Abstract
Background
Communities often experience relatively quiet periods disrupted by sudden surges of drug overdoses, suggesting that the risk of death can change substantially in a short period. Multiple substances are often involved in overdoses, and it is crucial to understand which are more frequently observed in spikes. This study aimed to investigate the involvement of specific substances in overdose spikes at the state and county levels in Massachusetts.
Methods
We applied a spike detection method to identify daily spikes among 9915 overdose fatalities in Massachusetts from 2020 to 2023. A day was identified as a spike if the number of overdose deaths was over two SDs of the adjusted moving average with a lag of 30 days. We used a general linear mixed model to compare the presence of cocaine, psychostimulants, fentanyl, heroin and prescription opioids in overdose deaths between spike and typical days. Finally, we analysed and compared the results for 8 of the 14 counties in Massachusetts with the highest maximum overdose deaths per day.
Results
At the state level from 2020 to 2023, no substances showed significant differences in spike and typical days. Cocaine was more prevalent in multiple counties during spikes. Fentanyl-involved overdose deaths were common on typical days in two counties. Psychostimulants were more prevalent on spike and typical days depending on the county. Heroin and prescription opioids only showed differences in one county each.
Conclusion
Our analysis shows high heterogeneity of involved substances in overdose death spikes across counties in Massachusetts. This heterogeneity reinforces the need for localised response strategies. This understanding can guide more targeted local resource allocation and inform strategies to improve public health practices, ultimately aiming to reduce the impact of future overdose spikes.
Keywords: Public Health, Epidemiology, Population Surveillance
WHAT IS ALREADY KNOWN ON THIS TOPIC
Overdose spikes account for a noticeable portion of all overdose fatalities.
Substance involvement in overdose deaths varies year by year, but little is known about their involvement during spikes.
WHAT THIS STUDY ADDS
Our study shows that the substances involved in overdose spikes, compared with typical days, can differ significantly by geography.
Demographics of overdose decedents, such as sex, race and age, do not influence the likelihood of a death occurring during a spike.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
We identify which specific substances are more likely involved in spikes, contributing to the development of more context-specific, collaborative community response strategies to enhance prevention, treatment and recovery services.
Introduction
The drug overdose epidemic in the USA is a national public health emergency, with the number of overdose deaths remaining at historically high levels.1 During a short time period, these increases often occur as sudden surges rather than a smooth, gradual ascent. People who use drugs, harm reduction providers, emergency responders and law enforcement officers frequently report that relatively quiet periods are disrupted by spikes in drug overdoses.2 3 Overdose spikes can account for 32% of all fatal overdoses annually,4 meaning that preventing them could have a disproportionate impact on reducing deaths.
Various factors may potentially trigger spikes, including abrupt changes in the illicit drug supply, limited access to behavioural or mental health services, shifts in access to healthcare due to pandemics and natural disasters or shifts in the availability of prescription medications (eg, pain medications).5 6 Prior research has also examined the impact of increased law enforcement activities, identifying a relationship between law enforcement arrests and seizures and subsequent abrupt changes in overdoses.7 However, to understand the phenomenon, it is critical to investigate the characteristics of the spikes, including the timing and geographic location of spikes and drug involvement.
Multiple studies have examined drug involvement in overdose deaths across multiyear periods,8,15 but there is a lack of research on the short-term trends of drug involvement. Differential drug involvement over a period of months or years versus days could point to different types of interventions for preventing deaths. Namely, some drugs may be more or less likely to be present during spikes as opposed to typical days, which could highlight future areas of investigation needed to identify the factors most likely contributing to spikes. For instance, if psychostimulants or cocaine are more present on spike days than typical days, this could suggest that the cocaine supply is particularly risky, or that people who are using these substances need additional harm reduction outreach to make them aware of these heightened risks. This insight could be paired with proactive spike alerts in overdose surveillance systems.16 Continually integrating this information as spikes happen and afterwards can increase stakeholders’ ability to prepare and more efficiently mitigate future deaths.5
In Massachusetts, up to one-fifth of days were identified as spikes in drug overdose deaths each year between 2017 and 2023.4 Furthermore, a separate study from 2005 to 2021 found that drug involvement in fatal overdoses varied substantially across geographical regions of Massachusetts over time.17 Particularly, western Massachusetts recorded significantly higher rates of fatal opioid-related overdoses involving prescription opioids, heroin and stimulants. Synthetic opioid-involved overdose death rates were significantly lower than the state average in western Massachusetts from 2012 to 2015 and the suburbs west of Boston from 2016 to 2021, while ubiquitous hotspots appeared in southeastern Massachusetts and Cape Cod from 2016 through 2021. These observations allude to the importance of analysing which drugs involve spikes per county and how they differ.
