Koen Y Blot, Jochen Bergs, Stijn I Blot, Dominique M Vandijck

Prevention of central line-associated bloodstream infections through quality improvement interventions: a systematic review and meta-analysis

Koen Y Blot, Jochen Bergs, Stijn I Blot, Dominique M Vandijck

Affiliations

Faculty of Medicine and Health Sciences, Ghent University, Ghent, Belgium (KY Blot BSc, Prof SI Blot PhD, Prof D Vandijck PhD); General Internal Medicine and Infectious Diseases, Ghent University Hospital, Ghent, Belgium (Prof D Vandijck PhD); and Health Economics and Patient Safety, Hasselt University, Diepenbeek, Belgium (J Bergs MSc, Prof D Vandijck PhD)

Corresponding author

Koen Blot, Ghent University, Faculty of Medicine and Health Sciences, 9000 Ghent, Belgium, , +32476992438

Keywords

Central line-associated bloodstream infection, catheter-related bloodstream infection, quality improvement intervention, compliance, meta-analysis.

Summary (39 of 40 words)

Nosocomial central line-associated bloodstream infections are associated with the use of central venous catheters. Quality improvement interventions aim to improve professional adherence to prevention measures. The results of the meta-analysis provide evidence that quality improvement interventions decrease infection rates.

Abstract (146 of 150 words)

Background

A systematic review and meta-analysis was performed to examine the impact of quality improvement interventions on central line-associated bloodstream infections in adult ICUs.

Methods

Studies were identified through Medline (1995–June 2012) and manual searches. Odds ratios (OR) and 95% CIs were calculated for infection rate reductions using a random-effects model. Subgroup analysisassessedthe impact of bundle/checklist interventions and high baseline rates on treatment effect.

Results

Forty-three studies were included.Meta-analysis of interrupted time series studies revealed an infection rate reduction three months post-intervention (OR 0.30; 95% CI 0.10–0.88; p=0.03). Likewise, before-after studies identified an infection ratedecrease(OR 0.39; 95% CI 0.33–0.46; p<0.001). This effect was more pronounced for before-after trials implementing bundle or checklist interventions (p=0.03).

Conclusion

This analysis suggests that quality improvement interventions contribute to the prevention of central line-associated bloodstream infections. Implementation of care bundles and checklists appears to yield stronger risk reductions.

1

Introduction

Central venous catheters, central lines that terminate at or close to the heart, are indispensable devices in the intensive care unit (ICU). Approximately half of patients admitted require a central line for infusion of medication, fluid, or blood products, hemodialysis, withdrawal of blood, or hemodynamic monitoring.However, use of these invasive devices predisposes for the development of central line-associated bloodstream infections (CLABSI). These serious, preventable complications increase patient morbidity, leading to increased length of hospitalization andresource use.[1,2]Moreover, CLABSI mightimpact mortality and compromise patient prognosis.[3-5]

Infection prevention measures during insertion or maintenance of central lines, such ashand hygiene, use of maximal sterile barriers upon catheter insertion, disinfecting the skin with chlorhexidine, optimal catheter site selection, and daily review of line necessity with prompt removal of unnecessary lines, have been proven to decrease the risk of CLABSI incidents.[6,7] The Institute for Healthcare Improvement (IHI) recommends use of the aforementioned items, in the form of a central line care bundle, to decrease CLABSI occurrence.Despite the presence of evidence-based interventions,summarized in guidelines,[8,9]CLABSI remain a substantial threat for hospitalized patients, with pooled estimated mean occurrence rates of 4.4 CLABSI per 100 devices inserted (95% confidence interval [CI] 4.1–4.9) and 2.7 CLABSI per 1000 catheter days (95% CI 2.6–2.9).[10]

In recent years it has become clear that the limiting factor to infection prevention is the implementation rather than publication of recommendations.[11]Introducing prevention measures may be hampered by multiple factors, such as lack of problem awareness, poor familiarity or non-agreement with the guidelines, low self-efficacy, inability to change previous practice, or lack of resources.[12,13]Quality improvement interventions such as personnel education, feedback reporting of infection rates, or use of central line care bundles and checklistsaim to decrease CLABSI incidence by improving adherence to prevention measures.[14] However, the efficacy of these interventions has not yet been fully assessed.

