Computer-based coding of free-text job descriptions to efficiently identify occupations in epidemiological studies

Occupational exposure to organic solvents and risk of bladder cancer

BACKGROUND

Bladder cancer has been linked to several occupations that involve the use of solvents, including those used in the dry-cleaning industry.

OBJECTIVES

We evaluated exposure to solvents and risk of bladder cancer in 1182 incident cases and 1408 controls from a population-based study.

METHODS

Exposure to solvents was quantitatively assessed using a job-exposure matrix (CANJEM). Exposure to benzene, toluene and xylene often co-occur. Therefore, we created two additional sets of metrics for combined benzene, toluene and xylene (BTX) exposure: (1) CANJEM-based BTX metrics and (2) hybrid BTX metrics, using an approach that integrates the CANJEM-based BTX metrics together with lifetime occupational histories and exposure-oriented modules that captured within-job, respondent-specific details about tasks and chemicals. Adjusted odds ratios (ORs) and 95% confidence intervals (95% CI) were estimated using logistic regression.

RESULTS

Bladder cancer risks were increased among those ever exposed to benzene (OR = 1.63, 95% CI: 1.14-2.32), toluene (OR = 1.60, 95% CI: 1.06-2.43), and xylene (OR = 1.67, 95% CI: 1.13-2.48) individually. We further observed a statistically significant exposure-response relationship for cumulative BTX exposure, with a stronger association using the hybrid BTX metrics (ORQ1vsUnexposed = 1.26, 95% CI: 0.83-1.90; ORQ2vsUnexposed = 1.52, 95% CI: 1.00-2.31; ORQ3vsUnexposed = 1.88, 95% CI: 1.24-2.85; and ORQ4vsUnexposed = 2.23, 95% CI: 1.35-3.69) (p-trend=0.001) than using CANJEM-based metrics (p-trend=0.02).

IMPACT

There is limited evidence about the role of exposure to specific organic solvents, alone or in combination on the risk of developing bladder cancer. In this study, workers with increasing exposure to benzene, toluene, and xylene as a group (BTX) had a statistically significant exposure-response relationship with bladder cancer. Future evaluation of the carcinogenicity of BTX and other organic solvents, particularly concurrent exposure, on bladder cancer development is needed.

Standard Occupational Classification Codes: Gaps in Federal Data on the Public Health Workforce

Objectives. To determine whether US Department of Labor standard occupational classification (SOC) codes can be used for public health workforce research. Methods. We reviewed past attempts at SOC matching for public health occupations and then used the National Institute for Occupational Safety and Health Industry and Occupation Computerized Coding System (NIOCCS) to match the actual job titles for 26 516 respondents to the 2021 Public Health Workforce Interests and Needs Survey (PH WINS) with SOC codes, grouped by respondents' choice of job category in PH WINS. We assessed the accuracy of the NIOCCS matches and excluded matches under a cutpoint using the Youden Index. We assessed the percentage of SOC matches with insufficient information and diversity of SOC matches per PH WINS category using the Herfindahl-Hirschman Index. Results. Several key public health occupations do not have a SOC code, including disease intervention specialist, public health nurse, policy analyst, program manager, grants or contracts specialist, and peer counselor. Conclusions. Without valid SOC matches and detailed data on local and state government health departments, the US Department of Labor's data cannot be used for public health workforce enumeration. (Am J Public Health. 2024;114(1):48-56. https://doi.org/10.2105/AJPH.2023.307463).

Artificial intelligence exceeds humans in epidemiological job coding

BACKGROUND

Work circumstances can substantially negatively impact health. To explore this, large occupational cohorts of free-text job descriptions are manually coded and linked to exposure. Although several automatic coding tools have been developed, accurate exposure assessment is only feasible with human intervention.

METHODS

We developed OPERAS, a customizable decision support system for epidemiological job coding. Using 812,522 entries, we developed and tested classification models for the Professions et Catégories Socioprofessionnelles (PCS)2003, Nomenclature d'Activités Française (NAF)2008, International Standard Classifications of Occupation (ISCO)-88, and ISCO-68. Each code comes with an estimated correctness measure to identify instances potentially requiring expert review. Here, OPERAS' decision support enables an increase in efficiency and accuracy of the coding process through code suggestions. Using the Formaldehyde, Silica, ALOHA, and DOM job-exposure matrices, we assessed the classification models' exposure assessment accuracy.

