Reducing Clostridium difficile infection in acute care by using an improvement collaborativeBMJ 2010; 341 doi: https://doi.org/10.1136/bmj.c3359 (Published 21 July 2010) Cite this as: BMJ 2010;341:c3359
- Maxine Power, national improvement adviser1,
- Neil Wigglesworth, specialty registrar in public health2,
- Emma Donaldson, quality improvement fellow3,
- Paul Chadwick, consultant microbiologist3,
- Stephen Gillibrand, pharmacist3,
- Donald Goldmann, professor of paediatrics; senior vice president45
- 1Department of Health, London SE1 6LH
- 2NHS East Lancashire, Nelson BB9 8AS
- 3Salford Royal NHS Foundation Trust, Salford M6 8HD
- 4Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115, USA
- 5Institute for Healthcare Improvement, 20 University Road, Cambridge, MA 02138
- Correspondence to: M Power
Problem In 2006, despite a focus on infection control, Salford Royal had the fourth highest rate of Clostridium difficile infection in north west England.
Design Interrupted time series in five collaborative wards (intervention group) and 35 non-collaborative wards (control group).
Setting University teaching hospital with 850 acute beds.
Key measures for improvement Number of cases of C difficile infection per 1000 occupied bed days.
Strategies for change In February 2007, a newly formed antimicrobial team led the implementation of revised guidelines in all wards and departments. From March to December 2007, five wards participated in an improvement collaborative. Since December 2007, the changes from the collaborative have been collated and implemented throughout the organisation.
Effects of change At baseline the non-collaborative wards had 1.15 (95% CI 1.03 to 1.29) cases per 1000 occupied bed days. In August 2007 cases reduced 56% from baseline (0.51, 0.44 to 0.60), which has been maintained since that time. In the collaborative wards, there were 2.60 (2.11 to 3.17) cases per 1000 occupied bed days at baseline. A shift occurred in April 2007 representing a reduction of 73% (0.69, 0.50 to 0.91) from baseline, which has been maintained.
Lessons learnt Careful use of antimicrobial drugs is important in reducing the number of cases of C difficile infection. A collaborative learning model can enable teams to test and implement changes that can accelerate, amplify, and sustain control of C difficile.
Outline of problem
Clostridium difficile infection is a major cause of diarrhoea associated with antibiotics. Symptoms vary from mild diarrhoea through to potentially fatal colitis.1 Spore bearing bacteria, shed in the stool, can survive for months in the environment and be transported on the hands after direct contact with infected patients or with surfaces.2 The control of C difficile is therefore multifaceted and relies on careful use of antibiotics, stringent infection control practices and thorough cleaning of the environment.3 4
Salford Royal NHS Foundation Trust is a university teaching hospital in the north west of England with 850 beds, which provides care for 320 000 inpatients a year. In addition to providing local services, the hospital provides tertiary care for renal medicine, neurosciences, complex spinal surgery, and intestinal failure. In 2006, Salford Royal had 350 cases of C difficile infection in patients aged over 65, the fourth highest rate of infection in north west England. 5 Despite system-wide changes in infection control and cleaning practice, infections rose, peaking at 115 cases during the first quarter of 2007.
We describe the effect of three measures: system-wide changes to guidelines and practices for using antimicrobial drugs; an improvement collaborative, in which five wards worked together on an ambitious aim to reduce the incidence of C difficile infection by 50% within one year; and the impact of spreading successful improvements from the collaborative throughout the hospital.
The primary study question was to determine whether teams participating in a Breakthrough Series improvement collaborative (a short term learning system for teams to work together to deliver improvements 6) could deliver accelerated and enhanced reduction in C difficile infection. We collated the number of cases of C difficile infection on five participating wards and 35 control wards for a baseline period (from April 2006) throughout the collaborative (March to December 2007) and for the one year scale-up (January to December 2008).
We used an interrupted time series design in which two groups of participants were observed repeatedly before, during, and after the intervention. Data were collected from five collaborative wards (intervention group) and 35 non-collaborative wards (control group).
During the six months that predated the collaborative, changes were made to infection control throughout the hospital. These included the introduction of a rapid response cleaning team, a deep clean programme, and a focus on hand hygiene and uniform protocols. The only system-wide change that coincided with the study period was the revision of antimicrobial policies and practice.
