Journal of Health Psychology 2015, Vol. 20(10) 1263 –1274 © The Author(s) 2013 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/1359105313510338 hpq.sagepub.com

Introduction

Employers have a compelling interest to reduce infectious disease including influenza, the com- mon cold, and gastrointestinal (GI) infections. They pay for the direct costs of absenteeism in employee wages and indirect costs for overtime pay, replacement staff, and reduced quality of services. US employers bear the costs of esca- lating health premiums and, if self-insured, the direct costs of health claims. The World Health Organization (WHO, 2013) estimates an annual global influenza attack rate of 5-10 percent for adults and 20–30 percent for children, while the National Institutes of Health (2011) approxi- mates an annual influenza rate of 5 percent–20 percent. In addition, the average annual number of common cold cases range between 2 and 4

for adults and 6 and 8 for children (Heikkinen and Järvinen, 2003; Monto et al., 2001). Outcomes include deaths, hospitalizations, out- patient visits, absenteeism from work, and lost productivity. US costs for seasonal influenza and the common cold are estimated at US$87.1 billion (Molinari et al., 2007) and US$40 billion (Fendrick et al., 2003), respectively. Although GI infections are less well quantified, roughly 210,000,000 cases occur each year in the United

Hand hygiene performance and beliefs among public university employees

Maggie Stedman-Smith, Cathy LZ DuBois and Scott F Grey

Abstract The workplace is an important location to access community members, and employers have a direct interest in employee well-being. A survey administered to a random sample of employees at a Midwestern US university tested the ability of a model informed by the theory of planned behavior to predict hand hygiene practices and beliefs using structural equation modeling. Questions demonstrated acceptable validity and reliability. Constructs predicted self-reported hand hygiene behaviors, and hand hygiene behaviors reduced the odds of reporting sickness from respiratory tract and gastrointestinal infections. The findings support multi-modal hand hygiene improvement interventions.

Keywords employees, hand hygiene, health promotion, infectious disease, theory of planned behavior

Kent State University, USA

Corresponding author: Maggie Stedman-Smith, Kent State University, Lowrey Hall, 750 Hilltop Drive, #305F, Kent, OH 44242, USA. Email: mstedman@kent.edu

510338 HPQ0010.1177/1359105313510338Journal of Health PsychologyStedman-Smith et al. 2013

Article

1264 Journal of Health Psychology 20(10)

States, of which, 64 percent are nonfoodborne in origin (Mead et al., 1999), while an estimated 9.4 million annual cases occur in the United Kingdom, translating to one out of five people, of which, 50 percent are nonfoodborne (Wheeler et al., 1999). One study found that those who reported experiencing GI infections had a six- fold greater risk of consultation with their phy- sicians at 3 months post infection and a sixfold increased risk of developing irritable bowel syndrome (Cumberland et al., 2003).

Hands contaminated with pathogens from surfaces or direct contact with humans or ani- mals are an important source of transmission for GI, respiratory, and skin infections since they come directly into contact with portals of entry through the mouth, nose, and conjunctiva of the eyes (Bloomfield et al., 2007). Two meta- analyses showed that hand hygiene improve- ment interventions reduced the risk of GI disease by 42 percent–47 percent (Curtis and Cairncross, 2003) and respiratory infection by 16 percent (Rabie and Curtis, 2006). The authors of these meta-analyses concluded that although the number and quality of studies were limited, the evidence showed a clear, consistent pattern of protection from hand hygiene improvement interventions and called for addi- tional trials with greater rigor. A third meta- analysis found that hand hygiene enhancement interventions reduced respiratory and GI infec- tions in community settings, such as schools, day care, and homes, by 21 and 31 percent, respectively (Aiello et al., 2008).

Hand hygiene and respiratory hygiene are recommended by the WHO as strategies to pre- vent pandemics and severe acute respiratory syndrome (Gostin, 2006; Lau et al., 2004). The CDC (CDC, 2013a) advocates a three-tiered approach in the workplace to reduce seasonal influenza: implementing vaccination cam- paigns, performing hand hygiene and respira- tory hygiene, and educating workers to recognize early symptoms of influenza and to stay home when sick. Vaccines among working populations have substantially reduced influ- enza-like illness episodes, along with related

work loss and health-care provider visits (Nichol et al., 2009). However, in the event of a pandemic, hand hygiene and respiratory hygiene are likely to become a first line of defense to slow the spread of disease until vac- cines become available (Bell et al., 2006; Bloomfield et al., 2007). Protective hand hygiene practices are needed as an additional measure to reduce the transmission of infec- tions that are not induced by the influenza virus, such as the common cold and GI infections.

