This study was conducted in 48 adult ICU beds in five public tertiary hospitals in Recife, in the Northeastern region of Brazil, from January/2018 to June/2019. These were clinical-surgical ICUs with an admission rate of around 1800 patients per year.
Study setting and design
In this quasi-experimental time-series study, interventions were carried and data was collected on a monthly basis for 18 months, including all patients admitted to the ICUs. The methodology was the BTS5 using the “Improvement model”.
Hospital teams were trained by Brazilian philanthropic hospitals of excellence (HE) in diagnostics, data collection and in developing cycles to improve quality and to prevent HAIs (VAP, CLABSIs and CAUTIs). In Brazil, according to Decree 8242 of 04/23/2014, HE are those certified as excellence and exempt from social security contributions, as long as part of their services be offered to SUS (Unified Health System, treated in Brazil by its Portuguese acronym SUS). These institutions are mostly private, offering assistance, teaching, and research activities, to qualify the public health system in exchange for the non-payment of taxes that should be collected.15 These face-to-face and online training sessions took place during periodic sessions for sharing questions, experiences and results. The hospitals received educational visits every 4 months together with online consultations with facilitators on the improvement model, patient safety, intensive care and infectious diseases.
The methodology at original project and this study included following instructional diagrams demonstrating the preventive measures for HAIs, implemented through PDSA (Plan-Do-Study-Act) rapid cycle testing.5 PDSAs are improvement tests when changes were first performed with a small group of patients and healthcare professionals, thereby enabling small-scale testing to result in learning and adaptations. Once the process was considered suitable for the local reality and the tests had achieved success, it was progressively implemented throughout the rest of the unit. The implemented improvements were monitored by indicators and the institutions received technical visits from the HE.
After each learning session with specialists in quality improvement and HAIs, with the presence of four representatives from each of the hospitals (local management team), periods of action were initiated, during which the teams returned to their organizations and tested the changes in their contexts.
Result indicators were monitored monthly: incidence densities (ID) of the HAIs, length of stay and mortality in ICUs and process indicators: the rate at which devices were used and adherence to the preventive measures (bundles).
The local teams were instructed to carry out systematic educational observations on the diagnoses and adherence to the bundles, with at least 20 monthly observations per indicator, in order to plan new PDSAs. The established bundles were: 1- VAP: oral hygiene, raised headboard (30 -45º), reduced sedation, verifying the possibility of extubation, maintaining the cuff pressure of the tracheal cannula (25-30cm of H2O or 20-22 mmHg) and adequate maintenance of the mechanical ventilation system. 2- CLABSIs: on insertion of the central venous catheter (CVC) – check indications, precautions for maximum barrier, skin antisepsis with chlorhexidine, optimal selection of insertion site, adequate dressing after insertion; maintenance of CVC - indication of permanence, aseptic technique in handling, maintenance of the infusion system, correct dressing technique. 3- CAUTIs: when inserting the urinary catheter (UC) – check indication, aseptic technique; maintenance of the UC - permanence of the closed system, correct technique during drainage manipulation, hygiene of the urethral meatus, check the need to maintain the UC.
The local teams monitored and shared the active PDSAs with the ICU team, on a weekly basis - through rounds -, and the indicators, on a monthly basis. The monthly data on the frequency of HAIs and adherence to bundles were recorded on a digital platform to be analyzed in order to direct the necessary actions to improve the team's performance.
The aggregate results of the 119 hospitals participating in the Collaborative until April 2019 have shown reductions of 41% in CLABSI, 48% in CAUTI and 28% of VAP.13
Surveillance of the HAIs was conducted by professionals trained in infection control, using the definitions of the US Centers for Disease Control and Prevention – CDC16 and their incidence was expressed as cases per 1,000 devices-day, obtained by the ratio of the monthly number of cases of infection by the number of patients using the device-day related to this infection.
The utilization rate of the devices was the percentage calculated by adding the number of patients using the device-day divided by the sum of the total number of patient-days in the same period.
The percentage of adherence to bundles was assessed by dividing the number of patients observed with 100% adherence to all items in the bundle by the number of patients observed with the device.
All isolates were identified by manual or automated methods and confirmed with the Vitek 2 system (bioMerieux Vitek, Inc., Hazelwood, MO).
This research was promoted and authorized by Brazilian Ministry of Health, carried out through the Institutional Development Program of the Integrated Health System PROADI-SUS12 and approved by the Ethics Committee of the Hospital das Clínicas - UFPE, under No. 3,307,293.
In the presentation of hospital characteristics, absolute and percentage frequency measurements were performed for categorical variables, and the mean and standard deviation were calculated, as well as the medians and interquartile ranges for quantitative variables. The hypothesis of normality for incidence densities (ID) was tested by the Shapiro-Wilks test, and the hypothesis of normality was accepted.
In the analysis regarding the length of stay, mortality, the IDs of VAP, CLABSIs and CAUTIs over time, Generalized Estimating Equation (GEE) model was applied for continuous variables, using the constant correlation (exchangeable) between assessments over time. The model estimated the average difference (β coefficient of the model) of the measures analyzed during two periods: a period in the year 2017 (prior to implementing the project) and in the years 2018 and 2019 (during the project).
The Spearman's correlation coefficient was estimated in the assessment of process indicators as explanatory variables of the behavior of the result indicators. The percentage of variation in the intervention period was based on the difference between the result indicator in January 2018 and June 2019. All tests of statistical significance were bilateral, with a significance level of 0.05 (p <0.05). All data analyzes were performed using STATA 14.