The Development and Validation of the Health Belief Model for Shift Workers (HBM-SW) Scale

ABSTRACT

Objective

Shift work is associated with circadian misalignment, sleep loss, and suboptimal health behaviors, which may contribute to longer term negative health outcomes. To inform future interventions for shift workers, the present study aimed to develop and evaluate the Health Belief Model for Shift Workers (HBM-SW) scale.

Methods

The HBM-SW development involved a seven-step process, including a literature review, expert panel analysis, cognitive interviews with shift workers, and a trial with a pilot sample of shift workers (n = 153). Utilizing exploratory factor analysis for factor identification and item reduction, the developed scale loaded on seven factors in line with the theoretical framework of the Health Belief Model: Perceived Threat, Perceived Severity, Perceived Benefits, Perceived Barriers, Cues to Action, Self-efficacy, and Health Motivation. Validation of the scale was conducted utilizing Pittsburgh Sleep Quality Index, International Physical Activity Questionnaire, and Food Frequency Questionnaire.

Results

The pilot sample had an average age of 34.0 (18.0) years, was majority female (54.2%), with an average of 8.0 (11.0) years shift work experience. The HBM-SW showed good – excellent (α = 0.74–0.93) internal consistency and moderate – good (ICC = 0.64–0.89) test re-test reliability. Using health behavior outcome measures, the HBM-SW scale showed meaningful correlations with sleep quality, sleep duration, diet quality and leisure time physical activity, and acceptable validity and reliability. Further testing should be conducted in a larger sample to facilitate confirmatory factor analysis.

Conclusions

The developed Health Belief Model for Shift Workers scale is likely beneficial for use in future studies of interventions for shift workers.

Introduction

Sleep, dietary intake and physical activity are all health behaviors which can be considered through an ecological lens (Bronfenbrenner, 1977; Grandner, 2019). Sleep, diet, and physical activity are influenced by a range of factors both internal (circadian rhythms, hormones, and biological changes) and external (social, cultural, and environmental) (Broussard et al., 2017; Potter et al., 2016; Short & Mollborn, 2015). Beyond these internal and external influences on health behaviors, there are additional influences from micro (i.e., gender, age, and genetics), meso (i.e., support, and occupation), and macro (health policies and frameworks) levels (Reynolds et al., 2021). One external factor that impacts both the internal and external factors of these health behaviors is shift work.

In our 24 h society, there is a need for around-the-clock availability of employees across various sectors, which requires a proportion of workers to participate in shift work. Shift work is the term used to describe “work arrangements outside conventional daytime hours, and can include fixed early morning, evening, and night work, as well as roster work and rotating three shift work” (Kecklund & Axelsson, 2016). Between 16% to 21% of the population are engaged in shift work schedules globally (Australian Bureau of Statistics, 2010; Eurofound, 2017; Reynolds et al., 2021; U.S. Bureau of Labor Statistics, 2019). While these schedules meet societal needs, engagement in shift work is known to be associated with negative chronic health outcomes for the workers, such as increased risk of cardiovascular disease, some cancers, and metabolic syndrome (Brown et al., 2009; Cheng et al., 2021; Kecklund & Axelsson, 2016; Wang et al., 2018, 2011; Ward et al., 2019).

As a diurnal (day-active) species, disruptions to certain lifestyle behaviors are common for shift workers. Disruption to diurnal behavior is thought to contribute to some of the negative health outcomes frequently observed in the shift working population (Baron & Reid, 2014; Kecklund & Axelsson, 2016). For example, working a night shift means that physical activity and food consumption occur during the biological night, and sleep occurs during the biological day. Together, these altered behaviors contribute to misalignment of our circadian system, the 24-hr rhythms that ensure the body’s processes are optimized. Broadly, this circadian misalignment has been linked to insufficient sleep, suboptimal health behaviors such as physical inactivity (e.g., not engaging in sufficient regular physical activity), and poor diet quality (e.g., higher intake of fat and sugar and lower calcium, fruits, and vegetables intake than recommended).

While shift work has been associated with negative consequences in the long term (Kecklund & Axelsson, 2016), it is important to also consider the healthy worker effect (McMichael, 1976). The healthy worker effect is a phenomenon whereby those who have fewer health complaints are more likely to work in jobs that require shift work (McMichael, 1976). The healthy worker effect means that, due to the potentially healthier population of workers, it is possible to underestimate the true effect of shift work on health. Furthermore, the healthy worker effect also means that people that develop coping strategies or may be biologically predisposed to “adapt” better to shift work tend to stay in the job and others will self-select out and thus, those in shift work may present with healthier beliefs.

Identifying and managing the consequences of shift work schedules for health is important to support both workers and employers in the long term. Circadian adjustment interventions (i.e., interventions that optimize circadian alignment with work schedule) have successfully improved sleep and other chronic disease risk factors in shift workers (Lowden et al., 2019). However, it is not always feasible or desirable to align with a shift work schedule, particularly when workers are not on a permanent shift pattern (i.e., rotating shift work) (Folkard, 2008). Therefore, it is important to look beyond circadian adjustment to other possible pathways to mitigate negative health outcomes for shift workers. Optimizing health behaviors is one consideration.

Before attempting to improve health behaviors in shift workers, we first must identify factors that may be impacting a shift worker’s ability to engage in optimal health behaviors. To do this, we require a purpose developed scale to understand and predict health behaviors in shift workers. The use of existing behavioral models to inform scale development and future interventions will likely be beneficial for improving health behaviors in shift workers, as interventions which are developed according to health behavioral frameworks are more effective than interventions that are not (Ammerman et al., 2002; Legler et al., 2002). One specific behavioral model which has been examined across a variety of populations and interventions, and which considers the various levels of behavioral influence (micro, meso, macro), is the Health Belief Model (Abood et al., 2003; Conner & Norman, 2005; Jeong & Ham, 2018; Mehta et al., 2014; Shmueli, 2021).

The health belief model

The health belief model proposes that engaging in a health behavior is a result of a combination of psychological and social influences (hereafter referred to as “psychosocial determinants”), which operate mostly subconsciously, and may change over time (Champion & Sugg Skinner, 2008). The Health Belief Model proposes that an individual’s health behaviors can be explained by, or predicted from, a set of psychosocial determinants that underpin engagement with health behaviors (Champion & Sugg Skinner, 2008; Janz & Becker, 1984; Rosenstock et al., 1988). Brief explanations of the psychosocial determinants proposed by the Health Belief Model are provided in Table 1.

Table 1. Psychosocial determinants proposed by health belief model with brief explanation.

