National Health Interview Survey (NHIS) is a cross-sectional in-person household population health survey conducted annually in the United States since 1957 by the National Center for Health Statistics (CDC). Data are collected on the civilian non-institutionalized population of the United States [Pleis et al., 2010]. The survey uses a multistage clustered sample design, oversampling blacks/African Americans, Hispanics, and Asians. NHIS produces nationally representative data on health related topics such as health care access and utilization, health status, and health behaviors. NHIS survey has been previously described in detail [Luckhaupt et al., 2013].

The 2010 NHIS was approved by the Research Ethics Review Board of the National Center for Health Statistics (Protocol #2009-16) and the U.S. Office of Management and Budget (Control #0920-0214). Oral consent prior to participation was sought from all 2010 NHIS respondents.

The study population was restricted to adult females who worked at any time within 12 months preceding interview, supplied work schedule information, and provided the frequency of and reason for breast, cervical, and/or colorectal cancer screening. These three cancer screening recommendations were selected for analysis because they are the only tests found by USPSTF to decrease cancer mortality. For the sake of analytic consistency across the three cancer screening recommendations, only females were included in the analyses. Females who received testing for diagnostic purposes (not screening purposes) were excluded from analyses.

The study populations were restricted based on age recommendations issued by USPSTF for each of the three cancer screening tests. For the breast cancer screening analyses, the study population was restricted to females, aged 50–74 years. Adherence to breast cancer screening recommendations was defined as having had a mammogram within 2 years preceding interview. For the cervical cancer screening analyses, the study population was restricted to females, aged 21–65 years, with no history of hysterectomy. Adherence to cervical cancer screening recommendations was defined as having had a pap smear within 3 years preceding interview. Although human papillomavirus (HPV) testing was added to USPSTF cervical cancer screening recommendations in 2012, HPV testing was not used to determine adherence to cervical cancer screening because NHIS did not collect this information. For colorectal cancer screening, the study population was restricted to females, aged 50–75 years. Adherence to colorectal cancer screening could be met in one of three ways: (1) colonoscopy within 10 years preceding interview, (2) sigmoidoscopy within 5 years and fecal occult blood test within 3 years preceding interview, or (3) fecal occult blood test within 1 year preceding interview.

These three study populations were not mutually exclusive because they were drawn from the same pool of subjects. For each of the three analyses, females who had never been screened or were not screened within the USPSTF recommended time frame for that particular cancer were classified as being non-adherent to cancer screening recommendations.

Adults who were currently employed or were employed within the last 12 months at the time of the interview were asked to describe the hours they usually worked at their current job, or, if not currently employed, at the job(s) they were employed at within the past 12 months. The question stated “Which of the following best describes the hours you usually work/worked?” Response choices included: (1) regular daytime schedule, (2) regular evening shift (4 pm to midnight), (3) regular night shift (midnight to 8 am), (4) rotating shift (combination of day, evening, or night shift), and (5) some other schedule. For some analyses, the five shift work categories were dichotomized as (1) workers on regular daytime shifts (regular daytime schedule) and (2) workers on alternative shifts (all other shifts/schedules).

Each employed sample adult provided his/her industry (employer’s type of business) and occupation (employee’s type of work) for the main job held in the 12 months preceding interview (i.e., most recently held job). These responses were reviewed by U.S. Census Bureau coding specialists who assigned 4-digit I&O codes based on the 2007 North American Industrial Classification System (NAICS) and 2010 Standard Occupational Classification (SOC) system. To allow for sufficient sample size for more reliable estimates, the less detailed 2-digit I&O recodes were used. The industry recodes consist of 21 simple industry categories, and 23 occupation categories.

Behavioral and personal factors such as obesity, smoking status, alcohol consumption, race, income, education, health insurance coverage, and marital status were considered to be potential risk factors. Each of these factors has been previously linked to working alternative shifts [Gordon et al., 1986; Trinkoff and Storr, 1998; Nabe-Nielsen et al., 2011; Barbadoro et al., 2013]. Each potential risk factor was checked for confounding and interaction by examining the associations of each confounder with both adherence to cancer screening recommendations and alternative shift work. Only variables that meet the definition of a confounder and changed the risk estimate by at least 10% for the majority of the models were included in the final model. Since the study population selected for the cervical cancer screening analysis covered a wide age range (21–65 years), age was considered as a potential confounder and effect modifier for the cervical cancer analyses.

Data analyses was conducted in 2012 using SAS ^® 9.2 callable SUDAAN^® Release 10.0.0 (Research Triangle Institute, Research Triangle Park, NC) to account for population sampling weights. The sampling stratum clusters and weights for all the analyses were provided by NHIS survey documentation. Poisson regression was used to estimate the adjusted prevalence ratios (PR) and 95% confidence intervals (CIs) for all analyses.