RESULTS

We show that, using expert-coded job descriptions as gold standard, OPERAS realized a 0.66-0.84, 0.62-0.81, 0.60-0.79, and 0.57-0.78 inter-coder reliability (in Cohen's Kappa) on the first, second, third, and fourth coding levels, respectively. These exceed the respective inter-coder reliability of expert coders ranging 0.59-0.76, 0.56-0.71, 0.46-0.63, 0.40-0.56 on the same levels, enabling a 75.0-98.4% exposure assessment accuracy and an estimated 19.7-55.7% minimum workload reduction.

CONCLUSIONS

OPERAS secures a high degree of accuracy in occupational classification and exposure assessment of free-text job descriptions, substantially reducing workload. As such, OPERAS significantly outperforms both expert coders and other current coding tools. This enables large-scale, efficient, and effective exposure assessment securing healthy work conditions.

Quest markup for developing FAIR questionnaire modules for epidemiologic studies

BACKGROUND

Online questionnaires are commonly used to collect information from participants in epidemiological studies. This requires building questionnaires using machine-readable formats that can be delivered to study participants using web-based technologies such as progressive web applications. However, the paucity of open-source markup standards with support for complex logic make collaborative development of web-based questionnaire modules difficult. This often prevents interoperability and reusability of questionnaire modules across epidemiological studies.

RESULTS

We developed an open-source markup language for presentation of questionnaire content and logic, Quest, within a real-time renderer that enables the user to test logic (e.g., skip patterns) and view the structure of data collection. We provide the Quest markup language, an in-browser markup rendering tool, questionnaire development tool and an example web application that embeds the renderer, developed for The Connect for Cancer Prevention Study.

CONCLUSION

A markup language can specify both the content and logic of a questionnaire as plain text. Questionnaire markup, such as Quest, can become a standard format for storing questionnaires or sharing questionnaires across the web. Quest is a step towards generation of FAIR data in epidemiological studies by facilitating reusability of questionnaires and data interoperability using open-source tools.

Occupational models from 42 million unstructured job postings

Structuring jobs into occupations is the first step for analysis tasks in many fields of research, including economics and public health, as well as for practical applications like matching job seekers to available jobs. We present a data resource, derived with natural language processing techniques from over 42 million unstructured job postings in the National Labor Exchange, that empirically models the associations between occupation codes (estimated initially by the Standardized Occupation Coding for Computer-assisted Epidemiological Research method), skill keywords, job titles, and full-text job descriptions in the United States during the years 2019 and 2021. We model the probability that a job title is associated with an occupation code and that a job description is associated with skill keywords and occupation codes. Our models are openly available in the sockit python package, which can assign occupation codes to job titles, parse skills from and assign occupation codes to job postings and resumes, and estimate occupational similarity among job postings, resumes, and occupation codes.

Prerequisite for Imputing Non-detects among Airborne Samples in OSHA's IMIS Databank: Prediction of Sample's Volume

INTRODUCTION

The US Integrated Management Information System (IMIS) contains workplace measurements collected by Occupational Safety and Health Administration (OSHA) inspectors. Its use for research is limited by the lack of record of a value for the limit of detection (LOD) associated with non-detected measurements, which should be used to set censoring point in statistical analysis. We aimed to remedy this by developing a predictive model of the volume of air sampled (V) for the non-detected results of airborne measurements, to then estimate the LOD using the instrument detection limit (IDL), as IDL/V.

METHODS

We obtained the Chemical Exposure Health Data from OSHA's central laboratory in Salt Lake City that partially overlaps IMIS and contains information on V. We used classification and regression trees (CART) to develop a predictive model of V for all measurements where the two datasets overlapped. The analysis was restricted to 69 chemical agents with at least 100 non-detected measurements, and calculated sampling air flow rates consistent with workplace measurement practices; undefined types of inspections were excluded, leaving 412,201/413,515 records. CART models were fitted on randomly selected 70% of the data using 10-fold cross-validation and validated on the remaining data. A separate CART model was fitted to styrene data.

RESULTS

Sampled air volume had a right-skewed distribution with a mean of 357 l, a median (M) of 318, and ranged from 0.040 to 1868 l. There were 173,131 measurements described as non-detects (42% of the data). For the non-detects, the V tended to be greater (M = 378 l) than measurements characterized as either 'short-term' (M = 218 l) or 'long-term' (M = 297 l). The CART models were complex and not easy to interpret, but substance, industry, and year were among the top three most important classifiers. They predicted V well overall (Pearson correlation (r) = 0.73, P < 0.0001; Lin's concordance correlation (rc) = 0.69) and among records captured as non-detects in IMIS (r = 0.66, P < 0.0001l; rc = 0.60). For styrene, CART built on measurements for all agents predicted V among 569 non-detects poorly (r = 0.15; rc = 0.04), but styrene-specific CART predicted it well (r = 0.87, P < 0.0001; rc = 0.86).