Selection of participants
We identified five wards within the directorate of medicine for the elderly that had a high incidence of C difficile infection, and we invited clinicians on these wards to attend an information workshop. We outlined the aim, ambition, and structure of the collaborative model and asked teams to discuss whether they would like take part. All five teams agreed to participate as pilot sites, and they recruited relevant staff (housekeepers, domestics, facilities, and nursing, medical, and pharmacy staff) from their clinical and ward areas.
The collaborative model
We ran a Breakthrough Series collaborative,6 in which teams were able to learn from each other and from recognised experts around a focused set of objectives. Teams committed to work together over a nine month period (mid-March to mid-December 2007 and attend three, two day learning sessions, which provided instruction in the theory and practice of improvement. In between the learning sessions, teams participated in action periods, in which they tested changes and during which the teams had ward visits from members of the hospital’s executive and direct access to the project director (the consultant nurse in infection control) and an improvement adviser, and to one another, via a web based portal (extranet).
Scale-up and spread of changes
In December 2007, we scaled up activity across the organisation. We encouraged all wards to implement the successful changes developed by the pilot teams (box) and to develop their own ideas using plan-do-study-act (PDSA) methods. Changes to prescribing had been introduced in February 2007 (box), coincident with changes in the statistical process control charts (see fig 1⇓). MP, NW, and SG supported the new teams, which were also encouraged to access expertise from peers in the original collaborative wards. Three learning sessions took place in February, April, and July 2008, led by team members from the original collaborative wards. Teams were visited monthly by a senior leader designated by the executive sponsors (associate directors of medicine, nursing, operations, and finance), using the format developed during the mentoring visits to the pilot wards. In September 2008, we introduced an assessment framework for quality standards for all wards; this is conducted by a matron at least once a year and provides assurance that changes have been implemented as agreed, and that new procedures continue to be followed.
Box 1 Changes introduced to all wards in February 2007
An antimicrobial management team comprising a senior pharmacist (1 whole time equivalent (WTE)) and a consultant microbiologist (0.1 WTE) began daily ward rounds to review patients receiving antimicrobials and advise medical teams about best practice for their use
Antimicrobial guidelines were amended to restrict the use of cephalosporins and quinolones in the first line treatment of pneumonia and urinary tract infection
Access to antimicrobials was restricted by removing third generation cephalosporins and oral quinolones from ward stocks. (In a small number of clinical exceptions antimicrobials were left in ward stocks and a system of monitoring by pharmacy and microbiology was implemented)
Antimicrobial dispensing (including cephalosporins and carbapenems) was restricted to prescriptions issued by consultant grade doctors or junior doctors and specialist registrars who had sought approval from microbiology
Policy was developed and implemented to assure the appropriate use of antimicrobials in surgical prophylaxis, including guidance on prescribing, review, and cessation
Identification and containment
A simple 10 point questionnaire was developed to identify key issues related to the nature of C difficile infection, its symptoms, the optimal response to a report of symptoms, modes of transmission, and treatment; the questionnaire was administered to all nursing, support, domestic, and clerical staff and re-administered after training until scores of 90% or greater were achieved
All staff were given focused and systematic education, targeting identified knowledge gaps from questionnaire, using a short, ward based teaching pack and written leaflets. Information was included in “sound bites” into the ward safety huddles
Local protocols were developed to clarify the responsibility of the ward teams for starting isolation at the point of suspected symptoms (loose stool, farmyard smell) before requesting tests or consulting the medical team or infection control
On admission, all patients were asked to report symptoms to their named nurse
Infection control passports were introduced to provide details of patients’ requirements on transfer to other units, and for cleaning of wheelchairs and trolleys by porters on return to the unit
Habits and patterns
Strict enforcement of hand hygiene on entry to ward. “No entry” posters developed and tested to ensure hand washing on entry to the ward
Weekly peer audits for compliance with hand hygiene were introduced. Patterns of errors were identified (for example, not washing hands after touching patient’s environment or after taking off gloves) and posted on ward notice boards, discussed at safety huddles, and shared widely with the teams, who then tested changes
Hand washing rounds for patients before meals and snacks were introduced to reduce the risk of ingesting spores. Doctors’ hand hygiene patterns during rounds were studied and improved. “Bare below the elbow” was strictly enforced for all staff
Self caring patients’ usual patterns of postmicturation hand hygiene were deduced and a poster campaign targeted at men’s bathrooms was developed to highlight risk in hospital; the poster’s position, message, and size were continuously modified until 95% of men were compliant
Single use soap and shower gel sachets were provided; all multiple use shower and bath products, including soaps and gels, were removed
Cultural assumptions were challenged—for example, members of the team were asked to assume that key pieces of equipment were “dirty” until proved clean when they moved equipment from stores to wards. Solutions included “clean seal equipment” (sealed, signed, and dated, with seals broken immediately before use)
Disposable washbowls were introduced for all infected patients and for cleaning the lower body of all patients
Systems were developed and agreed between wards and the facilities directorate for deep cleaning and curtain replacement, including review (and replacement or deep clean where necessary) of toilet roll holders, waste bins, taps, and soap dispensers
The bed area was cleaned between patients, and bed and other linens were stored centrally until the patient arrived at the bed area
Five key surfaces received daily decluttering and thrice daily wipe with chlorine disinfectant: table, locker, bed rails, buzzer, chair
Key measures for improvement
The primary outcome was the incidence of C difficile infection on the five collaborative (intervention) wards between March 2007 and December 2008, measured as cases of C difficile infection per 1000 occupied bed days. A secondary outcome was the impact of revised antimicrobial practices (from February 2007) and scale-up (from December 2007) on the whole hospital (non-collaborative (control) wards) from December 2007 to December 2008.