The present study investigates the specific drugs involved in overdose spikes in the state of Massachusetts and in the counties within the state that have experienced the highest maximum overdose deaths per day. We focus on cocaine, psychostimulants, fentanyl, heroin and prescription opioids, which are among the most prevalent drugs involved in fatal overdoses in Massachusetts.18
Methods
We examined overdose fatalities involving different substances such as stimulants, opioids and prescription opioids reported in Massachusetts’ vital statistics data. We included all 9915 deaths from January 2020 to December 2023. The Mass General Brigham Institutional Review Board exempted the study from review and waived informed consent because data were not obtained through participant interaction. We followed the Strengthening the Reporting of Observational Studies in Epidemiology reporting guideline.
To identify spikes in overdose deaths, we used a recent spike detection algorithm based on Z-scores, applied at both the state and county levels.4 19 20 Our model uses three key parameters: lag, SD and influence. The lag parameter determines the moving window for calculating the rolling average of historical data. The SD sets the threshold for when the distance between a data point and the rolling average is considered significant. The influence parameter controls the impact of detected spikes on subsequent rolling averages and SD. We set the parameters at a lag of 30 days, an SD of 2 and an influence of 0.3.4 After applying this algorithm, we defined our primary outcome measure as a binary variable indicating whether a death occurred on a spike day or typical day. The spike detection method may vary the days detected as spikes depending on the chosen parameters.4 In the existing literature, the commonly used lags are 7, 30 and 90 days, where also a threshold parameter is defined.4 16 21 22 Given this variability, we conducted a sensitivity analysis to explore how drug involvement in spikes changes with different lags.4
We used ICD-10-CM codes to identify the primary substances of interest: cocaine (T40.5), psychostimulants (T43.6), fentanyl and other synthetic opioids, excluding methadone (T40.4), which we will refer to as fentanyl, heroin (T40.1) and prescription opioids (T40.2 and T40.3). We define all substance use-related deaths in the analysis as overdoses, though we acknowledge that more accurate terms could be used to describe stimulant-related deaths, depending on the death-related circumstances.23 Nonetheless, the database does not offer sufficient information to identify such stimulant-related deaths. Additionally, we included the demographics of the decedent (ie, age, sex, race) to analyse whether these factors were associated with spikes. We also included the year of death to analyse if the prevalence of various substances involved in spikes followed a temporal trend.
To examine differences in substances involved on spike days compared with typical days, we first ran a generalised linear mixed model (GLMM) with a logit link function at the Massachusetts state level, including all observations and exploratory variables described above. We used the county (ie, location of death) as the analytical unit and added random intercepts for each county, where 82.6% of the deaths, between 2020 and 2023, had the same county of residence and 3.9% of the deaths had a different state of residence. We controlled for county-level characteristics, gathered from the 2020 US census,24 including median population age, percentage of white population and population size, as previous evidence suggests varying overdose risks across these subpopulations.25 26 Additionally, we gathered the median yearly household income from the 2022 American Community Survey.27 For example, by 2020, overdose death rates were the highest among young and middle aged (25–54 years old) white men and older black men.28 In 2023, almost half of the opioid-related overdose deaths in Massachusetts were between 25 and 44 years old.29 Studies have also shown that white and minority populations in Massachusetts have experienced distinct drug-specific overdose risks over time.30 For instance, minorities (eg, American Indian, black) had the highest rates of opioid-related overdose deaths in 2023,17 25 29 and non-Hispanic black and Hispanic residents experienced elevated opioid-related overdoses involving stimulants in 2020 and 2021.17 Additionally, in the USA, lower income households experience a higher rate of overdose losses versus higher income households (39% vs 26%).31 32 We included the population size, as rural areas in Massachusetts have higher rates of opioid overdose deaths compared with urban areas due to factors such as the available infrastructure and access to healthcare.28 29 33 We did not include the county-level percentage of the male/female population in our analysis as there is no statistical difference in the sex distribution across counties. The GLMM model requires that observations exist in both comparison groups, that is, typical and spike days. Therefore, we excluded counties that did not have any overdose deaths on typical days. Finally, to examine the heterogeneities of drug involvement across different counties, we stratified GLMMs by substance and added random slopes for each substance of interest.
