Spatiotemporal analysis of the association between residential eviction and fatal overdose in Rhode Island

Introduction

Drug overdose is the leading cause of injury-related death in the USA.1 Overdose mortality accelerated during the COVID-19 pandemic, reaching record highs in 2021.2–5 Structural barriers to overdose prevention services—upheld by persistent criminalisation and stigmatisation of people who use drugs (PWUD)—were reinforced by health and social policies in response to COVID-19 that disrupted healthcare and recovery services as well as patterns of substance use.6 For instance, policy ramifications of the pandemic shifted drug market dynamics and restricted access to harm reduction services.2 7 8 Simultaneously, the corresponding economic crisis left communities increasingly vulnerable to unstable housing situations.9

Eviction is a policy-sensitive driver of houselessness and a lasting effect of structural disinvestment in neighbourhoods and simultaneously of gentrification as well.10 11 To reduce evictions in the midst of unprecedented levels of economic vulnerability and housing instability during the COVID-19 pandemic, states and later the federal government implemented temporary eviction moratorium policies—blocking select stages of the eviction process such as notices, filings, hearings and/or enforcement.12 These policies dramatically reduced evictions but did not eliminate them entirely, producing a sudden shift in which demographic groups were targeted for evictions and the immediacy of displacement threat posed by an eviction filing.13–15 In the context of pandemic-era eviction moratoria, the dynamics between housing and health may have shifted as well. Several previous studies have identified associations between eviction moratoria and improved physical and mental health outcomes, yet none to our knowledge has examined a relationship with overdose mortality.16–19

Housing insecurity (a term used to describe limited or uncertain access to adequate housing20) is a known risk factor for fatal overdose and is especially detrimental for people who are penalised for their drug use (ie, through losing housing, being excluded from supportive housing or incarceration).21–24 Eviction functions as a structural and policy-sensitive cause of housing insecurity.11 Non-payment of rent, for instance, can initiate eviction proceedings, which often leads to removal from the home. The distribution or use of a controlled substance by tenants can also serve as legal justification for landlords filing for eviction, which puts PWUD at extremely high risk of losing housing—particularly those who experience an overdose that prompts law enforcement response.25 Further, qualitative evidence from Canada suggests that gaps in tenancy protections and landlord dispute resolution processes for PWUD often lead to unlawful evictions.26 However, little is known about the reverse association among communities of PWUD, that is, the impact of residential eviction on overdose risk. Eviction has the potential to disrupt patterns of drug procurement and consumption and at the same time could lead to increased neighbourhood blight, social fragmentation and socioeconomic marginalisation—stressors that amplify overdose risk.21 27–30 An earlier, county-level investigation found a positive association between eviction and drug and alcohol mortality in the USA.31 Eviction filings might increase overdose risk by (1) increasing psychosocial and financial stress17 32–34 and (2) displacing people from their homes.35 As a means of coping with the former, people may change their patterns of drug use or engagement with harm reduction strategies.22 The latter might disrupt people’s social networks,11 access to healthcare36 and harm reduction services,37 and drug supply.21 More broadly, residential eviction is also associated with stressors such as interpersonal violence among PWUD,38 mental distress34 39 and disengagement from healthcare services,36 40 which may further exacerbate overdose risk.

Rhode Island (RI) communities face risks that contribute to housing insecurity and overdose mortality, particularly because annual housing production in RI has not kept pace with state need, ranking last in the nation in 2021.41 In a previous study in RI, we found that the proportion of residents living in unaffordable housing (ie, tenants who pay more than 30% of their income for housing) was associated with excess overdose fatalities at the beginning of a month, suggesting that even the threat of eviction may result in higher intensity of substance use leading up to rent deadlines.42 Using data from RI census tracts, we sought to examine how residential eviction plays a role in spatial patterns of fatal overdose in the context of pandemic-related policy and an increasingly potent drug supply. We hypothesise that eviction rates will be positively correlated with risk of fatal overdose. During the COVID-19 pandemic and periods of statewide eviction moratoria in RI, we hypothesise that the magnitude of association between eviction and overdose mortality is further increased as eviction filings that occurred during this time (ie, a time of great stress imposed by the pandemic) were potentially even more disruptive, exacerbating a range of additional pandemic-related social, economic and health stressors.

Methods

Results

Estimated as average annual rates over the full study period for all census tracts included in the analysis, the median residential eviction rate was 3.26 (IQR 2.13–4.64) per 100 renter-occupied housing units and the median fatal overdose rate was 27.17 (IQR 13.65–44.11) per 100 000 person-years (table 1). The average annual eviction rates were higher prior to the COVID-19 pandemic and periods of eviction moratoria; over the prepandemic study period, the median was 3.73 (IQR 2.34–5.36) per 100 renter-occupied housing units, but during the pandemic, the median was 2.18 (IQR 1.29–3.31) per 100 renter-occupied housing units. Conversely, the average annual fatal overdose rates increased during the COVID-19 pandemic, as the medians were 24.99 (IQR 11.61–40.77) per 100 000 person-years over the prepandemic study period and 28.74 (IQR 12.30–51.34) per 100 000 person-years during the pandemic.

Residents of census tracts in higher eviction quintiles were generally younger with lower incomes and were more likely to be black or Hispanic, unemployed, uninsured or live in an unmarried household compared with those living in census tracts belonging to lower eviction quintiles (table 1). Census tracts in higher eviction quintiles also had proportionally more households who received SNAP, had children or had individuals with a disability and had greater proportions of housing units that were vacant and housing units that were rented compared with those in lower eviction quintiles (table 1). Notably, census tracts in higher quintiles of eviction rates, based on average annual estimates over the full study period, had increased rates of fatal overdose relative to those in lower quintiles of eviction rates (figure 1).

