Introduction

Healthcare-associated infections (HAI) are a significant cause of morbidity andmortality for hospitalized patients accounting for approximately 100,000 deaths yearlyin the United States [1]. Though intensive care unit (ICU) beds make up the minority of hospital bedsnationwide, they account for the highest burden of nosocomial infections [2]. Specifically, in the ICU, central line-associated bloodstream infections(CLA-BSI) account for much of the excess morbidity, health cost expenditures, andmortality associated with nosocomial infections [3–6].

The risk of developing a CLA-BSI depends on a variety of factors such as the duration ofcatheterization, location of catheter, and type of ICU to which a patient is admitted [4, 5, 7]. Evidence-based interventions effective in combating CLA-BSIs include: usingchlorhexidine skin preparation and maximal sterile barriers (MSB) during insertion ofcentral venous catheters (CVC), use of checklists for insertion, using the subclavian orinternal jugular vein instead of the femoral vein, and daily review of line necessity [5–13].

These techniques have been validated in the literature and put together in a 'bundle',which was installed by the Institute for Healthcare Improvement (IHI) to help providersdeliver more consistent care [14–18]. However, many of these studies have focused primarily on the insertion ofthe central line [15, 17–19] rather than ongoing line maintenance. Other studies using compliance coupledwith adherence to safe line maintenance standards and prompt removal, despiteimprovement in CLA-BSI rates, have not necessarily documented rates below the NationalHealth Safety Network (NHSN) benchmarks for CLA-BSI and certainly continue to show ratesof CLA-BSI above the ultimate goal of 'near-zero' [15, 16]. There have been suggestions that the high sensitivity of the CLA-BSIsurveillance definition by the Center for Disease Control (CDC)/National Health SafetyNetwork (NHSN) leads to categorization of positive blood cultures as CLA-BSI when theymay not actually be related to infections, rather contamination [20]. Some authors have made suggestions that zero CLA-BSI may not be realistic atall [21]. Regardless of the cause of these positive blood cultures, they have apotential to lead to increased antibiotic use, removal of catheters with placement ofnew catheters and even increased hospital length of stay.

We implemented a systematic team approach with very aggressive interventions surroundingthe IHI CLA-BSI bundle resulting in marginal success toward the target of reducinginfections to near-zero in our ICU. However, our innovative approach toward implementingthese interventions allowed us to reexamine the central line bundle efficacy and augmentour process improvement strategy with accessory interventions as our unit's journeyprogressed to a near-zero rate of CLA-BSI.

Materials and methods

Design overview

This was an observational cohort study that used historical controls. The project wasdeemed as quality improvement by the Institutional Review Board of this organizationand need for research approval and informed consent was waived.

Setting and participants

The study setting was a 25-bed medical ICU located in 1200-bed tertiary care academichospital at the Ohio State University Wexner Medical Center. All patients admitted ortransferred into the unit were included in the intervention. The patient populationdid not include surgical ICU patients. The nursing to patient ratio averaged 1:1.5and varied between 1:1 and 1:2 as patient acuity mandated.

Routine surveillance for CLA-BSI

We measured CLA-BSI rates during the following time periods: baseline (January 2008to December 2009), intervention year (IY) 1 (January to December 2010), andintervention year 2 (January to December 2011). All blood cultures obtained frompatients admitted to the ICU were reviewed by the infection preventionist and allsuspected CLA-BSI were confirmed by an epidemiologist/infectious diseases physicianutilizing the definition put forth by the CDC through the NHSN [22]. We defined a central line as a catheter that ends in the superior orinferior vena cava at or near the heart. Specific lines present in our populationincluded peripherally inserted central catheters (PICC), central venous catheters(CVC), and pulmonary artery catheters (PA). Arterial lines were not included inCLA-BSI surveillance per NHSN definition. However, we have not noticed any centralarterial line infections during our routine surveillance. Total patient days werecalculated daily by number of patients on the ICU service census at midnight. Thepresence of at least one central line in a patient was counted as one catheter day inaccordance with the NHSN guidelines [22]. There were no changes to the epidemiology staff during the interventionperiod, nor were there any changes to the CLA-BSI definition utilized over the courseof the study.

