Article Text

The Rhode Island ICU collaborative: a model for reducing central line-associated bloodstream infection and ventilator-associated pneumonia statewide
1. Vera A DePalo1,2,
2. Lynn McNicoll1,3,4,
3. Margaret Cornell4,
4. Jean Marie Rocha5,
6. Peter J Pronovost7
1. 1Department of Internal Medicine, Alpert School of Medicine at Brown University, Providence, Rhode Island, USA
2. 2Memorial Hospital of Rhode Island, Pawtucket, Rhode Island, USA
3. 3Rhode Island Hospital, Providence, Rhode Island, USA
4. 4Quality Partners of Rhode Island, Providence, Rhode Island, USA
5. 5Hospital Association of Rhode Island, Cranston, Rhode Island, USA
6. 6Rhode Island Quality Institute, Providence, Rhode Island, USA
7. 7Johns Hopkins University, Baltimore, Maryland, USA
1. Correspondence to Dr Vera A De Palo, Pulmonary, Critical Care, & Sleep Division, Memorial Hospital of Rhode Island, 111 Brewster Street, Pawtucket, RI 02860, USA; Vera_DePalo{at}brown.edu

## Abstract

Background Implementing bundles of best practices has been shown to provide patients with recommended care and reduce medical errors. Rhode Island's (RI) hospital leaders, quality organisations and insurers discussed the results of a quality improvement initiative in Michigan, the Keystone project, and explored the possibility of replicating these results statewide in RI.

Design Hospital executives and intensive care unit (ICU) staff, RI's quality organisations, RI Quality Institute, Quality Partners of RI, and Hospital Association of RI and consultants from Johns Hopkins University, worked together to implement evidence-based interventions and change safety culture in RI's ICUs.

Objectives The authors describe the RI ICU Collaborative, funded by insurers and hospitals, and report on statewide central line-associated bloodstream infection (CLABSI) and ventilator-associated pneumonia (VAP) rates between 1 January 2006 and 30 June 2008.

Subjects All adult ICU patients in RI.

Measures CLABSI and VAP rates.

Results 100% of 23 ICUs in 11 hospitals participated in the RI ICU Collaborative. The statewide mean CLABSI rate decreased 74% from 3.73 (median 1.95) infections per 1000 catheter days to 0.97 (median 0) in quarter (Q) 2 (March–June) 2008 (p=0.0032). The VAP rate fell 15% from 3.44 (median 0.58) to 2.92 VAPs (median 0) per 1000 ventilator days in Q2, 2008.

Conclusion The RI ICU Collaborative, a statewide quality improvement initiative, served as the platform by which multifaceted interventions were associated with reductions in CLABSI and VAP rates, and an increase in the use of evidence-based interventions. Completing Phase II, the RI ICU Collaborative continues to sustain these statewide reductions.

• ICU complications
• catheter infection
• ventilator associated pneumonia
• quality collaborative
• collaborative
• healthcare quality improvement

## Introduction

The Institute of Medicine estimated that medical errors result in up to 98 000 deaths/year.1 Recommended care is received only 50% of the time.2 Care complications, central line-associated bloodstream infection (CLABSI) and ventilator-associated pneumonia (VAP), lead to increases in morbidity and mortality, length of stay and healthcare costs.3–5 The Centers for Disease Control and Prevention (CDC) estimated that approximately 248 000 bloodstream infections and 250 000 cases of pneumonia occur annually in US hospitals.6 In intensive care units (ICUs), CLABSI and VAP are responsible for about 31 000 and 36 000 deaths, respectively, with each infection potentially averaging up to $40 000.6 7 Evidence-based elements have been shown to reduce risk and improve these outcomes, and, when bundled, led to even greater benefit than each of the strategies alone.8–12 It is generally difficult to disseminate quality improvement interventions. Little is known about factors that predict successful dissemination. As such, we sought to evaluate whether the reductions in CLABSI and improvements in culture achieved in the Michigan Keystone project could be extended to Rhode Island.13 In Rhode Island (RI), in 2005, 15 500 patients were admitted to the ICU, consuming 30% of RI's acute care costs of$558 million (RI's Discharge Database, 2005). RI is unique in that all the state's ICUs partnered with RI's quality organisations (RI Quality Institute (RIQI), Quality Partners of RI (QPRI), and Hospital Association of RI (HARI)) to create a quality improvement platform, the RI ICU Collaborative. While the RI ICU Collaborative project was described in 2005,14 and in 2009 results of the first 4 years were reported,15 the present work seeks to communicate to a broader audience the gains that implementing quality improvement initiatives can realise. The specific aims of this report are to outline the development and implementation of the RI ICU Collaborative, the first statewide ICU quality initiative, and to report with statistical analysis on the reductions in CLABSI and VAP that have been realised.

