JSTOR

Methods

In 2005, 26 hospitals registered for a collaborative quality improvement project (the collaborative) to reduce the rate of CVC‐associated BSIs. The collaborative was available only to member hospitals of Child Health Corporation of America (CHCA), a business alliance of 42 freestanding children’s hospitals. This was an observational study with historical controls limited to pediatric critical care patients hospitalized in freestanding pediatric ICUs or cardiac ICUs. The intent of the collaborative was to accelerate translation of scientific evidence to clinical practice, enhance patient safety, produce better clinical outcomes, and reduce the costs resulting from CVC‐associated BSIs. CVC‐associated BSI protocols were systematically developed, including catheter insertion and maintenance bundles (Table 1), daily review of CVC necessity, and daily goals. The primary goal of the intervention was to achieve either a 50% reduction in the CVC‐associated BSI rate or a rate of 1.5 CVC‐associated BSIs per 1,000 CVC‐days in each ICU at the end of a 9‐month improvement period. The occurrence of a CVC‐associated BSI was defined in accordance with CDC National Nosocomial Infections Surveillance system definitions.⁵ Additional goals were to double the number of days between CVC‐associated BSI occurrences, to use the CVC insertion bundle during CVC placement for 95% of patients, to use the maintenance bundle during CVC manipulation for 95% of patients, and to perform daily assessments for catheter necessity in 95% of patients with a CVC. Compliance with a bundle was defined as completion of all elements of the bundle. All ICU patients with a CVC (including peripherally inserted central catheters) were included. Data were collected using prospective surveillance, review of clinical documentation (eg, daily goals sheets), and in‐person observations. Data were entered into a secure online database designed by and located at the Institute for Healthcare Improvement. Following the 9‐month improvement period, centers were encouraged to continue to collect and report data by means of the same measures and protocols for an additional 12‐month “sustain” period.

Table 1.  Central Venous Catheter Insertion and Maintenance Bundles Developed for the Study

Open New Window

Design. The collaborative process was based on the model for improvement as developed by Associates in Process Improvement and adopted by the Institute for Healthcare Improvement. The collaborative was led by 2 improvement‐trained CHCA directors and a panel of nationally recognized experts in infectious diseases and critical care, 2 improvement advisors from the National Initiative for Children’s Healthcare Quality, clinicians from CHCA hospitals (2 pediatric infectious diseases physicians, a pediatric intensivist, a pediatric nurse, and an epidemiologist), and 2 performance consultants from Convergent HRS (Weston, Florida). The panel met to review the literature and develop the CVC‐associated BSI prevention intervention bundle. These interventions were evidence based and included recommendations from the CDC guidelines,¹⁴ published studies, and the Institute for Healthcare Improvement's Save 100,000 Lives Campaign. The expert team defined 7 measures to monitor progress (see Appendix, Table A). Because multiple improvement practices were implemented simultaneously, the impact of individual improvements could not be evaluated for effectiveness. Hospital teams were encouraged to collect data at various times of day and shifts and across other variables within the ICU.

Participants collected baseline data and shared their findings with the other teams at the initial meeting. Teams also attended a second meeting to review their progress, share their successes, and discuss challenges. Postintervention data were submitted monthly, including team self‐assessment scores. Additionally, teams submitted a qualitative report, explaining the month’s barriers, successes, lessons learned, and next steps. CHCA produced a hospital‐specific report each month, advising and challenging teams on improvements for the next phase. CHCA made the official determination of assessment scale ratings for each reporting period. In addition to hospital‐specific reporting, CHCA produced aggregate reports monthly to monitor and communicate the progress of the entire collaborative. CHCA also communicated progress to key stakeholders on an ongoing basis (eg, chief executive officers and quality and safety leaders). At the conclusion of the 9‐month improvement project, data submissions and conference calls were conducted quarterly for an additional 12 months to ensure that improvements were sustained.

CHCA obtained approval for data management and analysis from Western Institutional Review Board (Olympia, Washington). Each participating site obtained institutional review board approval before study participation, if deemed necessary by that hospital team. Some institutional review boards considered the work to be quality improvement and therefore not subject to their oversight. Because the interventions were not specific to individual patients, informed consent was not required.