Interdisciplinary team formation - December 2009

A multidisciplinary team that included an infectious diseases physician and infectionpreventionist, ICU medical directors (critical care physicians), nurse manager andclinical nurse specialists (CNS) was convened. Each individual's role in theperformance improvement process was clearly defined in the initial meetings.

Interventions related to central line bundle - January 2010

Several interventions focused on the central line insertion bundle, dressingmaintenance and line access practices were simultaneously introduced or reemphasizedto ICU physicians and nurses in January 2010. This marked the beginning of theintervention period.

1. At the start of each rotation, education was reinforced to all house staff to use the Vascular Access Selection Criteria to ensure proper selection of catheter site with emphasis on internal jugular or subclavian placement [23]. This education was part of a refresher course in the hospital simulation laboratory on line placement and sterile technique.

2. Lines in the ICU were placed by resident physicians, critical care or nephrology fellows, critical care attending physicians, or the hospital PICC insertion nursing team. Ultrasound was used to place nonemergent central lines.

3. The continued need for a CVC was reviewed daily, during interdisciplinary ICU rounds by the critical care fellow as part of the daily goals checklist, with removal of the catheter wherever possible [3–5]. In addition, the critical care fellow and the CNS reviewed line necessity during quality rounds that were conducted each afternoon.

4. Nurses placed peripheral intravenous catheters with ultrasound guidance wherever possible to avoid CVC placement and to facilitate removal.

5. Removal within 24 hours of all CVCs placed emergently, that is 'code lines' or any line placed without maximum sterile barrier precautions (sterile gown, sterile gloves, full-size sterile drape, face mask, cap, and chlorhexidine skin preparation solution). A label was used to identify these catheters as emergently placed central lines.

6. CNS led mandatory demonstration session for dressing change and proper line access on a manikin for all nursing staff at the beginning of the study. During this session, all nursing staff was assessed for competence in their dressing change and line access techniques. Chlorhexidine gluconate (CHG) Tegaderm™ dressings were used on all central lines from the beginning of IY1. Nursing performance was evaluated annually by unit management staff as part of annual mandatory education.

7. A CVC insertion checklist, with all requirements to comply with the sterile procedure for CVC placement, was attached to all central line kits. All CVCs placed were antimicrobial catheters. Nursing staff was instructed to use the checklist at the time of line insertion. All providers in the room were required to wear sterile cap, mask, and gloves. Nurses were empowered to stop procedures if sterile technique was not correctly employed. Arterial lines were placed in a similar fashion using full barrier precautions.

8.All CVC and PICC insertion trays were augmented with components to comply with the central line bundle including the use of chlorhexidine sponges for cleaning the skin.

9. The infection preventionist gave feedback to the unit on adherence to protocol based on the central line dressing maintenance audits performed each month and any CLA-BSI every week. These audits assessed compliance with keeping the dressings clean, dry and intact. In addition, staff nurses were encouraged to practice proper line access techniques on a monthly basis by the nurse manager via emails and during staff meetings.

Timely feedback on CLA-BSI occurrence and sentinel event investigation on CLA-BSIat the unit level - January 2010

Clinical Epidemiology provided weekly feedback on CLA-BSI occurrence to the unitleadership that included the ICU medical directors, CNS and nurse manager. The CNSresponsible for each investigation was able to evaluate practitioner variation,nursing variation, CVC access and blood culture collection techniques, and anatomicalCVC site of placement among other factors that may have contributed to each CLA-BSI.The nurse manager and medical directors shared the results of the investigation withthe unit's nursing and medical staff, respectively.