The objective of this study was to examine whether quality improvement strategies, defined as a combination of quality improvement interventions, reduce CLABSI rates in the adult ICU. Subgroup analyses were conducted to determine whether certain quality improvement interventions led to stronger infection rate reductions.
Methods

Search strategy

Medline was systematically searched (1995–June 2012) through PubMed MesH terms. Extra studies were identified by scanning reference lists of articles, manually and through Ovid and ScienceDirect databases. A combination of the following search terms was used: catheter-related infections/prevention and control; catheterization, central venous/adverse effects; catheters, indwelling/adverse effects; infection control/methods; infection control/standards; intensive care units; quality control; quality of health care; and bundle. (webappendix 1). The search strategy was developed by two authors (KB and DV) and performed by one investigator (KB).

Study selection

Eligible studies used before-after, interrupted time series (ITS), controlled before-after, non-randomized controlled trial, or randomized controlled trial study designs that complied with the Cochrane EPOC Group methodological criteria. ITSstudies report at least three data points before and after a defined point in time in which the intervention is implemented. Participants consisted of adult ICU patients with central line catheters. Trials implemented a quality improvement intervention aimed at increasing professional adherence to evidence-based infection prevention processes. The primary outcome measure was the CLABSI rate: the number of CLABSI per catheter days during pre- and post-intervention. Only English language papers were included. Medline search results were screened and excluded by their title or abstract (KB). Selected papers underwent a full-text assessment (KB) and eligibility issues were resolved between authors (KB, SB, DV).

Data extraction

Extracted data included author and year of publication, settings and study populations, study designs and periods, implemented quality improvement and preventive interventions in the baseline and intervention periods, number of CLABSI and catheter days, and applied CLABSI definitions. Secondary outcome measures such as length of stay, mortality, hospital costs and savings, and compliance reporting were likewise noted for the systematic review. Two authors (KB and DV) developed a data extraction sheet and one reviewer (KB) retrieved the data from included studies. Study authors were not contacted for additional data.To obtain effect sizes for ITS studies,infection rate data was extractedfrom figures published in original papers using the program PlotDigitizer.Results reported as a mix from both included and excluded study participants were included. When studies provided only two of three CLABSI rate parameters (number of CLABSI, catheter days, and CLABSI per 1000 catheter days), the third was derived through manual calculation. To avoid adding extraneous data, quality improvement interventions were classified under general headers (webappendix 2) and only preventive interventions described by the CDC guidelines[8] and applicable to the majority of ICU patients were noted.

Quality assessment

The Downs and Black checklist was applied to ascertain study methodological risk of bias (KB).[15]The tool consists of 27 questions concerning the reporting, external validity, internal validity, and power of non-randomized studies of health care interventions. Studies were scored based on whether they sufficiently fulfilled the items. The criteria for these individual checklist items were adapted for CLABSI studies following discussion between authors (KB, SB, DV). A power analysis was performed to determine whether studies had a sufficient sample size to detect an odds ratio (OR) of 0.50 with a p-value of less than 5%.

Statistical analysis

A random-effects meta-analysis was performed using the DerSimonian-Laird estimator to obtain the ORand95% CI for CLABSI rate reductions.Clinical heterogeneity was assessed by comparing study settings, CLABSI definitions, and intervention implementation methodology. The Higgins I2 test waspredefined to quantify heterogeneity(I2≤25% for low, 25%<I2<50% for moderate, and I2≥50% for high). Subgroup analysis by means of meta-regression was carried out for before-after study designsby comparing treatment effects between studies that did and did not implement bundle/checklist interventions and studies with a baseline rate above or equal to and below 4.0 CLABSI per 1000 catheter days. A funnel plot was constructed to assess publication bias and a sensitivity analysis was performed to identify heterogeneous studies that could influence the meta-analysis model. Monthly ITS studydata was standardizedby dividing the outcome and standard error (SE) by the pre-intervention standard deviation (SD). One study reported annual data points; these were used for the 12 and 24 month follow-up analyses.[16]The intervention effect was calculated through SPSS version 22, using segmented time series regression analysis,takinginto account time trend and autocorrelation among the observations. A negative change in level or slope indicates an intervention effect: an infection rate reduction.[17]A p-value less than 0.05 was used to denote statistical significance.