Statistical methods. The measures data were not normally distributed. Thus, the measures were presented on a quarterly basis by use of medians with interquartile ranges (IQRs; ie, first quartile to the third quartile), and nonparametric testing was used. The unit of analysis was the hospital. The number of days between infections can be analyzed only for individual hospitals (not in aggregate), so those results are not included in this article. Some hospitals had missing monthly data on the measures. Before computing the quarterly data, we imputed these missing values¹⁶ by means of a “last point carried forward” strategy. The imputation rates ranged from 1.04% (for CVC‐associated BSI rates) to 11.85% (for daily goals prevalence). A sensitivity analysis was conducted by excluding hospitals that did not report baseline data. Doing so produced no change in the results, so all hospitals were included in this study.

Where baseline data were available, we analyzed whether a statistically significant change occurred from baseline to the last 3 months of the study by means of the Wilcoxon rank sum test or the Fisher exact test. For process measures that did not have baseline data, we compared the last 3 months of the study to the first 3 months of the study by means of these statistical tests. The reported P values are 2‐sided, and a P value of <.05 was considered to indicate a significant difference. All analyses were performed using SAS, version 9.1 (SAS).

Results

Twenty‐six hospital pediatric ICUs participated in the collaborative. Approximately 80% of the pediatric ICUs were extracorporeal membrane oxygenation and transplantation centers, conducted cardiovascular surgery, admitted trauma patients, and were affiliated with residency training programs. Just over half (15 [58%]) of the pediatric ICUs were in academic medical centers, and 8 (31%) of the pediatric ICUs admitted burn patients. Hospitals were instructed to provide 2–9 months of baseline BSI rate data (with regard to the period immediately before implementation). The median number of baseline months submitted was 6 (range, 1–15). The collaborative baseline median CVC‐associated BSI rate was 6.3 CVC‐associated BSIs per 1,000 CVC‐days (IQR, 5.0–8.9 CVC‐associated BSIs per 1,000 CVC‐days). In the final quarter of the collaborative, the median rate decreased to 4.3 CVC‐associated BSIs per 1,000 CVC‐days (IQR, 2.6–7.6 CVC‐associated BSIs per 1,000 CVC‐days; ); this is a median proportional reduction of 32% (Table 2 and Figure). Seventeen (65%) of the 26 participants documented a decrease in their CVC‐associated BSI rate. Nine (35%) teams achieved or surpassed the established collaborative goal of at least 50% reduction from baseline.

Table 2.  Reduction in Rate of Central Venous Catheter (CVC)–Associated Bloodstream Infections (BSIs) From Baseline for Pediatric Intensive Care Unit Collaborative Participants

Open New Window

(24 KB)

Figure.  Graph of reduction in rate of central venous catheter (CVC)–associated bloodstream infections (BSIs) for pediatric intensive care unit collaborative participants over time.

Open New Window

The total number of CVC‐associated BSIs prevented during the collaborative was computed on the basis of the difference between the baseline CVC‐associated BSI rate and the CVC‐associated BSI rates reported during each month of the collaborative. Across all teams, an estimated 69 CVC‐associated BSIs were prevented. On the basis of pediatric studies that estimated a cost range of $39,219¹⁵ to $46,133⁶ per BSI, the total estimated CVC‐associated BSI cost avoided across all teams was approximately $2,940,000.

Process measures were analyzed by comparing the median rates for the first quarter of the intervention phase with the median rates for the last quarter of the intervention phase. Baseline data were unavailable and/or not applicable for recommended processes. The median daily goals prevalence rose from 84% (IQR, 45.3%–100.0%) to 97% (IQR, 85.9%–100%; ) (Table 3). Nine teams (35%) achieved or surpassed the 95% prevalence goal for daily goals. Median insertion bundle compliance increased from 82% (IQR, 53.1%–94.0%) to 94% (IQR, 85.7%–100.0%; ) (Table 3). Ten teams (38%) achieved or surpassed the 95% insertion bundle compliance goal. Median maintenance bundle compliance rose from 86% (IQR, 78.3%–100%) to 100% (IQR, 94.5%–100%; ) (Table 3). Fourteen teams (54%) achieved or surpassed the 95% maintenance bundle compliance goal. There was no change in pediatric ICU patient length of stay (4.4 vs 4.5 days). The prevalence of CVC use was higher at the end of the collaborative than at baseline (59% vs 62%; ) (Table 4). The interventions were expected to decrease (rather than increase) catheter use, suggesting that this increase in catheter prevalence may be related to patient factors.