Positive reinforcement strategy - June 2010

We engaged hospital leadership to establish milestones for CLA-BSI avoidance (100,200, and 365 days) and provide incentives upon achieving the milestones to sustainCLA-BSI improvement. These included coat pins indicating the days without a CLA-BSIand recognition breakfasts for the unit staff.

Clinical epidemiologic investigation - November 2010

We conducted an epidemiologic investigation at the end of IY1 when, despite a modestimprovement in CLA-BSI rates, there was a continued CLA-BSI burden in the face ofimplementation of and compliance with aggressive measures toward the CVC bundle andline maintenance practices. We investigated the types of organisms causing CLA-BSIsand the location of CLA-BSI patients within the ICU to evaluate for clustering ofinfections. Upon noting a surge and clustering of vancomycin-resistantEnterococcus (VRE) faecium CLA-BSIs from certain ICU rooms,Epidemiology performed environmental cultures of 42 high-touch surfaces (HTS) withinthe rooms where patients with VRE faecium CLA-BSI were identified. HTScultured included bed rails, bedside tabletops, and keyboards, call buttons, supplycart handles and television remote controls.

Molecular epidemiologic investigation - December 2010

Based on the results of the clinical epidemiologic investigation, a molecularepidemiologic investigation was conducted. Molecular typing of VRE faecium isolated from blood cultures and patients' environment was performed usingrepetitive extragenic palindromic sequence-based polymerase chain reaction (rep-PCR)DiversiLab kits (bioMérieux, Durham, NC, USA) following the Diversilab™Enterococcus kit package insert and previously described methods [24–26]. An ATCC™ E. faecalis 51299 strain was used as controlalong with another E. faecium control strain obtained from a patient's bloodculture from a different area in the hospital. The modified Kullback-Leibler distancemethod was used to create a pairwise percent similarity matrix, and a dendrogram wasgenerated using the unweighted pair group method of arithmetic averages. Isolatessharing greater than 97% similarity and/or indistinguishable (no band difference)were grouped for further analysis. The graph overlay feature was utilized to observesmall differences between isolates that were otherwise not apparent on virtual gelimages. Each new rep-PCR pattern identified was based on one peak difference and wasassigned a sequential numeric classification based on the overlay. VRE organisms wereconsidered similar if they had a one-peak difference.

Environmental decontamination and nursing staff education - end of December2010

Hospital policy already specified that all patients with VRE required contactisolation including a gown and gloves for anyone entering the room. Based on theresults of the above investigation, in addition to continuation of contact isolationprotocols an intensive terminal cleaning of all ICU rooms was conducted. Each pod ofthe ICU was emptied for a day at a time by moving patients into a different pod tofacilitate this cleaning. The walls, floors and all surfaces and equipment in therooms were spot cleaned and wiped down with a disinfectant. Environmental Services(EVS) personnel created a dedicated cleaning team that was specifically trained toclean the ICU rooms. Nursing staff was educated about proper technique for obtainingblood cultures by the clinical nurse specialists to reinforce their knowledge.

Follow-up environmental culturing - March 2011

We repeated environmental culturing to document decontamination of VRE following deepcleaning of ICU rooms. Over 200 HTS from all ICU rooms were cultured for VRE toassess for continued effectiveness of cleaning by our EVS staff that were unaware ofthis surveillance.

Chlorhexidine gluconate bathing - April 2011

We introduced CHG bathing of all patients in the ICU in response to one VRE CLA-BSIthat occurred three months after environmental decontamination. This intervention wasintroduced to reduce the potential bacterial burden on patients [4]. Patients admitted to the ICU underwent a CHG bath on admission and dailywith a diluted solution of 4% chlorhexidine gluconate in tap water based on previousstudies showing eradication of VRE and methicillin-resistant Staphylococcusaureus (MRSA) colonization at this dose [27, 28].