## Patients and methods

In RI, there are 11 acute care hospitals with one or more adult ICUs, totalling 23 units with a care spectrum from coronary, medical, surgical, respiratory, intermediate care, neurosurgery and trauma care.

### Establishing the collaborative

In September 2004, physician and nursing leadership of each ICU and the Chief Executive Officer (CEO) of each hospital came to the table with representatives from RIQI, QPRI and HARI, and insurers (Blue Cross & Blue Shield of RI (BCBSRI) and UnitedHealthcare of New England (UHCNE)) to learn of the quality improvement work by the Johns Hopkins' Quality & Safety Research Group (JHQSRG) and the outcomes in the Michigan Keystone13 project, and to explore its replication in RI. ‘The RI ICU Collaborative’ was established with leadership from the RIQI and co-management by QPRI and the HARI (appendix 1). The leadership team comprised representatives of the three organisations, the physician consultant to QPRI and an ICU physician consultant.

Each ICU was visited to assess data collection and utilisation. CEO support was formalised with a commitment letter. Each ICU identified a team of multidisciplinary champions and a senior executive partner.

Funding, provided by BCBSRI and UHCNE, paid for the project management, a shared database, the logistics and speakers of the learning sessions and face-to-face meetings, and the consultants. Hospitals provided an in-kind contribution for the costs of data collection, time for meetings, calls and learning sessions, Continued Medical Education hours and statistical analysis. Global project management included organising work structure, coaching teams and monitoring performance. The approach has focused on interhospital collaboration and customised consultation at each site with review of the adaptive and technical component progress. A bimonthly team-lead forum reinforced educational elements, shared best practices and discussed challenges.

### Data collection and interventions

Using a prospective cohort study design, individual ICUs implemented several patient safety interventions and monitored the impact of these interventions on specific safety measures. The initiative began with a learning session in November 2005 which focused on improving culture and outlined strategies to reduce CLABSI. Baseline, preintervention data collection for CLABSI and VAP began with the first quarter (January through March) of 2006. The first intervention consisted of education and implementation of best practice strategies to reduce central-line-associated bloodstream infections The JHQSRG served as consultants communicating content for the educational sessions and conference calls. The content and coaching calls reinforced strategies. By the end of the first quarter (Q), 2006, unit teams had educated their bedside staff to best practice strategies and implemented processes. Data from Q2, 2006 and after are considered postintervention (education and implementation) and continued focus and outcome assessment are ongoing. The CLABSI best-practice strategies targeted the clinician's use of five evidence-based behaviours recommended by the CDC, identified as being the most effective at reducing CLABSI.8 9 These behaviours included hand-washing, using full barrier precautions when inserting central access catheters, chlorhexidine skin-cleansing, avoiding the femoral site if possible and removing unnecessary catheters.

Education for VAP reduction occurred in March 2006, and best practice strategies included daily assessment for liberation from mechanical ventilation, elevation of the head of the bed, following commands as an indication of appropriate sedation, using deep venous thrombosis and gastric stress ulcer prophylaxes. For this bundle, teams submitted compliance data for each component of the bundle. Compliance with the ventilator bundle was determined by including in the numerator all ventilated patients with ‘Yes’ or ‘Not-indicated’ for all interventions and in the denominator all ventilated patients multiplied by 100. Any element missing or coded as ‘No’ classified the submission as ‘non-compliance’ with the ventilator bundle.