DISCUSSION

Among the limitations of our work is the fact that samples may have been collected on different workers and processes within each inspection, each with its own V. Furthermore, we lack measurement-level predictors because classifiers were captured at the inspection level. We did not study all substances that may be of interest and did not use the information that substances measured on the same sampling media should have the same V. We must note that CART models tend to over-fit data and their predictions depend on the selected data, as illustrated by contrasting predictions created using all data vs. limited to styrene.

CONCLUSIONS

We developed predictive models of sampled air volume that should enable the calculation of LOD for non-detects in IMIS. Our predictions may guide future work on handling non-detects in IMIS, although it is advisable to develop separate predictive models for each substance, industry, and year of interest, while also considering other factors, such as whether the measurement evaluated long-term or short-term exposure.

Evaluation of the updated SOCcer v2 algorithm for coding free-text job descriptions in three epidemiologic studies

OBJECTIVES

Computer-assisted coding of job descriptions to standardized occupational classification codes facilitates evaluating occupational risk factors in epidemiologic studies by reducing the number of jobs needing expert coding. We evaluated the performance of the 2nd version of SOCcer, a computerized algorithm designed to code free-text job descriptions to US SOC-2010 system based on free-text job titles and work tasks, to evaluate its accuracy.

METHODS

SOCcer v2 was updated by expanding the training data to include jobs from several epidemiologic studies and revising the algorithm to account for nonlinearity and incorporate interactions. We evaluated the agreement between codes assigned by experts and the highest scoring code (a measure of confidence in the algorithm-predicted assignment) from SOCcer v1 and v2 in 14,714 jobs from three epidemiology studies. We also linked exposure estimates for 258 agents in the job-exposure matrix CANJEM to the expert and SOCcer v2-assigned codes and compared those estimates using kappa and intraclass correlation coefficients. Analyses were stratified by SOCcer score, score distance between the top two scoring codes from SOCcer, and features from CANJEM.

RESULTS

SOCcer's v2 agreement at the 6-digit level was 50%, compared to 44% in v1, and was similar for the three studies (38%-45%). Overall agreement for v2 at the 2-, 3-, and 5-digit was 73%, 63%, and 56%, respectively. For v2, median ICCs for the probability and intensity metrics were 0.67 (IQR 0.59-0.74) and 0.56 (IQR 0.50-0.60), respectively. The agreement between the expert and SOCcer assigned codes linearly increased with SOCcer score. The agreement also improved when the top two scoring codes had larger differences in score.

CONCLUSIONS

Overall agreement with SOCcer v2 applied to job descriptions from North American epidemiologic studies was similar to the agreement usually observed between two experts. SOCcer's score predicted agreement with experts and can be used to prioritize jobs for expert review.

Occupational history associates with ALS survival and onset segment

OBJECTIVE

To identify associations between occupational settings and self-reported occupational exposures on amyotrophic lateral sclerosis (ALS) survival and phenotypes.

METHODS

All patients seen in the University of Michigan Pranger ALS Clinic were invited to complete an exposure assessment querying past occupations and exposures. Standard occupational classification (SOC) codes for each job and the severity of various exposure types were derived. Cox proportional hazards models associated all-cause mortality with occupational settings and the self-reported exposures after adjusting for sex, diagnosis age, revised El Escorial criteria, onset segment, revised ALS Functional Rating Scale (ALSFRS-R), and time from symptom onset to diagnosis. Multinomial logistic regression models with three categories, adjusted for age, assessed the association between occupational settings and exposures to onset segment.

RESULTS

Among the 378 ALS participants (median age, 64.7 years; 54.4% male), poorer survival was associated with work in SOC code "Production Occupations" and marginally with work in "Military Occupation"; poor survival associated with self-reported occupational pesticide exposure in adjusted models. Among onset segments: cervical onset was associated with ALS participants having ever worked in "Buildings and Grounds Cleaning and Maintenance Occupations," "Construction and Extraction Occupations," and "Production Occupations"; bulbar onset with self-reported occupational exposure to radiation; and cervical onset with exposure to particulate matter, volatile organic compounds, metals, combustion and diesel exhaust, electromagnetic radiation, and radiation.