Results
There was a total of 9915 overdose deaths in Massachusetts from January 2020 to December 2023. Across the counties, over one-half (53.6%) of the deaths occurred on spike days. Table 1 presents sample characteristics. The average age of the deceased was 44.8 years old (SD: 13.1); 71.5% were male and the majority were white (71.3%). Fentanyl and cocaine were the leading substances involved in overdose deaths, with fentanyl present in 84.3% of cases and cocaine in 46.1%. Heroin was the least commonly involved in overdose deaths (5.6%). We also present the p value for each variable, where heroin, the overdose race and ethnicity distribution, and the county distribution show statistical differences between typical and spike days.
Table 1. Overdose death characteristics in Massachusetts from 2020 to 2023, stratified by spike and typical days.
| Variables | Total | Typical days | Spike days | P value |
|---|---|---|---|---|
| Number of overdose deaths | 9915 | 4600 | 5315 | |
| Drugs (n, %)* | ||||
| Cocaine | 4569 (46.1) | 2149 (46.7) | 2420 (45.5) | 0.246 |
| Psychostimulants | 807 (8.1) | 365 (7.9) | 442 (8.3) | 0.512 |
| Fentanyl | 8361 (84.3) | 3921 (85.2) | 4440 (83.5) | 0.022 |
| Heroin | 554 (5.6) | 209 (4.5) | 345 (6.5) | <0.001 |
| Prescription opioids | 1340 (13.5) | 615 (13.4) | 725 (13.6) | 0.716 |
| Demographics | ||||
| Age (mean (SD)) | 44.8 (13.1) | 44.9 (13.0) | 44.7 (13.1) | 0.584 |
| Male (n, %) | 7092 (71.5) | 3316 (72.1) | 3776 (71.0) | 0.285 |
| Race and ethnicity (n, %) | <0.001 | |||
| Black | 1091 (11.0) | 553 (12.0) | 538 (10.1) | |
| Hispanic | 1482 (14.9) | 767 (16.7) | 715 (13.5) | |
| White | 7065 (71.3) | 3163 (68.8) | 3902 (73.4) | |
| Other | 277 (2.8) | 117 (2.5) | 160 (3.0) | |
| County (n, %) | <0.001 | |||
| Barnstable | 328 (3.3) | 0 (0.0) | 328 (6.2) | |
| Berkshire | 246 (2.5) | 0 (0.0) | 246 (4.6) | |
| Bristol | 1126 (11.4) | 514 (11.2) | 612 (11.5) | |
| Dukes | 20 (0.2) | 0 (0.0) | 20 (0.4) | |
| Essex | 1091 (11.0) | 541 (11.8) | 550 (10.3) | |
| Franklin | 110 (1.1) | 0 (0.0) | 110 (2.1) | |
| Hampden | 1006 (10.1) | 487 (10.6) | 519 (9.8) | |
| Hampshire | 133 (1.3) | 0 (0.0) | 133 (2.5) | |
| Middlesex | 1512 (15.2) | 789 (17.2) | 723 (13.6) | |
| Nantucket | 13 (0.1) | 0 (0.0) | 13 (0.2) | |
| Norfolk | 599 (6.0) | 228 (5) | 371 (7) | |
| Plymouth | 694 (7.0) | 334 (7.3) | 360 (6.8) | |
| Suffolk | 1736 (17.5) | 1054 (22.9) | 682 (12.8) | |
| Worcester | 1301 (13.1) | 653 (14.2) | 648 (12.2) |
Data: Overdose fatalities reported in Massachusetts’ vital statistics data from January 2020 to December 2023.
Note that multiple drugs could be involved in an overdose death; therefore, the sum of the percentages could exceed 1.
