Abstract
This cross-sectional study examines the travel patterns from home residences to locations of fatal drug overdoses in Cook County, Illinois, from August 1, 2014, to December 31, 2018.
Fatal drug overdoses increased in US cities from 1999 through 2017.1 Low-resource urban areas—popularly characterized as places where overdoses and drug trade thrive—are often considered self-contained, with residents considered the primary consumers of local drug markets.2 There has been less discussion regarding inbound travel to such drug markets from nonresidents and which neighborhood characteristics may correspond with travel decisions. Targeting residents of overdose hot spots for intervention may reinforce stereotypes while excluding nonresident people who use drugs from treatment screening and delivery. We examined travel patterns between locations where fatal drug overdoses occurred and the home residences of the people who died plus the neighborhood-level characteristics that may have differed across these locations. We also compared travel patterns specifically for overdoses involving fentanyl and heroin.
Methods
The Columbia University Medical Center institutional review board waived study review and also waived the need for patient informed consent because this research did not involve human subjects research. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. For this cross-sectional study, all fatal drug overdose records from the Cook County, Illinois, medical examiner were obtained for the period August 1, 2014, through December 31, 2018. The records included full toxicology reports; Global Positioning System (GPS) coordinates for overdose location; zip codes of home residence; and age, sex, and race/ethnicity (eMethods in the Supplement). Public transit data included Chicago Transit Authority L and Metra commuter rail stations because transit hubs provide access to drug markets for nonresidents.3
We assessed neighborhood disadvantage using the formula {[(c/10 + d/10) − (a/10 + b/10)]/4} with 5-year US Census percentages, where a represents adults 25 years or older with a college degree, b represents owner-occupied housing, c represents households with incomes below the federal poverty threshold, and d represents female-headed households with children. Neighborhood disadvantage scores ranged from −5 (very low or little disadvantage) to +5 (very severe disadvantage).4 We assessed segregation using the Index of Concentration at the Extremes5 by subtracting the number of non-Latino Black residents from the number of non-Latino White residents in a zip code and dividing by the zip code population (segregation ranges, −1 indicates 100% Black population; 0 indicates 50% Black, 50% White; and 1 indicates 100% White].
We calculated the euclidean distance from the home zip code centroid to GPS coordinates of the overdose location (eFigure in the Supplement). Because urban zip codes are compact and divided along lines that do not necessarily correspond with residents’ versions of their neighborhood,6 we designated overdoses that occurred in the same or contiguous zip codes as home zip code nontraveling and overdoses that occurred 2 or more zip codes away as “far” traveling (eFigure in the Supplement). We used logistic regression to assess individual- and neighborhood-level correlates of travel. Two-sided P < .05 was considered statistically significant. R software, version 3.4.1 (R Foundational for Statistical Computing) was used for statistical analysis.
Results
Of 3927 fatal overdoses, the mean (SD) age across all overdoses was 44.1 (12.6) years, 2972 (75.7%) were men, 1832 (46.7%) were non-Latino White, and 1596 (40.6%) were non-Latino Black. A total of 1171 individuals (30%) had traveled 2 or more zip codes beyond their home zip code (mean [SD] distance, 49.4 [262.4] km). Men (923 of 1171 individuals [78.8%]; P = .003) and younger individuals (mean [SD] age, 41.9 [12.2] vs 44.8 [12.6] years; P < .001) were significantly more likely to travel, and there were no differences by racial/ethnic subcategories. Decedents were more likely to travel far from zip codes with low to high neighborhood deprivation (adjusted odds ratio [AOR], 1.43; 95% CI, 1.27-1.60) and from zip codes that were predominantly non-Latino White to predominantly non-Latino Black (AOR, 2.13; 95% CI, 1.61-2.83) (Table 1). Travel was significantly associated with fentanyl-involved overdoses (AOR, 1.40; 95% CI, 1.20-1.63), but not with heroin-involved overdoses (AOR, 1.12; 95% CI, 0.96-1.29) after controlling for race/ethnicity, sex, neighborhood deprivation score, and transit hub in home zip code (Table 2).