CONCLUSION

Occupational settings and self-reported exposures influence ALS survival and onset segment. Further studies are needed to explore and understand these relationships, most advantageously using prospective cohorts and detailed ALS registries.

Automated Coding of Job Descriptions From a General Population Study: Overview of Existing Tools, Their Application and Comparison

OBJECTIVES

Automatic job coding tools were developed to reduce the laborious task of manually assigning job codes based on free-text job descriptions in census and survey data sources, including large occupational health studies. The objective of this study is to provide a case study of comparative performance of job coding and JEM (Job-Exposure Matrix)-assigned exposures agreement using existing coding tools.

METHODS

We compared three automatic job coding tools [AUTONOC, CASCOT (Computer-Assisted Structured Coding Tool), and LabourR], which were selected based on availability, coding of English free-text into coding systems closely related to the 1988 version of the International Standard Classification of Occupations (ISCO-88), and capability to perform batch coding. We used manually coded job histories from the AsiaLymph case-control study that were translated into English prior to auto-coding to assess their performance. We applied two general population JEMs to assess agreement at exposure level. Percent agreement and PABAK (Prevalence-Adjusted Bias-Adjusted Kappa) were used to compare the agreement of results from manual coders and automatic coding tools.

RESULTS

The coding per cent agreement among the three tools ranged from 17.7 to 26.0% for exact matches at the most detailed 4-digit ISCO-88 level. The agreement was better at a more general level of job coding (e.g. 43.8-58.1% in 1-digit ISCO-88), and in exposure assignments (median values of PABAK coefficient ranging 0.69-0.78 across 12 JEM-assigned exposures). Based on our testing data, CASCOT was found to outperform others in terms of better agreement in both job coding (26% 4-digit agreement) and exposure assignment (median kappa 0.61).

CONCLUSIONS

In this study, we observed that agreement on job coding was generally low for the three tools but noted a higher degree of agreement in assigned exposures. The results indicate the need for study-specific evaluations prior to their automatic use in general population studies, as well as improvements in the evaluated automatic coding tools.

Between the farm and the fork: job quality in sustainable food systems

Advocates for structural change in the food system see opportunity in alternative food systems (AFS) to bolster sustainability and equity. Indeed, any alternative to industrial labor practices is assumed to be better. However, little is known about what types of jobs are building AFS or job quality. Failing to understand job quality in AFS risks building a sustainable but exploitative industry. Using a unique and large data set on job openings in AFS, this paper narrows this gap by providing an assessment of labor demand and job quality for AFS in the United States between 2010 and 2019. Job advertisements are matched to 2018 Standard Occupation Codes to characterize work. Wages are compared to living wage standards and median incomes by occupation and local labor market. Considering living wage tests and local labor market competitiveness together, the potential for high job quality in AFS is mixed. Optimistically, higher prices in occupation that are close to consumers and experiencing significant labor demand, like food service and sales, saw more competitive wages. However, these roles frequently failed to offer living wages. Farm work occupations underperformed compared to local labor markets. In addition, uncompetitive senior-level jobs may indicate low-quality career pathways for leadership roles charting paths forward in AFS. These results suggest more institutional action are necessary to enhance labor quality within these spaces and more broadly across the food system. These results also raise questions about who is able to participate in AFS development and whether barriers to participate may replicate equity blind spots.

Associations of self-reported occupational exposures and settings to ALS: a case-control study

BACKGROUND

Environmental exposures contribute to the pathogenesis of amyotrophic lateral sclerosis (ALS), a fatal and progressive neurological disease. Identification of these exposures is important for targeted screening and risk factor modification.

OBJECTIVE

To identify occupational exposures that are associated with a higher risk of ALS using both survey and standard occupational classification (SOC) coding procedures, and to highlight how exposure surveys can complement SOC coding.

METHODS

ALS participants and neurologically healthy controls recruited in Michigan completed a detailed exposure assessment on their four most recent and longest held occupations. Exposure scores were generated from the exposure survey, and occupations were assigned to SOC codes by experienced exposure scientists.