There were large heterogeneities in overdose mortality rates across the counties, as shown in figure 1. In 6 of Massachusetts’ 14 counties (Barnstable, Berkshire, Dukes, Hampshire, Franklin and Nantucket), the maximum number of overdose deaths recorded on any given day did not exceed three (table 2). For these counties, the spike detection algorithm identified every day that had an overdose death as a spike. Therefore, we could not analyse if specific drugs differ between typical and spike days. For this same reason, we did not apply the spike detection algorithm at a subcounty level. Thus, for the county-level analysis, we only included and presented the results of the remaining eight counties (ie, Bristol, Essex, Hampden, Middlesex, Norfolk, Plymouth, Suffolk and Worcester) that exhibited larger numbers of overdose deaths in a day. We present the number of days and the percentage of days identified as spikes in online supplemental table A1. These counties tend to be outside of the west and south periphery of the state, as shown in figure 1. Notably, Suffolk County had one of the highest yearly average overdose mortality rates (54.4 deaths per 100 000 people from 2020 to 2023), which is the home to Boston, the largest city in Massachusetts, and has the largest percentage of non-white populations (55.8%) in the state. Yearly overdose death rate per county is found in online supplemental table A2.
Figure 1. Massachusetts counties with their yearly average overdose mortality rate (per 100 000 people), between 2020 and 2023. Data: Overdose fatalities reported in Massachusetts’ vital statistics data from January 2020 to December 2023.
Table 2. County-level characteristics.
| Counties | Maximum overdose deaths per day* | Population size (thousands)† | Median age† | Non-white population† | Median household income(thousands $)‡ |
|---|---|---|---|---|---|
| Included in the county-level analysis | |||||
| Bristol | 4 | 579.2 | 41.5 | 23.1% | 103.3 |
| Essex | 5 | 809.8 | 41.4 | 33.8% | 118.3 |
| Hampden | 5 | 465.8 | 40.2 | 40.2% | 85.0 |
| Middlesex | 5 | 1632.0 | 38.6 | 33.5% | 153.2 |
| Norfolk | 5 | 726.0 | 41.2 | 29.4% | 155.9 |
| Plymouth | 4 | 530.8 | 43.5 | 23.3% | 128.7 |
| Suffolk | 6 | 797.9 | 32.3 | 55.8% | 102.7 |
| Worcester | 6 | 862.1 | 40.2 | 30.1% | 112.5 |
| Not included in the county-level analysis | |||||
| Barnstable | 3 | 229.0 | 55.0 | 15.0% | 118.6 |
| Berkshire | 3 | 129.0 | 47.8 | 15.3% | 92.5 |
| Dukes | 1 | 20.6 | 49.3 | 21.2% | 121.4 |
| Franklin | 2 | 71.0 | 47 | 13.5% | 92.6 |
| Hampshire | 2 | 162.3 | 38.3 | 21.4% | 113.4 |
| Nantucket | 1 | 14.3 | 40.7 | 30.6% | 144.2 |
Overdose fatalities reported in Massachusetts’ vital statistics data from January 2020 to December 2023.
2020 US Census.
2022 American Community Survey.
Table 2 presents the county-level characteristics. The eight included counties had a higher non-white population, ranging from 23.1% to 55.8%, and are metropolitan areas with populations higher than 250 000.34
The state-level analysis using GLMM is shown in table 3. Substances with an adjusted OR above 1 suggest that these substances were associated with the occurrence of the spikes. On the other hand, an adjusted OR below 1 indicates that the substance was more prevalent on typical days, suggesting other factors might have contributed to the spikes. Overall, at the state level, there were no statistical differences in the drugs involved in deaths on spike days compared with typical days. Nonetheless, there were significant demographic differences, with men being less likely to die on spike days. Factors including race and ethnicity, age, county-level characteristics and year were not significantly associated with being spikes.
Table 3. State-level analysis of factors associated with overdose deaths on spike days† using multivariable GLMM.
