Table 4.
Regression analyses of intervention effects on smoking prevalence among Chinese Americans in New York Citya
| Independent variables | Odds ratio (95% CI) | P value | Marginal Effectb (95% CI) | P value |
|---|---|---|---|---|
| Interventionc = Interaction of Area * time | ||||
| Flushing post-intervention | NAd | _ | −0.028 (−0.03–−0.00) | 0.001 |
| Other | Referent | |||
| Time period | ||||
| Post- intervention | 0.77 (0.63–0.95) | 0.015 | −0.033 (−0.00–0.01) | 0.013 |
| Pre- intervention | 1.00 | Referent | ||
| Site | ||||
| Flushing | 1.33 (1.03–1.70) | 0.028 | 0.037 (0.00–0.07) | 0.032 |
| Sunset Park | 1.00 | Referent | ||
| Age | ||||
| 18 to 34 | 1.22 (0.80–1.85) | 0.358 | 0.026 (0.03–0.08) | 0.37 |
| 35 to 54 | 1.38 (1.01–1.89) | 0.041 | 0.041 (0.00–0.08) | 0.04 |
| ≥55 | 1.00 | Referent | ||
| Income | ||||
| < $20,000 | 1.51 (1.05–2.17) | 0.024 | 0.05 (0.00–0.09) | 0.024 |
| $20–40,000 | 1.74 (1.22–2.47) | 0.002 | 0.08 (−0.02–0.13) | 0.004 |
| >$40,000 | 1.00 | Referent | ||
| Education | ||||
| < HS | 1.42 (1.03–1.95) | 0.028 | 0.046 (−0.00–0.08) | 0.034 |
| = HS | 1.42 (1.03–1.97) | 0.034 | 0.048 (−0.00–0.09) | 0.047 |
| > HS | 1.00 | Referent | ||
| % years in US | ||||
| <20% | 1.31 (.95–1.82) | 0.100 | 0.036 (0.00–0.08) | 0.115 |
| 20–35% | 1.32 (.96–1.82) | 0.084 | 0.037 (0.00–0.08) | 0.094 |
| >35% | 1.00 | Referent | ||
(NYC Chinese Health Study 2002–2006)
Abbreviations: CI, confidence interval. Dependent variable: 0= never and former smokers and 1=current smokers.
N = 2165 (total nonmissing observations, baseline and follow-up)
Independent effects of predictors on smoking prevalence (marginal change in probability of smoking) were calculated using the STATA v9. However, the independent effect of the intervention on smoking prevalence was separately calculated following methods for calculating estimates of interaction terms in a nonlinear model. All values in the marginal effect column represent absolute percentages. (Ai and Norton 2001)
The effect of the intervention, measured as the interaction between pre-post time period (0,1) and control vs intervention community (0,1) is interpreted as the additional decline of smoking prevalence by an absolute 2.8% in Flushing relative to Sunset Park.
Odds ratio calculation algorithms do not apply to interaction terms (Ai and Norton, 2003).