Background: To enhance the usability of existing large population-based studies in epidemiologicresearch on office workers, we developed a procedure for a proxy variable on office worker in datawith open-ended responses on occupation.
Methods: Self-reported open answers on occupation (n=3738) from the LifeGene pilot study werelinked to a modified version of the Swedish Standard Classification of Occupation 2012 (SSYK12). TheSSYK12 includes 8946 job titles with 4-digit codes which were categorized to managers, white-collarand blue-collar workers. Managers and white-collar workers were used as a proxy for office workers.We then used fuzzy string matching in R to calculate the Jaro-Winkler distance between the LifeGenepilot data answers on occupation and the modified SSYK12 job titles. Zero distance indicated aperfect match, whereas distances above zero were checked manually to assess various job titles asoffice worker or non-office worker. Thereafter, the resulting procedure was applied to the wholeLifeGene study with data on occupation (n=23 525).
Results: We got perfect match against the modified SSYK12 job titles for 16 275 responses (69%) inthe large LifeGene data. Another 1721 responses (7%) matched occupations that we had manuallydefined as office worker or non-office worker in the pilot data set, and the remaining 5529 (24%)were unmatched. Among the matched occupations, 15 159 (84%) were office-workers, 2493 (14%)non-office workers, and 344 (2%) nondistinctive.
Conclusion: The procedure for a proxy variable on office worker allowed us to classify three quartersof the open-ended responses on occupation.