RESULTS

This study included 381 ALS and 272 control participants. ALS participants reported higher duration-adjusted occupational exposure to particulate matter (OR = 1.45, 95% CI 1.19-1.78, p < 0.001), volatile organic compounds (OR = 1.22, 95% CI 1.02-1.45, p = 0.029), metals (OR = 1.48, 95% CI 1.21-1.82, p < 0.001), and combustion and diesel exhaust pollutants (OR = 1.20, 95% CI 1.01-1.43, p = 0.041) prior to ALS diagnosis, when adjusted for sex, age, and military service compared to controls. In multivariable models, only occupational exposure to metals remained significant risk (OR = 1.56, 95% CI 1.11-2.20, p = 0.011), although in an adaptive elastic net model, particulate matter (OR = 1.203), pesticides (OR = 1.015), and metals (1.334) were all selected as risk factors. Work in SOC code "Production Occupations" was associated with a higher ALS risk. SOC codes "Building and Grounds Cleaning and Maintenance Occupations", "Construction and Extraction Occupations", "Installation, Maintenance, and Repair Occupations", and "Production Occupations" were all associated with a higher exposure to metals as determined using survey data.

DISCUSSION

Occupational exposure to particulate matter, volatile organic compounds, metals, pesticides, and combustion and diesel exhaust and employment in "Production Occupations" was associated with an increased ALS risk in this cohort.

A Narrative Literature Review of Natural Language Processing Applied to the Occupational Exposome

The evolution of the Exposome concept revolutionised the research in exposure assessment and epidemiology by introducing the need for a more holistic approach on the exploration of the relationship between the environment and disease. At the same time, further and more dramatic changes have also occurred on the working environment, adding to the already existing dynamic nature of it. Natural Language Processing (NLP) refers to a collection of methods for identifying, reading, extracting and untimely transforming large collections of language. In this work, we aim to give an overview of how NLP has successfully been applied thus far in Exposome research. Methods: We conduct a literature search on PubMed, Scopus and Web of Science for scientific articles published between 2011 and 2021. We use both quantitative and qualitative methods to screen papers and provide insights into the inclusion and exclusion criteria. We outline our approach for article selection and provide an overview of our findings. This is followed by a more detailed insight into selected articles. Results: Overall, 6420 articles were screened for the suitability of this review, where we review 37 articles in depth. Finally, we discuss future avenues of research and outline challenges in existing work. Conclusions: Our results show that (i) there has been an increase in articles published that focus on applying NLP to exposure and epidemiology research, (ii) most work uses existing NLP tools and (iii) traditional machine learning is the most popular approach.

Beware the Grizzlyman: A comparison of job- and industry-based noise exposure estimates using manual coding and the NIOSH NIOCCS machine learning algorithm

Recently, the National Institute for Occupational Safety and Health (NIOSH) released an updated version of the NIOSH Industry and Occupation Computerized Coding System (NIOCCS), which uses supervised machine learning to assign industry and occupational codes based on provided free-text information. However, no efforts have been made to externally verify the quality of assigned industry and job titles when the algorithm is provided with inputs of varying quality. This study sought to evaluate whether the NIOCCS algorithm was sufficiently robust with low-quality inputs and how variable quality could impact subsequent job estimated exposures in a large job-exposure matrix for noise (NoiseJEM). Using free-text industry and job descriptions from >700,000 noise measurements in the NoiseJEM, three files were created and input into NIOCCS: (1) N1, "raw" industries and job titles; (2) N2, "refined" industries and "raw" job titles; and (3) N3, "refined" industries and job titles. Standardized industry and occupation codes were output by NIOCCS. Descriptive statistics of performance metrics (e.g., misclassification/discordance of occupation codes) were evaluated for each input relative to the original NoiseJEM dataset (N0). Across major Standardized Occupational Classifications (SOC), total discordance rates for N1, N2, and N3 compared to N0 were 53.6%, 42.3%, and 5.0%, respectively. The impact of discordance on the major SOC group varied and included both over- and under-estimates of average noise exposure compared to N0. N2 had the most accurate noise exposure estimates (i.e., smallest bias) across major SOC groups compared to N1 and N3. Further refinement of job titles in N3 showed little improvement. Some variation in classification efficacy was seen over time, particularly prior to 1985. Machine learning algorithms can systematically and consistently classify data but are highly dependent on the quality and amount of input data. The greatest benefit for an end-user may come from cleaning industry information before applying this method for job classification. Our results highlight the need for standardized classification methods that remain constant over time.