Psychosocial Determinant	Explanation
Perceived Threat [PT]	An individual perception of whether they are at risk of a health outcome or disease
Perceived Severity [PS]	An individual’s perception of the seriousness of condition or disease
Perceived Benefits [BF]	An individual’s beliefs about the efficacy of a health behavior to reduce risk
Perceived Barriers [BR]	An individual’s beliefs about the cost of, or obstacles to, engaging in a health behavior
Cues to Action [CA]	Diverse triggers that can activate engagement in the health behavior
Self-efficacy [SE]	The belief that an individual has the capacity to engage in a health behavior
Health Motivation [HM]	An individual’s general motivation to look after their health

Note. Definitions of determinants in table are drawn from existing literature (Conner & Norman, 2005; Glanz et al., 2008).

Some of the psychosocial determinants proposed by the health belief model have previously been explored in shift working populations. Psychosocial determinants including benefits, barriers, and self-efficacy are associated with physical activity in health-care staff and students (Blake et al., 2017; Mo et al., 2011; Patra et al., 2015). In addition, a review of studies exploring food intake in shift workers found reports of many perceived barriers to healthy eating (Anstey et al., 2016; Persson & Martensson, 2006). Qualitative explorations of factors that influence healthy lifestyles in shift workers have identified multiple barriers and possible benefits for improving health behaviors (Nea et al., 2017). Specially, shift workers commonly reported that working time arrangements are a barrier to prioritizing sleep, eating well, and exercising (Nea et al., 2017). Specific information about the benefits, barriers, and self-efficacy in shift workers is now required to support meaningful interventions. In addition to these determinants, the shiftwork locus of control (consisting of elements of self-efficacy and health motivation) is associated with better health behaviors in shift workers (Lawrence Smith et al., 1995, 2005). Finally, when compared to daytime workers, shift workers are more likely to perceive a health risk associated with their work schedule (Crowther et al., 2020), indicating that the perceived threat aspect of the health belief model may also be relevant to shift workers. These findings support the use of the health belief model in studies of shift workers and a need for systematic exploration using the complete model in a validated metric.

Many of the psychosocial determinants that underpin the health belief model are applicable and relevant to shift workers, and interventions anchored in the health belief model have a greater chance of success within this population. A robust scale would allow for meaningful cross-sectional and longitudinal assessments of health beliefs in shift workers. The present study aimed to develop and psychometrically test a Health Belief Model for Shift Worker (HBM-SW) scale and investigate the health belief model factors that are associated with sleep, physical activity, and diet quality.

Methods

The development of this scale followed best practice guidelines proposed by Boateng et al. (2018) and DeVellis (1991). The Health Belief Model for Shift Workers (HBM-SW) scale was developed through seven overarching steps.

Step 1:

A literature review was conducted as Step 1 utilizing key words (e.g., “Health Belief Model” “Scale development” “Shift work” “Physical Activity” “Diet” “Nutrition” “Sleep” “Behavior change” “Health Belief Model”) to examine available literature on health outcomes, health behaviors, and applications of the health belief model in shift workers.

Step 2:

Following literature review, the initial items were developed using deductive methods (Morgado et al., 2018), including some adaptions of four previously developed health belief model questionnaires (Guvenc et al., 2011; Jeong & Ham, 2018; Moreira et al., 2020; Wu et al., 2020). The existing scales were evaluated for relevance to (i) our population of interest, and (ii) the outcomes of interest. Additional relevant sleep and shift worker scales were also used to inform wording of the developed HBM-SW scale (Cho et al., 2013; Murawski et al.). A total of 75 draft items were developed at this stage.

Step 3:

The use of expert panel judges, independent of the research team, is recommended to ensure items for scale development represent domains of interest and to avoid bias in assessment of items (Boateng et al., 2018; Haynes et al., 1995). Samples of between five and seven judges are recommended (Haynes et al., 1995); however, larger panels may result in more robust ratings (Boateng et al., 2018). International experts (n = 14) in the areas of occupational health, scale development, health behaviors, and health behavior models were invited by e-mail to participate in the assessment of initial items. The assessment was conducted by survey using REDCap software (Harris et al., 2009). Experts were asked to rate each individual scale item on:

1. Relevance to domain measurement (“How relevant is this item to the domain?” 1) not at all relevant, 2) A little bit relevant, 3) Somewhat relevant, 4) Very relevant),

2. Interpretability of item (“Is the item clear and understandable?” 1) Not at all clear/understandable, 2) A little bit clear/understandable, 3) Somewhat clear/understandable, 4) Very clear/understandable),

3. Meaningfulness of the item (“Is the item meaningful and/or significant to the domain?” 1) Not at all meaningful/significant, 2) A little bit meaningful/significant, 3) Somewhat meaningful/significant, 4) Very meaningful/significant).

Experts were also given the opportunity to provide qualitative feedback on each item (“Any comments about this item”), to allow for a more precise understanding of the modifications needed (Elangovan & Sundaravel, 2021). Content Validity Index (CVI) for each item was calculated by calculating the number of experts who scored the item a 3 (somewhat meaningful/significant) or 4 (very meaningful/significant) and dividing by the total number of experts who provided a response for that item. Thus, CVI maximum is 1 for each of the above response categories (relevance, interpretability, and meaningfulness). These three domains were then averaged to provide an average item-level CVI (Lynn, 1986). Item-level CVI scores of 0.79–1.0 were considered acceptable, 0.60–0.79 required modification and items with scores of <0.60 were removed (Zamanzadeh et al., 2015). Expert panel qualitative feedback was also considered for modification of items to ensure items accurately reflected the domain of interest and were relevant to the study population (Elangovan & Sundaravel, 2021).

Step 4:

The next phase of scale development engaged end-users (current shift workers) to ensure the developed scale was easy to answer and interpretable by the intended population. One of the most efficient ways of ensuring questions are well understood and relevant to the intended population is through cognitive interviews (Beatty & Willis, 2007; Boateng et al., 2018). The suggested number of cognitive interviews needed for scale development varies within the literature, with suggested samples ranging between five and 15 (Alaimo et al., 1999; Beatty & Willis, 2007; Boateng et al., 2018). Given these guides, a minimum of five participants was considered necessary for sufficient scale evaluation. Cognitive interviews involved asking current shift workers to read each scale item and verbally describe the process for answering the question, and took between 32 and 103 min to complete. Participants were told that it was not their answer to the scale item we were interested in, but rather how and why they would answer in that way.

Interviews were transcribed, and key findings regarding each scale item were coded. When considering the end-user feedback, items were either retained, removed, or modified following discussion within the research team. Items that were not easily interpreted by end-users were removed.

Step 5:

The draft scale was administered to a pilot sample of shift workers via Qualtrics Survey software (Qualtrics, 2020). Participants could take part in the pilot survey if they were currently working nonstandard schedule (i.e., outside 8am−6pm). The decision to include only those working nonstandard hours was due to the wording and intention of the scale being focused on those who were engaged in shift work and not those working regular hours.