| Explanatory variables | Adjusted OR | 95% CI |
|---|---|---|
| Drugs | ||
| Cocaine (yes vs no) | 1.04 | (0.93 to 1.17) |
| Psychostimulants (yes vs no) | 1.04 | (0.85 to 1.27) |
| Fentanyl (yes vs no) | 0.98 | (0.84 to 1.14) |
| Heroin (yes vs no) | 1.06 | (0.83 to 1.34) |
| RX opioids (yes vs no) | 0.94 | (0.80 to 1.11) |
| Demographics | ||
| Sex (male vs female) | 0.86 | (0.77 to 0.97)* |
| Race and ethnicity: | ||
| Black vs white | 1.12 | (0.93 to 1.34) |
| Hispanic vs white | 0.86 | (0.73 to 1.02) |
| Other vs white | 0.74 | (0.51 to 1.07) |
| Age, per year increase | 0.99 | (0.64 to 1.52) |
| Census | ||
| Population size (per one million increase) | 1.17 | (0.91 to 1.49) |
| Median income (per ten thousand increase) | 0.80 | (0.52 to 1.24) |
| Race (white vs others) | 1.00 | (0.99 to 1.01) |
| Median age, per year increase | 1.84 | (0.12 to 2.82) |
| Year | 1.00 | (0.95 to 1.05) |
P value<0.05.
A day was identified as a spike if the number of overdose deaths was over two SD of the adjusted moving average with a lag of 30 days.
GLMM, generalised linear mixed model.
Results of the county-level analysis are presented in figure 2. There were notable variations between the spike and typical days across counties and substances. Cocaine-related deaths were more prevalent on spike days for Bristol and Hampden, two of the top three counties with higher overdose mortality rates and the lowest median household incomes. Fentanyl-related deaths were more prevalent in typical days for Essex and Plymouth, two of the top three counties with the highest median age. Heroin-related deaths were more prevalent on spike days only in Plymouth, and prescription opioids were significantly more involved on spike days for Norfolk. Psychostimulants were the only drug that had contrasting outcomes in different counties, being more prevalent on typical days for Norfolk and more prevalent on spike days for Suffolk. Additionally, we present the results by county, year and substance in online supplemental figure A1. The year-by-year results are largely consistent with the primary analysis based on combined data across all years.
Figure 2. Substance involvement in overdose deaths comparing spikes and typical days by county between 2020 and 2023. (A) Massachusetts map indicating counties and substances with significant difference; (B) adjusted OR and 95% CIs for each substance. Note 1: Black line indicates a significant difference between spikes and typical days. Note 2: A day was identified as a spike if the number of overdose deaths was over two SD of the adjusted moving average with a lag of 30 days.
The demographics of the decedents were not associated with the likelihood of deaths occurring during spikes, as detailed in online supplemental table A3. Therefore, we did not identify certain demographics that had a higher chance of appearing in a spike versus a typical day. Regarding county-level characteristics, we only found that a higher median household income can affect the likelihood of a fentanyl-related death occurring on a spike day. Additionally, we also found that a larger white population increases the likelihood of a prescription opioids-related death occurring on a spike day.
Finally, our sensitivity analysis reveals that varying the temporal lag has little impact on the overall findings, with results largely consistent with the primary analysis. Differences in statistical significance were mainly observed in cases where the effect size was small (adjusted OR close to 1, see online supplemental table A4). For example, with all alternative lags in the sensitivity analysis, cocaine was not more prevalent on spike days in Bristol County, compared with the primary analysis. On the other hand, when the lag was 7 days (meaning the rolling average was calculated over that time period), cocaine was more prevalent on spike days in Norfolk County and fentanyl was more prevalent during spike days in Hampden County. The difference in behaviours shows the importance of engaging with local stakeholders to identify the lag that best suits their requirements. When the lag was 90 days, cocaine was more prevalent on typical days in Worcester. In Essex and Plymouth Counties, fentanyl’s prevalence was mixed based on varying lags. For instance, when the lag is 7–30 days, fentanyl is more prevalent in typical days for Essex, while there is no difference when the lag is 60–90 days. Finally, although statistical significance varied, the adjusted ORs for prescription opioids in the sensitivity analysis were similar compared with those in the primary analysis. Additionally, we examined whether drug involvement varied across demographic subgroups by testing interaction terms in the multivariable models. We did not identify statistically significant interaction terms, even with different lag values.
Discussion
Our study underscores the importance of investigating the characteristics of drug overdoses during spikes and at a local level (ie, the county level). Using daily fatal overdose data from Massachusetts, we demonstrated that while state-level data showed no variation in substances involved between spike and typical days, significant differences emerged at the county level, varying by the substance and county.
