Overdose mortality ratio in Milwaukee County
After preprocessing and cleaning the data, we were left with a total of 17,476 nonfatal and 1,985 fatal overdoses in Milwaukee County. The Overdose Mortality Ratio (OMR), the percentage of fatal overdoses out of all overdoses, was calculated with an average value of 12.14% across the county. The OMR showed slight yearly fluctuation, with a minimum of 11.05% in 2020 and a maximum of 13.02% in 2019.
Identification of high- and low-risk census tracts for overdose mortality
We used SEBS and Jenks Natural Breaks Classification to calculate the Excess Risk Factor (ERF) of overdose mortality and create a map based on five classes of SEBS ratios: Very Low, Low, Moderate, High, and Very High. There was significant spatial heterogeneity in mortality in Milwaukee (Fig. 1), with census tracts having OMRs as low as 4.61% and as high as 26.48%.
Geographic racial disparities in overdose mortality risk
Milwaukee has a high level of racial segregation. Thus, we divided Milwaukee into three regions based on the predominant racial population (Black, Hispanic, or White) and calculated average OMRs across these communities. Over the 4-year period, average OMRs in Black-, Hispanic-, and White-majority neighborhoods were 11.96%, 16.15%, and 11.45%, respectively. OMRs also varied across racial demographic majority communities over time. In predominantly Black neighborhoods, OMRs showed little variation over the years, ranging from 12.68% in 2018 to 12.16% in 2021. In contrast, predominantly Hispanic neighborhoods experienced more variation, with the highest OMR of 21.03% in 2019 and the lowest of 13.72% in 2020. Predominantly White neighborhoods exhibited a stable OMR trend, with a slight increase from 11.37% in 2018 to 11.79% in 2021 (Table 1). It should not be assumed that differential OMRs in neighborhoods classified according to racial demographic majorities translate to similar OMRs in members of those communities or in community members of the same racial demographics. Indeed, we have previously reported that many overdose deaths, particularly in Hispanic-majority neighborhoods, are geographically discordant (i.e., overdosing individuals are traveling from other communities of residence; [24]). Nonetheless, the findings are consistent with previous research documenting disparities in access to health care and social determinants of health across racial and ethnic communities [16].
Table 1
Overdose Mortality Ratios Across Racial Demographic Majority Communities in Milwaukee County.
| 2018 | 2019 | 2020 | 2021 |
Black | 12.68% | 13.00% | 9.99% | 12.16% |
Hispanic | 14.39% | 21.03% | 13.72% | 15.46% |
White | 11.37% | 11.39% | 11.25% | 11.79% |
Standardized mortality ratios
Next, we applied the INLA Bernardinelli model to calculate relative risks and expected overdoses, accounting for the complex spatial and temporal patterns of drug overdose mortality. We calculated SMRs using 12*4 strata, defined based on age, race, and gender categories. Categories included six age groups (under 18, 18–20, 20–30, 30–40, 40–50, and older than 50), four racial groups (White, Black, Hispanic, and other), and two gender groups (female/male). Results obtained from the model were used to estimate the expected number of overdoses in each stratum. Estimates were compared to the observed number of overdoses to calculate SMRs for each census tract. SMRs provide a measure of relative risk, indicating whether an area has higher, equal, or lower risk than expected from the standard population. An SMR of 1.0 indicates that the observed number of overdoses is the same as expected; an SMR greater than 1.0 suggests higher risk; and an SMR less than 1.0 indicates lower risk.
SMRs varied across Milwaukee (Fig. 2), demonstrating spatial heterogeneity in overdose mortality risk. Areas with the highest relative risk are concentrated in the central and southern parts of Milwaukee County, including downtown and predominantly Hispanic neighborhoods. By contrast, areas with the lowest relative risk of are generally located in the shoreline and western parts of Milwaukee, including suburban and predominantly White neighborhoods. Overall, the results suggest that overdose mortality risk is not evenly distributed across communities and that some communities and likely subpopulations are at higher risk than others. These findings are consistent with studies identifying disparities in overdose mortality rates across neighborhoods and demographic groups.
Socioeconomic and demographic characteristics of high and low SMR communities
Table 2 displays socioeconomic and demographic data for areas with higher- and lower-than-expected mortality risk and county-wide data. Areas with higher-than-expected relative risk had a lower proportion of White residents and a higher proportion of Black or Hispanic residents relative to lower-than-expected risk areas or Milwaukee as a whole. Higher-than-expected risk areas also had lower median age, educational attainment, and per capita income and greater unemployment and incarceration rates, as well as poorer mental and physical health and a higher digital divide index, indicating lower access to digital resources. Conversely, lower-than-expected risk areas had a higher proportion of White residents, were older, had higher educational attainment and per capita income, and lower unemployment and incarceration rates, as well as better mental and physical health and a smaller digital divide.