Process of information gathering
Diagnosis of C difficile infection
Tests for C difficile toxin were performed routinely when patients presented with loose stools. Samples were tested for C difficile toxins with a commercial ELISA kit (C difficile Tox A/B II Techlab, Blacksburg, VA).
Classification of hospital acquired C difficile infection
For this study, we restricted our analysis to cases classified as hospital acquired infection (the patient entered the institution without symptoms and then developed symptoms, and had positive results on a stool test) more than 48 hours after their index admission, in line with the Department of Health definition of “community-onset C difficile”).4 Cases with positive results from stool samples collected within 28 days of a previously positive sample were excluded from the analysis.
Occupied bed days
Rates of C difficile infection were calculated by using the denominator of total bed days for all patients on the wards—that is, the total number of nights spent in hospital by all patients in each setting from April 2006. Bed days were calculated using hospital episode statistics (http://www.hesonline.nhs.uk). Occupied bed days were defined as the arithmetic difference in days between the admission date and the date of discharge. These data were used to derive the denominators for the rate.
We used Chartrunner 3.0 (PQ Systems Europe Formby, Merseyside) to construct statistical process control charts for rate based data (U charts), where the numerator was a count of cases of C difficile infection and the denominator occupied bed days. We used the standard control chart method to interpret patterns of variation in data over time. Non-random patterns (special cause variation) were determined according to standard definitions (see bmj.com).7
Cases per 1000 occupied bed days were calculated for the baseline and subsequent shifts using the following equation:
No of cases in the timeframe × 1000 / Total number of bed days (exposure)
As cases of C difficile infection are rare events, we used analyses based on a Poisson distribution.
Reduction in C difficile infection
The percentage reduction in C difficile infection was calculated by comparing the cases per 1000 occupied bed days during time periods where the data were stable (showed only normal variation) with the baseline (table⇓). Baseline data were collected from April 2006 to the month immediately before the first episode of special cause variation, which occurred in April 2007 for the collaborative wards and August 2007 for the non-collaborative wards (fig 1⇓).
Strategies for change
In February 2007 we brought together an expert steering group from microbiology, infection control, nursing, elderly care medicine, pharmacy, and improvement to develop a project aim and a framework for change. The team examined epidemiological data and decided on a time limited, measurable stretch goal to reduce C difficile infection by 50% within one year in pilot wards. They examined the most recent literature on the control, containment, and management of C difficile. They identified four key drivers for change: use of antimicrobial drugs; environmental cleaning; early identification and containment of infection; and habits and patterns. Teams were encouraged to use the diagram (fig 2⇓) to structure their improvement work.
The collaborative was designed according to the guidance for Breakthrough Series collaborative model (fig 3⇓). At the first learning session, teams were made aware of the epidemiology and local context with respect to C difficile. They were presented with ward level data and a patient story. In addition, they learnt the model for improvement (including plan-do-study-act (PDSA) cycles), measurement, and the principles of reliability.8 Each team selected one of the four key drivers and developed their first test of change. At the second learning session, the teams reported back on their tests and received feedback from their peers and an improvement adviser, and decided whether their changes should be adapted, adopted, or abandoned. During the next action period, all teams implemented changes and monitored cases using a web based platform (extranet). This cycle of learning and testing was repeated in each action period—three times in the collaborative wards, then (after scale-up) three times hospital-wide.
