BRFSS Survey Questions

The BRFSS is an annual, CDC-funded, state-based, random digit-dialed, landline telephone survey of the non-institutionalized adult (≥18 years old) U.S. civilian population. BRFSS collects data on health-related behaviors and health conditions. In addition to the CDC Core questions, states may also select additional CDC modules or develop their own questions to include on their state’s BRFSS. BRFSS core questions collect data on age, gender, race/ethnicity, smoking status, weight, height, and current employment status.

Influenza vaccination status was determined by affirmative response to the question “A flu shot is an influenza vaccine injected into your arm. During the past 12 months, have you had a flu shot?” Or “During the past 12 months, have you had a flu vaccine that was sprayed in your nose? The flu vaccine sprayed in the nose is also called FluMist”.

The 2009 and 2010 Washington State BRFSS cooperation rate was 70.1% and 48.2% respectively, and the response rate was 68.9% and 47.5% respectively [13].

Reference group.

To identify occupations with higher prevalence of ILI, prevalence ratios (PRs) were calculated for all 29 individual occupations with more than 70 respondents compared to a reference group. The reference group was an aggregation of all other employed respondents in occupations with less than 71 respondents. The selection of the individual occupations was blinded to ILI status.

The reference group was significantly different (p<0.001) from the aggregate 29 occupations (see Table 1). The reference group had a lower mean age (42) than that of the aggregate occupations (43). The reference group had a higher percentage of males than the aggregated occupations; more Hispanic respondents and less respondents in the Other Race category than the aggregated occupations. The reference group also had lower rates of vaccination than the aggregated occupations (overall; and in both males and females).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Differences between the reference group and the aggregate 29 selected occupations.

https://doi.org/10.1371/journal.pone.0048806.t001

Analysis

Using the coded occupational responses and the BRFSS ILI questions, descriptive statistics, ILI prevalence by occupational groups [17], and prevalence ratios were estimated. The data were weighted to account for the BRFSS sampling design (weighting by CDC to age-,race-,sex-specific state population estimates and respondent’s probability of selection).

All analyses were performed using STATA software (STATA Version 8.0, Stata Corporation, College Station, TX, USA 2003). The Washington State BRFSS survey questions and protocols were approved by the Washington State Institutional Review Board.

Gender

Previous research suggests a higher percentage of ILI in women (9.0%) compared to men (7.1%) and our data confirmed this observation [12].

The results do reflect a higher prevalence of ILI in women and healthcare workers are nearly 80% women [20]. However, when looking at ILI stratified by sex, neither Registered Nurses nor Nursing Aides, Orderlies and Attendants showed a higher prevalence of ILI. Low prevalence of ILI for Registered Nurses (and to a lesser extent Nursing Aides, Orderlies, and Attendants) may be at least partly explained by higher vaccination rates.

Vaccination

The vaccination rate of the Washington employed population in our data was 20.7% (Table 2). These rates varied by occupation, with Registered Nurses having the highest vaccination rate, 58%, of the 29 selected occupations. There was a significant difference (p<0.001) in the vaccination rate between men (17.5%) and women (25%). Meta-analysis of predictors of seasonal influenza vaccination in health care workers (HCW) described male gender, age of 40 or older, and being a physician, as associated with a slightly increased chance of being vaccinated [21]. Knowing that the vaccine is effective, being willing to prevent influenza trasnmission, and believing that influenza is highly contagious, were among several beliefs on influenza that were identified as strong predictors of HCW being vaccinated [21]. These attitudes and beliefs may vary by education and occupation; there may also be differences in vaccination during pandemic influenza outbreaks as compared to seasonal influenza. Differences in vaccination rate by gender and occupation may account for some of the differences seen in ILI in the Washington State employed population as a whole, as well as between occupational groups in the Health care occupations.

Occupation and Social Contact

Health care occupations (such as Registered Nurses and Nursing Aides, Orderlies, and Attendants) and teaching occupations have potentially high levels of contact with groups such as ill persons or children who may have a higher prevalence of ILI. HCW in particular are often targeted for prevention; and school closures have been identified as a way of social distancing to prevent the transmission of pandemic influenza [22]. However, both groups showed lower prevalence of self-reported ILI when compared to the other occupations analyzed.

