Evidence has shown that residents of public housing developments (PHD) experience compromised health, but few studies make comparisons of health between PHD residents and non-PHD residents from the general population while accounting for sources of confounding that may influence the association between public housing residency and health. By combining a large convenience sample of adult PHD residents living in Boston with a large, complex survey sample of adult non-PHD Boston residents, the present study contributes to filling this gap by accounting for measured sources of confounding to model the association between PHD residency on select health behaviors.
We used a novel dataset combining data from two health surveys fielded in Boston, the Boston Behavioral Risk Factor Surveillance System survey, collected in 2015, and the Boston Housing Authority’ Resident Health and Wellness Survey, collected in 2016. We analyzed a sample (n=3,025) of PHD residents (n=791) and non-PHD residents (n=2,234) using propensity score techniques. Specifically, we fit a propensity score model of PHD residency as a function of age, sex, race/ethnicity, primary language, education, foreign-born status, number of years living in the United States, and neighborhood. This model generated inverse probability of treatment weights (IPTW), which work to synthesize as-good-as random assignment to PHD residency status in our sample. We estimated sample average treatment effects in weighted logistic regression for public housing residency and amount of water consumption, primary water source (tap vs. bottled water), sugar-sweetened beverage (SSB) consumption, and current smoking.
Propensity score weighting provided a sample balanced in terms of the measured factors that were fit in the propensity score model. Main results indicate PHD residents in comparison with non-PHD residents in our sample have significantly higher probabilities of worse health behaviors, including 9% lower probability of consuming tap water (vs. bottled), a 12% greater probability of consuming any SSB in a month (vs. none), and a 6% greater probability of currently smoking.
Conclusions and Implications
The use of propensity score weighting represents a valuable assessment of the association between PHD residency and select health behaviors. As such, these findings strengthen our understanding of the complex factors driving public housing residents’ health. However, to the extent that our propensity score model does not include some factors predictive of public housing residency, our results likely do not fully account for sources of confounding, particularly mental health and trauma. Broadly, our findings indicate that public housing residency may be negatively associated with residents’ health behaviors and therefore suggests the urgency of better understanding this association in order to promote the health of public housing residents. Social work researchers are well equipped to take up this line of inquiry and to inform actionable policy and practice recommendations to ensure public housing serves the wellbeing of residents, including the promotion of health and healthy behaviors.