Study design and setting
We conducted a quasi-experimental study (named as the ‘WardGel’ study) in which HH compliance and HAI rates were compared before and after the introduction of ABHR on 3 clinical departments in Mbale Regional Referral Hospital (MRRH), a government hospital in eastern Uganda. MRRH is one of the 14 governmental regional referral hospitals in Uganda, and serves over four million people in its catchment area of 15 local districts and beyond. The hospital has 12 wards with 550 beds. It also functions as a teaching hospital where there are 40 physicians (including interns), 170 nurses and students including medical students and nursing students. Before this study, ABHR was only used by a small number of senior HCPs such as nursing supervisors who carried portable ABHR bottles. This was mainly because of financial constraints that prevented its purchase by the hospital. As such, ABHR use was almost non-existent prior to this study. Typically, only one or no functional sinks/taps were available in each ward. Portable water bottles and basins were an alternative for hand washing. Gloves, even non-sterile ones, were rarely available. A ward was usually a single, open space without isolation rooms. The hospital followed a standard operating procedure for general environmental cleaning, which was unchanged during the study period.
We selected five wards across three departments as the study sites, namely; the acute and general pediatric wards (pediatric department), the gynecology and post-natal wards (obstetrics/gynecology (OBGYN) department), and the general surgical ward (surgical department). The two pediatric wards were selected because evidence of the impact of ABHR on HAIs among pediatric patients has been particularly scarce in resource limited settings [7]. The latter three were selected because there were relatively few cases of infection on admission on those wards, and it was thought that HAIs would be more accurately observed than at other wards where febrile illnesses such as malaria were more common on admission. Overall, the study was designed so that a wide variety of the patients were observed for multiple types of HAIs to enhance generalisability of the study.
The study was conducted over 24 weeks between October 2014 and April 2015, with the first 12 weeks of the baseline (pre-intervention) phase followed by 12 weeks of the intervention phase. There was a 4-week study interruption between December 2014 and January 2015 (after week 8) due to the holiday seasons in Uganda when the number of inpatients and HCPs were low. Inclusion criteria for the study were all the patients who were admitted on the above selected wards during the study period (hospital day 1 = day of admission).
During the baseline phase, the HH compliance rate among HCPs was assessed by direct observations performed by the trained research assistants on the wards, based on the WHO hand hygiene technical reference manual and the WHO five moments for HH, i.e. before touching a patient, before clean/aseptic procedure, after body fluid exposure risk, after touching a patient, and after touching patient surroundings [2]. Each trained research assistant on each ward observed HCPs during the day-times only for targeted ward activities including busy times such as ward rounds by physicians. The research assistants conducted observations openly, without interfering with the ongoing clinical work, but kept the identity of the HCP confidential, observing up to a maximum of three HCPs simultaneously provided there were no missing opportunities. One observation session lasted for 10-30 min; and only prolonged the sessions in situations of observing a care sequence to its end.
The HH compliance was calculated by the number of observed HH actions (using either ABHR or hand washing with soap and water) upon an opportunity divided by the total number of opportunities for HH actions. The amount of ABHR consumption from non-portable bottles was assessed as a supplemental indication for HH compliance through ABHR. The incidence of HAIs was also measured during the baseline phase through prospective follow-up of patients by research assistants (see below).
During the intervention phase, in addition to the study activities performed during the baseline phase, the one-litre ABHR bottles fitted in a locally-made metallic holder were mounted on the walls of the wards where the access to ABHR was thought to be convenient. Mobile bottles were also placed on the trolleys for ward rounds, on the reception area, and on the treatment area. Portable 40 ml hand-sized bottles were also provided to HCPs and kept available throughout the intervention phase. The ABHR used in the study was Alsoft V, ABHR locally made from sugar cane in Uganda by Saraya East Africa Co. Ltd. It manufactured locally, but according to international standards (Good Manufacturing Practice), and contained the recommended concentration of ethanol (76.9 to 81.4 vol%) [2, 10, 11]. An introductory training session on ABHR use was provided to all the HCPs on the target wards by the research team in week 12 with the help of staff trainers from Saraya East Africa Co. Ltd. In order to maintain the HH compliance, the introductory training was followed by the distribution of educational posters on HH in week 17 and follow-up training in week 18. Additional training was also conducted as required when new medical students and nursing students came to the wards. As the first training session was provided in the last week of the baseline phase, patients who were hospitalized in week 12 and week 13 were excluded from the final analysis of the incidence of HAIs to minimize study contamination between the baseline and the intervention phases.