Determining occupation for National Violent Death Reporting System records: An evaluation of autocoding programs

BACKGROUND

Suicide is a leading cause of death for working-age adults. Suicide risk varies across occupations. The National Violent Death Reporting System (NVDRS) collects information about violent deaths occurring in the United States. Occupation can be determined using autocoding programs with NVDRS data. The objective of this analysis is to determine the accuracy of autocoding programs for assigning occupations in the NVDRS.

METHODS

Deaths from suicide were identified in NVDRS for individuals age 16 and older from 2010 to 2017. Occupations were assigned after processing job description free text with autocoding programs. Job assigned by autocoding program were compared with the occupation code recorded on the death certificate.

RESULTS

Assignment of major occupation group had substantial agreement (Cohen's kappa > 0.7) for the two autocoding programs evaluated. Agreement of assigned code varied across race/ethnicity and occupation type.

CONCLUSIONS

Autocoding programs provide an efficient method for identifying the occupation for decedents in NVDRS data. By identifying occupation, circumstances of suicide and rates of suicide can be studied across occupations.

Procode: A Machine-Learning Tool to Support (Re-)coding of Free-Texts of Occupations and Industries

Procode is a free of charge web-tool that allows automatic coding of occupational data (free-texts) by implementing Complement Naïve Bayes (CNB) as a machine-learning technique. The paper describes the algorithm, performance evaluation, and future goals regarding the tool's development. Almost 30 000 free-texts with manually assigned classification codes of French classification of occupations (PCS) and French classification of activities (NAF) were used to train CNB. A 5-fold cross-validation found that Procode predicts correct classification codes in 57-81 and 63-83% cases for PCS and NAF, respectively. Procode also integrates recoding between two classifications. In the first version of Procode, this operation, however, is only a simple search function of recoding links in existing crosswalks. Future focus of the project will be collection of the data to support automatic coding to other classification and to establish a more advanced method for recoding.

Labour market competition for public health graduates in the United States: A comparison of workforce taxonomies with job postings before and during the COVID-19 pandemic

A strong public health workforce (PHW) is needed to respond to COVID‐19 and public health (PH) issues worldwide. However, classifying, enumerating, and planning the PHW is challenging. Existing PHW taxonomies and enumerations focus on the existing workforce, and largely ignore workforce competition for public health graduates (PHGs). Such efforts also do not utilize real time data to assess rapid changes to the employment landscape, like those caused by COVID‐19. A job postings analysis can inform workforce planning and educational program design alike. To identify occupations and industries currently seeking PHGs and contrast them with existing taxonomies, authors matched existing PHW taxonomies to standardized occupational classification codes, then compared this with 38,533 coded, US job postings from employers seeking Master's level PHGs from 1 July 2019 to 30 June 2020. Authors also analysed 24,516 postings from March 2019 to October 2019 and compared them with 24,845 postings from March 2020 to October 2020 to assess changing employer demands associated with COVID‐19. We also performed schema matching to align various occupational classification systems. Job postings pre‐COVID and during COVID show considerable but changing demand for PHGs in the US, with 16%–28% of postings outside existing PHW taxonomies, suggesting labour market competition which may compound PHW recruitment and retention challenges.

Usefulness of a job-exposure matrix 'MADE' as a decision tool for compensation of work-related musculoskeletal disorders

We aimed to assess the predictivity of the biomechanical job-exposure matrix 'MADE' using compensation data from the National Health Insurance for work-related disorders. Data were obtained from 2013 to 2015, area under curves (AUC), sensitivity, specificity and predictive values were calculated using compensation results as reference. We collected 163 128 cases data. AUC ranged from 0.64 for shoulders disorder to 0.82 for knee disorders. If two thresholds were considered, 28.7% of the sample fit under or over those. The matrix showed a fair predictivity. Such matrix cannot replace expertise but might be a tool used for improving compensation process.

Occupation Coding of Job Titles: Iterative Development of an Automated Coding Algorithm for the Canadian National Occupation Classification (ACA-NOC)

Background

In many research studies, the identification of social determinants is an important activity, in particular, information about occupations is frequently added to existing patient data. Such information is usually solicited during interviews with open-ended questions such as “What is your job?” and “What industry sector do you work in?” Before being able to use this information for further analysis, the responses need to be categorized using a coding system, such as the Canadian National Occupational Classification (NOC). Manual coding is the usual method, which is a time-consuming and error-prone activity, suitable for automation.