Since a culture of safety is necessary for implementing and sustaining improved clinical outcomes, a Comprehensive Unit Based Safety Program was introduced at the launch of the Collaborative. This programme included educating staff on the science of safety, identifying hazards, identifying senior executive partners, learning from defects and implementing teamwork tools.16 Team members were empowered with the ability to stop procedures if safety was compromised.

The main dependent variables are the rates of CLABSI and VAP per 1000-device-days/quarter. Discussion with the state's Infection Control Practitioners (ICPs) resulted in the uniform use of definitions according to National Nosocomial Infection Surveillance (NNIS) criteria.17 This caused a definition change for some hospitals in order to comply with the standards of the RI ICU Collaborative.

The main independent variable is time postintervention. Time periods were divided into quarters. We compared the baseline period with the postinterventions periods.

We report on data from 1 January 2006 to 30 June 2008. The first quarter of 2006 (January through March) data are used as baseline. Data were collected in a deidentified manner by counting device days during the ICU stay. At each hospital, ICPs collected deidentified VAP and CLABSI data, and judged their validity. The use of multiple catheters in one patient was counted as one catheter-day. The deidentified, aggregated data were sent to a central database for RI ICU Collaborative use. Throughout the study period, teams received feedback on infection rates by accessing database reports and review with the project manager.

Subcommittees outlined recommendations for central line care, developed guidelines to reduce VAPs and strategised to improve executive and physician engagement. Activities to boost engagement included a letter and quarterly conference calls for physician champions and a focus on executive and ICU Director participation at learning sessions with an agenda to best accommodate executives' time constraints.

### Data analysis

Univariate analysis of the number of infections (CLABSI and VAP) revealed the data to be highly non-normal. Summary tables are displayed with medians and interquartile ranges. Non-parametric methods were used to examine infection rates from the initial four quarters of surveillance (Q1–Q4, 2006) compared with the most recent four quarters (Q3–Q4, 2007 and Q1–Q2, 2008).

Examination of the number of infections over time was achieved using mixed effects models with a Poisson distribution. This method of analysis was chosen to account for both the repeated quarterly measures and the effect of multiple ICUs within hospitals. Additionally, hospital size and teaching status were incorporated into the models along with the number of device days. The number of device days was included to control for potential volume bias. All data analyses were conducted using SAS 9.1.3 (SAS Institute, Cary, North Carolina). Mixed models were analysed using PROC GLIMMIX within SAS.

The Institutional Review Board of the Memorial Hospital of RI has determined that these RI ICU Collaborative quality improvement activities do not constitute ‘human subjects research’ under the ‘Common Rule.’ Financial support for project management, database and Learning Session costs for the RI ICU Collaborative was provided by BCBSRI and UHCNE.

## Results

Data were collected from the 23 adult intensive care units (263 adult ICU beds) in the 11 adult acute care hospitals in RI. Seventy-five per cent of the beds are in the six teaching hospitals. Most hospitals have one ICU, one hospital has four units, and one hospital has 10 units. One hundred per cent of the units in 100% of the adult acute care hospitals in RI participated in the RI ICU Collaborative.

Assessment of the completeness of data for statewide CLABSI rates revealed that calculations were made on 634 of a possible 690 data-months (91.9% complete). For statewide VAP rates, calculations were made on 679 of a possible 690 data-months (98.4% complete).

The statewide mean CLABSI rate at baseline was 3.73 (median 1.95) infections/1000-catheter-days (Q1, 2006). There has been a progressive decline in the annual statewide CLABSI mean rate/1000-catheter-days from 3.11 for 2006 to 1.79 for 2007 (figure 1). The statewide CLABSI mean rate for Q1 and Q2, 2008 was 1.23 infections/1000-catheter-days. The Q2, 2008 mean rate of CLABSI was 0.97 (median 0).

Figure 1

Mean central line-associated bloodstream infection (CLABSI) rates.