The response options for all questions, except for Perceived Barriers, were as follows: 1. Strongly disagree, 2. Disagree, 3. Neither agree nor disagree, 4. Agree, and 5. Strongly agree. The response options for Perceived Barriers were as follows: 1. Never, 2. Not often, 3. Sometimes, 4. Often, and 5. Always. For all scale items, higher scores were indicative of greater health beliefs (i.e., higher levels of perceived barrier sub-scale are indicative of greater perceived barriers).

Participants were invited to participate in an online survey through social media (Twitter, Facebook, Instagram, and Reddit) for observational studies (Shatz, 2017; Topolovec-Vranic & Natarajan, 2016), paid social media advertising (Facebook and Instagram) and the Survey Circle website. Participants were eligible to participate if they were i) currently working shift work, and ii) could read/speak fluent English. There were no other inclusion/exclusion criteria.

Participants who completed the survey could opt in to be placed in the draw for a $30 gift card and were invited to indicate their interest in completing a shorter survey in 2 weeks’ time. The subsequent shorter survey contained only the draft scale (demographics and other sleep and health measures were not repeated) and was used to determine preliminary test re-test reliability of the developed scale.

Measures for pilot

Demographics

Participants were asked to report their age, gender, occupation, average weekly work hours, current shift work schedule, usual shift length, current contract condition (permanent, fixed-term contract, casual, other), and years of shift work experience. Participants were also asked to report whether they had been diagnosed with a chronic disease and/or whether a close family member/friend had been diagnosed with a chronic disease.

Pittsburgh Sleep Quality Index

The Pittsburgh Sleep Quality Index (PSQI) (Buysse et al., 1989) is a validated self-report measure of sleep quality. The PSQI contains seven subscales of sleep, with one global sleep score. PSQI sleep scores >5 are indicative of poor sleep quality. The PSQI also allows for measurement of self-reported habitual sleep duration. The PSQI was chosen due to its previous use in studies of shift workers (Lajoie et al., 2015; Sathvik et al., 2022) and its ability to reflect on a period >7 days, which is important for shift workers whose schedule may change between weeks.

International Physical Activity Questionnaire

The International Physical Activity Questionnaire Long Form (IPAQ) (Craig et al., 2003) is a validated self-report measure of physical activity. The IPAQ allows for calculation of job-related, transportation, domestic, and garden and leisure time physical activity, as well as typical sitting time. The IPAQ was scored in line with the recommended data cleaning and scoring guidelines (International Physical Activity Questionnaire, 2005). Continuous weekly leisure time Metabolic Equivalents (METs) were calculated and used for analysis.

Dietary Screener Questionnaire

The Dietary Screener Questionnaire (DSQ) (National Cancer Institute, 2015) was developed for use in the National Health and Nutrition Examination Survey. The DSQ contains 26 items which are based on known dietary risk factors. The DSQ captures intakes of fruits and vegetables, dairy/calcium, added sugars, whole grains/fiber, red meat, and processed meat. The scoring from DSQ provides a predicted daily intake of each dietary variable.

Analysis of pilot data

All analyses were conducted in R (R Core Team, 2021) and R Studio (R Core Team, 2019). Participants who had data missing from ≥60% of the first survey (i.e., pilot HBM-SW) were considered incomplete responses, and the cases were removed from all analyses. All data were examined visually for normal distribution and using the Shapiro Wilks test in “stats” package (R Core Team, 2021). Differences between participants who completed the entire survey (i.e., HBM-SW, demographics, PSQI, IPAQ, and DSQ) and those who completed at least the HBM-SW scale but not the subsequent PSQI, IPAQ, and DSQ were compared using Kruskal–Wallis tests and Chi-square test in the “coin” package (Hothorn et al., 2006) to determine any differences between groups related to likelihood of completion.

Step 6:

Testing for the HBM-SW scale was conducted to assess whether the remaining data were Missing Completely At Random (MCAR), using Little’s test of MCAR (Little, 1988) in “Naniar” (Nicholas Tierney et al., 2021). Suitability of the data for factor analysis was assessed in line with recommendations (Hair et al., 2009) to ensure adequacy of sample size (i.e., sample must be larger than n = 100), the Kaiser–Meyer–Olkin test of sampling adequacy (value must be >0.50), and Bartlett’s Test of Sphericity (significance <0.05) in “psych” (Revelle, 2021). Multiple imputations were conducted using “MIFA” in R (Vahid Nassiri et al., 2021), based on 50 dataset imputations and using the method proposed by Nassiri et al. (2018), with auxiliary variables of gender, age, average work hours, and shift work experience. Principal component analysis was subsequently conducted using promax rotation, as this assumes correlation between factors (Yong & Pearce, 2013). Factors were initially determined through analysis of eigenvalues, with values >1 retained (Kaiser, 1960) and using visual analysis of scree plots (Yong & Pearce, 2013). Following factor identification, items were reduced based on factor loading. Following guidelines suggested by Hair et al. (2009), accounting for our sample size of >150, items were considered to sufficiently load on a factor if the factor loading was >.45. Items that loaded less than ≤.45 (Hair et al., 2009), loaded on to alternative factors >30, or did not demonstrate a difference of 0.20 between their primary and alternative factors were removed (Howard, 2016) (Yong & Pearce, 2013).

Step 7:

Behavior-specific belief subscales

The developed HBM-SW is based on beliefs related to three health behaviors. As an individual’s engagement in each of these health behaviors may differ (i.e., someone may engage in sufficient physical activity but not sufficient sleep), behavior-specific belief subscales were proposed, in addition to the HBM-SW subscales (i.e., perceived threat, perceived severity, etc.) scores. The use of specific behavior-specific belief subscales (i.e., sleep, dietary intake, and physical activity subscales) will allow for a better understanding of how the HBM-SW may explain specific health behaviors independent of each other (i.e., sleep subscale without questions that mention physical activity or diet). Furthermore, the subscales were calculated as Benefits (Behavior-specific) minus Barriers (Behavior-specific), plus Cues to Action (Behavior-specific), Self-efficacy (Behavior-specific), and Health Motivation (Behavioral-specific) The calculation of subscales in this way allows for the Benefits to outweigh the Barriers as this is the proposed framework for predictive health behaviors in the Health Belief Model (Champion & Sugg Skinner, 2008; Orji et al., 2012), which is then combined with other determinants specific to behavior to understand how the model predicts specific health behaviors (Supplementary Material).

Reliability

Reliability of the scale was assessed using the Cronbach’s alphas (Cronbach, 1951) of each of the HBM-SW subscales, in line with best practice guidelines (Boateng et al., 2018). Cronbach’s alphas were calculated using “psych” R package (Revelle, 2021). Items that were shown to decrease the reliability of subscale were removed (Boateng et al., 2018). Test re-test reliability was measured using a two-way mixed effects model intraclass correlations in “psych” (Revelle, 2021), using the baseline and follow-up responses from participants who took part in both time points, as in previous studies (Koo & Li, 2016).