For the assessments, a paper-based surveillance form was created to record demographics, patient interventions (e.g. surgical interventions), vital signs, clinical findings, antibiotic use and patient’s outcomes. SSIs, urinary tract infections (UTIs), pneumonia, central nervous system (CNS) infections, gastroenteritis and episiotomy infections were selected as the HAIs measured in this study. Each definition of HAIs was modified from the 2014 version of United States Center for Disease Control and Prevention’s National Healthcare Safety Network (CDC/NHSN) surveillance definitions of HAIs, considering locally available resources [12]. Research assistants (mainly registered nurses but with one physician) were trained to fill out the surveillance form, to measure vital signs, to identify relevant clinical signs from patients’ medical records, and to record laboratory and imaging findings. After enrollment into the study, the patients were prospectively followed on a daily basis until discharge. Post-discharge follow-up calls were also attempted for all the patients around 1 month after their discharge. Individual patients’ data collected from patients’ medical records and vital signs measured by the research assistants were also reviewed by research supervisor for quality assurance of the data. All the data collected on the paper-based forms were entered into an Epi Info® database installed onto computers at the Sanyu Africa Research Institute (SAfRI). The only exception was the HH compliance data that was entered into the WHO-produced Microsoft Word® data collection sheet for analysis [13].
Outcomes
The primary outcomes were the HH compliance rates among HCPs and the incidence rates of HAIs before and after the intervention. The secondary outcomes included antibiotic usage, length of hospital stay and hospital mortality of the study participants.
During the run-in period, it was noted that medical documentation by physicians and clinical officers was not always sufficient to make a diagnosis of an HAI as defined for this study. Direct questioning of the physicians and clinical officers involved with the patient management was also difficult due to their busy schedule and difficulty in recall given the large volume of patients seen per physician. There was therefore a post-hoc change in the primary outcome from HAIs only to the composite outcome of HAIs and or criteria for systemic inflammatory response syndrome (SIRS) occurring on hospital day 3 or after (SIRS/HAI) (c.f. SIRS criteria for adult patients, two or more of: 1. temperature > 38 °C or <36 °C, 2. heart rate > 90/min, 3. respiratory rate > 20/min or Paco2 < 32 mmHg, 4. white blood cell (WBC) count >12,000/mm3 or <4000/mm3 or >10% immature bands. SIRS criteria for pediatric patients, two or more of the same four items with variable thresholds for age, at least one of which must be temperature or WBC) [14, 15]. Those with SIRS on hospital day 1 or 2 of their hospital stay were excluded from the composite primary outcome. Paco2 and WBC were rarely performed at MRRH, therefore vital signs measured by research staff were mostly used to determine whether SIRS criteria was met or not.
Statistical analysis
Means, standard deviations (SD) with t-tests, and proportions with chi-squared tests were calculated for continuous and categorical variables in bivariate analyses, respectively, in order to describe demographics and clinical variables of the study participants. Poisson regression analysis was used to compare HH compliance rates before and after the intervention. Linear regression analysis was used to compare HH compliance rate and ABHR consumption during the intervention. Relative risks and incidence rate ratios of SIRS/HAI were also calculated to compare risks and incidence rates before and after the intervention, respectively. Survival analyses with the cox proportional hazard model were performed to calculate hazard ratios (HRs) in order to describe the associated variables for SIRS/HAI. Multivariate survival analysis with backward selection and plausible causal interpretation was used to calculate adjusted HRs. Statistical significance was defined as a p value of <0.05 and 95% confidence intervals (CIs) were reported. All statistical analysis was conducted using Statistical Analysis System (SAS®) version 9.3 (SAS Institute, Cary, NC, USA).
There was no formal HAI rate reported in Uganda that we could use for the sample size calculation. We therefore used the WHO data for developing countries to estimate that 10% of the hospitalized patients would develop any type of HAI [16]. We estimated a 3% reduction in HAIs after the intervention. With the level of significance defined as α = 0.05, and statistical power as β = 0.80, we estimated 1356 patients each would be required before and after the intervention.
Role of the funding source
The study was conducted with funding provided by Saraya East Africa Co. Ltd., who also provided the ABHR made at their local factory in Uganda. Their staff helped the research team conduct HH training on the wards. Otherwise, the funder was not involved in the study design, data collection, data analysis, data interpretation, or manuscript writing.
The study was approved by the Mbale Regional Hospital Institutional Review Committee (MRHIRC)(REIRC IN – COM 098/2014) and registered to ClinicalTrials.gov (NCT02435719).