A designated sponsor from the hospital’s leadership team supported each of the five project teams. Executive mentoring visits were conducted six times (bimonthly) during each action period (a total of 18 visits). This sponsor provided support for the team leader and ward team in developing ideas for changes, testing changes, removing barriers, and facilitating rapid uptake of successful changes within key areas.
Changes that resulted in improvement
The top 20 changes developed during the collaborative (see box 1) were collated into a change package that included step by step guidance on how the collaborative teams collected baseline data, tested changes, and set up systems for prospective monitoring. For example, the aim of the team that developed hand hygiene protocols (on entry to wards) was that 95% of visitors would wash their hands on entry to the ward. The seven steps that resulted in improvement (box 2) provide an example of the approach to developing changes adopted by all teams.
Box 2 Developing changes: an example
Step 1: Collect baseline data
Observing the first 10 visitors entering the ward at the start of ward visiting times, staff were able to deduce the proportion of visitors who washed their hands on entry to the ward. Observations were repeated on three occasions and resulted in a baseline data set, which showed the compliance rate to be 40%.
Step 2: Strip away the old
Staff then looked critically at the size, placement, and message of posters that were already in place to prompt visitors to wash their hands and discussed how to improve them. They found an overwhelming number of reminders (on carpets, doors, walls) with conflicting and duplicative messages. All prompt materials were removed for one week and the compliance measurement was repeated; it remained constant at 40%.
Step 3: Develop a test
A single, large (A1 size) poster was developed and placed on an easel next to the sink. Data were collected as at baseline, and the compliance rate was found to be 80%.
Step 4: Learn from failure
Staff used observation and asking questions to study why the visitors were not washing their hands. Reasons given included not being aware they had to; no time; nowhere to put down items they were carrying; no one else was doing it. Teams identified the top three reasons (which accounted for 80% of the failures).
Step 5: Design for reliability
Staff implemented three changes to deal with the top three reasons:
A small shelf was placed at the side of the sink with a note encouraging visitors to use it for their belongings while washing their hands
Feedback posters placed above the sink congratulated visitors on compliance, and staff asked whether they could help visitors who were seen to be non-compliant
Staff were empowered to refuse entry to all routine visitors unless they washed their hands on entry.
Step 6: Measure
Once a week, data on compliance among a sample of 10 visitors were collected.
Step 7: Evaluate culture change
Comments from ward leaders indicated that working on a focused segment of hand hygiene brought secondary benefits. The ward staff felt better able to challenge all visitors, including other healthcare professionals.
Teams were encouraged to use plan-do-study-act cycles to develop questionnaires and checklists—for example, the C difficile questionnaire was developed using a test sequence. Firstly, on one shift, two nurses identified the staff groups they wished to survey with the questionnaire, developed 10 questions which they tested on one qualified nurse and one healthcare assistant, clarified ambiguous questions, and made changes to the format of the questionnaire, which was then typed by the ward clerk. Then the questionnaire was given to two more staff members and similar modifications made. The next phase consisted of an opportunity sample of all staff on the morning and afternoon rotas of a specified day, with a minimum of 10 questionnaires. Feedback was sought and these data were taken as the initial baseline. Minor revisions were made and the number of staff members who correctly answered all 10 questions was recorded (3/10) to give a baseline score of 30%. The questionnaire was run on two further shifts, with scores of 40% and 30%, which provided baseline data. The team began to develop and test their intervention, a fact sheet about C difficile infection.
Effects of change
The non-collaborative wards had 1.15 (95% CI 1.03 to 1.29) cases per 1000 occupied bed days at baseline. One episode of special cause variation (see appendix) occurred in August 2007, six months after the change to antimicrobial prescribing practices, signalled by eight consecutive data points below the baseline mean. This represented a reduction of 56% (to 0.51 (0.44 to 0.60 cases per 1000 occupied bed days) (fig 1⇑).
The collaborative wards had 2.60 (2.11 to 3.17) cases per 1000 occupied bed days at baseline. One episode of special cause variation occurred in April 2007, two months after the change to antimicrobial prescribing practices and one month after the start of the collaborative. This represented a 73% reduction (to 0.69, 0.50 to 0.91 cases per 1000 occupied bed days), which has been maintained (fig 1⇑).
The primary study question was to determine whether teams participating in the collaborative could reduce C difficile infection faster and better than non-collaborative teams. We showed three things:
Careful use of antimicrobials is key to reducing C difficile infection
Teams participating in an improvement collaborative can deliver more improvement sooner than with careful use of antimicrobials alone
Sustainable reductions of up to 73% are achievable when infection rates are high.