In June 2008, no CLABSI was reported by any unit. From 1 July 2007 to 30 June 2008, 61% (15/23 units) had no CLABSI for six consecutive months or more. Rates for the last six quarters (Q1, 2007 to Q2, 2008) have been consistently below the overall statewide mean. The median CLABSI rate was 0 for the last 9 months of the study period with decreasing interquartile ranges between units during the three quarters (table 1).

Table 1

Median rates and device days

Comparison of mean rates of the first four quarters (Q1–4, 2006) of the study period with the last four quarters (Q3–Q4, 2007 and Q1–Q2, 2008) shows a significant difference in CLABSI (p=0.0032). Using the mix model framework, there is a decreasing trend over time for CLABSI (p=0.0030).

RI ICU Collaborative statewide mean VAP rates also declined from a baseline of 3.44 (Q1, 2006) per 1000-ventilator-days to a mean of 2.86 VAPs/1000-ventilator-days for Q1 and Q2, 2008. Median rates for VAP in Q1 and Q2, 2008 are 0. Twenty-two per cent (5/23) of the units did not submit complete data for Q1, 2006. Four ICUs began collecting VAP data for the Collaborative in February 2006, and one unit began submitting data in March 2006. There was a slightly higher statewide VAP mean rate overall (4.02 VAPs/1000-ventilator-days) for 2006 compared with baseline, but this decreased to 3.36 for 2007 and 2.86 for Q1 and Q2, 2008.

Similar non-parametric analysis for VAP showed mixed results. There was no overall time trend (p=0.5982). The relationship with ventilator days was present (p<0.0001). Within the mixed model framework, however, the comparison of Q1–Q4, 2006 with the last four quarters of surveillance (Q3–Q4, 2007 and Q1–Q2, 2008) showed a decreasing trend (p=0.0750).

Although hospital size and teaching status were included as covariates, they did not achieve statistical significance for either CLABSI or VAP (table 2).

Table 2

Statewide median device rates by hospital size and teaching status

The ventilator bundle compliance data demonstrated a statistically significant increasing linear trend from Q1, 2006 (60%) to Q2, 2008 (78%) (p<0.0001). Additionally, there is also an improvement in the VAP compliance rate between the first four quarters of surveillance (Q1–4, 2006) to the last four quarters (Q3–4, 2007 and Q1–2, 2008) (p<0.0001).

Evaluation of the individual VAP bundle components at baseline revealed compliance with following commands to have the lowest compliance rate (68.6%). The compliance data for each of the other elements were above 87.52% at baseline. Eighty-seven per cent of the patients receiving mechanical ventilation were assessed daily for weaning at baseline. In the last two quarters (January to June 2008) of the study period, 75.8% of the vented patients followed commands. Compliance with all other VAP bundle components was above 90.35% (daily weaning assessment component).

VAP rates decreased as ventilator bundle compliance rates increased (figure 2).

Figure 2

Mean ventilator-associated pneumonia (VAP) rates and VAP bundle compliance.

## Discussion

The intravenous catheter and the mechanical ventilator are indispensable tools in the ICU. However, CLABSI and VAP result in increased morbidity and mortality to our patients and increased financial burden to our healthcare system.18 19 In US ICUs, more than 30 000 patients die annually of CLABSI, and estimated annual costs range from $296 million to$2.3 billion.6 20 21VAP is a leading cause of death in ICU patients with a 15% higher mortality than in those who do not develop VAP.22

Evidence-based interventions have been shown to decrease CLABSI and VAP rates, but the difficulty is in translating that research into true patient benefit.9 23 24 Education directed towards ICU nurses and physicians highlighting best practices results in reduced CLABSI rates25 26

The implementation of evidence-based strategies has been shown to significantly reduce CLABSI and VAP.15 27–29 In this study, the use of a multifaceted intervention was associated with substantial reductions in the rates of CLABSI and VAP. RI demonstrated significant statewide reductions in CLABSI (3.73 infections/1000 line-days to 0.97, a 74% reduction). The statewide VAP rate was reduced by 15% (3.44 infections /1000-ventilator-days to 2.92 infections/1000-ventilator-days). Ventilator bundle compliance increased from 60% to 78%.