Table 3.  Process Measures Over Time, April–December 2005

Open New Window

Table 4.  Catheter Prevalence From Baseline to Final Quarter

Open New Window

Twenty hospitals completed the additional 12‐month sustain period. For these hospitals, BSI rates at the end of the initial collaborative (4.0 infections per 1,000 CVC‐days) were sustained in the final quarter of the sustain period (3.5 infections per 1,000 CVC‐days; ). Eleven teams (55%) reported the same or lower BSI rate, 5 teams (25%) did not submit enough data to evaluate, and 4 teams (20%) reported an increase in BSI rate, compared with the end of the original improvement period. Three of the 4 teams with an increase in BSI rate during the sustain period reported sustained or improved process compliance; 1 hospital did not report process data. An estimated 198 infections were prevented during the sustain period, for an additional cost avoidance of approximately $8,450,000. Median compliance with process measures remained high (daily goals, 100%; insertion bundle, 90%; maintenance bundle, 94%) through the final quarter of the sustain period.

Discussion

The intent of this CHCA‐led collaborative was to reduce the rate of CVC‐associated BSIs in pediatric ICU patients. Through increased compliance with evidence‐based prevention bundles of simple, inexpensive process changes, two‐thirds of participants were able to reduce their CVC‐associated BSI rates, with a median reduction in CVC‐associated BSI rate of 32%. We did not document a reduction in mean length of stay, but the study was not designed to assess this outcome.

The first step in the pathogenesis of CVC‐associated BSIs is the acquisition of a biofilm on the internal and external surfaces of the CVC that facilitates bacterial attachment.¹⁷ Bacteria can colonize the catheter through inoculation from the skin surface, contamination of the catheter hub, hematogenous spread from infection at another site, or infusion of a contaminated fluid. The 2 most common routes, by far, are contamination of the skin or the hub, originating most often from the hands of healthcare personnel.¹⁸ Hence, interventions aimed at reduction of the rate of CVC‐associated BSIs should focus on prevention of contamination with skin organisms during catheter placement and hub manipulation.¹⁴

The interventions implemented in this collaborative therefore emphasized adequate skin preparation before catheter placement, optimal aseptic technique, hand hygiene with either alcohol‐based gels¹⁹ or antibacterial soap and water,²⁰ and full maximum barrier precautions (cap, mask, sterile gown, sterile gloves, and large sterile drape) during catheter insertion.²¹ Skin preparation was performed with 2% chlorhexidine (unless contraindicated), because it has been shown to be superior to other skin antiseptics for catheter insertion.^22,23 Adherence to the CDC guidelines for catheter care was required in order to comply with maintenance bundle elements.¹⁴ Finally, daily questioning of the necessity of continued catheter use was performed on clinical rounds.²⁴

Reports of single‐center studies have revealed a reduction in the CVC‐associated BSI rate in adult ICUs^7‐10 and pediatric ICUs^4,11,25 after the implementation of evidence‐based interventions. In addition, there are 2 multicenter interventions^12,13 that reduced CVC‐associated BSI rates in adult ICU patients. However, to our knowledge, this is the first report of a multicenter collaborative among pediatric ICU patients that resulted in a reduction in CVC‐associated BSI rates. In addition, our study was designed to document sustained improvement.

The impact of healthcare‐associated BSI is substantial to the patients, their families, and the healthcare system. This collaborative has avoided hospital and healthcare costs associated with CVC‐associated BSIs by an estimated $2.9 million in the initial improvement phase and $8.4 million during the sustain phase and has reduced the societal costs of family members’ inability to work, increased childcare expenses, and emotional stress.

When the project began, many pediatric ICU caregivers believed CVC‐associated BSIs were an acceptable and expected part of the care of critically ill and injured children. However, an overriding tenet of our collaborative was that healthcare‐associated CVC‐associated BSIs were largely preventable and could potentially be eliminated. While the acceptance of CVC‐associated BSIs was not measured explicitly, some teams noted a change toward zero tolerance of these infections.

Each team was asked to engage senior leaders whose primary role was to remove barriers to implementation. CHCA played the central administrative role, requiring monthly reports from institutions, assessing progress, and providing comparative reporting across the collaborative. The adoption of new clinical processes is difficult, especially when those processes may disrupt established workflows. However, the adoption and incorporation of the improvements introduced in this collaborative were likely achieved because of the perseverance of local leaders, the solidarity that resulted from collaborative efforts, and the visual display of measurable improvement.

There are several limitations of our study. First, we did not collect data on demographic or other factors that could influence BSI rates, such as type of line, location where the line was placed, the type of provider who placed the line, or its duration of use. We were not aware of any marked changes in these characteristics. Because risk adjustment has been shown to not change substantially over time,²⁶ it is likely that the groups remained similar. The initial postintervention period was only 9 months, half as long as that in the study by Pronovost et al.,¹³ so it was unlikely that there were significant changes during the study period.

