Abstract
Only 52% of the workforce began the program and only 24% (2606/10899) completed 1 year. (Only 12% have been included in the analysis at 2 years.) The results should be seen in this context of low ongoing participation and a very high degree of self-selection.
The latest report presents data from a subgroup. It is not clear how this group was identified. This is critical to any interpretation.
The demographics of the intervention group are unusual—more than half were at least 6 years older than the mean age, suggesting an uneven age distribution with an unusually large number of very young workers. It would be helpful to report a comparison of the demographics of the intervention group with the total workforce (and with those who did not complete the program).
The initial report did not use a control group. Although the intervention group was compared with a Natural Flow model this approach has shortcomings. The model does not appear to have been validated across different work settings and certainly not for this intervention group. The comparison model uses updated and unpublished data. The model describes changes in the whole population while the intervention group was a self-selected minority. This is likely to have exaggerated the beneficial effect of the intervention in the statistical analysis.
Although the original report described some successes this was not complete—at 1 year more subjects were smokers, more were overweight, and more used drugs for relaxation. These were absolute increases despite the intervention. In addition, although the initial assessment identified many conditions directly amenable to medical treatment it is a concern that so many subjects still were not being effectively treated after 1 year (eg, hypertension).
The latest data are very difficult to interpret. Percentages alone give no indication of impact. For example, in the initial report 7 more participants were smokers after 1 year (129 from 122; +6%). In the latest report the number of smokers had reduced by 22.6%. Similarly, an increase of obesity in the first year (1083 to 1113; +3%) was followed by an impressive reduction in the second (−16.6%). It is difficult to understand why so much would be achieved at a relatively late stage: These transformations in the second year warrant explanation.
Finally, the title of the recent letter suggests the intervention is effective in lowering costs. This claim is not supported by any data in the letter (or in the previous report).
This is an important and potentially influential project. The authors should be encouraged to make their data more accessible. A statistical analysis based on paired data would be very worthwhile as would an indication of program costs. Most importantly, the data deserve independent analysis with critical commentary on the success. It will then be clearer how much has been achieved and the extent to which this might be generalized.