Data collection

We measured CLA-BSI incidence per NHSN definition [29] before (baseline period) and after (post-intervention period) theimplementation of the 'line bundle' and subsequent process improvement methods. Thequarterly rate of infections was calculated as follows: (number of CLA-BSIs/number ofcentral line days) × 1,000 for each three-month period. Quarterly rates wereassigned to one of four categories based on when the study intervention wasimplemented: at baseline, during the early post-intervention period (year 1), or latepost-intervention period (year 2). We also collected data on the number of temporaryCVCs, including PICCs used in ICU patients over the study years. Device utilizationratio was calculated as follows in accordance with NHSN guidelines: number of deviceor catheter days/number of patient days [29]. Patient days were counted using the daily ICU census at midnight.

Outcome measures and study hypotheses

Primary outcome measure was quarterly CLA-BSI rate per 1,000 central line days.Secondary outcome measures were compliance with CVC insertion and dressingmaintenance practices. The primary study hypothesis was that the CLA-BSI rate wouldbe reduced by at least 50% after implementation of the study intervention as comparedto the baseline over a two-year intervention period. We did not evaluate the relativeeffectiveness of the separate components of the intervention.

Statistical analysis

As used in previous studies [5], and because CLA-BSIs are rare events, a Poisson regression analysis wasused to generate an incidence rate ratio (IRR) compared with baseline CLA-BSI rates(Stata software, version 10; Statacorp, College Station, TX, USA).

Results

CLA-BSI reduction

There were 2.65 infections per 1,000 catheter days (30 CLA-BSIs in 11,317 centralline days) in the ICU in the baseline period (Table 1). The netinfection rate at the end of the two-year intervention period decreased by 53% to1.24 infections per 1,000 catheter days (14 CLA-BSIs in 11,271 central line days)with IRR of 0.47 (95% CI = 0.25 to 0.88, P = 0.019) (Table 1). During IY1, the CLA-BSI rate was reduced to 1.97 per 1,000 catheterdays (11 CLA-BSIs in 5,589 central line days); the IRR was 0.74 (95% CI = 0.37 to1.65, P = 0.398). During IY2 (months 13 to 24 of the intervention), theCLA-BSI rate further reduced to 0.53 per 1,000 catheter days (3 CLA-BSI in 5,682central line days); the IRR was 0.20 (95% CI = 0.06 to 0.65, P = 0.008) with80% reduction compared to the baseline period. There were zero CLA-BSIs for the last10 months of the intervention period included in the analysis (Figure 1) and with zero CLA-BSIs for a total of 15 calendar months.

Table 1 CLA-BSI incidence rate/1000 patient days, incidence rate ratio (IRR) in thepost-intervention period compared to baseline period.
Full size table
Figure 1
figure 1

Central line-associated bloodstream infections, compliance with central lineinsertion and dressing maintenance during the study period. NHSN,National Health Safety Network.

Full size image

Compliance with insertion, dressing maintenance, and line removal

Compliance with CVC insertion practices was high based on the audits; compliance withCVC dressing maintenance increased steadily and remained high at 80 to 100% duringthe intervention period assessments (Figure 1). Despite a dailychecklist to remind the team of prompt removal of unnecessary CVCs, deviceutilization ratio (central line days/patient days) did not significantly changeduring the intervention (Table 2).

Table 2 Monthly CLA-BSI during the study period, device utilization ratio and organismscausing CLA-BSI in each month.
Full size table

Clinical epidemiologic investigation

Seven out of eleven CLA-BSIs in 2010 were caused by VRE faecium (Table 2). Four out of seven patients with a VRE CLA-BSI were located inpod B; of the other three patients, one was in pod A, one in pod C and one in pod D(Figure 2). Out of the four patients with a VRE CLA-BSI in podB, two were located in one room during different months; the other two were locatedin another room during different months. Selective environmental culturing for VRE ofHTS was performed initially only in the rooms that harbored patients with a VRECLA-BSI in 2010. Out of 40 HTS sites cultured in five patient rooms, 8/40 (20%) sitescultured positive for VRE faecium. The HTS that tested positive were bedrails (three cultures), supply cart handles (two cultures), computer keyboard (oneculture), call button (one culture), and bedside table (one culture). Following deepenvironmental cleaning, all cultures were negative for VRE from all previouslypositive VRE patient rooms. Additional follow-up cultures of HTS in all ICU roomsperformed four months later did not reveal any positive VRE.

