Study design and subjects
During a prospective sequential period, ITS analysis was conducted to evaluate the effects of daily CHG bathing in a 23-bed MICU at Korea University Anam Hospital, a 1048-bed tertiary care teaching hospital in South Korea, from September 2016 to December 2017. We employed a pre-intervention and intervention, quasi-experimental design with equivalent assessment timing to evaluate the effect of CHG bathing implementation after completion of the intervention. We checked the achieved effects and assessed the unexpected effects of the intervention. This enabled detection of changes in real-world data that might be attributable to the intervention.
The study comprised two periods: the pre-intervention period from September 2016 to February 2017 (26 weeks) and intervention period from July 2017 to December 2017 (26 weeks). A 6-month intervention with 2% CHG daily bathing was performed, whereas standard bed baths using soap and water were provided twice a week during the pre-intervention periods. An intervening period between two periods was excluded from the analysis, which corresponded to the preparation of the customized CHG bathing protocol and pilot trial (9 weeks), and the following wash-out period (8 weeks). The subjects included patients aged ≥ 19 years, who had been admitted to the MICU for > 72 h (Fig. 1). The primary outcome was a change in the acquisition rate of rectal VRE, as determined using active surveillance culture (ASC). The secondary outcome included a change in the incidence of VRE, MRSA, or CRAB, as determined using clinical cultures.
Customized 2% chlorhexidine daily bathing protocol
Infection control unit staff nurses developed the CHG bathing protocol based on a previously published CHG decolonization protocol [17]. Commercial CHG cloth-compatible products were not available in South Korea; therefore, cotton wipes impregnated with 2% CHG were prepared daily, using 5% CHG solution (Green Pharmaceutical Co., Ltd, Seoul, South Korea) and warm, sterile distilled water prior to use. Six clean CHG cloth wipes were used to bathe the whole body below the jawline, concentrating on each area of six body sites sequentially (neck, arms, groin/perineum, right leg/foot, left leg/foot, and back of neck, back, and buttocks). Partial bathing was performed for patients who had medical conditions in which they were unable to move easily or moved with pain. CHG bathing was performed once daily by two trained ICU assistants during the entire ICU stay. All the patients were tested for skin contact irritation or allergic reactions to CHG exposure prior to bathing. Patients with MDRO isolation were bathed last. Patients in critical condition, dermatitis, or abnormal skin conditions were also excluded from CHG daily bathing, depending on the patient’s condition.
CHG bathing compliance was assessed daily, based on documentation provided by an infection control practitioner. A patient receiving full or partial bathing was considered “compliant”. Compliance rate was calculated through dividing the total number of the “compliant” patient-days by the total number of patient-days.
Skin swab cultures of the body sites at high risk of MDRO acquisition were monitored in a small number of randomly selected patients prior to the pilot trial (n = 4) and during the 6-month intervention (two patients per month, n = 12), including patients who had a longer ICU stay of > 30 days and MDRO isolates from the clinical culture (Additional file 1: Table S1).
Routine infection control and prevention program in the ICU
The hospital runs a routine ICU infection control and prevention program, including ASC on ICU admission for rectal VRE and nasal MRSA, a follow-up weekly surveillance rectal culture for VRE acquisition, single room or cohort isolation of patients with VRE, monitoring of hand hygiene adherence for HCWs, monitoring of HAIs, and the electronic antibiotic stewardship program. Patients with a previous MDRO colonization were placed on contact precautions and cohort isolation on admission to the ICU. The infection control staff monitored all the measures on a monthly basis.
Data collection
Demographic and clinical data concerning patients in the ICU were collected through reviewing hospital electronic medical records. We collected data on age, sex, admission department, length of ICU and hospital stay, Acute Physiology and Chronic Health Evaluation (APACHE) II score, medical devices used (mechanical ventilator, urinary catheter, and central venous catheter), and recent surgery and clinical outcomes.