Notably, cocaine-related deaths were more prevalent on spike days than other drugs in two of the analysed Massachusetts counties with the top highest overdose mortality rate and lowest median household. Having these results implies local stakeholders can better focus their efforts on reducing spike occurrence. This observation aligns with the broader trends in rising cocaine-involved overdoses,35 a key characteristic of the current (ie, fourth) wave of the overdose epidemic.36 Throughout our study period in Massachusetts, cocaine was the second most prevalent drug in opioid-related overdose deaths, following fentanyl.17 33 Additionally, fentanyl was consistently the most involved substance in overdose deaths for all counties, and it was more prevalent on typical days in Essex and Plymouth. These results might suggest that between 2020 and 2023, the overall involvement in overdose deaths was not correlated with the occurrence of spikes. A potential factor may be the unpredictable dangerous effects of fentanyl due to its high potency and presence in other substances, often unknown to users.37 Even when aware, its potency makes it difficult for individuals to safely gauge their dosage.38 Furthermore, psychostimulants showed contrasting trends, being more prevalent on typical days in Norfolk and more prevalent on spike days in Plymouth. Finally, heroin and prescription opioids only showed differences in Plymouth and Norfolk, respectively. Nonetheless, these results highlight the presence of county-level heterogeneity, which may stem from differences in drug availability, access to care and the quality of healthcare infrastructure.5
Collaborating with local authorities and stakeholders to develop strategic, context-specific community response efforts is crucial. Engaging with local stakeholders to create a formal network of individuals and organisations can enhance harm reduction services, treatment recovery and other services to respond to spikes.5 16 An important part of this collaboration involves defining what constitutes a spike for effective resource allocation, as we observed that varying the temporal lag of the spike detection algorithm impacts the days identified as spikes4 and, consequently, what drugs are involved in overdose spike days.
This analysis has several limitations. To begin, the spike detection method we employed faces its own limitations, particularly since there are multiple definitions for what constitutes a spike. Depending on the applied method, the drugs involved during spikes can vary, as shown in the sensitivity analysis. Additionally, due to the nature of the spike detection algorithm used, a more granular (eg, subcounty) analysis was not performed. In the future work, alternative approaches could be explored to identify spikes, with model selection more tailored to local needs, such as data availability, overdose trends and stakeholders’ requirements. Second, we accounted for various factors including county-level demographic characteristics; nonetheless, there could be unadjusted confounders that could explain these differences, such as specific geographic location within the county, public infrastructure and access to treatment and healthcare. Third, we analysed all fatal overdoses, including unintentional suicide, homicide and child overdoses, which may need different intervention strategies. Fourth, we do not account for xylazine in our analysis due to current data availability. Given recent reports of increasing xylazine in the drug supply and its contribution to overdose deaths,33 39 including this drug in future analyses, can provide additional insights about spikes. Fifth, while medical examiner counties are associated with better identification of opioids, it is possible that even Massachusetts' centralised state medical examiner system40 might not fully account for all substances that contributed to a decedent’s death on death certificates. Sixth, our model does not consider the interactions between substances. 52% of deaths in Massachusetts involved more than one substance, and fentanyl was involved in 85% of all overdose deaths. Nonetheless, including interactions increases the model’s dimensionality, which can lead to issues such as multicollinearity and overparameterisation. Future analyses could focus on interactions between substances that provide additional insights into mixtures that influence spikes. Seventh, our analysis used the recorded place of death as the geographic location of overdose events. We acknowledge that this may introduce bias, particularly when deaths occurred in hospitals outside the decedents’ county of residence.41 However, data indicated that, in the majority of cases (83%), the county of death matched the county of residence, and only about 4% of decedents resided out of state. Therefore, we anticipated that the influence of misclassification bias was limited. Finally, our results are specific to Massachusetts, limiting the generalisability to other geographical areas.
To mitigate future overdose deaths, it is essential to extend our efforts beyond investigating the drugs involved in spikes. Obtaining a better understanding of the underlying causes of these spikes is crucial. This could involve leveraging additional datasets, including drug-related arrests or seizures and additional demographic information to understand the differences between counties. Our study underscores the importance of examining the drugs involved in drug overdose spikes, as these are not necessarily the same as drugs most often involved in overdoses (ie, fentanyl). The significant heterogeneity we observed at the county level highlights the critical need for localised response strategies.
Supplementary material
Footnotes
Funding: This work was supported in part by grant R01CE003358 from the US Centers for Disease Control and Prevention. This article reflects the views of the authors and does not represent the views or policies of the CDC or the US Department of Health and Human Services.