Request permission

In the time-updated models in which rates of eviction and fatal overdose were each discretised in 6-month periods, the association between eviction and fatal overdose was only significant during periods in which a statewide eviction moratorium was not in effect (tables 2 and 3, online supplemental table 1). Prior to the implementation of an eviction moratorium in response to the COVID-19 pandemic in RI, the average annual rate of fatal overdose in census tracts in the highest eviction quintile was 1.49 times (95% CrI 1.05 to 2.13) that of census tracts in the lowest eviction quintile (table 3, online supplemental table 1). However, there were also substantially more census tracts in the higher-evicting quintiles during this time period (online supplemental table 2). During periods in which an eviction moratorium was in effect, the average annual rate of fatal overdose in census tracts in the highest eviction quintile was 1.32 times (95% CrI 0.88 to 1.98) that of census tracts in the lowest eviction quintile (table 3, online supplemental table 1). Sensitivity analyses using quintiles of eviction hearings rather than eviction filings produced similar results (online supplemental tables 3–5).

Discussion

In this 6-year population-based study, we found a direct, dose–response, census tract-level association between residential eviction and fatal overdose rates in the absence of a statewide eviction moratorium (ie, during the 4 years prior to the COVID-19 pandemic). These results highlight a neighbourhood-level relationship between residential eviction and fatal overdose prior to the COVID-19 pandemic and suggest that pandemic-era policy interventions to prevent evictions may have had additional benefits vis-à-vis overdose risk. Enhanced investment in eviction prevention policies, such as rent relief and Just Cause rules that limit the circumstances under which landlords can file for eviction, may complement ongoing harm reduction efforts in the state to mitigate neighbourhood-level overdose inequalities. It may also be true that efforts to destigmatise drug use and address overdose mortality (eg, through harm reduction approaches) could serve to reduce housing insecurity.

Our findings contradicted our hypothesis that we would observe a stronger association between eviction and fatal overdose on implementation of an eviction moratorium. Instead, we did not find a significant association between eviction filings and fatal overdose in the periods covered by eviction moratoria. This is not to say that evictions during the pandemic did not cause harm. There are several potential mechanisms to explain the null association. First, moratoria during this period changed the meaning of eviction filings, reducing the immediate risk of displacement associated with a filing. If the association between filings and overdose operates largely via displacement, rather than the stress of a filing alone, moratoria could attenuate associations by intervening on a key mediator.21 Second, it is possible that the contrast between high-evicting and low-evicting census tracts became less meaningful during the moratorium. This could be true if, for example, landlords replaced formal eviction filings with informal or illegal evictions during the moratoria,13 such that rates of tenant stress and displacement were similar between the two. Third, the moratoria reduced levels of filings overall. With fewer census tracts in the top quartiles of eviction filings postmoratorium, we are potentially less powered to see associations. Lastly, elevated pandemic-era overdose mortality was driven by a complex combination of factors, including a range of psychosocial, health, and economic stressors and an increasingly potent drug supply5—all of which may have obscured the relationship between eviction and fatal overdose and any effects that the eviction moratorium may have had on overdose outcomes.

Future research should consider racial and ethnic inequities in the relationship between eviction and fatal overdose, and the extent to which law enforcement oversurveillance in black and Hispanic communities contributes to this. There is potentially a stronger correlation between eviction rates and risk of fatal overdose in minoritised communities that are disproportionately affected by policing and other mechanisms of structural racism, particularly as racialised policing and oversurveillance of PWUD may reinforce exclusionary housing practices.50 This work motivates further assessment of an individual-level association between eviction and fatal overdose, which may allow for the evaluation of how this relationship varies by race and ethnicity and other sociostructural factors.

Although our analysis offers preliminary findings, more robust and expansive data may add richness to our results by investigating additional factors that contribute to eviction and overdose risk in more generalisable contexts. There are opportunities for future work to examine other mechanisms at a more granular level, including how changes in employment status occurring throughout the pandemic may have precipitated eviction and/or served as a common cause of eviction and fatal overdose. As more recent fatal overdose data become available, it will also be critical to explore postpandemic patterns in eviction and fatal overdose following the expiration of the eviction moratorium.

This study is primarily limited by its ecological design and thus we cannot infer causality, temporality or individual-level relationships between eviction and fatal overdose. It is feasible that a fatal outcome on an individual level may actually precede the exposure of interest within the household, if, for instance, a household is evicted as a result of an overdose death in that household, and future research should better assess this temporal relationship. However, our effect modification findings by whether an eviction moratorium was in place support our hypothesis that intervening on neighbourhood-level eviction risk may have implications for overdose mortality. Additionally, we were limited in our ability to control for educational attainment due to model instability and to assess effect measure modification of the association between eviction and fatal overdose by race and ethnicity due to insufficient variation in the tract-level distributions of race and ethnicity. The RI population is segregated to the extent that the vast majority of the black and Hispanic populations live in the state’s relatively few urban census tracts. We attempted to assess the effect modification by race and ethnicity of overdose decedents to overcome this data constraint, but we encountered challenges with small sample sizes. We encourage future research to consider how racial segregation shapes the geospatial relationship between eviction and fatal overdose. Lastly, we do not account for unmeasured confounding and our measured covariates do not allow us to fully disentangle how structural racism, poverty and segregation shape neighbourhoods and affect both eviction and overdose mortality.

In conclusion and despite these limitations, our study offers a broad examination of the relationship between eviction filing and fatal overdose in RI and suggests that, in the absence of an eviction moratorium, neighbourhood-level exposure to early stages of the eviction process is associated with elevated overdose mortality.