Figure 2
figure 2

Intensive care unit layout; four patients with VRE central line-associatedbloodstream infection were located in two rooms of pod B - two in one roomand two in another room. One patient each was located in each of theother pods.

Full size image

Molecular epidemiologic investigation

The rep-PCR patterns of all VRE faecium isolates were depicted to be inPatterns 1 (Key numbers 5, 6), Pattern 3 (Key numbers 1, 2, 7, 8, 9, 15, 16), Pattern5 (Key numbers 3, 4, 13, 14), Pattern 9 (Key numbers 10, 11, 12, 17), Pattern 10 (Keynumber 18) and Pattern 11 (Key number 19). Based on the overlay, Pattern numbers 1,3, 5 and 9 had a single band difference and were related. Pattern numbers 10 and 11were different from Pattern numbers 1, 3, 5 and 9. The E. faecium controlstrain obtained from a blood culture from a patient in a different area (Pattern 9;Key number 17) had a one-band difference from Patterns 1, 3, and 5. It was related tothe patient and environmental samples under investigation. Control VRE faecalis strain ATCC™ 51299 (Pattern number 11; Key number 19) was unrelated to thepatient and environmental samples under investigation (Figure 3).

Figure 3
figure 3

Rep-PCR analysis of VRE isolates from blood cultures and environmentalcultures related to patients with VRE CLA-BSI. CLA-BSI, centralline-associated blood stream infections; P, pattern number; VRE,vancomycin-resistant Enterococcus faecium. Control (row 19) =E.faecalis ATCC™ 51299 control strain.

Full size image

Sentinel event investigation of residual CLA-BSI

All seven residual CLA-BSIs were caused by VRE faecium. All VRE werecultured from one bottle from a CVC with one or more negative culture bottles fromsimultaneous peripheral blood cultures in all patients. Scenarios where bloodcultures were obtained included: in response to hypotension caused bygastrointestinal bleed in two patients, pulseless electrical activity (PEA) in twopatients (one after a radical neck dissection, one after transjugular intrahepaticportosystemic shunt (TIPS) procedure), one temperature recording of 100.8 in apatient with deep venous thrombosis (DVT), in response to leukocytosis in a patientwho was afebrile and was treated with corticosteroids, and in response toleukocytosis in a patient with acute coronary syndrome. All seven patients had theircatheters removed. Three patients died within three days of diagnosis of bacteremiasecondary to their presenting diagnosis with none of the deaths attributed to theCLA-BSI. Four patients were treated for bacteremia for 14 days.

Discussion

CLA-BSIs continue to be a challenge in healthcare delivery, especially in the ICU, wherenearly 50% of patients will have a CVC inserted at some point in their care [30]. CLA-BSIs are responsible for significant morbidity resulting in extendedhospitalizations, hospital costs accounting for a loss of up to $26,000 per CLA-BSI [31], and increased mortality accounting for approximately 100,000 deaths a year [1]. Though multiple investigators have reported interventions surrounding thecentral line bundle to successfully reduce the burden of CLA-BSIs in their units, fewhave reported the results of continued process improvements and secondary interventionsthat can be utilized when, as is often the case, compliance with the CVC bundle alonehas not resulted in a significant reduction in these infections.