Content validity

The developed HBM-SW scale was compared to the sleep, exercise, and diet quality health outcome measures. To allow for as complete analysis as possible, participants were included for each outcome variable that they provided complete responses for (i.e., participants who completed HBM-SW and PSQI but not IPAQ or DSQ, would only be included in the HBM-SW and PSQI analysis). The content validity was measured using pairwise Pearson’s correlations for each HBM-SW subscale and outcomes of interest (global sleep quality, sleep duration, leisure time METs, and diet quality components). Kruskal-Wallis Rank Sum Tests were conducted using high, average, and low levels of HBM-SW behavioral specific subscales with continuous outcomes, to measure whether differing levels of HBM-SW behavioral subscales were associated with relevant behavioral outcomes, using “coin” in R (Hothorn et al., 2006).

Results

The findings from scale development are presented in the order of the development steps outlined above.

Expert panel review

Nine of the 14 invited experts (64%) completed the expert panel review of the draft scale of 75 items. All experts had a PhD-level education and were in areas of expertise including shift work research (22%), occupational health (33%), health behaviors (78%), behavioral models (44%), and scale development (33%). For the drafted items, CVI ranged from 0.33 to 1, with an average CVI of 0.85, representing a high level of content validity. Of the initial scale items, 20 items showed insufficient CVI (i.e., <0.60) and thus were removed from the scale.

Following the removal of items first based on CVI, a review of the qualitative feedback from the experts resulted in a further 45 items, with sufficient I-CVI, being modified in line with expert advice to ensure they were easily interpreted by the population. For example, the terms “healthy eating” and “regular exercise” were replaced with “range of nutritional food” and “regular physical activity”. For items in Barriers, Benefits, Self-efficacy, Cue to Action that used the term “Prioritising sleep” were modified to “get enough sleep”, in line with expert suggestions that health behaviors are better reflection by actually doing the behaviors (i.e., “Getting enough sleep) vs intending to do so (i.e., “Prioritising sleep”). Furthermore, sleep duration was suggested as the anchor in the sleep questions to avoid additional item complexity that may arise with terms such as sleep quality. Finally, 10 items remained as initially proposed. By completion of the expert panel review stage, 55 draft items remained for inclusion in cognitive interviewing.

Cognitive interviews

Cognitive interviews were conducted with nine shift workers. Of the nine shift workers 67% were male (n = 6). The average age of interview participants was 34.7 ± 7.7 years (Range: 25–54 years). Interviews took on average 60.4 ± 22.7 min (Range: 32–103 min). The majority of the interview participants (n = 7) indicated that they worked a rotating shift arrangement. All participants reported shift lengths of between 8 and 12 hours, and four participants reported being on call at times for their working arrangements. The average duration of shift work experience was 14.4 ± 9.6 years (Range: 3–37 years). Following cognitive interviews, 14 of the 55 draft items from the expert panel level were removed, modifications to wording were made for a further 13 draft items in line with shift workers’ feedback. Overall, 28 draft items were retained unchanged. Items were modified, in line with end-user suggestions, to ensure they were easily understood and relevant to the population. For example, the original item CA1. “As a shift worker, I am more likely to get enough sleep if my family or friends remind me” was modified to “As a shift worker, I am more likely to get enough sleep if my family or friends encourage me”. The item was seen as interpretable by shift workers, but they felt that they were more likely to be responsive to being encouraged by their family/friends rather than reminded.

Administration of the scale

A total of 273 participants commenced the survey, of whom 23 (8%) did not meet eligibility criteria. Of the remaining 250 potential participants, 24 did not provide any data beyond eligibility (9%), 153 participants (61%) completed the developed HBM-SW scale, while 122 (49%), 99 (40%), and 97 (39%) participants completed the PSQI, IPAQ, and DSQ, respectively. Finally, 36 participants (14%) completed the second survey a minimum of 2 weeks later for test–retest reliability. The demographic characteristics of all participants who completed the HBM-SW at time point 1 are presented in Table 2. To gauge any group differences in those who completed the HBM-SW (n = 153) and those who started but did not complete HBM-SW (n = 73) were compared. Those who did not complete HBM-SW were more likely to not report a chronic disease (p = 0.014) and to not have a family member or friend with a chronic disease (p = 0.039) (Table S1, Supplementary Material). We also examined any differences between those who partially completed the whole study (HBM-SW alone versus full study), we compared participants who completed the DSQ (as this was the final survey presented to participants and thus completion of this measure was indicative of survey completion) (n = 97) and those who did not (n = 55), and found they did not significantly differ on any demographic measures or any HBM-SW subscale or behavioral scale scores (Table S2, Supplementary Material).

Table 2. Demographics of sample that completed HBM-SW scale.

Characteristic	n (%)
Age (yrs.)*	34.0 (18.0)
Gender
Male	64 (41.8%)
Female	83 (54.2%)
Non-binary	3 (0.0%)
Missing	3 (0.0%)
Shift arrangements
Rotating shift work – Yes	93 (60.8%)
On call – Yes	29 (19.0%)
Shift work experience (yrs.)*	8.0 (11.0)
Weekly work hours*	38.0 (10.0)
Occupation
Health care	43 (28.1%)
Trades and Services	26 (17.0%)
Emergency Services	19 (12.4%)
Care assistant or Support worker	14 (9.2%)
Retail	9 (5.9%)
Hospitality	7 (4.6%)
Transport	7 (4.6%)
Administration	6 (3.9%)
Cleaner	5 (3.3%)
Technology services	5 (3.3%)
Management	2 (1.3%)
Education	2 (1.3%)
Other	4 (2.6%)
Not reported	4 (2.6%)
Chronic Disease
Chronic disease – Yes	52 (34.0%)
Family/friends with chronic disease – Yes	72 (47.1%)
Behavioral Outcomes
Global PSQI (>5 = poor sleep)*	9.0 (5.0)
Sleep duration (hrs. per night)*	6.5 (2.0)
Fiber (mg per day)*	16.4 (4.1)
Calcium (mg per day)*	955.3 (222.5)
Wholegrain (ounce per day)*	0.7 (0.4)
Sugar (tsp per day)*	15.0 (5.4)
Dairy (cup per day)*	1.5 (0.5)
Vegetables (cup per day)*	1.6 (0.4)
Fruit (cup per day)*	0.9 (0.5)
Sugar sweetened beverages (tsp per day)*	5.6 (3.9)
Occupational Physical Activity (METs per week)*	3168.0 (6755.0)
Leisure time Physical Activity (METs per week)*	396.0 (1980.0)
Sitting time (hrs per week)*	11.0 (15.0)

Note: Data reported as n(%), unless indicated with * where data are reported as median and interquartile range.