Hand hygiene community guidelines have been issued in the United States (CDC, 2013b). Community members can be reached in the workplace since employees are a captive audi- ence, spending up to half of their waking hours at work. In addition, close proximity during working hours may contribute to the spread of infections. However, to date, only three peer- reviewed articles have reported hand hygiene interventions with general employees, beyond workers who are required to perform hand hygiene to minimize risk to themselves and the public. One study found a 10% reduction in teacher absenteeism as a secondary aim in a hand hygiene improvement intervention that was designed to reduce illness among elemen- tary students (Hammond et al., 2000). A second study among workers in a German public administration setting found 65 percent reduced odds of self-reported illness from the common cold (Hubner et al., 2010). While both of these interventions utilized alcohol-based hand sani- tizer in addition to soap and water, a third study among Finnish workers found a 6.7 percent reduction in infections among workers in the intervention arm that consisted of education and soap and water (Savolainen-Kopra et al., 2012).

Hand hygiene behavior change is complex, and interventions are more likely to be effec- tive if they are informed by behavioral motiva- tions (Glanz et al., 2008; Pittet, 2004). A paucity of information exists about the hand hygiene motivations and behavior of workers and how, if at all, these may differ from the general public. The Theory of Planned

Stedman-Smith et al. 1265

Behavior (TPB) has been used successfully to guide health behavioral interventions among the public since the mid-1980s (Ajzen, 1991; Montano and Kasprzyk, 2002). More specifi- cally, the theory has been used to gain under- standing about hand hygiene performance and motivations among health-care professionals (The Joint Commission, 2009) and profes- sional caterers (Clayton and Griffith, 2008). The TPB postulates that beliefs underlie atti- tudes, subjective norms, and perceived behav- ioral control, which together influence intention and behavior. These beliefs include beliefs about the outcome of the behavior (behavioral beliefs, corresponding to attitudes); beliefs about the perception of expectations regarding the behavior (normative beliefs, corresponding to subjective norms); and beliefs about the extent to which individuals can control the per- formance of the behavior (control beliefs, cor- responding to perceived behavioral control).

Stedman-Smith et al. (2012) administered a pilot survey based on a modified TPB model to workers from 39 bank branches to test the capacity of the TPB to guide understanding about knowledge, beliefs, and practices of self- reported hand hygiene behavior as a prelude to a planned worksite hand hygiene intervention. Findings revealed that behavioral beliefs and normative beliefs were predictive of self- reported hand hygiene performance. However, since these findings were derived from a rela- tively small convenience sample of 159 work- ers, generalizability was limited.

Purpose

The purpose of this study is twofold: (1) to uti- lize a larger, random sample of workers in a public university setting to determine whether a modified model of the TPB generates under- standing about hand hygiene beliefs and prac- tices, as well as predicts self-reported hand hygiene behavior, self-reported infectious dis- ease and related absenteeism and (2) to inform the development of interventions among employees in similar public work settings.

Methods

Procedure and measures

A voluntary, anonymous survey was adminis- tered online to a randomly selected sample of 1600 full-time employees from a Midwestern US university of 5504 workers. Employees received an email that encouraged participa- tion. A hyperlink led to the opening screen of the survey that provided informed assent. At the end of the questionnaire, an additional link took participants to a secure website on a different server to enter a drawing for a US$10 gift card. The survey was open from 7 March 2012 to 2 April 2012; data analysis was performed in the summer and fall of 2012. The study was approved by the University’s Institutional Review Board.

Four major TPB-related constructs were measured using 5-point Likert scales: response options were strongly agree (5), agree (4), some- what agree (3), disagree (2), and strongly disa- gree (1). These included self-reported hand hygiene practices (behavior) and four distinct beliefs: benefits of performing hand hygiene (behavioral beliefs), normative expectations of others (normative beliefs), and behavioral con- trol (control beliefs). Because we wanted to understand workers’ perceptions pertaining to when hand hygiene is needed, a fourth construct, beliefs about protective practices, was measured. Four out of these five constructs have previously shown satisfactory validity and reliability (Stedman-Smith et al., 2012). Questions were informed by United States Centers for Disease Control and Prevention (US CDC) (2013b) com- munity hand hygiene guidelines and scientific literature (Boone and Gerba, 2007). Additionally, one question assessed intention. Intention was not directly measured as a complete construct with three or more items due to constraints on the length of the survey. Consistent with the model used by Sax et al. (2007), it was indirectly meas- ured through the outcome of self-reported hand hygiene behaviors.

Hypotheses were threefold. First, construct validity would be demonstrated for all major

1266 Journal of Health Psychology 20(10)

latent constructs. Second, beliefs would predict self-reported hand hygiene behaviors. Third, hand hygiene behaviors would predict self- reported symptoms of the common cold/ influenza-like illness or GI infections during the past 30 days.