Provenance and peer review: Not commissioned; externally peer-reviewed.
Patient consent for publication: Not applicable.
Data availability free text: Raw data used in this analysis can be obtained through a request to the Massachusetts Department of Public Health. Analysis codes are available in the supplementary materials of this article.
Map disclaimer: The depiction of boundaries on this map does not imply the expression of any opinion whatsoever on the part of BMJ (or any member of its group) concerning the legal status of any country, territory, jurisdiction or area or of its authorities. This map is provided without any warranty of any kind, either express or implied.
Patient and public involvement: Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Ethics approval: The Mass General Brigham Institutional Review Board exempted the study from review and waived informed consent because data were not obtained through participant interaction.
Data availability statement
Data may be obtained from a third party and are not publicly available.
References
- 1.Thombs RP, Thombs DL, Jorgenson AK, et al. What Is Driving the Drug Overdose Epidemic in the United States? J Health Soc Behav. 2020;61:275–89. doi: 10.1177/0022146520939514. [DOI] [PubMed] [Google Scholar]
- 2.Cuda A. Emergency responders fighting hard against overdose deaths. 2016. [28-May-2024]. https://www.stratfordems.org/2016/02/07/emergency-responders-fighting-hard-overdose-deaths/ Available. Accessed.
- 3.Smiley-McDonald HM, Attaway PR, Richardson NJ, et al. Perspectives from law enforcement officers who respond to overdose calls for service and administer naloxone. Health Justice. 2022;10:1–13. doi: 10.1186/S40352-022-00172-Y/TABLES/1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Lee H, Otero-Leon D, Dong H, et al. Uncovering Patterns in Overdose Deaths: An Analysis of Spike Identification in Fatal Drug Overdose Data in Massachusetts, 2017-2023. Public Health Rep. 2024:333549241299613. doi: 10.1177/00333549241299613. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Overdose spike response framework for communities and local health departments. 2021. [27-Apr-2024]. https://www.naccho.org/uploads/full-width-images/OVERDOSE-SPIKE-RESPONSE-FRAMEWORK-FOR-COMMUNITIES-LHDS-1.pdf Available. Accessed.
- 6.Friedman J, Gjersing L. Increases in drug overdose deaths in Norway and the United States during the COVID-19 pandemic. Scand J Public Health. 2023;51:53–7. doi: 10.1177/14034948221075025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Ray B, Korzeniewski SJ, Mohler G, et al. Spatiotemporal Analysis Exploring the Effect of Law Enforcement Drug Market Disruptions on Overdose, Indianapolis, Indiana, 2020–2021. Am J Public Health. 2023;113:750–8. doi: 10.2105/AJPH.2023.307291. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Dai Z, Abate MA, Groth CP, et al. Fentanyl and other opioid involvement in methamphetamine-related deaths. Am J Drug Alcohol Abuse. 2022;48:226–34. doi: 10.1080/00952990.2021.1981919. [DOI] [PubMed] [Google Scholar]
- 9.Love JS, Levine M, Aldy K, et al. Opioid overdoses involving xylazine in emergency department patients: a multicenter study. Clin Toxicol (Phila) 2023;61:173–80. doi: 10.1080/15563650.2022.2159427. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Moses J, Korona-Bailey J, Mukhopadhyay S. Exploring trends in benzodiazepine-positive fatal drug overdoses in Tennessee, 2019-2021. Ann Med. 2023;55:2287194. doi: 10.1080/07853890.2023.2287194. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Han B, Compton WM, Jones CM, et al. Methamphetamine Use, Methamphetamine Use Disorder, and Associated Overdose Deaths Among US Adults. JAMA Psychiatry. 2021;78:1329–42. doi: 10.1001/jamapsychiatry.2021.2588. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Hall KF, Korona-Bailey J, Mukhopadhyay S. Phentermine involvement in fatal drug overdoses in Tennessee, 2019-2022. Int J Drug Policy. 2023;119:104151. doi: 10.1016/j.drugpo.2023.104151. [DOI] [PubMed] [Google Scholar]
- 13.Ford JA. Prescription Opioid Misuse Among Adolescents. Pediatr Clin North Am. 2019;66:1099–108. doi: 10.1016/j.pcl.2019.08.005. [DOI] [PubMed] [Google Scholar]