Our unit's culture in the baseline period likely mirrored what may be the standardculture in many institutions. Physicians and nurses were aware of the dangers ofCLA-BSIs and educated on the expectation for strict adherence to sterile techniques forinsertion and access of central lines. Physicians were expected to utilize full-barrierprecautions and ultrasound guidance for CVC insertion. However, as with many academicinstitutions, there was a high turnover of new trainees into the environment making asustained cultural change more difficult. Into this environment, our epidemiology teamchampioned a systematic team-oriented approach to optimally reduce CLA-BSI to anear-zero rate. The implementation of a central line 'bundle' resulted in excellentcompliance with sterile insertion techniques and substantial improvements in centralline dressing care. However, despite improved compliance with the bundle, our initialefforts resulted in only a modest improvement of our CLA-BSI rate and the use of CHGTegaderm™ dressings may or may not have contributed to this modest improvement; wewere still far from our goal.

Our innovative team approach engaged hospital epidemiologists and clinicians with weeklyCLA-BSI surveillance followed by immediate, unit-level, root cause analysis, whichfacilitated feedback to clinicians in a timely fashion when they still remembered thecircumstances surrounding the infection. In the majority of our CLA-BSIs in IY1 wesuspected potential blood culture contamination with VRE coupled with clinicaldeterioration from a noninfectious etiology as a likely cause of persistence of CLA-BSIsdespite our compliance with bundle elements. Our interdisciplinary approach withinfection prevention experts, critical care physicians and nursing staff working as oneteam facilitated an investigation that identified geographic clustering of VRE cases inour ICU. As VRE can cause both monoclonal and polyclonal outbreaks [32–35], we performed environmental cultures coupled with a molecular epidemiologicinvestigation based on our finding of geographic clustering and our suspicion of bloodculture contamination. This documented environmental contamination with VRE anddemonstrated the genetic similarity between environmental VRE and the VRE associatedwith these CLA-BSIs. We then implemented a very aggressive intervention of cleaning ourICU by systematically emptying one ICU pod at a time and engaging our EnvironmentalServices team in our process improvement strategy. This investigation also recognizeddeficiencies in the structure of our environmental cleaning plan and identified the needto have more highly trained cleaning staff that was dedicated to an ICU. As a result, wehave now employed environmental cleaning teams that are dedicated to ICUs and aretrained to clean around ICU equipment in our institution. Studies have shown eliminationof VRE colonization events by meticulous attention to environmental cleaning [36, 37], but very few have utilized environmental decontamination as an interventionto achieve a reduction in VRE CLA-BSIs. We observed a decline in our VRE CLA-BSI uponimplementation of meticulous environmental cleaning. We chose not to conduct admissionand/or weekly surveillance cultures for VRE colonization followed by contact isolationas an intervention as it is an expensive strategy. Instead, based on our investigationand findings, we focused on improving environmental cleaning and reducing potentialbacterial burden on our ICU patients [4, 27, 28].

Based on our experience, there can be an underlying cause for reminiscent CLA-BSIs afterimplementation of a process improvement project emphasizing bundle compliance. How todeal with these residual infections is not frequently reported in the literature. Thereare many potential interventions including patient cohorting, purpose-made cathetersterilizing devices, dedicated line teams or, as in our case, environmentaldecontamination to name a few. Deciding which intervention to use next could bedifficult and using all or some interventions randomly can be costly. We suggest thatunit leaders should investigate the root cause of their residual infections rather thanimplementing further measures piecemeal.

This work demonstrates the success of this positive approach to the issue of CLA-BSIs,investigating the underlying cause of reminiscent CLA-BSI thereby sustaining CLA-BSIreduction. We have demonstrated that with an iterative team approach and by eliminatingunderlying causes of residual CLA-BSI and basing our approach on internal evidence forthe need for further interventions, CLA-BSI reduction is possible and sustainable. Ourefforts to reduce CLA-BSI incidence demonstrate a synchronized model formultidisciplinary teams, which included a hospital epidemiologist and administrativeleaders to increase compliance with bundle elements and to decrease blood culturecontamination. We engaged hospital leadership to establish milestones and recognize theunit at the institutional level upon milestone achievement. This strategy resulted in apositive effect on morale, and facilitated staff compliance with the new standards ofcare. At the same time, it promoted the culture of safety: no one wanted to be thepractitioner that started the clock back to zero days since the last CLA-BSI. We believethat these two strategies played an important role in sustaining our success withCLA-BSI reduction.