Pilot sample demographics

As shown in Table 2, more female (n = 83) than male (n = 64) shift workers participated in the pilot sample testing. All participants regularly worked shift work (outside 8am-6pm). The majority of participants reported working in a rotating shift work arrangement (60.8%), while the remaining participants reported early morning, afternoons, night shifts, or split shifts, the median weekly work hours were 38.1 (IQR = 10.0) hours. The occupations represented within the current sample varied, with most participants working in health (28.1%), trade (17.0%), and emergency services (12.4%). Over one-third of the sample reported being diagnosed with a chronic disease (34.0%) and almost half of the sample reported having a close family member or friend diagnosed with a chronic disease (47.1%). The medians for each health behavior outcome included are also provided in Table 2.

Factor analysis

Little’s test of MCAR was not significant (p = .98) and thus, data was considered to be missing at random, meaning multiple imputation for missing values was appropriate. The KMO score was 0.77, and Bartlett’s test of Sphericity was significant (p < .001), indicating adequacy for factor analysis.

Through factor analysis, five items were dropped due to insufficient loading or unacceptable cross loading. Principal component analysis resulted in a seven factor solution, which explained 61% of variance within the sample. The loading structure of all final 36 items is presented in Table 3. The seven factors extracted aligned with the proposed Health Belief Model factors: Perceived Threat [PT], Perceived Susceptibility [PS], Benefits [BF], Barriers [BR], Cues to Action [CA], Self-efficacy [SE], Health Motivation [HM]. The final HBM-SW scale is presented in Supplementary Material 4.

Table 3. Final scale items and results of factor anaylsis for Health Belief Model for Shift Workers Scale (HBM-SW).

Item Code. Item	Benefits	Threat	Barriers	Cues to Action	Self-efficacy	Severity	Health Motivation
PT1. Working shift work makes it more likely that I will develop a chronic disease	0.09	0.93	−0.10	0.03	−0.13	−0.05	−0.04
PT2. Working shift work makes it more likely I will develop chronic diseases compared to someone else my age who is not working shifts	0.03	0.86	0.01	0.03	0.04	0.01	−0.04
PT3. As a shift worker, my chances of developing a chronic disease in the next few years a high	−0.12	0.77	0.02	0.07	−0.17	0.06	0.04
PT4. Working shift work increases my chances of developing a chronic disease, compared to if I wasn’t working shift work	0.06	0.86	−0.13	0.07	−0.13	0.10	−0.01
PT5. I am concerned about the negative effects of developing a chronic disease because of working shift work	0.04	0.70	0.13	−0.04	0.00	0.09	0.08
PS1. If I develop a chronic disease because of working shift work, my whole life would change	0.19	0.07	−0.09	−0.11	0.07	0.81	−0.09
PS2. Developing a chronic disease because of working shift work, would impact my relationship with my important family members or friends	−0.09	0.12	0.12	0.00	0.04	0.67	−0.02
PS3. If I had a chronic disease my whole life would change for the worse	0.00	0.00	0.00	0.04	−0.09	0.73	0.02
BR1. My shift work schedule does not leave me enough time to get enough sleep	0.02	0.26	0.67	−0.07	0.12	−0.12	0.10
BR2. My shift work schedule does not leave me enough time purchase and/or prepare a range of nutritious foods	−0.09	−0.13	0.88	0.00	−0.01	0.03	−0.05
BR3. My shift work schedule does not leave me enough time to do regular physical activity	0.11	−0.05	0.86	0.04	−0.01	−0.10	−0.09
BR4. Working shift work makes it difficult for me to get enough sleep	0.05	0.23	0.61	−0.09	0.10	0.04	0.06
BR5. Working shift work makes it difficult for me to eat a range of nutritious foods	−0.02	−0.11	0.74	0.08	−0.22	0.13	0.00
BR6. Working shift work makes it difficult for me to do regular physical activity	0.07	0.00	0.70	0.03	−0.15	0.08	−0.10
BF1. As a shift worker, getting enough sleep will help prevent the development of chronic disease	0.70	0.23	−0.04	0.01	0.13	0.04	−0.03
BF2. As a shift worker, eating a range of nutritious food will help prevent the development of chronic disease	0.84	0.15	0.02	0.07	0.11	−0.17	−0.13
BF3. As a shift worker, doing regular physical activity (such as walking, bike riding or exercise) will help prevent the development of chronic disease	0.83	0.08	−0.01	0.03	0.01	−0.01	−0.02
BF4. As a shift worker, getting enough sleep is good for my health	0.85	−0.09	0.00	−0.01	−0.15	−0.01	0.07
BF5. As a shift worker, eating a range of nutritious food is good for my health	0.85	−0.08	0.03	−0.05	0.06	0.06	0.04
BF6. As a shift worker, doing regular physical activity is good for my health	0.68	−0.16	0.07	0.01	0.04	0.21	0.08
SE1. I am sure I can eat a range of nutritious foods, even when working shift work	0.04	0.09	−0.16	0.01	0.74	0.03	−0.05
SE2. I am sure I can do regular physical activity, even when working shift work	−0.02	−0.12	−0.04	0.00	0.77	0.01	0.09
SE3. When working shift work, if I really want to I can eat a range of nutritious food	0.14	−0.14	0.02	0.06	0.79	−0.11	−0.08
SE4. When working shift work, if I really want to I can do regular physical activity	0.01	−0.28	0.04	0.14	0.70	0.08	0.00
CA1. As a shift worker, I am more likely to get enough sleep if my family or friends encourage me	0.09	−0.06	0.04	0.73	−0.08	−0.06	0.05
CA2. As a shift worker, I am more likely to eat a range of nutritious foods if my family or friends encourage me	0.05	0.02	0.06	0.62	−0.11	−0.19	0.01
CA3. As a shift worker, I am more likely to get do regular physical activity if my family or friends encourage me	0.11	0.07	−0.09	0.63	−0.02	−0.01	−0.12
CA4. As a shift worker, I am more likely to try to get enough sleep after I hear/see a story about sleep on TV, Radio or Social Media	−0.20	0.06	0.05	0.67	0.15	0.05	0.04
CA5. As a shift worker, I am more likely to eat a range of nutritious foods after I hear/see a story about healthy eating on TV, Radio or Social Media	−0.15	0.07	0.06	0.68	0.17	0.05	0.01
CA6. As a shift worker, I am more likely to do regular physical activity after I hear/see a story about physical activity on TV, Radio or Social Media	−0.07	0.07	0.01	0.67	0.17	−0.01	0.03
CA7. As a shift worker, I am more likely to look after my health after health experts remind me	0.10	−0.06	−0.07	0.57	0.01	0.19	0.07
HM1. I prioritize my sleep for my health	0.22	−0.04	−0.04	0.00	−0.02	−0.07	0.55
HM2. I eat a range of nutritious food for my health	−0.11	0.10	0.03	−0.03	0.24	−0.02	0.59
HM3. I do regular physical activity for my health	−0.13	0.09	−0.11	−0.05	0.18	−0.02	0.65
HM4. Maintaining good health is extremely important to me	0.25	0.04	0.03	−0.08	0.00	−0.02	0.55
HM5. I look for new information about my health	−0.05	−0.09	−0.01	0.12	−0.22	0.04	0.60
Eigen Values	4.27	4.08	3.63	3.18	3.00	2.01	1.90
Variance Explained	12%	11%	10%	9%	8%	6%	5%

Note: PT: Perceived Threat, PS: Perceived Severity, BR: Barriers, BF: Benefits, SE: Self-efficacy, CA: Cues to Action, HM: Health Motivation. Loading onto factor in denoted by bold font. The authors recommend the scale be presented in a randomized order in limit any order effects.

