The NHIS is a cross-sectional in-person household survey conducted continuously since 1957 by the NCHS, the Centers for Disease Control and Prevention, and is used to monitor the health of the nation. Data are collected on the civilian noninstitutionalized population of the United States and exclude persons in long-term care facilities (eg, nursing homes), correctional facilities, active-duty Armed Forces personnel (although civilian family members are included), and US nationals residing in foreign countries.^8,9 The survey uses a multistage clustered sample design with an over-sampling of Black, Hispanic, and Asian persons. Black, Hispanic, and Asian adults aged 65 years or older are also over-sampled to complete the sample adult module, which, as described later, is one of the four main NHIS modules.

The NHIS questionnaire consists of two sets of questions: (1) a core set of questions that remain relatively unchanged from year to year and (2) supplemental questions that vary from year to year to collect data pertaining to current health issues of national importance. In 2010, the survey instrument had four main modules: household, family, sample child, and sample adult. The first two modules collected health and sociodemographic information on each member of each family residing within a sampled household. Within each family, additional information was collected from one randomly selected adult (the “sample adult”) aged 18 years or older and from the parent or guardian of one randomly selected child (the “sample child”) younger than 18 (if the family had children). In rare instances when a sample adult was physically or mentally unable to respond, proxy responses were accepted (< 1.5% of sample). Interviews were conducted in-person (some telephone follow-up is allowed) using computer-assisted personal interviewing. A total interview lasted, on average, about 1 hour. In 2010, NHIS interviews were conducted in 34,329 households, accounting for 89,976 persons in 35,177 families. The estimates presented in this article are based on data collected from current/recent workers among 27,157 sample adults. The household response rate was 79.5%, the conditional sample adult response rate (ie, the response rate for those sample adults identified as eligible) was 77.3%, and the final sample adult response rate (ie, the response rate that takes into account both the conditional sample adult response rate and the household/family response rate) was 60.8%.

Information regarding current and usual I&O of employment was obtained from core and Occupational Health Supplement questions included in the Sample Adult module. Demographic characteristics were obtained from questions asked in the Household and Family modules. Open-ended responses were obtained from each employed sample adult respondent regarding his or her industry (employer’s type of business) and occupation (employee’s type of work) for their current job and usual job. The current job was defined as their main job held in the week preceding interview, or, if not employed in the week preceding interview, their main job held in the 12 months preceding interview. The usual job was defined as the job that the respondent had held the longest, compared with all other jobs the respondent ever held. Specific questions used to collect this information included the following: “For whom did you work at your main job or business?”/”Thinking about the job you held the longest, for whom did you work? (Name of company, business, organization or employer),” “What kind of business or industry was this? (For example: TV and radio management, retail shoe store, State Department of Labor),” “What kind of work were you doing? (For example: farming, mail clerk, computer specialist.),” and “What were your most important activities on this job or business? (For example: sells cars, keeps account books, operates printing press.).” Respondents were also asked which of the following best described the job in question: “Employee of a private company for wages”; “A federal government employee”; “A state government employee”; “A local government employee”; “Self-employed in own business, professional practice or farm”; or “Working without pay in a family-owned business or farm.”

These responses were reviewed by US Census Bureau coding specialists who assigned four-digit I&O codes. The data were coded using US Census codes on the basis of the 2007 North American Industrial Classification System and 2010 Standard Occupational Classification (SOC) system. To allow for more reliable estimates, NCHS recodes I&O Census codes into simple and detailed I&O groups. In addition, conventional broad US occupational categories were assigned for the current and usual jobs and were partitioned into “white-collar,” “service,” “farming, fishing & forestry,” and “blue-collar” as previously described.¹⁰

Race was stratified into White, Black, and Other. Other included American Indian, Asian, Alaska Native, race group not releasable and multiple race. Ethnicity was stratified into Hispanic and non-Hispanic.

The 2010 NHIS was approved by the Research Ethics Review Board of the NCHS (Protocol #2009-16) and the US Office of Management and Budget (Control #0920-0214). Written consent for participation in the 2010 NHIS was not received, but instead all 2010 NHIS respondents provided oral consent prior to participation.