We observed a stability of the device utilization ratio, which may be due to the factthat a patient with multiple lines; for example, a triple-lumen CVC, and dialysiscatheter counts as one line day, even with discontinuation of one of the lines thepatient would still count as a central line day. Thus the device utilization ratio mayunderrepresent how many lines were being removed even with aggressive removal based on adaily goals checklist. While the burden and exposure may be reduced, the deviceutilization ratio will not account for this. This may argue for counting the number oflines or even lumens for a lumen/patient/day count. Unfortunately, this may beunfeasible in most settings leading to inaccurate counts. However, improved line daycounts might be possible in future studies as electronic medical record tracking ofvascular access days in ICUs increases.

Our experience also validates the existent concerns that all CLA-BSI may not bepreventable given the high sensitivity and low specificity of CDC/NHSN definition, assome of these positive blood cultures with significant pathogens may be related tocontamination or catheter colonization [20]. This may especially be true in units colonized with resistant bacteria suchas VRE or MRSA. Eventually, these units may be penalized financially if, as planned,healthcare providers in the United States receive only limited reimbursement from theCenters for Medicare and Medicaid Services (CMS) for any CLA-BSI acquired in thehospital, since this is now accepted as 'preventable' [21]. An acceptable compromise may envision CMS accounting the reasons why thereminiscent CLA-BSIs were not preventable on a case-by-case basis (for example, an'infection' per definition may not be an infection at all or the 'infection' may be asecondary bloodstream infection with another potential primary site) instead of basingreimbursement merely on CLA-BSI rate per institution.

Our study does have several limitations. First, we studied one particular ICU populationwith a homogeneous group of nurses and house staff. Thus, our results may not beextrapolated to surgical, cardiac, or other ICU populations. However, the principle ofour process is the continuous reporting of CLA-BSIs and refinement of unit-specificinterventions, which should be reproducible in other ICU populations. Second, we did notcontinuously monitor central line insertion and dressing maintenance throughout theintervention period. Even at baseline, the compliance with sterile technique forinsertion was extremely high and in addition, these audits were labor intensive and wefelt that intermittent auditing was an acceptable alternative to continuous audits oncecompliance with the sterile line insertion techniques and dressing maintenance had beenachieved. In addition to intermittent audits, we continue to conduct educationalsessions with periodic (yearly) reinforcement of education related to central linemaintenance and access. Our CNSs continued to investigate any new CLA-BSIs includingassessment of line access practices and insertion technique. In addition, house stafftrainees are given a half-day training session in our simulation laboratory emphasizingcentral line placement prior to each rotation in the ICU. Third, we only cultured theICU rooms where patients were diagnosed with VRE CLA-BSI during the first round ofenvironmental culturing. However, we felt the high prevalence of VRE on HTS in the roomscultured justified the deep cleaning of every room in the unit. In addition, we culturedall HTS in all ICU rooms during our follow-up environmental culturing to documentdecontamination of the rooms post environmental cleaning and to assess continuedeffective cleaning by our environmental staff. Fourth, we monitored hand hygiene uponentry and exit of the rooms through audits performed by the unit staff, which showedvery high compliance of greater than 90% during early months of the study, but we didnot monitor whether staff cleaned hands prior to accessing central lines. Doing suchaudits anonymously for accurate data gathering would be a challenge but we believe thatwith increased awareness of CLA-BSI prevention, this practice improved over time duringthis study. Last, we did not evaluate the relative effectiveness of the separatecomponents of the intervention related to central line bundle. However, our goal was toattain maximal improvement of patient safety in our ICU; this quality improvementinitiative was designed to optimize the use of well-documented best practices andoffered the greatest probability of reducing CLA-BSI incidence. We saw a stepwisedecline in CLA-BSIs as we strategically introduced accessory interventions beyond thecentral line bundle. It remains unknown whether implementation of CHG bathing alone orenvironmental cleaning alone (given enough time) in the absence of the other would haveeliminated reminiscent sporadic CLA-BSI in our ICU. However, we believe that deepenvironmental cleaning along with correction of deficiencies in the daily environmentalcleaning for ICUs was necessary to reduce bacterial burden in the environmentsurrounding the patients in our ICU. We saw a decline in VRE CLA-BSIs after thisintervention but extended the decontamination approach to patients with anotheroccurrence of VRE CLA-BSI since our follow-up cultures did not reveal VRE on multipleHTS. Our approach demonstrates our zero tolerance for CLA-BSI occurrence in our ICU. Wealso validated the currently existing thoughts about the high sensitivity and poorspecificity of CDC/NHSN definition for CLA-BSI. We showed that multidisciplinary effortstoward clinical epidemiologic investigation could actually lead to a near-zero rate evenin a complex ICU in a tertiary care hospital such as ours despite these concerns.