Reliability

As shown in Table 4, all HBM-SW subscales showed sufficient reliability with Cronbach’s alpha varying between Good – Excellent (0.74–0.93). Measures of test re-test reliability showed both moderate (Perceived Threat, Perceived Severity, Perceived Benefits, and Self-efficacy) and good (Perceived Barriers, Cues to Action, and Health Motivation) test re-test reliability.

Table 4. Measures of reliability outcomes presented by subscale.

Subscale	Cronbach’s alpha	Test re-test^a
Perceived Threat [PT]	.93	0.70
Perceived Severity [PS]	.82	0.64
Perceived Barriers [BR]	.90	0.89
Perceived Benefits [BF]	.92	0.69
Self-efficacy [SE]	.82	0.72
Cues to Action [CA]	.85	0.76
Health motivation [HM]	.74	0.75

*Test re-test measured using longitudinal sample (n = 36).

^bIntraclass correlation coefficients were based on two-way mixed effects model.

The initial item generation produced 75 items, each step resulted in the removal of between 1 and 20 items, with the final scale containing 36 items. A table outlining the removal and modification of items at each step is provided in the supplementary material (S5, supplementary material).

Validity

Measures of validity suggest that the outcomes of each HBM-SW are associated with some behavioral outcome measures, except for Perceived Severity which was not significantly associated with any of the behavioral outcomes. All correlations between HBM-SW subscales and outcome measures are presented in Table 5.

Table 5. Correlations between HBM-SW subscales and outcome measures.

	Sleep		Diet Quality								Physical activity
HBM-SW Subscale	Sleep quality	Sleep duration	Fiber intake	Calcium intake	Wholegrain intake	Sugar intake	Dairy intake	Vegetable intake	Fruit intake	Sugar beverages	Leisure Time METs
Perceived Threat	0.19*	−0.15	0.06	0.06	0.01	−0.22*	0.06	0.02	0.03	−0.19	0.02
Perceived Severity	0.09	−0.10	0.00	0.04	0.07	0.04	−0.04	−0.07	0.02	0.10	−0.09
Barriers	0.46***	−0.44***	−0.17	−0.01	−0.06	0.09	0.05	−0.23*	−0.03	0.15	−0.10
Benefits	0.21**	−0.14	−0.11	−0.13	−0.15	0.06	−0.13	0.04	0.07	0.07	0.06
Self-efficacy	−0.33***	0.26**	0.10	−0.02	0.20	−0.01	−0.04	0.16	−0.08	−0.07	0.15
Cues to Action	−0.17	0.21**	0.20*	0.10	0.18	0.00	0.00	0.02	−0.05	−0.18	0.07
Health Motivation	−0.15	0.16	0.21*	0.03	0.14	−0.33***	−0.01	0.22*	0.10	−0.44***	0.29**

Notes. *p-value <.05, **p-value ≤.01, ***p-value ≤.001.

Significant outcomes are denoted in bold, with a positive value indicating that as score on the HBSM subscale increases, so does the outcome measure. A negative value indicates that as the score on the HBSM subscale increases, the outcome measure decreases.

Perceived threat was significantly associated with poorer sleep and lower sugar intake. Perceived barriers were associated with poorer sleep quality, shorter sleep duration, and lower vegetable intake. Perceived benefits were associated with poorer sleep quality. Self-efficacy was associated with better sleep quality and longer sleep duration. Cues to Action were associated with increased sleep duration and higher fiber intake. Health motivation was associated with increased fiber and vegetable intake, lower sugar and sugar-sweetened beverage intake, and increased physical activity. The significant relationship between HBM-SW subscales and outcomes are presented visually in Figure 1.

Figure 1. Visual representation of significant correlations between HBM-SW psychosocial determinants and health behaviors.

Note. This figure depicts significant correlations only. Left side of zero represents negative relationship: Higher levels of psychosocial determinant are associated with lower levels of health behavior. Right side of zero represents positive relationship: Higher levels of psychosocial determinant are associated with higher levels of health behavior. Beneficial health relationships (i.e., higher levels of positive health behaviors) are denoted with green bars, and poorer health relationships (i.e., higher levels of negative health behaviors) are denoted with red bars. Length of bars indicates strength on relationship (i.e., further from zero equals stronger relationship) *Higher scores of sleep quality indicate poorer sleep quality.

As shown in Table 6, sleep duration and sleep quality significantly different by levels of the calculated sleep subscale (scoring algorithm supplied in supplementary material) (p < .001). Sleep quality was poorer and sleep duration shorter in those with lower levels of sleep subscale. Leisure time METs significantly differed by levels of physical activity subscale (p < .001). There was also a significant difference in sugar-sweetened beverage intake between the levels of HBM-SW scores (p = 0.032). There were no other significant differences between the level of diet subscale and any diet quality.

Table 6. Kruskal–Wallis tests between behavioral specific subscales (by low, average, high) and relevant outcomes of interest.

Scale	Category	Score range	n	Median outcome score	Chi-Square	P-value
Sleep outcomes x Sleep subscale
Sleep quality					14.7	<.001
	Low	4–10	47	10.0
	Average	11–13	62	9.0
	High	14–24	44	6.0
Sleep Duration					23.2	<.001
	Low	4–10	47	6.0
	Average	11–13	62	6.3
	High	14–24	44	7.3
Diet quality outcomes x Diet subscale
Fiber					3.1	.22
	Low	3–16	32	15.6
	Average	17–20	30	16.3
	High	21–28	35	16.9
Calcium					1.2	.56
	Low	3–16	32	916.9
	Average	17–20	30	987.5
	High	21–28	35	950.5
Wholegrain					0.1	.95
	Low	3–16	32	0.7
	Average	17–20	30	0.7
	High	21–28	35	0.7
Sugar					3.3	.20
	Low	3–16	32	15.7
	Average	17–20	30	14.9
	High	21–28	35	14.7
Dairy					0.7	.70
	Low	3–16	32	1.4
	Average	17–20	30	1.5
	High	21–28	35	1.5
Vegetables					1.1	.57
	Low	3–16	33	1.6
	Average	17–20	30	1.6
	High	21–28	35	1.7
Fruit					3.9	.14
	Low	3–16	33	0.9
	Average	17–20	30	0.7
	High	21–28	35	1.0
Sweetened beverages					6.9	.032
	Low	3–16	33	6.5
	Average	17–20	30	5.8
	High	21–28	36	4.7
Physical Activity outcomes x Physical activity subscale
Leisure METs					18.7	<.001
	Low	3–15	35	0
	Average	16–19	27	396.0
	High	20–28	39	1039.5

Note. Significant interactions are denoted by bold font.