Table 2
Socioeconomic, demographic, and health characteristics in areas with higher and lower than expected mortality risk in Milwaukee County.
| Equal | Higher | Lower | Milwaukee County |
White | 42.31% | 47.63% | 59.40% | 55.67% |
Black or African American | 44.23% | 35.12% | 28.17% | 30.64% |
Asian | 6.75% | 3.91% | 3.99% | 4.09% |
American Indian and Alaska Native | 0.47% | 0.70% | 0.52% | 0.56% |
Hispanic or Latino (of any race) | 10.09% | 20.64% | 12.74% | 14.63% |
born outside U.S. | 2.49% | 3.63% | 1.87% | 2.34% |
Households with an Internet Subscription | 74.88% | 75.56% | 81.49% | 79.70% |
Single-Parent Households | 44.70% | 47.17% | 38.19% | 40.75% |
Homeownership | 37.75% | 34.97% | 45.82% | 42.72% |
Median Household Income | $45,017 | $44,310 | $55,900 | $52,486 |
Per Capita Income | $23,613 | $22,687 | $31,361 | $28,823 |
Families Living Below Poverty Level | 21.35% | 23.19% | 14.40% | 16.93% |
Poor Physical Health: 14 + Days | 14.65% | 15.49% | 12.87% | 13.61% |
Poor Mental Health: 14 + Days | 17.40% | 17.86% | 15.39% | 16.10% |
Self-Reported General Health Assessment: Poor or Fair | 23.05% | 24.93% | 19.20% | 20.82% |
Adults without Health Insurance | 16.84% | 20.08% | 14.79% | 16.22% |
Adults Ever Diagnosed with Depression | 20.08% | 20.65% | 20.03% | 20.19% |
Renter-occupied housing units | 56.98% | 60.32% | 50.30% | 53.13% |
Median age (years) | 32.79 | 32.41 | 35.55 | 34.63 |
Full-time_ year-round civilian employed population 16 years and over | 44.00% | 44.85% | 50.69% | 48.91% |
People 25 + with a bachelor’s degree or Higher | 24.15% | 20.76% | 32.30% | 29.02% |
Renters Spending 30% or More of Household Income on Rent | 50.78% | 51.01% | 48.21% | 49.03% |
Spatiotemporal analysis of overdoses
While time-aggregated data provide important information regarding spatial distribution, the analysis of spatiotemporal trends is more important for guiding community responses. We used a Time-Space Cube analysis to determine distinct spatiotemporal patterns of overdoses. Due to the small numbers of fatal overdoses within each time-space cube, we were unable to conduct analyses of fatal overdoses or overdose mortality. We instead chose to examine total (fatal and nonfatal) overdoses using this approach. Twelve unique patterns were revealed: Consecutive Cold and Hot Spots, Diminishing Cold and Hot Spots, Intensifying Cold and Hot Spots, New Hot Spots, Oscillating Cold Spots, Persistent Cold and Hot Spots, Sporadic Cold and Hot Spots (Fig. 3). Overall, with some exceptions, Hot Spot neighborhoods were localized to areas in central and north-central Milwaukee, while Cold Spot neighborhoods were found along the lakefront and in the western suburban communities in Milwaukee. While many Hot Spots were Consecutive/Persistent, New and Intensifying Hot Spots were identified in north-central and south-central Milwaukee, respectively.
Demographic characteristics of hot and cold spot communities
There were large differences in demographic composition and socioeconomic indicators across communities defined according to spatiotemporal overdose patterns, indicating racialized disparities. Disparities were particularly evident when comparing Consecutive Cold and Hot Spots (Table 3). Consecutive Cold Spot census tracts for overdoses were overwhelmingly (86.81%) White, only 3.52% Black, and 8.62% Hispanic. By contrast, Consecutive Hot Spots census tracts were only 29.87% White, 52.72% Black, and 23,47% Hispanic. In comparison, Milwaukee County as a whole is 55.67% White, 30.64% Black, and 14,63% Hispanic. Communities experiencing sporadic overdoses were disproportionately Black and Hispanic. Intensifying Hot Spots were predominantly (71.48%) Hispanic, while new Hot Spots were primarily Black (30.64%) or White (61.18%). Another notable trend is the disproportionate representation of immigrant populations in communities identified as intensifying hotspots.
