Multilevel propensity score model (PSM) to predict a retail violation of sales to minors (RVSM) in 2017 (129 911 inspections clustered in n=26 131 zip codes)*
| Predictive variables | OR (95% CI) | P value | aOR (95% CI) | Adjusted p value |
| Zip code-level inspections and retailers | ||||
| Number of inspections 2015–2016† | 1.00 (1.00 to 1.00) | 0.6658 | 0.988 (0.987 to 0.99) | <0.0001 |
| Number of violations 2015–2016† | 1.02 (1.02 to 1.03) | <0.0001 | 1.07 (1.06 to 1.08) | <0.0001 |
| Number of penalties 2015–2016† | 1.06 (1.05 to 1.08) | <0.0001 | 0.97 (0.95 to 0.99) | 0.0011 |
| Number of tobacco retailers‡ | 1.21 (1.09 to 1.35) | 0.0007 | Not included | |
| Zip code-level characteristics | ||||
| Population§ | 1.00 (1.00 to 1.00) | 0.0045 | 1.00 (1.00 to 1.01) | 0.0011 |
| Metropolitan (vs non-metropolitan) | 1.22 (1.17 to 1.28) | <0.0001 | 1.3 (1.22 to 1.38) | <0.0001 |
| Race/ethnicity¶ | ||||
| Caucasian (%) | 0.95 (0.94 to 0.96) | <0.0001 | Not included** | |
| African American (%) | 1.08 (1.07 to 1.09) | <0.0001 | 1.04 (1.03 to 1.06) | <0.0001 |
| Hispanic (%) | 1.01 (1.00 to 1.03) | 0.1577 | 0.99 (0.98 to 1.01) | 0.5383 |
| Asian (%) | 0.90 (0.86 to 0.94) | <0.0001 | 0.94 (0.89 to 0.98) | 0.0101 |
| American Indian (%) | 1.19 (1.10 to 1.28) | <0.0001 | 1.19 (1.10 to 1.29) | <0.0001 |
| Age distribution (years)¶ | ||||
| 10–17 (%) | 0.85 (0.79 to 0.93) | 0.0002 | 0.80 (0.73 to 0.88) | <0.0001 |
| 18–20 (%) | 1.10 (1.03 to 1.18) | 0.0062 | 1.04 (0.95 to 1.15) | 0.3729 |
| 21–24 (%) | 1.15 (1.07 to 1.24) | 0.0001 | 0.99 (0.89 to 1.10) | 0.8253 |
| Education and income¶ | ||||
| Bachelor’s degree or higher (%) | 0.96 (0.94 to 0.98) | <0.0001 | 0.97 (0.95 to 1.00) | 0.0188 |
| Poverty (%) | 1.14 (1.11 to 1.17) | <0.0001 | 1.02 (0.98 to 1.07) | 0.2504 |
| County-level characteristics | ||||
| Adult smoking prevalence (%)¶ | 1.28 (1.20 to 1.36) | <0.0001 | 1.13 (1.03 to 1.23) | 0.0068 |
| Zip code-level Tobacco 21 policies†† | ||||
| No | Reference | Reference | ||
| Local | 0.72 (0.64 to 0.81) | <0.0001 | 0.70 (0.63 to 0.79) | <0.0001 |
| State | 0.49 (0.40 to 0.61) | <0.0001 | 0.50 (0.4 to 0.62) | <0.0001 |
| State-level characteristics | ||||
| Cigarette tax ($)††† | 0.97 (0.95 to 0.99) | 0.0011 | Not included** | |
| Tobacco retail licensing†† | ||||
| E-cigarettes or other tobacco products | 0.58 (0.54 to 0.62) | <0.0001 | 0.66 (0.62 to 0.71) | <0.0001 |
| Tobacco products other than e-cigarettes | 0.81 (0.77 to 0.85) | <0.0001 | 0.83 (0.78 to 0.87) | <0.0001 |
| No license required | Reference | Reference | ||
*PSM development: multilevel propensity score was developed using generalised estimating equation model, where the dependent variable was RVSM (yes vs no) at each inspection in 2017 and the explanatory variables included all variables listed in the first column as fixed effects along with the random effects shown in equation (1).ORs from univariate analysis and aORs from the multivariable analysis are reported.
†Per 1 unit increase.
‡Per 100 units increase.
§Per 1000 people increase.
¶Per 10% increase.
**Due to multicollinearity with other predictors, Caucasian (%) and state cigarette tax were not included in the multivariable regression.
††As of December 2016.
aOR, adjusted OR.