Role of the funding source

There was no funding source for this study. SB holds a research mandate of the Specific Research Fund at Ghent University. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.
Results

The search algorithm identified 633 abstracts (627 in PubMed and six in Ovid and ScienceDirect). Forty-threestudies, published in English between January 1995 and June 2012 consisting of 584ICUs, were included for meta-analysis (figure 1).Two studies[18,19] continued their quality improvement initiatives and republished their old data with new results.[7,20] The older study by Coopersmith et al.[18]was included for the meta-analysis of ITS studies and the paper by Pronovost et al.[19]wasaccessed to supplement information. One trial could not be included for meta-analysis because although baseline and intervention period interventions were qualitatively different, no new intervention types were implemented.[21]Another study included multiple sets of data, for which the set with the longest follow-up period was chosen.[22] Eleven studies could not included for ITS analysis because they implemented interventions in a step-wise manner.[20,21,23-31]

The majority of the 43 studies were prospective before-after intervention trials in medical-surgical ICUs (webappendix 3).Most trials implemented quality improvement interventions without simultaneously introducing new prevention measures and used CDC definitions for CLABSI diagnosis.

The 584 included ICUsconsisted of 564 adult, 11 pediatric,[22,32,33] and nine neonatal units.[22] Four studies, reported the number of adult ICUs studied, but did not specify the ICU type (n=270).[33-36] The remaining 294 adult ICUs were primarily mixed medical-surgical (n=135), medical (n=51), and surgical (n=61).

The analysis consisted of 35 before-after,[6,20,22,24-26,31-59]seven interrupted time series,[16,18,38,60-63] and one controlled before-after study.[64] Five of the ITS studies were eligible for inclusion in the meta-analysis consisting of primarily before-after study designs.[16,38,61-63]Duration of study periods ranged from nine[56] to 180 months,[16] with a mean length of 26.75 months.

Multiple quality improvement interventions were implemented in combinations (figure 2).Up to 14 different types of interventions were reported: personnel education, feedback reporting of infection or compliance rates, use of care bundles or checklists, prepackaged catheter carts or kits, surveillance of compliance with infection prevention processes, posters that raise awareness or report infection or compliance rates, designation of leaders or champions for CVC prevention measures, empowering nurses to stop incorrectly performed procedures, clinical reminders, distribution of fact sheets or leaflets, personnel training, organizational changes, and installation of hand-rub dispensers (figure 3).

There were levels of variability between implemented quality improvement interventions, the detail with which they were reported, and the methods used to apply them. Nearly all trials implemented educational interventions (n=33) consisting of single, monthly, quarterly, or yearly sessions on CLABSI prevention. Feedback reporting of infection or compliance rates to personnel occurred at monthly or quarterly intervals. Likewise, surveillance of compliance with preventive interventions was implemented daily, periodically, or at random intervals.

Half of trials implemented bundle or checklist interventions (n=20). Trials either initiated bundle interventions without checklists (n=2),[6,43] only checklists because bundles were used during the baseline period (n=9),[36,42,50,51,53,55,57,60,61] or both bundle and checklist interventions (n=9).[25,26,31,33,35,39,41,45,52]

Differing amounts of care items were grouped together to form a care bundle or checklist. Two trials[50,57] did not report which care items comprised their bundle and one used a checklist for one sole care process.[51] Other trials used either all five (n=7),[6,25,35,36,41,52,61] four (n=5),[26,31,39,45,60] three (n=3),[33,42,53] or two (n=2)[43,55] IHI care items in their bundle or checklist.

Four studies concomitantly targeted other healthcare-associated infections such as ventilator-associated pneumonia (VAP),[34] both VAP and catheter-associated urinary tract infections,[26,57] or VAP and surgical site infections.[32] Eight studies simultaneously initiated new prevention measures alongside their quality improvement interventions.[24,37,38,40,41,46,50,51]

The total number of CLABSI and catheter days were provided by 32 studies and manually calculated for eight.[6,16,34,39-41,48,53] The baseline CLABSI incidence varied between studies: rates ranged from 2.1[32] to 46.3 CLABSI per 1000 catheter days.[44]Sixteen trials[24,25,32-35,41,49-51,55-57,59,60,62]reported baseline infection rates equal to or lower than 5 CLABSI per 1000 catheter days and six trials above 15 CLABSI per 1000 catheter days.[16,22,39,44,47,58]