In social contact modelling schools are considered a major source of influenza transmission [22], [23], [24], [25]. However, the studies focus primarily on transmission between students, and from students to their household contacts, not necessarily from students to teachers. One possible reason for the lower ILI prevalence found in teachers might be the level of contact with students. In one description of a pandemic flu outbreak in a London school, the only staff members who became ill were those who had cared for ill students – no other staff members became infected – presumably because they did not have such close contact with the infected [25]. In another UK outbreak involving teenage students in a summer program, those staying in university accomodations had the highest attack rate, with “no association found between illness and classes in specific teaching groups or classrooms” [24]. A description of an outbreak in a New York City high school identified self-reported ILI through an online questionnaire in 35% of student respondents but only in 10% of employee respondents [26]. A modelling study of social contacts and infectious disease [23] did identify young people (5–19 years old) as being expected to have a high incidence of a simulated emerging infection – but their contacts were highly assortative (within their own age and grade); overall, the authors concluded that contact “at home, school, or leisure were more likely to be physical than contacts at the workplace or while travelling” and “more-intimate contacts are likely to carry a greater risk of transmission… the most-intimate contacts occur at home or in leisure settings…” [23]. Teachers may indeed be around a population of young people with expected high incidence of ILI or other infections, but not in “close” or “intimate” contact with their students.

Socially isolative professions such as Farming, and Truck Driving had a lower prevalence of ILI than other occupations. Though we cannot estimate what kind of exposure Technicians, n.e.c. have to higher ILI prevalence populations, it could be that because they are involved in working with tools, machines, or other such technical work, that they have less close social contact with other persons at work or work more independently.

The occupations with the highest prevalence of ILI were Secretaries, and Janitors and Cleaners. Secretaries may have higher social contact with an infected person in a work setting. Shared office space has been shown to increase the risk for common colds [27], which is consistent with our observation that Secretaries have an increased prevalence of ILI. Janitors and Cleaners may have increased contact with contaminated surfaces, and may face a higher prevalence of ILI due to the close-contact and hands-on nature of their work, and high exposure to contaminated surfaces and materials (fomite transmission).

Other Factors

Differences in PPE use, education, and training in infection control practices may partly explain the increased prevalence of ILI in Janitors and Cleaners. There may also be underlying demographic, socioeconomic, and employment-related factors contributing to differences in ILI prevalence [7], including job stress [28], [29], [30], [31], [32], job insecurity [33], availability of paid leave/benefits [34], and effort-reward imbalance (ERI) (which causes psychological distress that can negatively affect mental and physical health) [35].

Psychological stress can alter susceptibility to infections [28], [29], [30], is a risk factor for upper respiratory infections (URI) [28], [30], and can impair immune response to vaccinations [29]. One study found that individuals classified as “high stress” experienced signficantly more episodes of URI and more days with URI symptoms than individuals classifed as low stress or intermediate [30]. A study of stress and adherence to preventive measure for influenza in university students found that while higher levels of percieved stress did not affect facemask or hand hygiene compliance, higher levels of stress were significantly associated with a 25% increase in ILI incidence [31]. Percieved stress has also been associated with higher ILI reporting [32]. Job insecurity has also been associated with common infections and health complaints, with a 39% increased risk of self-reported ILI [33].

A study of sickness absence and work factors found sickness absence distributed on an occupational class gradient, with (male) manual workers and clerks having higher relative ratios of abscences for respiratory illnesses (4.21, 3.61 respectively) than engineers (1.14) and managers (reference group, 1.0), and the pattern was similar in women [36]; sickness absences for all causes was found to be attributable in part to work conditions such as ergonomic constraints and work stress in both men and women [36].

A lack of access to paid leave may also negatively affect workers in certain occupations. An evaluation of the impact of workplace policies fond that the absence of policies such as paid sick leave contributed a population-attributable risk of 5 million additional cases of ILI, and in particular, 1.2 million of those among Hispanics (also found in the same study to have higher percieved job insecurity) [34]. Of the WA BRFSS respondents, Janitors and Cleaners had the second highest proportion of Hispanic workers (25.8%, Table 3), behind farm workers. Analysis of BLS Employee Benefits surveys found that sick leave coverage among occupations varied considerably – of those in service occupations, only 37% had access to paid sick leave, and some occupational groups had even less, specifically handlers, equipment cleaners, helpers, and laborer occupations (35%) and machine operators, assemblers and inspectors (29%) [37]. In comparison, 73% of executive, administrative and managerial occupation workers had paid sick leave coverage, professional/technical occupations 71%, and administrative support and clerk occupations, 68% [37]. Access to paid sick leave was also found to be “largely restricted to workers in the top three wage quartiles… within almost every industry…” and this lack of access was noted to be particulary burdensome to women, who make up 60% of minimum-wage workers [37].

Of WA BRFSS respondents, female janitors and cleaners had the highest prevalence of ILI (21.4%, Table 3). A study of Las Vegas hotel room cleaners found that hotel room cleaners (99% female, 84% immigrant) had a high prevalence of effort-reward imbalance (ERI), which was significantly associated with the population’s high prevalence of poor or fair (self-rated) general health [35].