Table 1. Univariable and Multivariable Logistic Regression Results for Odds of Far Travel Among 3927 Fatal Drug Overdoses in Cook County, Illinois, August 1, 2014, to December 31, 2018a.
| Variable | Unadjusted OR (95% CI) | P valueb | Adjusted OR (95% CI)c | P valueb |
|---|---|---|---|---|
| Age | 0.98 (0.98-0.99) | <.001 | 0.99 (0.98-0.99) | <.001 |
| Race/ethnicityd | ||||
| Non-Latino | ||||
| White | 1 [Reference] | |||
| Black | 0.91 (0.78-1.05) | .20 | ||
| Latino | 1.10 (0.88-1.37) | .39 | ||
| Other | 0.63 (0.30-1.19) | .17 | ||
| Sex | ||||
| Female | 1 [Reference] | 1 [Reference] | ||
| Male | 1.28 (1.09-1.52) | .003 | 1.24 (1.03-1.48) | .02 |
| Change in neighborhood disadvantage score from home to overdose zip codese | 1.94 (1.80-2.08) | <.001 | 1.43 (1.27-1.60) | <.001 |
| Change in ICE from home to overdose zip codesf | 0.20 (0.17-0.24) | <.001 | 0.47 (0.35-0.62) | <.001 |
| Transit hub in home zip code | ||||
| No | 1 [Reference] | 1 [Reference] | ||
| Yes | 0.35 (0.30-0.41) | <.001 | 0.19 (0.15-0.24) | <.001 |
| Transit hub in overdose zip code | ||||
| No | 1 [Reference] | 1 [Reference] | ||
| Yes | 1.27 (1.06-1.53) | .01 | 3.91 (3.03-5.08) | <.001 |
Abbreviations: ICE, index of concentration at the extremes; OR, odds ratio.
The data source was the Cook County Open Data Medical Examiner Case Archive when accessed April 29, 2019 (https://datacatalog.cookcountyil.gov/Public-Safety/Medical-Examiner-Case-Archive/cjeq-bs86).
P values were 2-sided, with values less than .05 considered significant.
Adjusted for other covariates in the column (Akaike information criterion, 4068.1).
There were no significant differences for racial/ethnic subcategories in univariable analysis, and these subcategories were not included in the adjusted model.
Direction of travel dictates interpretation (eg, if a person traveled from a zip code of low deprivation [−5] to high deprivation [+5], the odds of far travel increased).
Direction of travel dictates interpretation (eg, if a person traveled from a predominantly Black [−1] to a predominantly White [+1] zip code, the odds of far travel decreased).
Table 2. Univariable and Multivariable Logistic Regression Results for Odds of Fentanyl- and Heroin-Involved Fatal Overdoses, Cook County, Illinois, August 1, 2014, to December 31, 2018a.
| Variable | Type of Overdose | |||||||
|---|---|---|---|---|---|---|---|---|
| Fentanyl-involved (n = 1917)b | Heroin-involved (n = 2232)c | |||||||
| Unadjusted OR (95% CI) | P valued | Adjusted OR (95% CI)e | P valued | Unadjusted OR (95% CI) | P valued | Adjusted OR (95% CI)e | P valued | |
| Age | 0.99 (0.99-1.00) | .14 | 0.99 (0.99-0.99) | .006 | 0.99 (0.98-0.99) | .02 | ||
| Race/ethnicityf | ||||||||
| Non-Latino | ||||||||
| White | 1 [Reference] | 1 [Reference] | 1 [Reference] | |||||
| Black | 1.57 (1.37-1.79) | <.001 | 1.69 (1.47-1.94) | <.001 | 1.01 (0.89-1.16) | .84 | ||
| Latino | 1.62 (1.32-2.00) | <.001 | 1.62 (1.32-2.00) | <.001 | 0.93 (0.76-1.15) | .52 | ||
| Other | 0.50 (0.26-0.91) | .03 | 0.50 (0.26-0.92) | .03 | 0.62 (0.35-1.08) | .09 | ||
| Sex | ||||||||
| Female | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] | ||||
| Male | 1.31 (1.13-1.51) | <.001 | 1.31 (1.13-1.52) | <.001 | 1.21 (1.04-1.40) | .01 | 1.20 (1.03-1.39) | .02 |
| Far travel | ||||||||
| No | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] | ||||
| Yes | 1.59 (1.39-1.83) | <.001 | 1.40 (1.20-1.63) | <.001 | 1.22 (1.06-1.40) | .006 | 1.12 (0.96-1.29) | .15 |
| Change in neighborhood deprivation score from home to overdose zip codesg | 1.18 (1.11-1.24) | <.001 | 1.13 (1.07-1.20) | <.001 | 1.09 (1.03-1.15) | .003 | ||
| Change in ICE from home to overdose zip codesh | 0.70 (0.61-0.80) | <.001 | 0.79 (0.69-0.90) | <.001 | 0.83 (0.72-0.96) | .009 | ||
| Transit hub in home zip codei | ||||||||
| No | 1 [Reference] | 1 [Reference] | 1 [Reference] | |||||
| Yes | 0.84 (0.73-0.97) | .02 | 0.84 (0.72-0.98) | .03 | 0.89 (0.77-1.03) | .12 | ||
| Transit hub in overdose zip codej | ||||||||
| No | 1 [Reference] | 1 [Reference] | ||||||
| Yes | 1.05 (0.90-1.23) | .55 | 0.88 (0.74-1.03) | .11 | ||||
Abbreviations: ICE, Index of Concentration at the Extremes; OR, odds ratio.