Objective

This study aims to facilitate automated coding by introducing a rigorous algorithm that will be able to identify the NOC (2016) codes using only job title and industry information as input. Using manually coded data sets, we sought to benchmark and iteratively improve the performance of the algorithm.

Methods

We developed the ACA-NOC algorithm based on the NOC (2016), which allowed users to match NOC codes with job and industry titles. We employed several different search strategies in the ACA-NOC algorithm to find the best match, including exact search, minor exact search, like search, near (same order) search, near (different order) search, any search, and weak match search. In addition, a filtering step based on the hierarchical structure of the NOC data was applied to the algorithm to select the best matching codes.

Results

The ACA-NOC was applied to over 500 manually coded job and industry titles. The accuracy rate at the four-digit NOC code level was 58.7% (332/566) and improved when broader job categories were considered (65.0% at the three-digit NOC code level, 72.3% at the two-digit NOC code level, and 81.6% at the one-digit NOC code level).

Conclusions

The ACA-NOC is a rigorous algorithm for automatically coding the Canadian NOC system and has been evaluated using real-world data. It allows researchers to code moderate-sized data sets with occupation in a timely and cost-efficient manner such that further analytics are possible. Initial assessments indicate that it has state-of-the-art performance and is readily extensible upon further benchmarking on larger data sets.

The Future of Careers at the Intersection of Climate Change and Public Health: What Can Job Postings and an Employer Survey Tell Us?

Climate change is acknowledged to be a major risk to public health. Skills and competencies related to climate change are becoming a part of the curriculum at schools of public health and are now a competency required by schools in Europe and Australia. However, it is unclear whether graduates of public health programs focusing on climate change are in demand in the current job market. The authors analyzed current job postings, 16 years worth of job postings on a public health job board, and survey responses from prospective employers. The current job market appears small but there is evidence from job postings that it may be growing, and 91.7% of survey respondents believe the need for public health professionals with training in climate change may grow in the next 5–10 years. Current employers value skills/competencies such as the knowledge of climate mitigation/adaptation, climate-health justice, direct/indirect and downstream effects of climate on health, health impact assessment, risk assessment, pollution-health consequences and causes, Geographic Information System (GIS) mapping, communication/writing, finance/economics, policy analysis, systems thinking, and interdisciplinary understanding. Ensuring that competencies align with current and future needs is a key aspect of curriculum development. At the same time, we recognize that while we attempt to predict future workforce needs with historical data or surveys, the disruptive reality created by climate change cannot be modeled from prior trends, and we must therefore adopt new paradigms of education for the emerging future.

Development of a Coding and Crosswalk Tool for Occupations and Industries

Introduction

Job coding into a standard occupation or industry classification is commonly performed in occupational epidemiology and occupational health. Sometimes, it is necessary to code jobs into multiple classifications or to convert job codes from one classification to another. We developed a generic tool, called CAPS-Canada (http://www.caps-canada.ca/), that combines a computer-assisted coding tool covering seven International, Canadian and US occupation and industry classifications and an assistant facilitating crosswalks from one classification to another. The objectives of this paper are to present the different functions of the CAPS-Canada tool and to assess their contribution through an inter-rater reliability study.

Method

The crosswalk assistant was built based on a database of >30,000 jobs coded during a previous project. We evaluated to what extent it would allow automatic translation between pairs of classifications. The influence of CAPS-Canada on agreement between coders was assessed through an inter-rater reliability study comparing three approaches: manual coding, coding with CAPS-Canada without the crosswalk assistant, and coding with the complete tool. The material for this trial consisted of a random sample of 1000 jobs extracted from a case-control study and divided into three subgroups of equivalent size.

Results

Across the classification systems, the crosswalk assistant would provide useful information for 83-99% of jobs (median 95%) in a population similar to ours. Eighteen to eighty-one percent of jobs (median 56%) could be entirely automatically recoded. Based on our sample of 1000 jobs, inter-rater reliability in occupation coding ranged from 35.7 to 66.5% (median 53.7%) depending on the combination of classification/resolution. Compared with manual coding, the use of CAPS-Canada substantially improved inter-rater reliability.

Conclusion

CAPS-Canada is an attractive alternative to manual coding and is particularly relevant for coding a job into multiple classifications or for recoding jobs into other classifications.

Harmonizing work history data in epidemiologic studies with overlapping employment records

BACKGROUND

Work history data often require major data management including handling of overlapping jobs to avoid overestimating exposure before linkage to job-exposure matrices (JEMs) is possible.