Discussion

The present study developed and validated a scale based on the health belief model, to measure psychosocial determinants of health behaviors in shift workers. The developed scale was informed through a multiple stage development process according to existing literature, which allowed for input from content experts (expert panel review) and shift workers with lived experience (cognitive interviews). The final 36-item HBM-SW scale presented in Table 3 loaded on seven factors, in accordance with psychosocial determinants in the health belief model. The final HBM-SW subscales are: Perceived Threat, Perceived Severity, Barriers, Benefits, Cues to Action, Self-efficacy, and Health Motivation.

Each of the subscales demonstrated good to excellent measures of reliability and moderate to good test–retest reliability. While test–retest reliability showed an acceptable level of reliability, it is important to acknowledge that it was conducted with a small sample (n = 36). The moderate, rather than good, level of reliability in two subscales (Perceived Severity and Perceived Benefits) may indicate consistency with previous research that found attitudes related to the impact of some psychosocial determinants may be dynamic, even over short periods of time (Matheson, 2019). Nonetheless, these measures of reliability support the developed HBM-SW scale as a suitable measure for assessing health behaviors in a shift working population. For measures of validity, there were significant associations between six of the subscales and at least one meaningful health outcome measure.

In terms of sleep quality, higher scores of Perceived Barriers were associated with poorer sleep quality. Thus, suggesting that those who perceive barriers associated with shift work are more likely to experience poorer sleep. Interestingly, higher scores of Perceived Threats and Perceived Benefits were also associated with poorer sleep quality (Figure 1), suggesting that those who are experiencing poorer sleep quality may perceive benefits of sleep (Perceived Benefits) and a risk of shift work on health (Perceived Threat) but this does not necessarily translate to better sleep. Thus, there is a need to better understand the interplay between all HBM-SW determinants in a larger sample. In contrast, higher scores of Self-efficacy were associated with better sleep quality, indicating those who felt they had a level of control over health behaviors experienced better sleep quality, which is consistent with previous sleep hygiene research (Murawski et al., 2021).

In terms of sleep duration, higher scores of Self-efficacy and Cues to Action were associated with self-reported longer sleep duration, while higher scores of Perceived Barriers were associated with shorter sleep duration. These findings are consistent with the previous findings that self-efficacy is associated with better sleep behaviors (Murawski et al., 2021). This suggests that strategies which heighten an individual’s self-efficacy and utilization of external cues (Cues to Action) may improve sleep in shift workers. However, individuals who reported barriers to health behaviors (Perceived Barriers) had shorter sleep durations. These barriers may be due to the difficulties of attempting to sleep during daytime (i.e., light, noise, internal wake drive, etc.) (Kecklund & Axelsson, 2016) or a lack of time, which is implicated as a barrier for sleep in shift workers (Nea et al., 2017). This is an important for consideration in the development of strategies, as identifying and addressing modifiable barriers may assist in supporting healthy sleep duration in the context of shift work.

In terms of dietary components, the relationships between the HBM-SW and diet quality components differed by dietary outcome. Higher levels of Perceived Threat and Health Motivation were associated with lower levels of sugar intake and sugar-sweetened beverage intake (Figure 1). Higher levels of Barriers were associated with lower vegetable intake, while higher levels of Health Motivation were associated with increased vegetable intake. The finding that higher Health Motivation was associated with healthier diet quality factors (lower sugar and sugar-sweetened beverage intake and higher vegetable intake) is consistent with existing studies of psychosocial determinants of diet quality in shift workers (Bonnell et al., 2017; Farías et al., 2020; Huggins et al., 2022) Previous research indicates that shift work is associated with insufficient fruit and vegetable intake (Farías et al., 2020; Ronda-Pérez et al., 2020). It is plausible that interventions targeting health motivation may be a useful strategy for improving diet in shift workers. Finally, higher levels of Health Motivation were associated with higher levels of Leisure Time METs, which is consistent with previous findings that higher motivation is associated with higher physical activity levels (Cortis et al., 2017). These findings are important for the shift working population given the concern for lack of physical activity in shift workers (Hulsegge et al., 2017). Irrespective of the health behaviors, health motivation appears to be an important determinant to target in shift workers.

When utilizing the behavior-specific belief subscales of the HBM-SW (sleep, diet, and physical activity), those who scored lower on sleep-specific subscale had significantly poorer sleep quality and shorter sleep durations. Individuals who scored higher on the physical activity-specific subscale had significantly higher weekly Leisure Time METs, compared to those who scored lower. Finally, there was a significant difference in sugar-sweetened beverage consumption between high and lowlevels of the dietary-specific subscale (i.e., individuals who scored higher on the diet subscale had lower sugar-sweetened beverage consumption). The behavioral-specific subscales (see section behavior-specific belief subscales), scored via the developed scoring algorithm based on the original model theory (Rosenstock, 1974), explains some group differences in health behaviors. Importantly, these findings indicate that higher levels of the HBM-SW behavioral-specific scales are associated with more positive health behaviors, providing support for use of this scale in shift working populations.

Importantly, except for sugar-sweetened beverages, none of the dietary components showed differences between levels of HBM-SW dietary-specific subscale. It is possible that this reflects the complex nature of dietary intake (Lowden et al., 2010). The lack of significant correlations between some diet components and HBM-SW is not surprising given the complexity of measuring diet quality (Williams, 2021) and the small sample size within the current study. Furthermore, it is important to recognize that factors influencing shift workers’ diet extend beyond the dietary components explored by the current study, with existing literature also suggesting the importance of meal timing and social aspects of food intake (Costa, 2010; Heath et al., 2019). It would be beneficial to consider the use of another measure of dietary intake, such as ASA24 food diaries (National Cancer Institute, 2016), to reexamine validity of the HBM-SW scale for dietary quality in the future.

The HBM-SW Perceived Severity scale did not show any significant correlations with behavioral outcomes. As the developed scale asks broadly about “chronic disease” rather than a specific disease itself, it is possible that participants had difficulty perceiving any severity. Therefore, future studies may consider whether specifying a particular disorder allows for a better understanding of Perceived Severity in shift workers.