Second, we did not collect data on preintervention process measure compliance, and this lack of data decreased our ability to measure the full impact of the bundles, although it is unlikely that the precollaborative performance was markedly different than that observed in the initial months of the project.

Third, it is possible that there was variability in the measurement of process compliance across collaborative hospitals. Standard observation worksheets were provided by CHCA to increase consistency.

Fourth, although we used the CDC definitions of CVC‐associated BSI, it is known that there is considerable variation in the application of these definitions. Although this may impact interhospital CVC‐associated BSI rate comparisons, it should not impact intrahospital rate comparisons, where the same personnel continually apply the definitions. We are not aware that participation in the collaborative resulted in any changes in methods or definitions.

Fifth, although we continued to collect data for an additional 12 months to assess the ability of participating hospitals to sustain their improvements, a few hospitals elected not to participate in the sustain project, and several of the participating hospitals did not submit complete process data. This was attributable to competing priorities and/or lack of perceived value of continued comparative reporting. We were not able to evaluate what role, if any, the ongoing quarterly conference calls and data submissions had on the ability of the hospitals to sustain their performance.

Finally, the time required to improve processes and effect culture change will vary from institution to institution. This effect could be magnified among hospitals that are just beginning to have cultural change and that have good process compliance but that have not yet seen a reduction in CVC‐associated BSI rate. It should be noted that variation in the ability of different centers to effect change within a relatively short improvement project is a common trait of multicenter improvement collaboratives. The collaborative goals were aggressive and were not met by all participants.

In conclusion, CVC‐associated BSIs are an important cause of morbidity and mortality in pediatric ICU patients, and they increase the cost of care and the length of the pediatric ICU stay. Several recent reports have shown CVC‐associated BSIs may be largely preventable in adult ICU patients through the use of evidence‐based interventions that are easily implemented and inexpensive.

The application of these interventions in 26 pediatric ICUs resulted in a median reduction in the CVC‐associated BSI rate of 32% across all centers, from 6.3 to 4.3 CVC‐associated BSIs per 1,000 CVC‐days. Two‐thirds of the centers showed a decrease in CVC‐associated BSI rates, and 35% reduced rates by at least 50% from baseline rates. Also, hospitals were able to sustain their improvements for 12 months after the conclusion of the original 9‐month improvement collaborative. An estimated $2.9 million in costs were avoided. The sustain period contributed an additional $8.4 million in estimated cost avoidance.

We conclude that a substantial and sustainable reduction in pediatric ICU CVC‐associated BSI rates and related costs can be realized by use of evidence‐based prevention interventions. Application of these interventions should be generalizable to all patients in pediatric ICUs everywhere and should result in substantial benefits to seriously ill children.

Participating Hospitals

Children’s Hospital Medical Center of Akron, Akron, OH; Children’s Healthcare of Atlanta, Atlanta, Georgia; Children’s Hospital, Boston, Massachusetts; Children’s Hospital of Buffalo, Buffalo, New York; Cincinnati Children’s Hospital Medical Center, Cincinnati, Ohio; Columbus Children’s Hospital, Columbus, Ohio; Children’s Medical Center of Dallas, Dallas, Texas; The Children’s Medical Center of Dayton, Dayton, Ohio; Children’s Hospital of Michigan, Detroit, Michigan; Cook Children’s Health Care System, Fort Worth, Texas; Arkansas Children’s Hospital, Little Rock, Arkansas; Children's Hospital Los Angeles, Los Angeles, California; Le Bonheur Children’s Medical Center, Memphis, Tennessee; Miami Children’s Hospital, Miami, Florida; Children’s Hospitals and Clinics of Minnesota, Minneapolis–St. Paul, Minnesota; Monroe Carell Jr. Children’s Hospital at Vanderbilt, Nashville, Tennessee; Children’s Hospital, New Orleans, Louisiana; Morgan Stanley Children’s Hospital of New York–Presbyterian, New York, New York; Children’s Hospital, Omaha, Nebraska; Children’s Hospital of Orange County, Orange, California; Phoenix Children’s Hospital, Phoenix, Arizona; Children’s Hospital of Pittsburgh, Pittsburgh, Pennsylvania; St. Louis Children’s Hospital, St. Louis, Missouri; All Children’s Hospital, St. Petersburg, Florida; and Children’s Hospital and Regional Medical Center, Seattle, Washington. Note: Rady Children’s Hospital and Health Center, San Diego, California, was a participating hospital but chose to focus on a neonatal intensive care unit; therefore, none of Rady’s data is included in the analysis.