Factors Associated With Hand Hygiene Compliance at a Tertiary Care Teaching Hospital

Benjamin Kowitt, MD; Julie Jefferson, RN, MPH; Leonard A. Mermel, DO, ScM


Infect Control Hosp Epidemiol. 2013;34(11):1146-1152. 

In This Article


Study Population and Setting

Rhode Island Hospital (Providence, RI) is an urban academic medical center licensed for 719 beds. It is a level I trauma center. The campus includes 16 buildings with clinical units or departments covering approximately 2 million square feet with over 2,700 alcohol hand rub dispensers and 3,000 soap dispensers. There are 1,767 affiliated physicians and 7,400 employees working at the hospital. In this setting, we analyzed data collected by observers of hand hygiene compliance of hospital staff across all inpatient units from July 2008 through December 2012.

Data Collection

Light-duty hospital staff (ie, hospital staff, such as nurses, who were unable to perform their regular duties but who were able to perform hand hygiene surveillance) were employed as observers and received a standardized set of instructions for obtaining hand hygiene observations. They were instructed to collect a minimum of 10 observations per inpatient unit per week. They were advised to introduce themselves as "collecting hand hygiene observations for Infection Control" when appropriate but to otherwise be unobtrusive. Additionally, they were asked to obtain the names of observed noncompliant staff. Data were hand recorded on tracking sheets and subsequently entered into a web-based database. Instructions included the definition of hand hygiene compliance and how to complete the tool. Light-duty staff were then initially sent out to the units with another experienced observer. Periodically, infection control staff accompanied the observers to check for interrater reliability.


The outcome variable was hand hygiene compliance, defined as using alcohol-based hand rub or hand washing with soap and water according to the Rhode Island Hospital infection control policy, based on the Healthcare Infection Control Practices Advisory Committee hand hygiene guidelines. Two opportunities for compliance were observed during each patient encounter: before entering and after leaving a patient room. Each of these opportunities was counted as a separate observation in our analysis in a binary fashion. If a staff member performed hand hygiene when leaving a patient room and immediately entered another patient room, they were not expected to perform hand hygiene again.

Other variables recorded by observers included unit, time and date, shift (day, evening, or night), healthcare worker type, whether the patient was under contact precautions, gloves worn, gown worn, and the healthcare worker's full name, if obtained.

For our analysis, nursing staff comprised registered nurses, licensed practical nurses, certified nursing assistants, unit assistants, student nurses, and intravenous team nurses; physician staff included physicians, house officers, medical students, physicians assistants, nurse practitioners, and dentists; and technical staff included respiratory therapists, rehabilitative and therapeutic service staff, radiology technicians, radiology students, radiology aids, dieticians, phlebotomists, dental hygienists, dental assistants, case managers, social workers, and dialysis technicians. Support staff included tray passers, patient transporters, environmental services, pastoral care, and personal care attendants. Patient care units were grouped as follows: medical units, surgical units, pediatric units, ICU, and other. The category "other" included the emergency department, specialty clinics, psychiatry units, endoscopy unit, radiology suites, dialysis unit, postanesthesia care unit, dental clinic, and overflow units, among others.

Statistical Analysis

Data were entered into a Microsoft Excel database. Analysis was performed with Excel and Stata/SE 12.1 (StataCorp). Frequency tables were tabulated with proportional confidence intervals for compliance. A χ2 test was used for comparison between those proportions (differences were deemed statistically significant in the results section if P <. 05). Bivariate odds ratios (ORs) were calculated with 95% confidence intervals. Multivariate analysis was estimated using a logistic model. The model was constructed with dummy variables for each variable studied, with the exception of a continuous month counter variable that spanned the entire observation period. All ORs in the multivariate models were estimated by exponentiation of the coefficients of the logistic regression, and 95% confidence intervals were used for those estimates.