METHODS

In a case-cohort study of 1825 male Norwegian offshore petroleum workers, 3979 jobs were reported (mean duration 2417 days/job; maximum 8 jobs/worker). Each job was assigned to one of 27 occupation categories. Overlapping jobs of the same category (1142 jobs) were collapsed and overlapping jobs of different categories (1013 jobs) were split. The resulting durations were weighted by a factor accounting for the number of overlapping jobs.

RESULTS

Collapsing overlapping jobs within the same category resulted in 3295 jobs (mean 2629 days/job). Splitting overlapping jobs of different categories increased the number to 4239 jobs (mean 2043 days/job), while the total duration in days dropped by 10%.

CONCLUSIONS

We demonstrated that overlapping employment data structures can be harmonized in a systematic and unbiased way, preparing work history data for linkage to several JEMs.

Efficiency of autocoding programs for converting job descriptors into standard occupational classification (SOC) codes

BACKGROUND

Existing datasets often lack job exposure data. Standard Occupational Classification (SOC) codes can link work exposure data to health outcomes via a Job Exposure Matrix, but manually assigning SOC codes is laborious. We explored the utility of two SOC autocoding programs.

METHODS

We entered industry and occupation descriptions from two existing cohorts into two publicly available SOC autocoding programs. SOC codes were also assigned manually by experienced coders. These SOC codes were then linked to exposures from the Occupational Information Network (O*NET).

RESULTS

Agreement between the SOC codes produced by autocoding programs and those produced manually was modest at the 6-digit level, and strong at the 2-digit level. Importantly, O*NET exposure values based on SOC code assignment showed strong agreement between manual and autocoded methods.

CONCLUSION

Both available autocoding programs can be useful tools for assigning SOC codes, allowing linkage of occupational exposures to data containing free-text occupation descriptors.

Capture and coding of industry and occupation measures: Findings from eight National Program of Cancer Registries states

BACKGROUND

Although data on industry and occupation (I&O) are important for understanding cancer risks, obtaining standardized data is challenging. This study describes the capture of specific I&O text and the ability of a web-based tool to translate text into standardized codes.

METHODS

Data on 62 525 cancers cases received from eight National Program of Cancer Registries (NPCR) states were submitted to a web-based coding tool developed by the National Institute for Occupational Safety and Health for translation into standardized I&O codes. We determined the percentage of sufficiently analyzable codes generated by the tool.

RESULTS

Using the web-based coding tool on data obtained from chart abstraction, the NPCR cancer registries achieved between 48% and 75% autocoding, but only 12-57% sufficiently analyzable codes.

CONCLUSIONS

The ability to explore associations between work-related exposures and cancer is limited by current capture and coding of I&O data. Increased training of providers and registrars, as well as software enhancements, will improve the utility of I&O data.

Smoking status, usual adult occupation, and risk of recurrent urothelial bladder carcinoma: data from The Cancer Genome Atlas (TCGA) Project

PURPOSE

Tobacco smoking and occupational exposures are the leading risk factors for developing urothelial bladder carcinoma (UBC), yet little is known about the contribution of these two factors to risk of UBC recurrence. We evaluated whether smoking status and usual adult occupation are associated with time to UBC recurrence for 406 patients with muscle-invasive bladder cancer submitted to The Cancer Genome Atlas (TCGA) project.

METHODS

Kaplan-Meier and Cox proportional hazard methods were used to assess the association between smoking status, employment in a high-risk occupation for bladder cancer, occupational diesel exhaust exposure, and 2010 Standard Occupational Classification group and time to UBC recurrence.

RESULTS

Data on time to recurrence were available for 358 patients over a median follow-up time of 15 months. Of these, 133 (37.2%) experienced a recurrence. Current smokers who smoked for more than 40 pack-years had an increased risk of recurrence compared to never smokers (HR 2.1, 95% CI 1.1, 4.1). Additionally, employment in a high-risk occupation was associated with a shorter time to recurrence (log-rank p = 0.005). We found an increased risk of recurrence for those employed in occupations with probable diesel exhaust exposure (HR 1.8, 95% CI 1.1, 3.0) and for those employed in production occupations (HR 2.0, 95% CI 1.1, 3.6).

CONCLUSIONS

These findings suggest smoking status impacts risk of UBC recurrence, although several previous studies provided equivocal evidence regarding this association. In addition to the known causal relationship between occupational exposure and bladder cancer risk, our study suggests that occupation may also be related to increased risk of recurrence.