The Told true group here aimed to mimic effects of endorsement by health promotion agencies. All analyses in the 2 x 2 mixed model ANOVAs contained the predicted interaction of Group and Rating (all p < .01), reflecting an increase in motivation ratings for the Told true group c.f. . Table 1a provides mean ratings (SE) for each message and summarizes the effects of post-hoc Bonferroni corrected t-tests. For purposes of comparison, Table 1b shows results for five different messages from the same interviews, the data for which were not presented previously for each individual message .
Agreement ratings were low for all messages that concerned calorific expenditure or weight relative to the other messages, irrespective of whether the other messages were general (daily exercise) or outlined consequences (keep healthy, exercise heart, keep fit). Formal comparisons between the messages on agreement ratings confirmed this (all prob. <.001). For the group Told nothing, agreement ratings were greater than motivation ratings for all messages not involving calorific expenditure as reported previously . Finally, informing respondents that the messages were true increased their apparent motivational properties. Hence, there were no differences between agreement and motivation ratings for messages not explicitly related to weight control. Further, motivation ratings were significantly greater than agreement ratings for all messages relevant to weight control, suggesting this type of message could be effective for stair climbing if endorsed by health promotion agencies.
Effects of intervention
Preliminary inspection of the data with box-plots revealed that pedestrian traffic levels greater than 20 min-1 were outliers and any such time points were excluded from further analysis (2.9% of observations). Thus, 28,854 and 29,352 separate choices between the stairs and lifts remained for analysis of ascent and descent respectively.
Table 2 summarises the omnibus analyses. Significant effects of the intervention on stair climbing contrasted with no effect on stair descent. As the intervention effect on stair climbing interacted with the site, separate analyses were computed for each site (all p < .001 unless stated). There was a greater overall effect of the Poster + Stairwell messages campaign (+12.3%; Odds Ratio (OR) = 1.52, 95% Confidence Interval (CI) = 1.40-1.66) than with the Posters alone (+7.2%; OR = 1.24, CI = 1.15-1.34). Figure 4, which depicts the percentage of individuals climbing stairs at the two sites over successive weeks, suggests that incremental changes occurred. Follow-up analyses revealed an increase between baseline and week one (Poster + Stairwell messages: OR = 1.36, CI = 1.23-1.50; Poster alone: OR = 1.12, CI = 1.03-1.22, p = .007) and a further increment between weeks one and two (Poster + Stairwell messages: OR = 1.33, CI = 1.18-1.50; Poster alone: OR = 1.18, CI = 1.07-1.31). Any incremental effect had dissipated by week three (Poster + Stairwell messages: OR = 0.88, CI = 0.78-1.00, p = .04; Poster alone: OR = 0.98, CI = 0.88-1.08). For stair descent, Table 2 reveals no effects of the intervention, nor any interaction between the interventions and site (Poster + Stairwell messages; baseline 66.4%, intervention 64.2%; OR = 0.99, CI = 0.92-1.06: Poster alone; baseline 70.8%, intervention 71.6%; OR = 1.03, CI = 0.96-1.11).
Alongside the effects of the intervention, Table 2 reveals that there were overall differences between the sites in stair usage. Stair climbing was more common at the Poster + Stairwell messages site (62.3% vs. 41.0%; see Figure 4) whereas stair descent was more common at the Poster alone site (71.3% vs. 65.5%). Finally, there were effects of pedestrian traffic on both ascent and descent such that less people used the stairs at higher rates of traffic volume.
The follow-up questionnaire was returned by 123 (24.6%) and 165 (13.8%) employees at the sites respectively. This low response rate suggests the data should be treated with some caution despite the reasonable numbers for statistical comparisons; only employees with complete data were included in analyses. While there were no differences between the gender composition of those returning the questionnaire between sites (Poster + Stairwell messages = 53.4% female; Poster alone = 58.6% females), respondents from the Poster + Stairwell messages site were younger than those from the Poster alone site (mean years = 34.0 SE = 1.3 vs. mean years = 42.2 SE = 0.9; t251 = 5.17, p < .001).
Respondents reported greater encouragement to use stairs for the Poster + Stairwell messages campaign (mean = 3.58, SE = 0.15) than for Posters alone (mean = 3.08, SE = 0.12: t251 = 2.31, p = .02). Intentions to use the stairs in the future were positive overall, with no differences between sites (Poster + Stairwell messages mean = 3.97, SE = 0.16, Posters alone mean = 3.82, SE = 0.13: t251 = 0.75, p = .46). Table 3 summarises agreement with four messages displayed in the stairwell at the Poster + Stairwell messages site, as well as the unused comparison messages, and the results of Bonferroni corrected t-tests. For the messages displayed at the Poster + Stairwell messages site, Table 3a shows no differences between the worksites for agreement with the ‘daily exercise’ or ‘helping weight control’ messages. In contrast, agreement with both calorific expenditure messages were higher at the Poster + Stairwell messages site than at the site not displaying them. As shown in Table 3b, there were no significant differences between the worksites in agreement with statements about stair climbing that were not presented in the stairwell.
Finally, exploratory analysis of intentions to use the stairs more in the future used the predictor variables of motivation by the campaign, agreement with the four additional stairwell messages contained in the questionnaire and the interaction terms reflecting potentially better agreement with these measures at the Poster + Stairwell message site which displayed them. All data were mean centred prior to analysis . To control for multi-collinearity of the predictor variables, particularly the stairwell statements, stepwise regression was employed with age, gender and worksite forced into the model. Table 4 summarises the final model accounting for 41.9% of the variance of future intentions to use the stairs more (F5,247 = 36.89 p < .001). Gender, encouragement by the campaign and greater agreement that regular stair climbing would help weight control at the Poster + Stairwell messages site relative to the Poster alone site contributed to intention. A model with just the two campaign contributors accounted for 40.1% of the variance of intentions. In follow-up analysis, females (mean = 4.17 SE = 0.15) had more positive intentions than males (mean = 3.69 SE = 0.13: t251 = 2.44, p = .02) but did not differ on any of the other variables (all prob. >.12).