During medical investigation, occupational health physicians collect huge amounts of scattered data about their patients. Recently, these data have been integrated into a single data warehouse. In the current study, this data warehouse has been addressed to study the evolution of BMI, and investigate how this trend is effected by the occupation of the employee, controlled for gender and age.Methods
Around 78 000 employees were followed-up from 1993 onwards, with following variables: BMI, age, sex, and isco encoded occupation. Multilevel analyses was performed to study the evolution of BMI, with time and time2 as time-varying variable, and sex, age at start of measurement and occupation as time-independent variables. Including time2 allows for modelling curvilinear trends over time. Random effects were allowed for the intercept and time-varying variables.Result
Over the 23 years of follow-up, we observed a high drop-out of around 90%. Nontheless, all occupations were still represented at the last measurement. 52.8% was male, mean age was 34.34 years (SD=9.43) at the start of the study. Analyses showed that BMI increases curvilinear with age: the younger, the steeper the curve; Males have higher BMI than females. Occupation also has an effect: Highest BMI was observed with Service personnel of machinery and installations assemblers; the increase of BMI was highest for Leading functions. Random effects showed large inter-individual differences in BMI at starting point and on effect of time.Discussion
We’ve illustrated how the data warehous can be accessed to answer substantive research questions. Differences in evolution of BMI seems to be related to occupation. The strong curvilinear effect probably indicates healthy worker effect. The high dropout might be explained by employees changing companies, companies changing external service, and/or the reach of retirement age.