Original article

Scand J Work Environ Health 2015;41(3):268-279    pdf full text

https://doi.org/10.5271/sjweh.3492 | Published online: 19 Mar 2015, Issue date: 01 May 2015

Developing register-based measures for assessment of working time patterns for epidemiologic studies

by Härmä M, Ropponen A, Hakola T, Koskinen A, Vanttola P, Puttonen S, Sallinen M, Salo P, Oksanen T, Pentti J, Vahtera J, Kivimäki M

Objectives Epidemiological studies suggest that long working hours and shift work may increase the risk of chronic diseases, but the “toxic” elements remain unclear due to crude assessment of working time patterns based on self-reports. In this methodological paper, we present and evaluate objective register-based algorithms for assessment of working time patterns and validate a method to retrieve standard payroll data on working hours from the employer electronic records.

Methods Detailed working hour records from employers’ registers were obtained for 12 391 nurses and physicians, a total 14.5 million separate work shifts from 2008–2013. We examined the quality and validity of the obtained register data and designed 29 algorithms characterizing four potentially health-relevant working time patterns: (i) length of the working hours; (ii) time of the day; (iii) shift intensity; and (iv) social aspects of the working hours.

Results The collection of the company-based register data was feasible and the retrieved data matched with the originally published shift plans. The transferred working time records included <0.01% missing data. Two percent were duplicates that could be easily removed. The 29 variables of working time patterns, generated for each year, were stable across the follow-up (year-to-year correlation coefficients from r=0.7–0.9 for 23 variables), their distributions were as expected, and correlations of the variables within the four main dimensions of working hours were plausible.

Conclusion The developed method and algorithms allow a detailed characterization of four main dimensions of working time patterns potentially relevant for health. We recommend this method for future large-scale epidemiological studies.

See 2016;42(1):97-98 for a correction.