Table 3
Socioeconomic, demographic, and health characteristics in Hot and Cold Spot Communities for Overdoses in Milwaukee County.
| Consecutive Hot Spot | Sporadic Hot Spot | Intensifying Hot Spot | New Hot Spot | Consecutive Cold Spot | Milwaukee County |
White | 29.87% | 37.54% | 38.81% | 61.18% | 86.81% | 55.67% |
Black | 52.72% | 41.77% | 9.95% | 28.14% | 3.53% | 30.64% |
Asian | 3.59% | 2.71% | 1.75% | 4.39% | 3.80% | 4.09% |
American Indian and Alaska Native | 0.43% | 0.64% | 1.28% | 0.56% | 0.64% | 0.56% |
Hispanic or Latino (of any race) | 23.47% | 35.11% | 71.48% | 6.68% | 8.62% | 14.63% |
born outside the U.S. | 3.46% | 3.52% | 13.42% | 1.57% | 1.14% | 2.34% |
Households with an Internet Subscription | 68.33% | 69.93% | 65.68% | 76.45% | 87.50% | 79.70% |
Single-Parent Households | 52.97% | 48.01% | 46.59% | 66.40% | 22.27% | 40.75% |
Homeownership | 24.61% | 28.72% | 23.38% | 6.45% | 58.57% | 42.72% |
Median Household Income | $35,812 | $35,219 | $30,824 | $39,001 | $74,446 | $52,486 |
Per Capita Income | $19,736 | $16,873 | $14,233 | $18,121 | $40,543 | $28,823 |
Families Living Below Poverty Level | 29.22% | 30.24% | 36.02% | 40.75% | 4.85% | 16.93% |
Poor Physical Health: 14 + Days | 17.37% | 17.15% | 18.63% | 10.80% | 10.40% | 13.61% |
Self-Reported General Health Assessment: Poor or Fair | 29.35% | 29.53% | 34.74% | 17.00% | 13.29% | 20.82% |
Adults without Health Insurance | 23.09% | 25.89% | 36.62% | 16.40% | 9.75% | 16.22% |
Adults Ever Diagnosed with Depression | 20.09% | 20.38% | 21.21% | 22.25% | 20.17% | 20.19% |
Poor Mental Health: 14 + Days | 19.48% | 18.90% | 20.37% | 21.05% | 12.60% | 16.10% |
Renter-occupied housing units | 68.16% | 65.09% | 74.28% | 93.66% | 38.72% | 53.13% |
Median age (years) | 30.47 | 30.29 | 28.77 | 24.55 | 39.84 | 34.63 |
Percent of Households Divorced | 10.57% | 10.52% | 8.05% | 6.55% | 10.69% | 10.77% |
Fulltime Employment | 36.06% | 40.15% | 40.51% | 18.46% | 60.51% | 48.91% |
Educational Attainment | 9.01% | 6.23% | 4.68% | 13.75% | 31.44% | 19.26% |
Renters Spending 30% or More of Household Income on Rent | 60.63% | 63.46% | 58.86% | 60.40% | 41.43% | 51.55% |
Socioeconomic and health characteristics of hot and cold spot communities
Compared to Consecutive Hot Spots and County-wide numbers, Consecutive Cold Spots were more economically affluent with more than twice the median household and per capita incomes and approximately eight times fewer families living below the poverty line. Employment in Consecutive Cold Spots was 60.51% and educational attainment (bachelor’s degree or higher) was 31.44% compared to 36.06% and 9.01% for Consecutive Hot Spots. Housing stability (homeownership, low renter occupied units, rent burden) was greater in Consecutive Cold Spots relative to Consecutive Hotspots. Moreover, Consecutive Cold Spot communities had better overall, physical and mental health relative to Consecutive Hotspots. Community members in Consecutive Cold Spots were more likely to have health insurance. Mean age in Consecutive Cold Spots was higher than in Consecutive Hot Spots (39.84 vs. 30.47). Overall, indicators in Sporadic and Intensifying Hotspots were similar to those in Consecutive Hot Spots. New Hot Spots tended to have higher educational attainment but lower fulltime employment, more renter-occupied housing units and less home ownership, and different health burden than other Hot Spot areas. Notably, mean age in New Hot Spot communities was much lower (24.55) than all other communities and 10 years less than the county mean (34.63). These results suggest a recent expansion of overdose risk that may include new subpopulations of community members previously less affected by the opioid crisis.