In review of our article (Kouvonen et al: BMC Public Health 2006, 6:251) we found an error in the equation of our measurement of trait anxiety.
Our data are comprised of two sub-samples, the 10-Town Study and the Hospital Personnel Study. Similar methods of data collection were used in both sub-samples. However, the direction of response alternatives for one of the scales (trait anxiety) differed between the two sub-samples.
That is to say that the response alternative "almost never" was coded as "1" in the first sub-sample whereas in the second sub-sample it was coded as "4". Therefore we can also deduce that the responses of the participants from the different sub-samples were coded into the different direction.
This error is corrected by recoding the variable so that in both sub-samples the higher score indicates higher trait anxiety. We re-conducted the analyses and found only minor differences compared to the original figures given and we also found that the error did not alter the main results or conclusions drawn.
The following corrections should be incorporated into any future analysis of our original article .
Within the Methods section, paragraph 10 (Trait anxiety), the Cronbach's Alpha for Trait Anxiety should be 0.85 instead of 0.88.
Within the Results, paragraph 6 (2nd paragraph within Validity), the last sentence should now read as "The associations with trait anxiety and magnitude of changes in work were weaker." rather than "In contrast, the associations with trait anxiety and magnitude of changes in work were much weaker."
Within the Results, paragraph 7 (3rd paragraph within Validity), the 3rd sentence should read "Further adjustment for personality trait anxiety attenuated the association in both sexes, but the results remained statistically significant." rather than "Further adjustment for personality factor trait anxiety had no effect on the associations."
Within the original Table 3 (The corrected version is available with this article as Table 1), the following two figures should be revised; Trait anxiety women should be changed from β = -0.07 to β = -0.19; and for men should be changed from β = -0.14 to β = -0.22.
Table 1. Corrected table 3: Associations between social capital measure and other constructs (GLIMMIX)
And finally, within the discussion, paragraph 5 (paragraph beginning "We assessed criterion") The 3rd sentence should read "The significant association in multilevel models together with a weaker association with one personality measure (trait anxiety) can indicate that the construct of social capital might be more than an aspect of an individual's personality." rather than "The significant association in multilevel models together with the substantially weak association with one personality measure (trait anxiety) can indicate that the construct of social capital might be more than an aspect of an individual's personality."
The pre-publication history for this paper can be accessed here: