Job exposure matrices (JEMs) are an important approach for estimating occupational exposures retrospectively. Researchers often rely on expert raters to assess exposure histories retrospectively when exposures cannot be directly measured, and historical exposure records and matrices do not exist. Expert opinion is considered more accurate than self-report for broad job groupings (Kromhout et al., 1987; Teschke et al., 2002). JEMs can be linked to job histories to facilitate assessment of relations between these exposures and risk of cancer and other adverse health outcomes in study populations (Fischer et al., 2017; Peters et al., 2016; Goldberg and Imbernon, 2002). A NIOSH project designed to link state cancer registry data to potential occupational exposures included the development of JEMs for solar and artificial ultraviolet radiation (UVR). The JEMs were developed by a panel of Certified Industrial Hygienists (CIHs) who assigned exposure ratings to standardized industry and occupation (I&O) codes.

Solar and artificial UVR were of interest because of uncertainties about the health effects of exposure. Adverse outcomes that have been reported for UVR exposure are malignancies, including non-melanoma (basal cell carcinoma, squamous cell carcinoma), with estimates of four million incident cases in the 2012 U.S. population (Rogers et al., 2015), and both cutaneous and ocular melanoma, with incidence of the former projected at just over 100,000 new cases in 2020 (SEER, 2020) and the latter accounting for nearly 4% of incident melanoma cases (Jovanovic et al. 2013).These estimates reflect disease resulting from both occupational and non-occupational exposures. Positive effects of UVR exposure include enhanced production of vitamin D, which is necessary for calcium homeostasis (Lukas and Wolf, 2019) and protection against seasonal affective disorder (WHO, 2006). Studies of the effects of UVR on the immune system are less uniform, with evidence of protection against multiple sclerosis, mixed results for different lymphomas, deleterious effects related to skin cancer, and suggestions that the immunologic effects may be dose-dependent (Lu et al., 2017; Liu-Smith et al., 2017; Hart et al. 2011; van der Mei et al., 2003; Cartwright et al., 1994).

While JEMs have been developed for solar and artificial UVR (Freedman et al., 1997; Guenel et al., 2001; Lutz JM, 2005; Peters et al., 2016; Peters et al., 2015; CANJEM, 2019; Vested et al., 2019), most focus on occupation (and do not account for industry), are limited to broad I&O categories, or were developed for case-control studies (Freedman et al., 1997; Guenel et al., 2001; Lutz et al., 2005) and therefore do not cover a wide range of jobs. No published JEM was found that included exposure ratings for solar or artificial UVR across a wide range of I&O combinations or pairs. Pilot testing by the panel determined that prevalence ratings could be assigned for both solar and artificial UVR; therefore, JEMs were constructed for both exposures.

Summary distributions for solar UVR across P and F categories and for artificial UVR across P, F, and I are shown in Table 2.

JEMs that incorporate industry and occupation can be used to identify and protect workers at risk of adverse outcomes due to high UVR exposure. The exposure assignments can also be used with surveillance data and in etiologic studies to investigate relations among UVR exposed workers and health outcomes.

Strengths of this project include the incorporation of both I&O in the JEMs and the wide range of U.S. based I&O combinations incorporated. The data for both JEMs are contained in Excel files and rank ordered from highest to lowest P, F, and I exposure ratings, and then by occupation code. The data can be quickly sorted by other variables including industry code, industry title, and occupation title. We did not combine exposure ratings for P, F and I, so this can be done by individual researchers based on research needs. For example, the solar UVR JEM did not incorporate 1) regional differences, 2) seasonal variation, 3) protective clothing and exposure mitigation practices, 4) shiftwork, or 5) non-occupational solar UVR; keeping the exposure parameters separate facilitates application of modifying factors specific to the study population and research objectives. Moreover, some researchers will be interested in more or less refined categorizations for specific parameters; in the case of the former, these JEMs offer a starting point and, for the latter, categories can be collapsed.

Several caveats should be considered when using the solar and artificial UVR JEMs. It was not uncommon for the raters to initially assign discrepant exposure ratings to the I&O pairs. The raters were selected for overall industrial hygiene experience rather than for knowledge of specific industries or hazards. Solar UVR was rated before artificial UVR. While the raters clearly understood the frequency (and for artificial UVR, intensity) parameters, prevalence for solar UVR was initially problematic, as evidenced by the particularly low inter-rater agreement. Discussions of the reasons for assigning prevalence ratings identified some confusion about whether raters were to identify the prevalence of all solar UVR-exposed workers who were in a specific industry, or to identify the prevalence of all workers employed in the industry who were exposed to solar UVR; the latter was the intended rating target. Following intensive concordance discussions, this confusion abated. This clearer understanding of the scope of this parameter likely explains much of the higher concordance rating for prevalence for artificial UVR. The probability of misclassification of ratings between adjacent categories is substantially higher than for non-adjacent categories. Therefore, the exposure ratings are perhaps most suitable for contrasting industries with high vs. low prevalence, frequency, or intensity of exposure, rather than a fully populated continuum of exposures

In addition, UVR exposures change over time, and these changes should be considered for application of the JEMs to exposure periods other than 2000–2009 (the focus of this project). Intensity of solar UVR changes with calendar time, season, and location, so the intensity parameter was not included in the solar UVR JEM. Growing public awareness of the hazards of solar UVR exposure over time could also affect exposure prevalence and frequency. In contrast, use of artificial UVR is evolving rapidly, with use likely increasing as an antimicrobial/disinfection method but perhaps decreasing in other applications due to increased hazard awareness (OR Today, 2018; Katara et al., 2008); increasingly, UVR is used in some scenarios where robots, not people, are present. Therefore, application of the artificial UVR JEM to exposure periods other than 2000–2009 requires consideration of these changes.

A number of other JEMS have been developed for solar and artificial UVR. The majority of these JEMs incorporated measurement data, an advantage over the reliance on expert-opinion only in the current project. The lack of measurement data for UVR exposures for industry-occupation combinations across large numbers of industries would have precluded consistent application of measurement data to our JEMs, and is the reason the previous JEMs that were comprehensive with respect to job types, such as CAREX, accounted for occupation alone. However, evaluation of the distribution of occupation ratings across industries suggests that for some exposures, including UVR (both solar and artificial) JEMs based solely on occupation, may provide insufficient information for JEMs intended to cover a wide range of industries. This difference in approach precludes direct comparison between these new JEMs and those developed previously. Between-country differences in both jobs and solar UVR exposure also hinder comparisons. Some of the variation in ratings is likely due to misclassification, and some might be ameliorated using a more detailed occupation coding scheme that narrows the range of industries corresponding to each occupation. However, the finding that at least one third, and in the case of artificial UVR prevalence, more than half of occupations, spanned multiple ratings categories across industries suggest the need to carefully consider the question of differences within occupational groupings when constructing JEMs.