To account for the complex sampling design of the NHIS, the 2010 data were analyzed using SAS 9.3 survey procedures and SUDAAN 11. The NHIS sample adult record weights provided by NCHS were also used in all analyses. Estimates based on cell sizes of 10 or fewer are not reported.

The kappa statistic (κ) was used to measure agreement between current job and usual job and was presented as a percentage. Kappa is defined as “the agreement beyond chance divided by the amount of possible agreement beyond chance.”¹¹ We calculated concordance across all industry (or occupation) categories by deriving kappa values from n by n tables, where n is the number of categories that current and longest-held industry (or occupation) was divided into (eg, 20 for simple industry groups). We did not weight kappa values to take into account the “closeness” of the various industry (or occupation n) categories. Because there is not a standard way to measure the “closeness” across all I/O categories, we presented unweighted kappa statistics. Nevertheless, we show the effects of the closeness of some of the detailed I/O categories by presenting results for both detailed and simple categories, where the simple categories lump together the “close” detailed categories.

To assess whether certain I/O categories had stronger concordance between current and usual jobs than other I/O categories, we also calculated concordance stratified by I/O categories. We did this by treating concordance between current industry (or occupation) and usual industry (or occupation) with regard to category x as a dichotomous variable. For example, we calculated agreement between whether the respondent’s current job was in the manufacturing industry category versus another industry category, and whether the respondent’s usual job was in the manufacturing industry category.

Finally, we stratified our analyses of overall I/O concordance and of simple I/O category-specific concordance by demographic factors, to assess whether there are demographic subgroups of workers for which the assumption that current job is a good surrogate for usual job is weaker than for other demographic subgroups.

Kappa values are rated on the following scale¹¹:

93 to 100 = excellent agreement

Statistical significance was assessed by the z test at α ≤ 0.05. To correct for multiplicity when conducting multiple comparisons, the alpha value was obtained by dividing 0.05 by the number of comparisons. For example, statistical significance for race comparisons, White versus Black and White versus Other Race, was each tested at α ≤ 0.025.

Although the findings for the simple I&O groups were stratified by sex, race, and ethnicity, it was beyond the scope of this report to provide this stratification for the detailed I&O groups.

Of the 27,157 sample adults who participated in the 2010 NHIS, 16,905 (62%) were included in the analysis of concordance by industry. Of these, 15,095 respondents had a job during the week prior to the survey, and 1810 did not have a job in the prior week but did work in the 12 months preceding interview. Sixteen of these current or recent workers were missing data for current and/or usual occupation, so the analyses by occupation categories are based on 16,889 sample adults. Excluded from all analyses were 10,252 (38%) sample adults. Of these, 7867 had no job in the last week or past 12 months, 1702 had never worked, and 64 had missing data for these items (past week or past 12 months). The remainder (619) had missing current and/or usual industry or were military.

Overall kappa values for simple I&O groups were indicative of good agreement ranging from 74.5 ± 0.5 (industry) to 76.3 ± 0.4 (occupation) (Table 1). Overall kappa values for detailed I&O groups were slightly lower ranging from 72.4 ± 0.5 (industry) to 73.7 ± 0.5 (occupation). On the contrary, the kappa value for broad occupational categories was higher (80.4 ± 0.6).

Overall kappa values for simple I&O groups were higher for men (industry: 75.1 ± 0.6; occupation: 77.0 ± 0.6) than for women (industry: 72.9 ± 0.7; occupation: 74.7 ± 0.6) (Table 2). The lowest kappa values were observed among those respondents with less than 1 year in their current job and ranged from 44.3 ± 2.0 for male industry listings to 48.9 ± 2.0 for male occupational listings (Table 2). Nevertheless, these kappa values were indicative of fair agreement. In contrast, those with 21+ years in their current job had the highest kappa values indicating excellent agreement. Indeed, kappa values increased monotonically for both I&O by length of current employment. With respect to worker age, kappa values were relatively stable staying within the good agreement range for both men and women (Table 2). Nevertheless, kappa values were lowest among workers aged 65 years or older. The kappa values were similar between Whites, Blacks, and Other race and reflect good agreement.