Microbiological data were collected from hospital electronic medical records. Data on MDRO surveillance and ICU-acquired device-associated infections (DAIs) were also obtained from the infection control unit database. The acquisition of rectal VRE using ASC and incidence of VRE, MRSA, and CRAB in clinical cultures were collected weekly for ITS analysis.
Definitions
The compliance rate with the CHG bathing intervention was calculated as the total number of CHG bathing days divided by the total number of patient-days as denominator and expressed in percentage. Since the patients with a stay ≤ 72 h in ICUs were excluded, the patient-days for these patients were subtracted from the denominator. VRE acquisition was evaluated for all the patients who had no previous VRE isolation, a negative ASC result for VRE on ICU admission, and a positive follow-up weekly ASC result for VRE. The incidence of VRE, MRSA, and CRAB was defined as newly acquired positive clinical cultures obtained > 48 h after ICU admission. Clinical cultures were obtained depending on a patient’s clinical status, as determined by the medical staff. Only the first isolate from a single body site per patient was included in the analysis. Identification and susceptibility determination of MDROs were undertaken in the hospital microbiological laboratory. The laboratory definition of each MDRO was determined according to the Clinical and Laboratory Standards Institute criteria [18].
The definitions of MICU-acquired DAIs, including central line-associated bloodstream infection (CLA-BSI), catheter-associated urinary tract infection (CA-UTI), and ventilator-associated pneumonia (VAP), were used based on those of the KONIS national surveillance program, as our hospital has been participating in the ICU module of the program since 2006 [19,20,21].
Statistical analyses
We compared continuous variables, expressed as median and inter-quartile range (IQR), using Student’s t- or the Mann–Whitney U tests, depending on their distribution. Categorical variables were analyzed using χ2- or Fisher’s exact tests. The VRE acquisition rate and the incidence rates of MDROs were calculated as the total number of episodes divided by the total number of patient-days as denominator and expressed per 1000 patient-days. DAI rates were calculated as the total number of cases per 1000 device-days.
ITS was used to statistically measure the changes in the level and slope of the trends over time for the acquisition of rectal VRE and incidence of MDROs between the pre-intervention and intervention periods. We calculated the final multivariable model with a significant parameter (the use of mechanical ventilator) using stepwise backward variable selection. The outcome and offset parameters were the weekly number of cases and the logarithmized patient days, respectively. Overdispersion was tested for the distribution of cases, including potential confounding parameter such as the use of mechanical ventilators. Since there was no overdispersion, we calculated the change in the level and slope of the series using a segmented Poisson regression model. The model used for the analysis is as follows:
Log[E(Y∣X1,X2,X3,X4)] = β0 + β1log(X1) + β2log(X2) + β3log(X3) + β4log(X4) + et
(X1: time, X2: intervention, X3: time after intervention, and X4: covariates), where log[E(Y∣X1,X2,X3,X4)] was the VRE acquisition and MDRO incidence per patient-days, time was the number of weeks starting in September 2016, β0 estimated the intercept at the beginning of the time series, β1 estimated the log-linear trend of the pre-intervention period, where X1 was a continuous variable indicating the time in weeks at time t from the initiation of the study period, β2 estimated the level changes in incidence, where interventiont = 0 was pre-intervention, and interventiont = 1 was intervention, β3 estimated the change in weekly trend from pre-intervention to intervention, β4 estimated the effects of significant covariates, and et was the random error at time t.
Incidence rate ratio (IRR) and the percent change as two standardized effect sizes were obtained as a ratio of change in level and trend between the pre-intervention and intervention periods [22]. A change in level represented how the outcome level had changed from the last observation prior to the intervention to the first one after, based on the Poisson model predictions and not on a difference between observed versus predicted levels. Moreover, a change in level represented a direct change, whereas a change in slope represented a sustained effect (if slope goes down) or unsustainable effect (if slope goes up, after level change down) [23]. A negative change in the level and slope indicated a reduction in the infection rates. A p-value of < 0.05 was interpreted as a significant association with intervention efficacy. ITS analysis was performed using package R version 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria) software.