Study design and randomisation
Interventions to promote HH on intensive care units have been conducted at our hospital in the past. On non-intensive care (i.e. peripheral) wards of our hospital, interventions as performed in this trial have not been undertaken previously. We therefore decided to perform a cluster randomised controlled trial in 20 peripheral wards of the Charité-University Medicine hospital in Berlin, a large tertiary care university hospital with three separate sites and approximately 3000 beds. Each participating ward represented a cluster.
In the process of ward selection, we decided to exclude palliative and paediatric wards for ethical reasons. Moreover, intermediate care units were excluded. Since one of our objectives was to assess the effect of the intervention on the incidence of positive blood cultures (see below), we decided to exclude wards with low blood culture sampling frequencies. To achieve that, we evaluated the blood culture sampling frequency of all non-intensive care wards in the year 2016. Wards with a number of blood cultures per 1000 patient-days above the median were included in the randomisation. Of these wards, 20 were randomly selected and allocated to either group (intervention: n = 10; control: n = 10) with a computer-generated sequence (https://www.randomizer.org/). The randomisation was performed by the study coordinators. Participation of wards was on a voluntary basis. The decision for or against participation was made by the head nurse and physician of the respective ward. Recruitment of wards was performed between September and November 2017. All wards were free to end their participation in the study at any stage without stating a reason. The study was approved by the ethics commission of Charité-University Medicine Berlin (EA4/123/17).
Description of the intervention
The intervention of this study consisted of three aspects. First, quarterly interdisciplinary team meetings of ward staff with the study team including data on HH compliance observations, and goal setting. Second, distribution of training materials (see description below), and third, distribution of flexibly mountable alcoholic hand rub dispensers. The selection of the individual elements of the intervention was done in alignment with available literature [10,11,12,13, 16]. These intervention measures were undertaken and provided solely for wards of the intervention group, not the control group. The intervention period was from the beginning of January 2018 until the end of December 2018.
Hand hygiene compliance observations
After randomisation of wards, 5 cycles of HH compliance observations according to the recommendations of the WHO [17], were performed in all 20 wards. The first cycle was performed in December 2017 and served as a baseline assessment against which the effects of the intervention were measured. The following 4 cycles were performed quarterly over the intervention period of 1 year. A minimum of 150 observations with a minimum of 30 observations before clean or aseptic procedure per cycle was required. While staff in control wards received no feedback on HH observation data, wards in the intervention group received regular feedback on compliance data. The feedback was embedded in interdisciplinary team meetings, where the study team presented results of the compliance observations that were then discussed with the ward staff.
HH compliance observations were executed by student employees, which were trained in HH observation methodology by experienced local infection control staff in a specifically organised workshop. Observations were recorded using the Observe app from HARTMANN. In order to increase the quality of observations, the student employees were supervised by an experienced infection control nurse.
Intervention materials
Wards randomised into the intervention group were introduced to the intervention during a “kick-off meeting” (i.e. feedback on the first cycle of HH compliance observation), where they received materials to promote HH and infection prevention that had been developed by the study team prior to the study. Wards were encouraged during the kick-off meeting and following quarterly team meetings, as well as by email reminders, to implement these materials into their routine patient care. Among the materials were 10 step-by-step pictogram checklists for selected aseptic procedures, two explanatory films, and flexibly mountable alcoholic hand rub dispensers. The materials were chosen based on the fact that HH is an integral part of many aseptic procedures in routine patient care.
The 10 step-by-step checklists addressed the following topics:
-
insertion of peripheral venous catheters (PVCs)
-
use of a sterile extension set for PVCs
-
management (incl. dressing change) of PVCs
-
management (incl. dressing change) of central venous catheters (CVCs)
-
management of central venous ports
-
preparation of intravenous injections
-
application of intravenous injections
-
preparation of intravenous infusions
-
application of intravenous infusions
-
disconnection of intravenous infusions
The two animated explanatory films contained additional information on the insertion of PVCs, and preparation as well as handling of intravenous infusions. All intervention materials were handed out during the kick-off meeting, where applicable, both in printed and electronic form.
Observed outcomes
As our primary outcome, we selected HH compliance of healthcare workers in participating wards. Compliance was determined by direct observations following the WHO Five Moments of Hand Hygiene model [18].
As a secondary outcome, we selected the incidence of device-associated BSIs for which prospective surveillance was conducted over a 12 months period (11 months during the intervention period, 1 month post-intervention). Due to technical difficulties, only 19 of the 20 participating wards were included in this part of the study. BSIs were defined as a blood culture with a recognised pathogen, or two positive blood cultures with a common skin contaminant, drawn at least 48 h after admission. The two positive blood cultures with a common skin contaminant had to be from two separate blood samples within a five-day period. BSIs were considered as device-associated, if an intravenous catheter was present either on the day of sampling, or the day before in case the catheter was removed on the day of sampling. Furthermore, the catheter had to be in place for a minimum of 3 days before the sample was taken. Where more than one catheter fulfilled the criteria for a device-associated BSI, an association was made to the catheter being inserted in the larger blood vessel (e.g. CVC was chosen over PVC), except where local signs of infection indicated otherwise. Information to make this allocation was gathered from the ward staff, primarily the treating physicians.
Additionally, we evaluated the frequency of positive blood cultures taken at least 48 h after admission of the patient to the ward. For this, all positive blood culture sets of all 20 wards were collected for a one-year baseline period, the one-year intervention period, and a three-month follow-up period. Isolated microorganisms from blood cultures were divided into nine categories: coagulase-negative staphylococci, Staphylococcus aureus, Streptococcus spp., Enterococcus spp., Enterobacteriales, non-fermenting bacteria, Candida albicans, other Candida spp., and others.
All observed outcomes pertained to the cluster level.
Statistical analysis
HH compliance was calculated for every group (control vs. intervention) and every cycle. HH opportunities and actions were analysed descriptively. To compare the baseline period (cycle 1) with the intervention period (cycles 2–5), the absolute change in compliance was calculated. Additionally, odds ratios (ORs) were calculated with a generalised linear mixed effect model including ward as a random effect. To minimise the potential distortion by confounders and investigate the effect of the intervention on HH compliance more accurately, a multivariable logistic regression analysis was conducted, taking into account other known factors of influence [4, 9, 14, 19]. For this purpose, ORs and 95% confidence intervals (CI95) were calculated. Parameters included in the model were: period (baseline vs. intervention), professional group, WHO-moment, and ward specialty.
Mean infection rates per 1000 patient-days were calculated with CI95 and mid p-values. To compare isolates per patient-days between baseline, intervention, and follow-up periods, incidence rates and incidence rate ratios (IRRs) with CI95 and mid p-values were calculated. Incidence rates were calculated as Poisson rates. CI95 and p-values for IRRs were calculated by median-unbiased estimation. A p-value of less than 0.05 was considered significant. All analyses were performed with SAS and R [20]. Graphics were done with ggplot2 [21].