Analyses

Descriptive statistics measured demographics, usual self-reported hand hygiene behaviors, and perceived beliefs about hand hygiene. To estab- lish construct validity, exploratory factor analy- sis (EFA) and confirmatory factor analysis (CFA) were employed. First, EFA was con- ducted using all survey questions. Since the latent factors are expected to be related, Promax rotation, an oblique rotation method that per- mits factors to correlate, was utilized. The num- ber of common factors to extract was determined by examining the scree plot of the eigenvalues for each factor to identify the “breakpoint” where the curve flattens out (Costello and Osborne, 2005).

Several item loading tables of extracted factors from near the breakpoint were exam- ined. The best factor structure was defined as the table in which the largest number of items were loaded onto single factors strongly (val- ues above 0.30), along with the smallest num- ber of cross-loading items. Next, CFA was conducted to establish the convergent and dis- criminant validity of these factors. Factor loadings represented a measure of convergent validity of the survey items onto latent factors, while correlations between the factors served as measures of discriminant validity by dem- onstrating enough independence between the factors to declare each factor to be a separate latent construct (Kline, 1998). Evidence of convergent validity was seen in statistically significant item loadings with standardized values greater than 0.40. Discriminant validity was observed in the correlations among latent constructs that did not exceed 0.80 (Floyd and Widaman, 1995). Finally, Cronbach’s alpha was calculated to determine the internal

consistency reliability of each remaining construct.

Relationships between constructs were mod- eled using structural equation modeling (SEM). In the structural part of the model, paths were specified where all hand hygiene belief con- structs and demographic characteristics were regressed on hand hygiene behaviors, and hand hygiene behaviors and demographic character- istics were regressed in a logistic model on the dichotomous variable that combined participant self-reported cold/influenza-like illness and GI infection during the past 30 days. Within SEM, statistical tests and standardized coefficients for continuous latent variables and odds ratios for the dichotomous illness outcome variable were calculated. Evaluation of the goodness of fit of both the CFA and SEM models was assessed using the Tucker–Lewis index (TLI) and the root mean square error of approximation (RMSEA) (Hu and Bentler, 1999).

Missing data from survey respondents var- ied substantially across variables, with no miss- ing data for initial survey items (0%–18%) and increasing substantially among the demo- graphic variables at the end of the survey (5%– 51%). To assess the missingness mechanism of the data, the above analyses were conducted three times. The first was conducted with sub- jects who had no missing data, effectively assuming that data were completely missing at random (CMAR). The second analysis was done with 10 imputations of the data set created using a sequential regression multiple imputa- tion method (Raghunathan et al., 2001) that assumes data were missing at random (MAR). The third was done using a pattern mixture modeling approach that assumes the data were not missing at random (NMAR) but conditional upon a pattern of missingness (Little, 1993). In this approach, subjects were divided into two data sets reflecting two patterns of missing data (low and high), and 10 imputations of each data set were created, merged together, and analyzed (Thijs et al., 2002). The MAR and NMAR anal- yses did not differ substantially, while the CMAR showed noticeably fewer statistically

Stedman-Smith et al. 1267

significant results due to smaller sample size. The MAR results are presented as they appear most appropriate.

Results

From a base of 5504 workers, 1600 email invi- tations were randomly generated. Of these, 28 were not operative and 19 had opted out of receiving any questionnaires from the univer- sity. Approximately 23 percent participated (n = 361/1553). Most participants were female (72%) and had completed graduate education (60%); of which, 14 percent self-identified as non-White. Over 1/3 reported raising children, and nearly 1/3 employed were taking classes. Greater than 1/3 reported symptoms of a cold or flu-like illness in the last 30 days, while 12 per- cent reported experiencing symptoms of a GI

infection. When combined, 39 percent (141) self-identified as having one or both illnesses in the past 30 days; of those, 34 percent (41) were absent from work due to these conditions, with most employees missing no more than 1 day (Table 1). The top three beliefs about protective hand hygiene practices for which participants reported always or usually were as follows: after using a urinal or toilet (97.2%); before eat- ing or handling food (79.2%); and after blowing their nose, coughing, or sneezing (60.6%).

All hypothesized theoretical constructs had factor loadings of 0.35 or greater in SEM. Control beliefs were separated into two distinct constructs that reflected environmental access and time. Results from CFA demonstrated sta- tistically significant loadings greater than 0.40 for all items. All correlation coefficients for dis- criminant analysis were substantially below 0.80 (Table 2). The construct of hand hygiene behaviors contained the largest number of items that stayed in the model, losing only one item due to lack of variation from a nearly unani- mous affirmative response (performing hand hygiene after using a urinal or toilet). A total of 96.8 percent of participants answered “always or usually” to the question “When I intend to clean my hands I actually follow through and do it.” Thus, this question dealing with inten- tion did not load in the model due to lack of variance.

