Study setting
This study was designed as before-after and quasi-experiment. The baseline data were collected from 1st March 2018 to 30th September 2018, before the intervention. The intervention was carried out from 1st October 2018 to 30th September 2019. In this study, a multimodal intervention treated A&F and benchmark as key components for MDROs IPC was implemented in three public hospitals in Wuhan, central China. We allocated the hospitals into monthly feedback group (A hospital) and quarterly feedback group (B and C hospital), based on the requirement of those hospitals. The reason for allocation was that the compliance of measures fluctuated unsteadily in one month in B and C hospital due to less MDROs infected patients and limited observed opportunities, which reduced the conviction of feedback for clinical staff [23].
Immediate verbal feedback was given to clinical departments after each audit on the spot in the three hospitals. In the monthly feedback group, monthly written feedbacks were delivered to clinical staff and the IPC committees of departments. The written feedback included the mean level of a hospital as the benchmark and department individual level of 40 A&F implementation indicators [24]. In quarterly feedback group, written feedbacks were delivered quarterly, with the other measures the same as the monthly feedback group. Along with feedbacks, several suggestive approaches and educations aiming at improving the low level of A&F indicators were given to clinical departments and IPC committees of departments. The hospital A had 4 000 beds, providing healthcare services to about 190 000 inpatients and 4.8 million outpatients in 2018. The hospital B and C had 615 beds and 998 beds, respectively. The characteristics of the three participating hospitals were described in Table 1.
A&F multimodal intervention design
In the current study, we adapted the A&F approach to enhance the implementation of multimodal interventions on MDROs IPC. The design process was based on the theoretical model of A&F proposed and recommended by the Healthcare Quality Improvement Partnership and the National Institute for Health and Care Excellence [25, 26].
The dean, leaders of the clinical department, clinical staff, and nosocomial department staff were included in an expert committee to establish the implementation plan. Measures of intervention were chosen from WHO guidelines, China guidelines, and China National health industry standards [3, 9]. The main nine components included hand hygiene, surveillance, contact precaution, patient isolation, environmental cleaning, specimen collection and transport, terminal disinfection and medical waste disposal, standard operating procedure, and appropriate use of antibiotics. A&F implementation indicator pool was developed based on the nine components above and the responsibilities of clinical staffs in MDROs IPC implementation. 40 implementation indicators were confirmed by two rounds of focus discussions based on the indicator pool. (Supplement) Besides, an audit checklist was developed based on those indicators for convenience of use. It is worth mentioning that nurses implemented most measures who were also responsible for reminding, directing, and even supervising environmental cleaners and technicians to implement IPC measures.
Assessment of implementation
The implementation level was measured by the compliance of indicators. A total of 40 training auditors led the A&F multimodal intervention implementation. The auditing was performed twice a week from 8:00-12:00 a.m. based on the audit checklist developed. The paired auditors independently observed IPC behaviors of clinical staff and fulfilled the checklist in one observation. A further mini-interview was also performed to confirm that doctors and nurses did such an operational procedure as usual with patients to lessen the Hawthorne effect. If there was any difference between the two auditors, the consensus was required via discussion with a third person to improve the validity.
Outcomes
Primary outcomes were the monthly incidence of nosocomial MDROs infection of three hospitals per 10 000 patient-days which was detected by clinical culture and acquired more than 48 hours after admission (namely, nosocomial MDROs infection). In our study, we included Carbapenem-resistant Enterobacteriaceae (CRE), Acinetobacter baumannii (CRAB), Pseudomonas aeruginosa (CRPsA), and Methicillin-resistant Staphylococcus aureus (MRSA) into the targeted MDROs, which were considered to need urgent and serious attention of the public [27].
Statistical analyses
Based on the baseline incidence of nosocomial MDROs incidence of 5.55 cases per 10 000 patient-days in the previous study [12], we estimated that both groups would have at least 851 157 patient-day with the assumption of 90% power to show 20% reduction in the incidence of MDROs, with a two-sided type I error of 0.05.
The characteristics and incidence of nosocomial MDROs infected patients were compared by t-test or t’-test and chi-square test [12]. The extent of the intervention changing the incidence of nosocomial MDROs was examined. Poisson segmented linear regression analysis was used to evaluate the effect of the A&F multimodal intervention. As the incidence tended to get a peak in a different season, consequently, we controlled seasonality in our model using harmonic functions [28].
Relationships between the implementation compliance and the incidence were examined using mixed-effect regressions. As data were clustered at the department level, mixed-effect regression analysis was conducted in the monthly feedback group, because only the monthly implementation compliance was available in the monthly feedback group (namely hospital A). There were no missing data in our model. Department-level data were controlled as a random effect. Time elapsed was also considered as the covariate. Descriptive data analysis was performed using IBM SPSS Version 20.0 (IBM, New York, NY, USA). Poisson segmented linear regression and mixed-effect regression analysis were performed using R (Team R Development Core: http://cran.r-project.org/).
The indicators, which were objectively obtained by direct observation or asking patients and relatives, were chosen in our statistical analysis. (Table 2) Other indicators were not included because they were difficult to observe or to ask proper people during the actual implementation. After all, asking nurses or associated medical staff might overestimate the implementation level because of “social desirability” [29]. Nevertheless, they were still treated as implementation indicators for education during the intervention.