Total scores on the Downs and Black quality assessment tool ranged from 15[57] to 26,[20,22,47] with a median score of 21 (webappendix 4). The checklist revealed that two studies did not describe their CLABSI definition,[38,57] nine did not sufficiently describe their quality improvement interventions,[22,31,41,44,51,54,55,57,63] and 27 measured compliance with preventive intervention.[6,20,22,25,31,33-36,38,39,42-44,46-48,51,53-58,61,64] Twenty-two reported confounding factors such as duration of catheterization, patient characteristics or injury severity,[18,20,22,24-26,37,39,41,44,48-52,55,56,59,61,64,65]which were comparable between baseline and intervention in 18 trials.[18,20,22,25,26,37,44,48,50-52,55,56,59,61,64,65] However, two trials corrected for the measured differences in patient characteristics.[39,49]The power analysis revealed that studies tended to have either low (n=24) or high power (n=13).[6,16,20,22,24,32,35,36,45,47,50,54,63]

Nearly all trials demonstrated CLABSI rate decreases, only ten did not achieve statistical significance.[34,37,39,48,49,53,55,57,62,64] Two studies, reporting outcomes from ICUs separately, revealed non-significant results for neurosurgical, neurological, cardiothoracic, and coronary care units, yet decreased their total CLABSI rate.[44,45]

Meta-analysis was performed on 41before-after and seven ITS study designs to assess the impact of quality improvement interventions on the occurrence of CLABSI. The before-after trials showed a reduction in the CLABSI rate (OR 0.39; 95% CI 0.33–0.46; p<0.0001, figure 4) with high statistical heterogeneity (I2=85.4%). Analysis of six ITS studies, involving 11 ICUs, identified a change in level for the CLABSI rate at 3 months post-intervention (OR 0.30; 95% CI 0.10–0.88; p=0.028, figure 5) with low heterogeneity (I2=24.5%). Changes in infection rate slope (OR 0.81; 95% CI 0.59–1.13; p=0.216) and levels at 6 (OR 0.36; 95% CI 0.11–1.19; p=0.094), 12 (OR 0.17; 95% CI 0.02–1.27; p=0.084), and 24 months post-intervention (OR 0.052; 95% CI 0.003–1.02; p=0.051) trended towards reductions, yet were not significant.

Subgroup analysisof before-after trials revealed that the CLABSI risk reduction was significantly stronger (p=0.026; figure 4) in trials implementing care bundles or checklists (OR 0.34; 95% CI 0.27–0.41) than in those implementing other interventions (OR 0.45; 95% CI 0.36–0.55) (webappendix 5). Further analysis revealed that studies with baseline rates above 4.0 CLABSI per 1000 catheter days (OR 0.37; 95% CI 0.33–0.46) did not demonstrate significantly better risk reductions (p=0.18) compared to studies below this infection rate (OR 0.49; 95% CI 0.37–0.66).

Funnel plots displayed asymmetrical patterns for before-after, but not ITS study designs (webappendix 6).The results of the sensitivity analysis of before-after study designs suggest that two studies contribute to residual heterogeneity; removing them from the meta-analysis model would reduce variability between studies.[47,50] However, doing so does not affect the results,therefore the choice was made to retain these studies(webappendix 7).
Discussion

The results of the meta-analysis of 43 studies, involving 584 ICUs, provide evidence that quality improvement interventions reduce CLABSI rates in adult ICUs. The effect size of 41 studies was significant yet highly heterogeneous. This infection rate decrease was more pronounced in studies using bundles or checklists, suggesting that implementation of these interventions as part of a quality improvement strategy lead to stronger rate reductions.The change in infection rate level for six studies at 3 monthsfollow-upalso demonstratesthe beneficial impact ofquality improvementinterventions, but with a low level of heterogeneity.Changes in levels at the 6, 12, and 24 month periodwere not significant yet trended towards decreases. Reflecting the importance of multifaceted initiatives to improve patient safety culture,studies often implemented multiple quality improvement interventions, as recommended by guidelines.[8]

Strengths of this study include the comprehensive search strategy encompassing various quality improvement interventions, the methodological quality assessment of trials, and the random-effects model analysis with multiple studies and interrupted time series study designs. It is however hampered by certain limitations: a lack of randomized or controlled study designs, inconsistent reporting of prevention measure compliance, and heterogeneity. Before-after studies run a higher risk of bias due to their liberal study design: they hamper the ability to recognize phenomena that influence the CLABSI rate such as virulent epidemic outbreaks or spontaneous regression to the mean.[14] There is some evidence to suggest that the effects of quality improvement interventions are overestimated when based on before-after studies.Time series designslimit this risk of bias byattempting to detect whether an intervention has had an effect significantly greater than the underlying baseline trend.[66]However, since these designsrequire initiatives to begin at a well-defined point in time, 11 studies withstep-wise intervention implementation had to be excluded.