It may be argued, in light of the fact that the majority of our residual CLA-BSIs weresuspected to be contaminants, that our improvement comes at a significant resource costand that we implemented aggressive interventions to try to curtail nonexistentinfections in the setting of contamination; however, we submit that these werereasonable interventions since we suspected central line contamination which, in thesetting of high environmental burden, reflects the risk of impending invasive infectionsunless action is taken. In addition, it is difficult for physicians to choose not totreat a patient who has a positive blood culture with a significant pathogen.Retrospectively, we suspected that the majority of these cultures were contaminants;however, the treating team presented with the culture results on critically ill patientsin real time had to treat the culture as a real result. Thus the elimination of thesepotentially erroneous CLA-BSIs has the added effect of reducing unnecessary antibioticuse. Unnecessary antibiotic use, however, can lead to increased hospital length of stayamong other complications thus making it important to eliminate blood culturecontamination via both compliance with line insertion bundle and environmentaldecontamination. We encouraged clinicians to avoid drawing blood cultures from theexisting catheters at the same time as we instituted these interventions. However, wesaw improvement in CLA-BSI before there was a change in this behavior. We are slowlyseeing a change in this behavior and are continuing to work on changing the culture ofdrawing blood cultures from existing CVCs. We felt that it was important to address therecognized deficiencies in the structure of our environmental cleaning plan and addressthe environmental contamination as soon as we identified the problem.

Conclusions

In conclusion, strict adherence to the central line bundle is essential to prevention ofCLA-BSIs, but may not completely eliminate these infections as blood culturecontamination contributes to CLA-BSIs that are detected by CDC/NHSN surveillancedefinition. Efforts to further reduce residual CLA-BSIs require a strategicmultidisciplinary team approach focused on epidemiologic investigations of practitioneror unit-specific etiologies. Continuous process improvement can then be targeted atlocal factors contributing to a CLA-BSI, such as environmental contamination in ourcase, with evidence-based interventions. Sustained reduction of CLA-BSIs requireslongitudinal support of hospital and unit leadership to continue to improve the care ofour most vulnerable ICU patients.

Key messages

Our study methods and findings include:

• Use of a multidisciplinary team including Clinical Epidemiology for root causeinvestigations.

• With reeducation and reemphasis on the central line bundle we achieved modestreductions in our rate of central line-associated bloodstream infections (CLA-BSIs)during our first year.

• During the second year, we utilized a root cause analysis approach toinvestigating our CLA-BSIs leading to the discovery of environmental contamination anderadication of CLA-BSIs with patient chlorhexidine bathing and deep environmentalcleaning.

• Use of PCR techniques to confirm environmental contamination as the source ofCLA-BSIs.

• Discussion of the CDC definition of CLA-BSI that includes inclusion ofpotentially contaminated line cultures in the calculated rate of CLA-BSIs for theunit.