Accuracy of a Single Item on Mentally Tiring Work as Proxy Measure of Job Demands and Efforts in the Gazel Cohort
To the Editor:
Psychosocial models and scales are widely used in occupational research.1,2 The two most commonly used models are the Demand-Control (DC) and the Effort Reward Imbalance (ERI) models.3,4 In line with a previous study about the accuracy of proxy measures on physically tiring work,5 we examine whether an item about mentally tiring work may approximate the DC and ERI models.
We used data from the Gazel cohort6 from the years 1997, 1998, and 1999, where DC (1997 and 1999) and ERI (1998) scales were administered, as well as a question about mentally tiring work (MTW).7,8 The question about MTW was: “do you find your work mentally tiring” with response options ranging from 0 (not at all) to 8 (very). We divided the scores into two categories (1 to 4 and 5 to 8) and three categories (1 to 3, 4 to 5, and 6 to 8) for examining a dose–response relationship. With regard to the DC model, we defined high demands as a score greater than 23 points (range 9 to 36), low job control as a score less than 72 points (range 24 to 96), and job strain as the simultaneous exposure to high demands and low job control.9 Regarding the ERI model, we defined high effort as a score greater than 12 points, low reward as a score less than 18 points, and ERI as a ratio between effort and rewards greater than 1.4 We calculated the correlation and accuracy (sensitivity, specificity, predictive values, likelihood ratio)10 for MTW in relation to the DC dimensions in 1997/1999, and MTW in relation to ERI in 1998. Temporal stability analyses were also carried out between 1997 and 1999 among those with or without high demands. Sensitivity analyzes included gender stratification and adjustments for age (DC and effort).
For the years considered, 5706 (1998) to 11,304 (1997) workers had responded to the questionnaires. Table 1 summarizes the results. The demands and effort dimensions were fairly well correlated with MTW, with good sensitivity (greater than 0.8), negative likelihood ratio (less than 0.33), and area under curves (greater than 0.7), and a possible dose–response relationship. Specificity was low (0.42 to 0.49) and the positive likelihood ratio was low (less than 2), albeit statistically significant. We found medium temporal stability between 1997 and 1999 with 0.36 of discrepancies between subjects with or without high demands. No statistically significant changes were found when age and sex were taken into account. Job control and ERI were not captured by MTW, and reward only partly.
MTW seems to be a satisfying proxy measure for psychological demands and effort but not for job control, ERI, and only partly for reward, especially for screening purpose. A major strength of the study is the use of a large and diversified occupational cohort, which, however has somewhat limited generalisability as it is based on a single national gas and electricity company. Other factors and dimensions related to general health may have affected the responses. Each year's questionnaire included both the MTW and the validated scales, but the MTW question was at the beginning of the questionnaire. The low specificity and positive likelihood ratios might be due to mentally tiring work including other dimensions than demands, efforts, job control, and rewards, for example violence and harassment at work, bullying, organizational injustice or role conflicts, to name just a few examples.