Results for the simple industry groups are shown in descending order by kappa values in Table 3. The simple industry group with the highest kappa value was “Construction Industries” at 82.0 ± 1.1. For men, the highest kappa value was also found for “Construction Industries” at 81.8 ± 1.2, whereas for women it was for “Health Care and Social Assistance Industries” at 78.4 ± 0.9. The highest kappa value for Whites was for “Construction Industries” at 82.1 ± 1.2; for Blacks, “Utilities” at 91.3 ± 4.7; and for Other Race, “Transportation and Warehousing Industries” at 90.5 ± 3.0. For ethnicity, the simple industry group with the highest kappa value was “Construction Industries” for Hispanics at 86.7 ± 1.9 and “Utilities” for non-Hispanics at 80.8 ± 3.4. The industry with the overall lowest kappa value was “Retail Trade Industries” at 67.2 ± 1.3.

Results for the simple occupation groups are shown in descending order by kappa values in Table 4. The simple occupational group with the highest kappa value was “Healthcare Practitioners and Technical Occupations” at 85.8 ± 1.0. For men, the highest kappa value was found for “Legal Occupations” at 88.9 ± 3.3 whereas for women it was for “Architecture and Engineering Occupations” at 87.8 ± 3.1. The highest kappa value for Whites was “Healthcare Practitioners and Technical Occupations” at 85.3 ± 1.2; for Blacks, “Legal Occupations” at 87.7 ± 6.2; and for Other Race, “Healthcare Practitioners and Technical Occupations” at 90.5 ± 2.5. For Hispanics, the simple occupational group with the highest kappa value was “Construction and Extraction Occupations” at 87.8 ± 1.8 and for non-Hispanics it was “Healthcare Practitioners and Technical Occupations” at 85.7 ± 1.1. The occupation with the overall lowest kappa value was “Sales and Related Occupations” at 68.5 ± 1.2.

The agreement between current and usual industry was statistically significantly higher for men versus women in three simple industry groups. Agreement was significantly lower for Whites versus Other Race in two simple industry groups. Agreement was significantly higher for Hispanics versus non-Hispanics in seven simple industry groups (Table 3).

For the simple occupation groups, the agreement between current and usual occupation was statistically significantly higher for men than for women in four occupation groups, and in one group it was higher for women than for men. There were no statistically significant differences for Whites versus Blacks. Nevertheless, Whites versus Other Race had significantly lower agreement in two occupation groups. Hispanics had significantly higher agreement than non-Hispanics in 10 occupation groups.

Results for the detailed groups of industries and occupations are shown in descending kappa values order in Tables 5 and 6, respectively. For industry, “Petroleum and coal products manufacturing” displayed the highest kappa value at 85.8 ± 6.6 and “Textile mills” displayed the lowest kappa value at 44.1 ± 11.4. Of the 73 detailed industry groups with cell sizes more than 10, 65 had a kappa value of 61 or higher, suggesting good agreement between current and usual industry. For occupation, “Air transportation workers” had the highest reportable kappa value at 89.8 ± 4.8 and the lowest reportable kappa value was for “Supervisors, transportation and material moving workers” at 54.7 ± 10.0. Of the 81 detailed occupation groups with cell sizes more than 10, 78 had a kappa value of 61 or higher.

Agreement by broad US occupational classification is shown in Table 7. Overall, agreement did not differ between blue-collar (83.1 ± 0.7) and white-collar (81.4 ± 0.6) workers (P > 0.0167) but agreement was greater for blue-collar workers than for service workers or farming, fishing, and forestry workers (P < 0.0167). In the farming, fishing, and forestry category, Hispanics tended to remain employed in this category whereas non-Hispanics seemed to change jobs (ie, non-Hispanics were more likely to have left employment in farming, fishing, and forestry, and were currently employed in another occupation). The age categorization shows that the majority of the movement out of farming, fishing, and forestry occurred in the 45- to 64-year age range (data not shown). It should be noted that the findings for farming, fishing, and forestry are based on small numbers compared with the other three categories.

