I: Linear probability models, age adjusted | II: Linear probability models, fully adjusted | III: alternative outcome* | IV: Logistic regression models | V: Negative control analysis† | |
Sample size | 63 234 | 63 234 | 63 234 | 63 234 | 60 727 |
Primary exposure | |||||
Exposed to programme | 0.106 (0.054 to 0.158) | 0.075 (0.042 to 0.109) | 0.066 (0.034 to 0.099) | 0.044 (0.016 to 0.072) | −0.023 (−0.055 to 0.010) |
Secondary exposure | |||||
Programme reach‡ | 0.269 (0.184 to 0.353) | 0.173 (0.112 to 0.234) | 0.149 (0.090 to 0.208) | 0.103 (0.050 to 0.156) | −0.004 (−0.087 to 0.080) |
Notes: All models except for column I were fully adjusted for place of birth (native, non-native), educational attainment (postsecondary education, no postsecondary education), sex, age (in 5-year age groups, except 18–24 years), survey and municipality fixed effects. Estimates in cells reflect probability differences in the proportion of ice cleat users between respondents exposed and eligible to a municipal ice cleat distribution programme versus those who were not; 95% CI (clustered by municipality) are presented in parentheses. Coefficients from the logistic regression reflect marginal effects (ie, probability differences) estimated via the margins command in Stata. The primary and secondary exposures were modelled separately.
*Respondents who replied seldom using ice cleats are also considered ice cleat users in this definition.
†Analyses coding respondents between 15 years and 1 year under the age threshold for programmes as the ‘treated’ age group. Sample size is smaller because those who were actually exposed are excluded from the analysis.
‡Distributed ice cleat pairs per age-eligible citizen, coded as zero for unexposed respondents.