died within a 6-month follow-up period after CIED implantation; 8 of 11 patients who died had received prolonged antimicrobial prophylaxis, including 4 in whom the prophylactic antibiotic was the only antibiotic exposure in the 6 months prior to CDI diagnosis (Appendix 1). In addition to the 11 deaths, 1 patient who received prolonged prophylaxis developed toxic megacolon requiring colectomy.
Overall, 19 of 27 (69%) CDI patients received prolonged postprocedural antibiotics. Among the CDI patients, antimicrobials used for prophylactic purposes included cephalexin (n=9), clindamycin (n=4), and doxycycline (n=3), among others (n=3). Of the CDI patients, 17 patients, including 10 patients who received prolonged post-CIED antibiotics, had received other antibiotic therapy in the 6 months prior to CDI diagnosis for a variety of indications. In 9 of 26 patients, the prophylaxis was their only antibiotic exposure (Appendix 1). The use of clindamycin for prolonged prophylaxis appeared to be a particularly large driver of prophylaxis-related CDI; in patients receiving clindamycin, the prolonged prophylaxis was their only antimicrobial exposure during the 6-month period prior to the CDI diagnosis.
Acute kidney injury (AKI)
Among the 6,497 unique patients who had a CIED procedure, 2,797 (43%) patients had pre- and postprocedural creatinine measurements available for analysis. In this group, prolonged prophylaxis with a single agent was not associated with increased odds of AKI (OR, 1.26; 95% CI, 0.84–1.91). However, prolonged postprocedural prophylaxis with regimens containing multiple antimicrobials (≥2) was associated with increased AKI odds (OR, 4.16; 95% CI, 2.50–6.90). The association was particularly strong among patients who received combinations containing vancomycin (OR, 8.41; 95% CI, 5.53–12.79).
Sensitivity analyses
We repeated the statistical models after excluding the observation that was misclassified as a device procedure. The results did not change substantially for any outcome.
Discussion
Although there is no evidential basis to support that prolonged prophylaxis reduces CIED infections, this practice is common across many clinical settings, including VA hospitals, the private sector, and academic medical centers.Reference Branch-Elliman, Stanislawski and Strymish 15 , Reference Mehrotra, Gupta and Strymish 16 In this national, multicenter cohort of VA patients undergoing CIED procedures, prolonged prophylaxis was associated with increases in preventable patient harm, including CDI. Prolonged prophylaxis with combination regimens, particularly those containing vancomycin, were also associated with increases in AKI. Similar results have been found following traditional surgical procedures.Reference Branch-Elliman, Ripollone, Strymish, Itani and Gupta 13 , Reference Branch-Elliman, Ripollone and O’Brien 14 , Reference Harbarth, Samore, Lichtenberg and Carmeli 25 – Reference Bamgbola 27
Studies examining the impact of prolonged duration of prophylaxis following traditional surgeries demonstrate no reduction in the incidence of SSI.Reference McDonald, Grabsch, Marshall and Forbes 11 , Reference Kriaras, Michalopoulos, Turina and Geroulanos 28 This lack of efficacy resulted in the integration of a set measure for discontinuing postsurgical prophylactic antibiotics into SCIP. The goal of this measure was to promote discontinuation of prophylactic antibiotics within 24 hours of surgery end time (48 hours for cardiac surgery).Reference Bratzler and Houck 22 This measure was highly effective and was discontinued at the end of 2016 after compliance reached >98% among procedures included under its umbrella. Notably, outpatient SCIP measures did not include an assessment of the duration of antibiotic prophylaxis, and prolonged duration continues to be common in this setting.Reference Remmelts, Meine and Loh 29
Increasing the duration of antimicrobial exposure is strongly associated with CDI across many studies, including in surgical settings.Reference Poeran, Mazumdar and Rasul 26 , Reference Stevens, Dumyati, Fine, Fisher and van Wijngaarden 30 Older age and medical comorbidities also increase the risk of CDI. The population of patients undergoing CIED placement is enriched among older patients; increasing age was also associated with increased propensity to receive prolonged antimicrobial courses.Reference Branch-Elliman, Stanislawski and Strymish 15 In our study, prolonged postprocedural antimicrobial use was associated with a 3-fold increase in the odds of developing CDI compared to patients who received guideline-concordant regimens. Given the frequency with which VA patients use providers outside of the VA system, our estimates of harm from prolonged prophylaxis likely underestimate the true burden of these adverse events. The association between prophylaxis and postprocedural CDI was particularly striking among patients who received clindamycin as their prolonged prophylactic agent, and these findings are consistent with other studies demonstrating that clindamycin is a strong driver of CDI.Reference Vardakas, Trigkidis, Boukouvala and Falagas 31 , Reference Deshpande, Pasupuleti and Thota 32
Although some patients who developed CDI had additional antimicrobial exposures, prolonged prophylaxis adds to unnecessary cumulative exposure. The substantial morbidity associated with CDI underscores the importance of stewardship efforts for improving clinical outcomes.Reference Kwon, Olsen and Dubberke 33
In addition to CDI, this study also measured antimicrobial-associated risk of AKI. We found combination prophylaxis regimens—particularly those containing vancomycin—to be strong drivers of post-CIED AKIs. Reference Branch-Elliman, Ripollone and O’Brien 14 Given the retrospective nature of the study, we were not able to ascertain whether this was due to the nephrotoxicity of the agent, the patient population who received the vancomycin, or some combination of these factors. However, despite these limitations, these findings suggest that shortening the duration of periprocedural antibiotic prophylaxis—particularly when prescribing vancomycin—to preoperative and intraoperative doses may improve clinical outcomes by reducing AKIs without increasing CIED infections.Reference Harbarth, Samore, Lichtenberg and Carmeli 25 Additional consideration might also be given to limiting vancomycin use specifically to patients with known MRSA colonization or severe β-lactam allergy. This approach follows the evidence regarding the beneficial effect of preoperative antibiotic prophylaxis and would reduce the harms caused by longer antimicrobial exposure.