The data source was the Cook County Open Data Medical Examiner Case Archive when accessed April 29, 2019 (https://datacatalog.cookcountyil.gov/Public-Safety/Medical-Examiner-Case-Archive/cjeq-bs86). Records for 1085 fatal overdoses that involved both heroin and fentanyl were included in both sets of analyses.
For fentanyl-involved overdoses, the neighborhood deprivation score and ICE showed collinearity in multivariable analysis (variance inflation factor >3); when the neighborhood deprivation score and ICE were both in the model, ICE was no longer significant (P = .47) and the model including ICE showed a poorer fit (Akaike information criterion [AIC], 5321 vs 5318).
For heroin-involved overdoses, the neighborhood deprivation score and ICE showed collinearity (variance inflation factor >3), and neither was significant when both were present in the model. The model including ICE showed better fit statistics (AIC, 5350 vs 5352).
P values were 2-sided, with values less than .05 considered significant.
Adjusted for other covariates in the column (AIC for fentanyl, 5318; AIC for heroin, 5350).
There were no significant differences for racial/ethnic subcategories in univariable analysis for heroin-involved overdoses, and these subcategories were not included in the adjusted model.
Direction of travel dictates interpretation (eg, if a person traveled from a zip code of low deprivation [−5] to high deprivation [+5], the odds of far travel increased).
Direction of travel dictates interpretation (eg, if a person traveled from a predominantly Black [−1] to a predominantly White [+1] zip code, the odds of far travel decreased).
There were no significant differences for transit hub in home zip code in univariable analysis for heroin-involved overdoses, and these were not included in the adjusted model.
There were no significant differences for transit hub in overdose zip code in univariable analyses for fentanyl- or heroin-involved overdoses, and these were not included in the adjusted model.
Discussion
Thirty percent of decedents traveled far from their home to the location of the fatal overdose. Decedents tended to travel to more resource-deprived and segregated neighborhoods compared with their home neighborhood. Those who traveled were more likely to have fentanyl in their system at the time of death. This cross-sectional study was limited to fatal overdoses in 1 US mixed urban-suburban county. We did not have access to narratives for how or why people traveled to their overdose location. Additional narrative information is needed to provide context into how place and travel contribute to overdose.
People who use drugs to fatal ends may reside far distances from where they consume drugs. Nonresidents of overdose hot spots should be a focus of treatment screening and delivery.
eMethods. Supplementary Methods
eFigure 1. Example of Distance Calculation for Far Travel From Zip Code of Home Residence to Location of Fatal Overdose
eReferences.
References
- 1.Hedegaard H, Miniño AM, Warner M. Urban-rural differences in drug overdose death rates, by sex, age, and type of drugs involved, 2017. NCHS Data Brief 345. 2019;345(345):1-8. [PubMed] [Google Scholar]
- 2.Cooper HL, Tempalski B. Integrating place into research on drug use, drug users’ health, and drug policy. Int J Drug Policy. 2014;25(3):503-507. doi: 10.1016/j.drugpo.2014.03.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.McCord ES, Ratcliffe JH. A micro-spatial analysis of the demographic and criminogenic environment of drug markets in Philadelphia. Aust NZ J Criminol. 2007;40(1):43-63. doi: 10.1375/acri.40.1.43 [DOI] [Google Scholar]
- 4.Ross CE, Mirowsky J. Neighborhood disadvantage, disorder, and health. J Health Soc Behav. 2001;42(3):258-276. doi: 10.2307/3090214 [DOI] [PubMed] [Google Scholar]
- 5.Krieger N, Waterman PD, Spasojevic J, Li W, Maduro G, Van Wye G. Public health monitoring of privilege and deprivation with the index of concentration at the extremes. Am J Public Health. 2016;106(2):256-263. doi: 10.2105/AJPH.2015.302955 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Coulton CJ, Korbin J, Chan T, Su M. Mapping residents’ perceptions of neighborhood boundaries: a methodological note. Am J Community Psychol. 2001;29(2):371-383. doi: 10.1023/A:1010303419034 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods. Supplementary Methods
eFigure 1. Example of Distance Calculation for Far Travel From Zip Code of Home Residence to Location of Fatal Overdose
eReferences.