Computational protocol: Allostatic Load and Effort-Reward Imbalance: Associations over the Working-Career

Similar protocols

Protocol publication

[…] The main analytical sample in this study (see ) was derived from the ELSA wave 6 nurse visit sample (n = 7699), out of which 5617 respondents had their allostatic load index score measured at wave 6. Among this group 2826 respondents were an employee at least once between waves 2 and 6 and had a measurement of ERI when they were an employee. The main analytical sample further reduced to 2663 when missing covariates were deleted from the sample. From this main analytical sample, a subsample of 1020 respondents had their allostatic load measured at wave 2.As the allostatic load index and the component systems are a count of biomarker risk indicators, the appropriate regression models to model count data include negative binomial regression and Poisson regression models. The allostatic load index-dependent variable was over-dispersed (the variance of allostatic load was greater than the mean), so negative binomial regression models were used to estimate the association between effort-reward imbalance and allostatic load after controlling for covariates. Poisson regression models were used to model the association between the effort-reward imbalance and the neuroendocrine, immune, cardiovascular, and inflammatory systems. A logit model was used for the anthropometric system as the count of the two risk factors (waist-height ratio risk quartile and underweight) reduced to a binary variable in the main analytical sample.Wave 6 cross-sectional nurse visit survey weights (derived by the ELSA study team) were used to examine the association of cumulative ERI with AL in all the regression models []. The wave 6 blood sample survey weights were not used as some of the wave 6 respondents provided a hair sample (for the cortisol and cortisone analytes) but did not provide a blood sample. The longitudinal weights were not appropriate as these have been derived only for core ELSA members from wave 1, and their use would have deleted refreshment sample members from the analysis. All statistics were calculated using the “survey (svy)” command in Stata version 14 (StataCorp., College Station, TX, USA) [], which takes account of sample selection, non-response bias and the complex survey design for point estimates and variance estimation. […]

Pipeline specifications

Software tools ERI, Stata
Application Protein sequence analysis