A common argument for continuing to administer prolonged prophylaxis is that lack of evidence of benefit does not equate to evidence that the practice is not effective. However, a simple decision analysis based on known adverse events demonstrates that no plausible effect estimate for CIED infection reduction due to prolonged prophylaxis would make the intervention attractive. Based on the attributable harm of prolonged prophylaxis (CDI risk difference, 0.29%), the number needed to harm (NNH) to cause 1 additional CDI case following receipt of prolonged postprocedural antibiotics is 335. Given an estimated incidence of CIED infection of 1%,Reference Greenspon, Patel and Lau 3 , Reference Voigt, Shalaby and Saba 4 prolonged postprocedural antimicrobials would have to reduce the absolute incidence of CIED infections by >30% to outweigh the preventable harm from additional CDI cases. An effect estimate of this size is highly unlikely given the existing data.Reference McDonald, Grabsch, Marshall and Forbes 11 Furthermore, these effect estimates do not weigh the mortality rate among the patients who developed CDI (42%) versus patients who developed CIED infections (4.6%–11.3%).Reference Sohail, Henrikson, Braid-Forbes, Forbes and Lerner 34 Accounting for this difference would drive the necessary effect estimate for CIED risk reduction even higher for the intervention to yield a clinical benefit. In addition, these effect estimates did not consider other adverse events, such as AKI and antimicrobial resistance, which further bias the decision against prolonged prophylaxis.
A major barrier to improving antimicrobial use in the electrophysiology laboratory may be a systematic bias in how feedback about adverse events is delivered to providers. Because the management of CIED infections requires repeat procedures and interventions, electrophysiologists are aware of the severity and harms of this outcome. However, electrophysiologists are often not informed about other types of adverse events, such as severe CDI, which can result from other aspects of the intervention, including the type and duration of antimicrobial prophylaxis. Thus, the clinical decision to prescribe postprocedure antimicrobials is based on high-quality feedback regarding one of the adverse outcomes (eg, CIED infections) but on low-quality information about others (eg C. difficile infections). Thus, one strategy to improve care might be to improve the surveillance feedback loop regarding the prevention of antimicrobial-associated patient harms, particularly in cases where no clinical benefit is expected.
This study has several limitations. First, CART-EP does not capture all electrophysiology laboratory procedures, but rather a subset based on voluntary entry. Thus, it is possible that this sample is not representative of all CIED procedures. Procedures and adverse events that occurred outside of the VA would not be captured in the VA CDW and therefore would not have been identified. However, this is unlikely to have significantly changed our findings. Recent studies suggest that in outpatient surgeries among dually eligible VA-Medicare patients, very few return to non-VA emergency departments or hospitals.Reference Mull, Gellad and Gupta 35 , Reference Mull, Rosen and O’Brien 36
In addition, VA pharmacy databases were used to identify antimicrobial prescriptions. Because of our reliance on VA data, it is possible that not all antimicrobial prescriptions may have been captured, particularly those that may have been written by non-VA providers. Thus, it is possible that not all antimicrobial orders were identified and that some of the observed cases of CDI were attributable to unidentified antimicrobial exposures. However, recent studies suggest that the vast majority of systemic antimicrobial orders in the postprocedural period are captured in the VA EMR.Reference Pindyck, Gupta and Strymish 37 Our findings are also strengthened by the manual reviews—including scanned-in paper records from outside hospitals—of the records for all laboratory-confirmed CDI patients.
Our power to adjust for multiple cofounding variables was limited by the small absolute number of CDI cases. However, we attempted to address this limitation with detailed manual review of all CDI cases to improve transparency. A major limitation of our AKI analysis was that pre- and postprocedural creatinine measurements were not available for all patients. This limitation may have introduced selection bias into this sample; it is possible that the patients who received multiple measurements were at inherently higher risk of both AKI and antimicrobial exposure. However, the association between AKI and vancomycin exposure has been well described, as has been the association between combination antimicrobial regimens and AKI, which puts our findings into a larger context.
In conclusion, prolonged postprocedural prophylaxis—particularly with combination regimens—may be associated with increased odds of preventable harm. These harms should be considered when weighing the risks and benefits of different antimicrobial prophylaxis strategies and durations. Future iterations of quality improvement measures, such as SCIP, should encompass a broader array of clinical settings, including the electrophysiology laboratory. Consideration should be given to improving surveillance and feedback to cardiac electrophysiology providers to include not only procedure-related harms but also harms associated with prevention interventions (eg, the downsides of unnecessary antimicrobial use). Also, these providers should be considered for inclusion in multidisciplinary stewardship initiatives. Expanding the scope of antimicrobial stewardship and infection prevention programs is essential as clinical care is increasingly delivered outside of inpatient settings of care and invasive procedures are performed outside of traditional operating rooms.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/ice.2018.170
Acknowledgments
This work was made possible by the help and assistance from the VA Clinical Assessment Reporting and Tracking program.
Financial support
The work was supported by American Heart Association Institute for Precision Cardiovascular Medicine (grant no. 17IG33630052). W.B.E. is supported by a Veterans Integrated Service Network (VISN)-1 Career Development Award.
Conflicts of interest
All authors report no conflicts of interest relevant to this article.