A challenge for all policy makers, government officials, clinicians, and healthcare leaders is to determine how much healthcase associated infection (HCAI) is avoidable. We have shown that many more cases of C difficile infection can be avoided than existing reports suggest. Our study shows the importance of careful use of antibiotics on the incidence of C difficile infection, and also that simple changes—developed, owned, and applied by frontline teams—can accelerate and amplify reductions.
More complex is the cause and effect relation between the interventions and their impact on outcomes. It would seem that antibiotic stewardship and collaborative participation combined resulted in a 73% rate reduction within three months in the five collaborative wards. In the non-collaborative wards, introducing antibiotic stewardship alone had an impact later, but with less effect and without additional benefit from scaling-up of the changes during 2008. These observations clearly require careful interpretation, and it is impossible with this experimental design to apportion the impact of the two interventions. Antimicrobial interventions probably had an important role overall in the sustained reductions.
In 2006, Salford Royal had a higher than average rate of C difficile infection, and by the end of the study this had fallen to lower than average for both the health region and England as a whole. Before April 2007, the national reporting scheme did not include reports of C difficile infection in patients under 65 but did include community acquired cases. This makes it difficult to compare our results, which included C difficile infection in adults under the age of 65 and excluded patients who developed symptoms within 48 hours of admission. However, we can compare the overall rate of C difficile infection in patients over 65 for the calendar year 2006 to the financial year 2007-8 (for national data) and to the calendar year 2008 (for regional data). At Salford Royal rates fell by 66% overall (from 2.67 to 0.91 cases per 1000 occupied bed days). This compared favourably with a decrease of 49% in England from (2.45 to 1.26 per 1000 occupied bed days) and of 15% in the region (from 1.97 to 1.67 per 1000 occupied bed days).5 9
The documentation of changes was a critical success factor to our work, but the changes we made (see box 1) are not prescriptive It is impossible to determine which of the changes had the greatest impact. Some of the interventions described in our change package lack any scientific basis—for example, the introduction of “bare below the elbow” for all clinical staff. This change generated considerable debate, but we felt it was an important visible marker of values and commitment to patients’ safety as a priority. Our “package” of interventions may be applicable in other healthcare settings, but other organisations must beware of including changes that may have little effect in a different setting.
We believe that ongoing data collection, reporting, and celebration are essential to sustain improvement. We are developing a robust method to communicate our success to staff, patients, and visitors. We have now switched our ward measurement system to a celebration of the days between events. Automating this feedback into wards’ reporting systems is an important piece of work for the future. Integration of the key process changes from the collaborative (box 2) into the ongoing governance arrangements (ward assurance framework) and assurance of high levels of reliability is essential to supplement the existing focus on outcomes. In line with the national Patient Safety First leadership intervention (www.patientsafetyfirst.nhs.uk), we have also introduced “executive walk rounds” for patient safety and have integrated updates on maintenance of reductions of C difficile infection into this process.
Given the overlap of hospital acquired and community acquired disease, working in partnership with our primary care trust to reduce the incidence of C difficile infection across the whole health economy is an important next step. We are also working to document the clinical and financial benefits to patients. Complications (in particular, readmission rate) may be reduced.
A collaborative model is an effective means of reducing C difficile infection in an NHS hospital. In this quality improvement project, pilot wards achieved early, large, and sustainable improvement. The changes developed in this collaborative can be used by other ward teams and may be an important tool for rapidly reducing the incidence of C difficile infection.
Cite this as: BMJ 2010;341:c3359
We thank E Francis Cook, Harvard School of Public Health, Boston, and the departments of infection control and performance, Salford Royal Hospitals NHS Trust, for their help and support.
Contributors: MP and NW were involved in the design and delivery of the programme and analysis of data; ED and SG were involved in analysis of the data and management of knowledge; PC was involved in the design and delivery of the antimicrobial interventions and analysis of data; DG was involved in the design of the programme and analysis of data. All authors were involved in manuscript preparation. MP and DG are guarantors.
Funding: Salford Royal NHS Foundation Trust.
Competing interests: All authors have completed the unified competing interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare no support from any organisation for the submitted work; no financial relationships with any organisation that might have an interest in the submitted work in the previous three years; and no other relationships or activities that could appear to have influenced the submitted work.
Ethics approval: Not needed.
Provenance and peer review: Not commissioned; externally peer reviewed.