SEM revealed that all hypothesized con- structs were significantly associated with hand hygiene performance. Demographic analyses revealed significantly fewer hand hygiene behaviors practiced by those who completed graduate education and those who were younger. No differences were found in the pre- diction of hand hygiene behaviors by gender or raising children (Table 3).

Hand hygiene behaviors significantly reduced the odds of reporting sickness by 45 percent from the common cold/influenza-like illnesses and GI infections during the past 30 days, when combined. Those who self-identified as non- White had over twofold higher odds of experi- encing these infections (Table 4).

Table 1. Demographic characteristics of participants.

Characteristic N %

Female 247 72 Raising children at home 125 37 School-aged children in home 62 19 Have a graduate degree 204 60 Employee taking classes 110 32 Non-White race 47 14 Received a flu shot 138 40 Missed work due to illness 64 19 30-day cold/flu illness 115 34 30-day GI illness 41 12

Mean SD

Age, years 43.20 12.64 Household income, US dollars 84,451 87,896 Estimated time spent working with public

54.57 29.75

Number of work days missed 0.39 1.00 Number of household wage earners

1.77 0.61

SD: standard deviation; GI: gastrointestinal. For dichotomous characteristics, the percentage values are calculated from the total number of responses, which may be less than the total number of respondents due to missing data.

1268 Journal of Health Psychology 20(10)

Table 2. Confirmatory factor analysis of hypothesized hand hygiene constructs.

Construct/Item Est. SE Construct correlations Cronbach’s alpha

1 2 3 4a 4b 5

1. Hand hygiene behavior 1.00 0.89 Before I eat or handle food 0.44 0.04 After I handle money 0.86 0.05 After coughing/sneezing/blowing

nose 0.63 0.05

After I use a shared keyboard/pad 0.97 0.05 After I share pens 0.84 0.05 After picking up something 0.79 0.06 After shaking hands coughed/

sneezed 0.65 0.05

After touching public surfaces 0.89 0.05 2. Behavioral beliefs 0.53 1.00 0.92 I can get sick if I don’t 0.75 0.04 Protecting the health of my family 0.70 0.03 Protecting my health 0.66 0.03 Protecting the health of my

coworkers 0.71 0.03

I can get sick if I rub my eyes/nose

0.57 0.04

3. Normative beliefs 0.35 0.32 1.00 0.80 My employer expects me to 0.58 0.07 My colleagues do 0.79 0.06 Our customers do 0.48 0.04 People I work with think it’s

important 0.95 0.07

The clientele that I serve do 0.54 0.04 4a. Control beliefs: access to

hand sanitizer 0.43 0.40 0.38 1.00 0.74

It is convenient for me to use 0.74 0.05 Hand sanitizer makes it easier 0.72 0.05 I carry hand sanitizer with me 1.20 0.07 I have hand sanitizer on my desk 1.25 0.07 4b. Control beliefs: convenience

of hand hygiene 0.21 0.23 0.16 0.13* 1.00 0.83

It is convenient for me to get to a sink

0.67 0.06

I have sufficient time 0.67 0.05 I do not perceive any barriers to

hand hygiene at work 0.54 0.05

5. Beliefs about protective hand hygiene practices

0.48 0.40 0.17 0.26 −0.01* 1.00 0.93

After using a shared keyboard/ pad

0.71 0.03

After sharing pens 0.91 0.04 After picking up from floor 0.79 0.04 After touching public surfaces 0.73 0.04

SE: standard error; TLI: Tucker–Lewis index; RMSEA: root mean square error of approximation. Model fit: TLI = 0.909; RMSEA = 0.06. All factor loadings and construct correlations are significant (p < 0.05) except correlations marked with “*.”

Stedman-Smith et al. 1269

Discussion

The findings from this study, which employed a model drawn from key components of the TPB,

revealed that all major constructs loaded in EFA; CFA demonstrated good convergent and discriminant validity, and Cronbach’s alpha

Table 3. Structural equation model of hypothesized hand hygiene motivators predicting hand hygiene behavior.

Regressor Standardized coefficient 95% confidence interval

Modified theory of planned behavior constructs Behavioral beliefs 0.30 (0.18, 0.41)** Normative beliefs 0.13 (0.02, 0.24)* Control beliefs: Access to hand sanitizer 0.16 (0.04, 0.28)* Convenience of hand