Using 2010 data, we found good concordance between I&O of workers’ current jobs and I&O of their usual jobs. This suggests that in most cases where detailed occupational histories are not available, current I&O are reasonable surrogates for usual I&O. As expected, overall concordance was higher for simple I&O categories than for more detailed I&O categories. The overall kappa statistic based on only four broad occupational categories approached the “very good” range.

The assumption that current I&O are reasonable surrogates for usual I&O does seem to be more strongly supported for some subgroups of workers than for others. That kappa values were among the lowest for workers in their most recent job for less than 1 year as compared to the near-perfect concordance for individuals who worked in their current jobs a minimum of 21 years is consistent with previous literature.⁵ That is, both this study and previous studies found that the longer a worker was in a current job, the more likely that current job and usual job were in concordance. These results suggest that higher job tenure will minimize ascertainment bias when using current job as a surrogate for usual job. In contrast, age-related values of concordance were relatively stable staying within the good agreement range but declined to a low numerical value for 65+, which may be reflective of retirement from the usual job followed by acquisition of another job in a different I&O. Levels of agreement do not seem to differ much by worker race, but Hispanic workers seem to have higher concordance than non-Hispanic workers, both overall, and within several simple I&O categories.

Kappa values for the vast majority of simple and detailed I&O groups showed good agreement or better between current job and usual job. Our overall highest concordance estimates among the simple I&O groups (ie, Construction Industries; Healthcare Practitioners; and Technical Occupations) were of very good agreement, which is greater than the highest kappa value (ie, 71.2 ± 0.4, for professional specialty) found in a previous study of 13 broad occupational groups.⁴ Unlike this study, which found no overall kappa for any simple I&O group to indicate slight agreement or worse (ie, kappa value of 40 or less), the Gomez-Marin et al.⁴ study found one simple group (ie, “Handlers, equipment cleaners, helpers, laborers”) with slight agreement.⁴ In this study, the simple I&O groups with the lowest overall kappa values (ie, “Retail Trade Industries” and “Sales and Related Occupations) still showed good agreement. The higher concordance obtained using 2010 data may reflect a propensity for workers to stay within a particular industry/occupation rather than risk unemployment in a tighter job market than was present in the 1980s.

Several previous studies have found increased concordance for higher-paying jobs and decreased concordance for lower-paying jobs.⁴ Although we did not assess income/salary in this study, those I&Os considered to have lower income/salary tended to have lower concordance than those I&Os traditionally considered to have higher income/salary. For example, “Health diagnosing and treating practitioners” and “Lawyers, judges, and related workers” were among the top with a kappa value of 89.3 ± 1.1 and 86.4 ± 3.6, respectively, whereas at the bottom were groups such as “Printing workers” at 59.2 ± 7.3 (Table 6). Further evidence of an association between pay and concordance is the lesser (albeit good) agreement for service workers and farming, fishing, and forestry workers as opposed to white- and blue-collar workers who displayed very good agreement.

The results presented herein are robust in that they are derived from a relatively large data set that is nationally representative. Nevertheless, the results are limited in that they are based on a sample of one year of data collection and collected at a time when the country was recovering from a recession. Moreover, respondents did not provide a complete occupational history, and no effort was made to verify the accuracy of the self-reported usual job. Also, extrapolations to occupations and industries with an estimated population size of fewer than approximately 250,000 would not be reliable because of NHIS sample size limitations.

This study provides the most recent assessment of the concordance between current I&O and usual I&O. Overall and for a vast majority of simple and detailed I&O groups, good or better agreement between current job and usual job was found, suggesting that current job is a reasonable surrogate for usual job in epidemiologic studies. Nevertheless, those with low tenure in their current job have substantially decreased concordance, suggesting the need for greater caution when using such low tenure employment as a surrogate for usual employment. Concordance was also relatively low for workers aged 65 years or older, whereas it was relatively high for Hispanic workers. Among major I&O groups, we found consistently high concordance for industries and occupations related to health care and construction, but the lowest concordance for workers in both industries and occupations related to retail/sales. Given that these data were collected in 2010 when the nation was emerging from a recession, additional data are needed to confirm whether the observed patterns hold when the nation is under different economic conditions.