Together, the findings of the present validation study suggest that health behaviors in shift workers are influenced by psychosocial determinants such as Perceived Threat, Barriers, Benefits, Cues to Action, Self-efficacy and Health Motivation. Considering the socio-ecological approach, these influences may exist at multiple levels for example; a micro level (i.e., individuals’ beliefs about health, cues to action, perceived benefits, etc.), a meso level (i.e., social support to provide cues to action, exposure to negative health outcomes which trigger perceived threat, barriers that impact engagement in behaviors), and a macro level (i.e., workplace legislation for break timing and availability rostering, etc.). The use of a model which considers multiple levels of influence is a strength of the present scale. Conversely, we recognize that the present study did not evaluate between worker differences including shift practices, recovery between shifts, etc. Future research may use the developed scale to explore these between worker and shift-type differences.

While not the primary aim of this study, it is important to acknowledge the health behavior findings within the current sample in the context of existing literature. The median sleep quality score according to the PSQI (9.0, IQR = 5.0) was higher than the clinical cutoff for poor sleep (>5), and the average sleep duration within the present sample was reported to be 6.5 hr per night (IQR = 2.0). Thus, the shift workers who responded to our survey were largely experiencing insufficient and poor sleep, which is consistent with existing literature (Kecklund & Axelsson, 2016). Furthermore, when considering the diet quality of the current shift worker sample, many of the average intakes were suboptimal compared to existing guidelines with >11 tsp sugar (15.0 tsp, IQR = 5.4, <1.0 cups of fruit (0.7 cups, IQR = 0.5), and >3 tsp sugar in sugar-sweetened beverages (5.6 tsp, IQR = 3.9). Furthermore, the present study suggests that shift workers may be getting insufficient leisure time physical activity with a median of less than the recommended >1000 METs per week (396.0 METs, IQR = 1980.0) (Ainsworth et al., 2011). Therefore, the current study further highlights the sleep difficulties, suboptimal diet quality, and physical inactivity in shift workers.

Given the suboptimal health behaviors observed in the present sample, in addition to the existing literature on suboptimal health behaviors in shift workers (Crowther et al., 2021), there is a need for further consideration of the interrelationships between shift work, psychosocial determinants (e.g., benefits and barriers), and behavioral outcomes. It is possible that a bi-directional relationship exists between the biological influence of shift work (i.e., circadian misalignment) and the psychosocial determinants. Future studies may consider adopting a path analysis approach to examining the relationship between shift work, psychosocial determinants of behavior and behavioral outcomes, to inform prevention and mitigation strategies for shift workers.

Limitation

We undertook a rigorous, multi-stage development and evaluation process to develop our scale. However, it is important to note that the application of the current scale was limited by the number of participants that completed both the scale and the validity measure. The small sample size meant that both an exploratory and confirmatory factor analysis, as is suggested to be best practice (Boateng et al., 2018), were not possible. Additionally, assessment of divergent and convergent validity was not conducted due to the additional participant burden that would be necessary and thus, should be considered at the time of confirmatory factor analysis (Rönkkö & Cho, 2020). Pleasingly, the present scale demonstrates acceptable psychometric properties, but it would be beneficial to be applied to a second sample to allow for analysis using confirmatory factor analysis and assessment of divergent and convergent validity, to further confirm the properties of the developed scale.

These results should also be considered in light of the limited availability of gold standard measurement tools for concurrent validity. There are no existing gold standard measures for psychosocial determinants of these health behaviors and thus, none could be utilized within the present study. Therefore, concurrent validity for the present study is based on behavioral outcome scales (PSQI, IPAQ, and DSQ) to measure associations between the developed scale and self-report behaviors, rather than measures of psychosocial determinants. Additionally, it is important to note that due to the complex nature of the studied health behaviors, the present study only evaluated certain dimensions of behaviors (sleep quality, sleep duration, certain dietary intake and leisure time physical activity). There may be other relationships that exist between the health belief model and other dimensions of health behaviors such as daytime functioning and sleep efficiency (Buysse, 2014). It is important that future studies of psychosocial determinants in shift workers consider additional dimensions of sleep, dietary intake, and physical activity.

The present study is also limited by the use of self-report behavioral measures for sleep and physical activity. Future validation studies should consider the use of actigraphy and sleep and physical activity diaries to allow for more robust estimates of sleep and physical activity. Furthermore, the current study also did not consider the potential associations between behavioral beliefs and knowledge (e.g., sleep hygiene, sleep beliefs, and dietary knowledge) and health behaviors. Given the prior research which suggests sleep hygiene is associated with sleep behaviors (Murawski et al., 2021), future studies may benefit from considering whether health behavior knowledge may influence aspects of health belief (e.g., benefits, barriers, and perceived susceptibility).

When considering our sample, individuals who started, but did not complete the HBM-SW were significantly less likely to report having a chronic disease, in themselves of family or friends, compared to those who did complete the whole HBM-SW. These differences suggest that those with exposure to a chronic disease may have been more likely to complete the HBM-SW survey, possibly due to the scale seeming more relevant to their situation.

Finally, the current scale was limited in the conclusions that could be drawn about psychosocial determinants of diet quality. While all health behaviors are complex and impacted by many factors, the measurement and assessment of diet quality is particularly challenging in self-report questionnaires. A small sample size in combination with the use of a food frequency questionnaire may have limited our ability to find associations between HBM-SW factors and diet quality. The use of repeated 24-hr food assessments for comparison with the developed HBM-SW may be beneficial in future.

Conclusion

The present scale is the first known scale based on the health belief model to be developed and evaluated for specific use in shift working populations. The rigorous use of an expert panel and cognitive interviews with shift workers enhanced the development of the HBM-SW scale through the use input of individuals who are familiar with the material and have lived experience of shift work. Further, the assessments of reliability and validity showed acceptable psychometric properties of the scale. These findings present a useful scale which is sufficiently acceptable and demonstrates associations between psychosocial determinants and key health behaviors, in particular sleep and physical activity. Further investigation of a larger sample would allow for the use of confirmatory factor analysis and thus, further validation of the usefulness of the developed metric.

Acknowledgments

The authors wish to acknowledge the support of Associate Professor Alex M T Russell (Central Queensland University) in setting up the Qualtrics survey for the current study. The authors also wish to acknowledge the time and expertise provided by the expert panel who assisted in the development of the current scale.

Disclosure statement

ACR reports research funding from MSD and Carers Australia through the Sleep Health Foundation, Flinders Foundation, Compumedics Ltd, Safework SA, Sydney Trains and Vanda Pharmaceuticals, outside of the submitted work.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/15402002.2023.2265520.

Additional information

Funding

The first author is supported by an Australian Government Research Training Program scholarship.