The key to these statewide successes has been the RI ICU Collaborative. Having all stakeholders present during the learning phase laid the groundwork for successful funding requests. In Phase I, the RI ICU Collaborative was funded with $850 000 by BCBSRI and UHCNE. An opportunity calculator30 (SM Berenholtz, personal communication 2006), a tool developed by the Johns Hopkins group, can provide estimates of preventable complications, length of stay, mortality and costs of care associated with mechanical ventilation and nosocomial infection based on performance on ventilator bundle quality measures, rates of VAP and CLABSI and on the efficacy of the interventions and average attributable morbidity, mortality and costs published in the literature.31–40 Inputting data to compare Q1 and Q2, 2008 to baseline (Q1, 2006), the tool calculated that the RI ICU Collaborative prevented 42 CLABSIs and 9 VAPs, achieved 10 saved-lives and reduced ICU stay by 608 days. The cumulative cost savings were estimated at$2 016 592.

The CLABSI and VAP rates at the start of this Collaborative (Q1, 2006) were at the lower end of the range of published CLABSI and VAP rates following implementation of multiple strategies to reduce ICU adverse events, nosocomial infections and costs.41 The present study shows that even with lower infection rates, the implementation of evidence-based improvement strategies is still able to reduce care complications, improve safety and decrease cost. As quality improvement work continues, it becomes more difficult to demonstrate statistically significant differences and return on investment when working to change practice. Future project selection needs to consider improvement gains and cost–benefit relationships. Additional funding was committed to fund Phases II and III of the RI ICU Collaborative for work in improving outcomes in sepsis through collaboration with Emergency Medicine colleagues and addressing palliative care in the ICU setting. Understanding who realises the financial benefits of these programmes is important to guide who should support them. We believe that these types of programmes offer a unique opportunity for insurers to partner with provider organisations to realise improvements in patient outcomes.

While a potential limitation of these data could lie in their validation, Infection Control practitioners, different from the ICU bedside staff and RI ICU Collaborative team leads, confirmed the presence of CLABSIs and VAPs through each hospital's own infection control process. Another limitation is that the reductions seen in infection rates may not have been solely due to the implementation of the CLABSI and VAP bundles. During the course of this project, three ICUs added the use of antibacterial-impregnated catheters to their CLABSI-reducing strategies. These were solitary units in hospitals with fewer than 200 beds. One unit in a mid-size hospital and one unit in the smallest hospital began using silver-coated endotracheal tubes to help reduce VAP during the course of the project. These interventions impacted only a small number of patients. The RI ICU Collaborative began as a quality improvement project and as such was not designed to randomise ICUs to intervention and non-intervention groups. At the start of the project, all units wanted to achieve improved patient outcomes.

The RI collaborative built upon and learnt from the successful Michigan (MI) project,13 balancing several tensions and advancing the science of quality improvement. Both MI and RI used the same model to put evidence into practice and improve culture through CUSP.42 The Michigan Keystone collaborative recognised that it is much more efficient and effective to centralise the technical work (measures, data collection and evidence summaries) and encourage local modification of the adaptive work (how to implement that evidence, informed by state and hospital context). To achieve the current projects goals, RI needed valid data to answer ‘Are we safer?’ with scientific integrity. As such, RI used the measures and interventions from MI, and modified and improved upon the implementation in several ways. MI started with paper data collection for infection rates, these data were entered into a database, and teams were sent reports. RI used a web-based data collection tool and built a data quality control into the project early and in real time. For example, early in the MI project, it had large amounts of missing data, and staff had to call hospitals to obtain these data. From the beginning, RI checked for missing data monthly. RI improved upon the training materials, making them more focused and specific. For example, in MI many of the training materials were intentionally at a concept level, for fear that hospitals were being told what to do. In RI, we learnt that teams welcomed unambiguous and explicit plans, and the training materials provided these. These explicit training materials often give teams comfort, and they do little to help them implement the adaptive work that requires changes in culture, work that is often messy and for which there is no script. RI had earlier and greater involvement of the infection prevention community (IP). The IP community was so engaged that at times we risked removing accountability from the ICU teams. While MI was led by a state hospital association, RI was led by the RI Quality Institute and incorporated the Hospital Association of RI, the quality improvement organisations and the state health department, thus strengthening the infrastructure to support the project. RI is a small state, and all hospitals participated, though competition among hospitals was often significant. RI strategically sought early and visible support from congressmen, which likely helped with recruiting hospitals. The lessons from RI continue to inform the programme, a programme now being implemented in all 50 states, all of England, Spain and Peru.

