Hand hygiene campaign

Surveillance data on MRSA bacteremia, following participation in EU/EEA surveillance networks, elicited a previously absent sense of urgency to address the problem. This coincided with the launch of the World Health Organization’s “Clean Care is Safer Care” Global Patient Safety initiative, promoting improved hand hygiene (HH) through increased alcohol hand rub (AHR) use at critical moments of patient contact. ^{Reference Pittet, Allegranzi, Storr and Donaldson7} A HH campaign was launched in 2008 and has been described elsewhere. ^{Reference Abela and Borg8} It focused heavily on extensive audits of HH compliance and monitoring of AHR use, which we have previously shown to correlate well with assessment of HH performance through visual observations at MDH. ^{Reference Borg and Brincat9}

Root cause analysis

At around this time, the successful efforts of the United Kingdom to address its hitherto high MRSA prevalence were being highlighted. ^{Reference Duerden, Fry, Johnson and Wilcox10} A cornerstone of this initiative was the requirement to perform a root cause analysis (RCA) for each case of HA-MRSA-B, conducted by the patient’s clinical team. The protocol required early gathering of data to find out what happened, generating an action plan to address the key issues identified and the implementation and monitoring of an action plan which would feed learning into an organization’s governance to help reduce the chances of a repetition. ¹¹ In 2010, an attempt was made to leverage positive deviance from this experience and implement a similar approach in MDH. However, initial experiments to replicate the UK administrative model were unsuccessful, primarily due to lack of ownership by front-line professionals. RCAs only materialized when the hospital’s Infection Prevention and Control (IPC) team took over their full management. IPC performs the preliminary review of the case and invites all the key stakeholders involved in the patient’s care to a meeting, led by the IPC Lead. The clinical team provides the background which is then followed by a discussion, asking the “five why’s?” to identify possible factors. ^{Reference Venier12} Every effort is made to ensure that the meeting takes place in a safe, non-punitive environment and that everyone can provide their input, keeping the focus on the event and related processes. At the end of the meeting, based on RCA conclusions, action items are agreed and minuted. The minutes and corrective actions are sent to all meeting participants and copied to senior management. The IPC team reviews implementation of agreed corrective actions and raises any identified lack of progress with senior hospital management.

An assessment of the RCAs held in 2011, the first full year of the initiative, showed that the absolute majority of cases were attributed to intravenous (IV) devices. Most were linked to renal dialysis (36%) where almost all patients had had a non-tunneled double-lumen hemodialysis catheter (NTDLHC) at the time of the HA-MRSA-B. A further 25% of HA-MRSA-B were linked to peripheral vascular cannulas (PVC) and 22% to central venous catheters (CVC). In practically all of these RCAs, clear practice gaps were identified and evidence-based corrective actions instituted hospital-wide (Table 1).

Table 1. Main factors identified from root cause analysis (RCA) reviews held in 2011 and corrective actions implemented.

Intravenous device interventions

The corrective actions included a mixture of policy development/updating, education and training, positive and negative motivators as well as—essentially—system change. In renal dialysis, targets were established to reduce the lag period from the start of dialysis until arterio-venous fistula surgery. NTDLHCs were only allowed as an emergency interim measure for no longer than 3 weeks. Within this period, and in all elective cases, a cuffed tunneled line needed to be inserted for hemodialysis. ^{Reference Gallieni, Brenna, Brunini, Mezzina, Pasho and Giordano13} Whereas 30.1% of patients on hemodialysis had a NTDLHC in March 2011, this reduced to 0.5% in March 2017, despite a 64% increase in the hemodialysis population during the same period. This major change was only possible following the engagement of two specialist radiologists with experience in such insertions. Maintenance of dialysis lines was also improved through adoption of dressings with higher moisture vapor transmission rates (MVTR) and training of nurses in care bundles.

The Visual Infusion Phlebitis (VIP) score became a required mandatory daily assessment for each patient with a PVC. ^{Reference Morris and Heong Tay14} This policy modification was only possible after another system change: the introduction of PVC dressings with transparent visibility windows. The implementation of a 72-hour cutoff for PVC duration was more contentious, since publications at the time had shed doubt on its effectiveness. ^{Reference Rickard, McCann, Munnings and McGrail15} Nevertheless, it was determined that the level of PVC monitoring and care in the centers where the studies were held, had not yet been achieved in MDH; therefore the literature could not necessarily be transposed to the hospital situation. CVC initiatives mirrored the recommendations of the Institute of Health Improvement and were based heavily on the work of Berenholtz et al. ^{Reference Berenholtz, Pronovost and Lipsett16} They were however modified to reflect local culture. In particular, ICU nurses raised objections to supervising intensivists during insertion and, even more, to stopping the procedure if they noted any deviations from operating procedures; neither was adopted. In addition, the checklists were used more as an aide memoire than the more stringent application in the reference publication.

Universal MRSA admission screening

The third key intervention was introduced in 2014 with the commencement of universal MRSA admission screening for practically all patients admitted to the hospital, other than in pediatrics and obstetrics. ^{Reference Borg, Suda, Scicluna and Brincat17} This intervention was also not without some controversy since consensus at the time advocated primarily risk-based MRSA screening strategies. ^{Reference Robotham, Deeny, Fuller, Hopkins, Cookson and Stone18} Nevertheless, as we have already described previously, ^{Reference Borg, Suda, Scicluna and Brincat17} it was felt that the hyper-prevalence of MRSA colonization in admitted patients at the time (exceeding 13%) as well as local issues which made consistent risk assessment at ward level improbable, were sufficient reasons to trial a universal strategy. This decision was vindicated by a significant reduction in all MRSA infections (including HA-MRSA-B) at very reasonable annual cost of €1058 per QALY gain per year. ^{Reference Borg, Suda, Scicluna and Brincat17} Once again a centralized process was adopted, managed completely by care assistants employed within the MDH IPC team. They identify new daily admissions from the hospital’s patient administration system, visit the wards to perform the screening themselves directly onto culture media, and then coordinate the decolonization of all positive cases by a 5-day regimen of nasal mupirocin and chlorhexidine bathing. ^{Reference Septimus, Weinstein, Perl, Goldmann and Yokoe19}

In order to assess the impact of the interventions on HA-MRSA-B rates, we used multiple change-point analysis using the R package changepoint (R Studio Ver 2023.03.0, United States) to detect significant shifts in the mean of the data, without prior setting of any intervention points in the time series model. ^{Reference Killick and Eckley20} Due to the retrospective nature of the data, the offline type was utilized. We adopted the binary segmentation technique proposed by Scott and Knott and avoided overestimation of the number of change points by applying penalties utilizing the Schwarz information criterion (SIC), Bayesian information criterion (BIC), Akaike information criterion (AIC) and Hannan–Quinn information criterion (HQIC). ^{Reference Scott and Knott21} A one-tailed Mann–Whitney test was used to determine whether overall incidence in the segments following each change point was significantly different from that in the segment preceding it. To justify the use of the Mann–Whitney test, we ensured that the data within segments were stationary and independent using the Dickey–Fuller test and Ljung–Box test, respectively. A Shapiro–Wilk test was also conducted to check for normality within each segment. The non-parametric Mann–Whitney test was chosen because the normality assumption was violated on more than one occasion. R Studio was also used to conduct this analysis.