Health Care Workers

Compared with other occupations, certain health care professionals (Registered Nurses, and Nursing Aides, Assistants and Orderlies) did not have a higher prevalence of ILI. These results are consistent with a previous study [8], where HCW in four hospitals did not face an increased risk for H1N1 influenza during the 2009 pandemic when compared to a non-clinical control population. Compliance with use of personal protective equipment in HCW such as PPE, attention to hand hygience, and other infection control precautions were suggested as an explantion for this observation. Lack of compliance with PPE and infection control precautions were also noted to possibly play a role in transmission of pandemic influenza to HCW [38]. Increased attentiveness to infection control precautions may also explain some differences in HCW prevalence of pandemic influenza outbreaks when compared to seasonal influenza.

Other studies have shown an increased prevalence of seasonal influenza for all health care workers (HCW) [9]. In a post-hoc analysis, we compared prevalence of ILI among different health care occupations. There was a difference in ILI prevalence between Health Diagnosing and Treatment occupations (Physicians, Registered Nurses, Health Care Managers, and Dentists) and that of Health Technologists, Health Technicians, Nursing Aides, Orderlies, and Attendants. Workers in the Health Diagnosing and Treatment occupations group had an ILI prevalence of 4.9%, whereas workers in the other HCW occupations group had an ILI prevalence of 8.9% (p<0.001). However, the PR for the comparison between Health Diagnosis and Treatment occupations such as Physicians and Registered Nurses (used as the reference group), and that of Nursing Aides, Orderlies and Attendants was not significant (PR = 1.7; 95% CI = 0.9, 3.2, p<0.08).

Given the differences in ILI prevalence by race and gender seen in our small sample, future work should further explore the potential reasons for these differences. Occupational segregation by race or gender might partially explain the higher prevalence of ILI in non-white and Hispanic populations, although our sample was too small to test this. Future work could also benefit from some measure of the contact rates in social and occupational networks; O*NET [39] provides a ranking of occupations by “Physical Proximity”, which assigns a score according to “what extent does this job require the worker to perform job tasks in close physical proximity to other people.” However, the occupations reported by the WA BRFSS respondents generally had higher scores of contact (when crosswalked to the O*NET titles; data not shown) and the ranking was of limited value.

Limitations

While there are some differences in the distribution of the population between the reference group (those occupations with ≤70 respondents) and the distribution of all of the employed, there was no difference in ILI PR overall between the reference group and our aggregate 29 presented occupations (Table 1).

Multivariate modeling by occupation would be preferable but would have required a larger surveyed population than what we had available. However, the observed results seem biologically plausible in terms of higher or lower exposure (direct social contact or via fomites) resulting in higher or lower risk of ILI by occupation when vaccination rates are considered.

Some possible bias towards the null may have been introduced by assignment of nonspecific narrative responses to larger occupational groups, such as those occupations not elsewhere -classified (n.e.c.).

Another important limitation is that the ILI questions measure non-specific symptoms attributable to many other illnesses like the common cold. Teachers and other school workers have been found to have a higher prevalence of head and chest colds during the school year than all other workers [40]. A study using clinical case definitions of flu-like illness, as fever with cough, deemed it “imperfect” and found a positive predicitive value of 86.8%, sensitivity of 77%, and considered it helpful only when influenza is known to be circulating [41]. Individuals may also percieve and self-report their illnesses differently, and this might also vary by occupation (some HCW for example, who might be more knowledgeable about ILI, may be more or less likely to self-report with ILI). However, because of the widespread nature of the flu, and that most people do not seek treatment, using such a definition of ILI in a telephone survey is an important surveillance tool to better estimate otherwise unrecognized ILI.

Since characterizing the degree of social contact necessary to complete work is difficult, OSHA guidance to prepare workplaces for an influenza pandemic recommends limiting “close contact (within 6 feet)” whenever possible, even for occupations at lower and medium exposure risk [42].

Strengths

We are not aware of any other published study analyzing self-reported ILI prevalence by occupation. This analysis provides valuable insight into the occurrence of ILI in a working population. Additionally, our data are representative of employees in Washington State, with the weighted estimates of the employed Washington population in BRFSS being consistent with estimates of the Washington employed population from other data sources [43].

BRFSS is a national-scale validated survey system that has been used for many years to collect health and behavioral information and set policy. The use of occupational data in this manner is a unique approach to differentiate ILI prevalence within an employed population, and could extend to assisting in pandemic flu planning. If industry and occupation data are collected in more states, or nationally, more occupations could be analyzed and the results would have improved power and the ability to explore differences by race and gender. Even at this limited level, however, one can differentiate ILI prevalence by occupation in the employed Washington population, demonstrating that further research using occupation data to characterize ILI is warranted.

Conclusions

Some occupations appear to have higher prevalence of ILI than others. Some of these differences may be explained, at least in part, by differing levels of social contact and other factors such as stress and access to leave by occupation. Given the serious impact of the flu on the working population, targeting of prevention resources should take occupational exposures into account. Routine collection of industry and occupation information in BRFSS and other surveillance instruments would help in identifying industries and occupations with higher prevalence of ILI or other outcomes of interest and tailoring prevention strategies accordingly.