Recently, other pressures have come to bear on efforts to improve safety and quality. Pay-for performance, regulatory measures and legislatively mandated public reporting are common mechanisms for improving quality. Their influences through the years of the RI ICU Collaborative cannot be discounted. However, even if these have added motivation for improvement, the RI ICU Collaborative has been RI's vehicle by which there has been a statewide improvement in safety, providing a platform for collaboration, networking and sharing resources.

In this project, multifaceted interventions were associated with substantial reductions in the incident rate of CLABSI and VAP, and an increase in the use of evidence-based interventions. These statewide improvements have been sustained, demonstrated by mean rates of CLABSI at Q2, 2009 that are 62% reduced from 2006 baseline, a statewide VAP mean rate (Q2, 2009) that is 21% reduced from baseline and a VAP bundle compliance rate that is at 88.4%. At the end of the second phase of the RI ICU Collaborative, there still remains 100% participation of all adult ICUs statewide.

## Acknowledgments

We acknowledge the hard work and the leadership, both clinical and administrative, at each of the participating hospitals of the RI ICU Collaborative. Without their dedication and commitment to their patients, this project would not have been possible. We thank the staff of Kent County Memorial Hospital, Landmark Medical Center, the Miriam Hospital, the Memorial Hospital of RI, Newport Hospital, Rhode Island Hospital, Roger Williams Medical Center, St Joseph's Hospital, South County Hospital, Department of Veterans Affairs Medical Center and the Westerly Hospital. We acknowledge the leadership of C Duquette, P McBride, M Claflin and DR Gifford in the initial phase of the RI ICU Collaborative. We thank M Roberts and the Center for Primary Care and Prevention of Memorial Hospital of RI for the statistical analysis.

## Appendix I Steps to forming the Rhode Island Intensive Care Unit Collaborative

 Steps Description 1. Bringing stakeholders together Initial meeting of ICU care leaders, hospital executives, insurers, and quality organizations 2. Determining the leadership team Quality organisations take the lead:RIQI(project leader), QPRI and HARI (project managers)hospital physician liaison and ICU physician leader 2. Survey of ICUs in RI Outlines the existent scope of care and data collection in ICUs across the state 3. Funding acquisition RIQI lobbies insurers to support project; funders at stakeholders meeting and learning sessions; other funding options pursued 4. Faculty and content expertise Phase I: Partnering with Johns Hopkins Quality and Safety Research Group and supplemented by local clinical experts;Phase II Local experts and some invited speakers 5. Commitment from hospitals CEOs present at stakeholders meeting and engaged through HARI meetings; commitment letters signed by each CEO 6. Surveying culture Culture assessment tool used to survey the culture of each ICU, administered at the start of the project and annually thereafter 7. Bundle implementation Engagement, Education, Execute and Evaluate: technical work (eg, putting together a line cart) and adaptive work (eg, empowering nurses to stop physicians who do not follow the checklist)(quality improvement cycles of change) 8. Culture building and quality improvement work Continued individual team efforts with customised project management support

## Footnotes

• Funding This projected was funded by Blue Cross & Blue Shield of Rhode Island (Providence, Rhode Island) and United Health Care of New England (Warwick, Rhode Island). Additional support for education was provided by Neighbourhood Health Plan of Rhode Island. While the authors individually have no financial conflicts of interest to disclose, the RIQI and QPRI participate in other projects funded by BCBSRI and UHCNE.

• Competing interests None.

• Ethics approval Ethics approval was provided by the Institutional Review Board of the Memorial Hospital of Rhode Island.

• Provenance and peer review Not commissioned; externally peer reviewed.

## Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.