The chi-square test was used to assess potential confounding effects of behavioral and personal risk factors on alternative shift work and cancer screening. Variable selection for the final model was based on statistical results, prior knowledge of confounders, and pursuit of a parsimonious model that best fit all three screening tests.

The “loglink” procedure in SUDAAN^® was used to calculate the PR and 95% CIs for each I&O category with sufficient sample size (i.e., no empty cells). For each I&O category, adherence to cancer screening recommendations was assessed for all workers and for workers employed on alternative shifts. In the analyses examining workers employed on alternative shifts, the comparison group consisted of all workers employed on day shifts. In the analyses examining all workers stratified by I&O categories, “Public Administration” was the comparison group for industry and “Business and Financial Operations” was the comparison group for occupation. These were selected as the comparison groups based on the following criteria: the category had more than 100 eligible individuals and had a high adherence rate for each of the three cancer screening recommendations.

Relative standard error (RSE) was calculated based on the prevalence of females who adhered to cancer screening recommendations. Risk estimates with a RSE ≥50% were considered too unstable to report. Statistical significance was defined as a P-value <0.05. Correction for multiple comparisons was not done.

Workers on alternative shifts were 35% (P <0.001) more likely to be non-adherent to breast cancer screening recommendations and 10% (P =0.048) more likely to be non-adherent to colorectal cancer screening recommendations (Table II). There was no difference in adherence for cervical cancer screening (Table II), apart from the age interaction described above.

Findings varied by specific types of alternative shift. For breast cancer screening recommendations, workers on any type of alternative shift had higher non-adherence than those employed on regular daytime shifts. However, these findings were statistically significant for those employed on regular evening shifts and on rotating shifts only. Similarly, for the cervical cancer screening recommendations, those employed on regular evening shifts were significantly more likely to be non-adherent compared to those employed on regular daytime shifts. For the colorectal screening recommendations, those employed on regular night and on rotating shifts were significantly more likely to be non-adherent (Table II).

Among workers on all shifts, the industries associated with significantly reduced adherence to screening recommendations for all three cancers were the “manufacturing” and “retail trade” industry categories, compared to workers employed in “public administration.” The occupations with significantly reduced screening adherence for all three cancers were “food preparation and serving,” “personal care and service,” “sales and related”, and “production” compared to workers employed in “business and financial operations.” Furthermore, significantly lower screening adherence for two of the three cancers was found for employment in five industry categories (“agriculture, forestry, fishing, and hunting,” “wholesale trade,” “transportation and warehousing,” “arts, entertainment, and recreation,” and “accommodation and food services”), and seven occupation categories (“education, training, and library,” “arts, design, entertainment, sports and media,” “healthcare support,” “building and grounds cleaning and maintenance,” “office and administrative support,” “farming, fishing and forestry,” and “transportation and material moving”; Table III).

This study has several limitations. First, this study is based on cross-sectional data, making it impossible to determine the direction of the association between alternative shift work and adherence to cancer screening (e.g., those prone to non-adherence to cancer screening may be more likely to work alternative shifts). Second, small sample size limited analyses by I&O categories and by specific type of alternative shift work. This was especially true for the construction, mining, and management industries. Third, USPSTF general cancer screening recommendations were used to restrict the age ranges of study populations and to identify testing frequencies required to meet the adherence definition. Thus, results are not generalizable to females with a family history of cancer or with an elevated risk of cancer who may require screening at an earlier age or with increased frequency. Fourth, this study is limited by the information collected by NHIS. NHIS did not collect data concerning HPV testing, duration of shiftwork, or lifetime history of work pattern. Similarly, the question on shiftwork was only asked if an individual was employed at any time within the last 12 months. Hence, this study was not able to include those not employed in the 12 months preceding interview (e. g., the unemployed, retired, and homemakers). Finally, cancer screening is self-reported and not independently verified.

One strength of this study is that NHIS is a large population-based survey whose findings are representative of the U.S. working population. Another strength is that NHIS collects data on a wide variety of health and socio-demographic variables which permitted us to assess potentially important confounders.

Among other things, the Affordable Care Act was designed to increase the population’s access to cancer screening [U.S. Departmant of Health and Human Services, 2012]. It will eliminate out-of-pocket expenses for breast, cervical, and colorectal cancer screening, because all new private insurance policies must cover their full cost. Enhanced marketing to inform the public that this screening is effective and involves no out-of-pocket expenses may increase overall adherence to cancer screening recommendations.

Another way to reduce screening barriers is to increase workers’ access to preventive health care opportunities outside of regular daytime working hours (e.g., scheduling mammography screening vans to be present at workplaces during alternative work shifts). Study findings suggest that outreach to the industries of “manufacturing” and “accommodation and food services,” and occupations of “food preparation and serving,” “personal care and service,” and “production,” may be most beneficial because of the low adherence to cancer screening recommendations among their workers on alternative shifts. Additional research is also needed to enhance the fundamental understanding of how working alternative shifts affects adherence to cancer screening recommendations.