Table 3. State-level alcohol characteristics and health outcomes.
| State-level Twitter variablesa | ||
|---|---|---|
|
| ||
| Percent tweets about alcohol | Percent Yelp listing, bars & pubsa | |
|
| ||
| State-level adult health outcomes | Beta (95% CI)b | Beta (95% CI)b |
|
| ||
| All-cause mortality per 100,000 | -43.30 (-63.56, -23.04)** | -21.60 (-40.51, -2.68)* |
| Homicide per 100,000 | -0.85 (-1.59, -0.12)* | 0.05 (-0.57, 0.67) |
| Suicide per 100,000 | 0.22 (-0.93, 1.38) | -0.50 (-1.45, 0.44) |
| Unintentional injury death | -2.17 (-4.63, 0.30) | -1.45 (-3.52, 0.62) |
| Percent poor/fair self-rated health | -1.66 (-2.32, -1.01)** | -1.23 (-1.80, -0.66)** |
| Percent binge drinking | 1.86 (0.85, 2.88)** | 2.30 (1.63, 2.96)** |
| Percent heavy drinking | 0.75 (0.40, 1.10)** | 0.62 (0.33, 0.91)** |
| Percent current smoking | -1.39 (-2.31, -0.48)** | 0.09 (-0.75, 0.93) |
a
Twitter-derived variables (independent variables in regression models) were standardized to have a mean of 0 and standard deviation of 1. N=49. States in the contiguous United States, including District of Columbia
b
Adjusted linear regression models were run for each outcome separately. Models controlled for state-level demographics: median age, % non-Hispanic white, median household income. Data sources for health outcomes: 2013 National Vital Statistics Reports, 2014 Behavioral Risk Factor Surveillance System
*
p<0.05;
**
p<0.01