Predictors of change in self-rated health in a 'Healthy worker' group: A longitudinal study

Aims: Self-rated health (SRH) has attracted research attention both as a predictor of mortality and health impairment, and as a variable which is itself influenced by social/biomedical factors and health behaviours; the present study sought to extend existing findings from community and clinical samples by assessing predictors of five-year change in SRH in a medically-fit 'healthy worker' group. Method: The data were obtained in 1995 and 2000 from 283 oil industry personnel working offshore during this period. A five-point single-item SRH measure was used. Job activity levels (sedentary; active; strenuous) were based on consensus ratings; other information was self-reported. Demographic factors, job activity level, smoking, body mass index (BMI), and negative affectivity were used to predict SRH change. Hypothesized linear and curvilinear/interactive relationships were tested. Results: Overall, SRH decreased (p < 0.005) and BMI increased (p < 0.001) between 1995 and 2000. In regression analyses, SRH change was treated as the dependent variable, with control for baseline SRH. In the main effects model, the significant predictors were job activity level (p < 0.05), change in BMI (controlled for initial BMI) (p < 0.001), negative affectivity (p < 0.05), and smoking habits (p < 0.02). Over and above this additive model, curvilinear/interactive terms involving age and BMI change contributed incrementally (p < 0.05) to SRH change: among younger participants (-1SD for age), decreased BMI predicted increased SRH, while among older participants (+1SD for age), increased BMI predicted decreased SRH. Conclusions: In this 'healthy worker' group, five-year decrease in SRH was associated with strenuous jobs, with starting or continuing smoking, with negative affectivity, and with increased BMI, but this latter effect depended on age. These results highlight the need to consider complex models of personal and occupational factors (including non-linear relationships) in SRH research.