- 14.Spencer MR, Miniño AM, Garnett MF. Co-involvement of opioids in drug overdose deaths involving cocaine and psychostimulants, 2011-2021 key findings data from the national vital statistics system. 2011. https://www.cdc.gov/nchs/products/index.htm Available. [PubMed]
- 15.Griffith J, Chambers LC, Hallowell BD, et al. Examination of the Accuracy of Existing Overdose Surveillance Systems. JAMA Netw Open. 2023;6:e2320789. doi: 10.1001/jamanetworkopen.2023.20789. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Lasher L, Hallowell BD, Chambers LC, et al. Using Timely Overdose Data to Address a Spike in Nonfatal Overdoses and Inform a Coordinated Community-Level Response in Rhode Island, 2019. Public Health Rep . 2021;136:24S–30S. doi: 10.1177/00333549211012407. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Stopka TJ, Larochelle MR, Li X, et al. Opioid-related mortality: Dynamic temporal and spatial trends by drug type and demographic subpopulations, Massachusetts, 2005-2021. Drug Alcohol Depend. 2023;246:109836. doi: 10.1016/j.drugalcdep.2023.109836. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Hedegaard H, Bastian BA, Trinidad JP, et al. Drugs Most Frequently Involved in Drug Overdose Deaths: United States, 2011-2016. Natl Vital Stat Rep. 2018;67:1–14. [PubMed] [Google Scholar]
- 19.Rykov Y, Thach TQ, Bojic I, et al. Digital Biomarkers for Depression Screening With Wearable Devices: Cross-sectional Study With Machine Learning Modeling. JMIR Mhealth Uhealth. 2021;9:e24872. doi: 10.2196/24872. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Kim M, Hargrove LJ. A gait phase prediction model trained on benchmark datasets for evaluating a controller for prosthetic legs. Front Neurorobot. 2023;16:1–13. doi: 10.3389/fnbot.2022.1064313. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Alter A, Yeager C. ODMAP; 2020. [20-Feb-2025]. The consequences of COVID-19 on the epidemic: overdoses are increasing.https://www.odmap.org:4443/Content/docs/news/2020/ODMAP-Report-May-2020.pdf Available. accessed. [Google Scholar]
- 22.Mukherjee S, Becker N, Weeks W, et al. Using internet search trends to forecast short term drug overdose deaths: a case study on connecticut. Proceedings - 19th IEEE international conference on machine learning and applications, ICMLA 2020, institute of electrical and electronics engineers inc; 2020. pp. 1332–9. [Google Scholar]
- 23.Black F, McMahan VM, Chang YSG, et al. Thematic analysis of medical examiner narratives to understand the socio-spatial context, recency of drug use, and likely mechanism of stimulant toxicity deaths. Drug Alcohol Depend. 2025;272:112700. doi: 10.1016/j.drugalcdep.2025.112700. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.U.S. Census Bureau . U.S. Census Bureau; [05-Feb-2025]. 2020 census demographic profile.https://www.census.gov/data/tables/2023/dec/2020-census-demographic-profile.html Available. Accessed. [Google Scholar]
- 25.DiGennaro C, Garcia GGP, Stringfellow EJ, et al. Changes in characteristics of drug overdose death trends during the COVID-19 pandemic. Int J Drug Policy. 2021;98:103392. doi: 10.1016/j.drugpo.2021.103392. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Garcia G-GP, Stringfellow EJ, DiGennaro C, et al. Opioid overdose decedent characteristics during COVID-19. Ann Med. 2022;54:1081–8. doi: 10.1080/07853890.2022.2067350. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.US Census Bureau American community survey (ACS) [14-May-2025]. https://www.census.gov/programs-surveys/acs.html Available. Accessed.
- 28.Monnat SM. Demographic and Geographic Variation in Fatal Drug Overdoses in the United States, 1999-2020. Ann Am Acad Pol Soc Sci. 2022;703:50–78. doi: 10.1177/00027162231154348. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Conti K. Massachusetts Department of Public Health; 2024. [09-Oct-2024]. DPH report massachusetts opioid-related overdose deaths decreased 10 percent in 2023.https://www.mass.gov/news/dph-report-massachusetts-opioid-related-overdose-deaths-decreased-10-percent-in-2023 Available. Accessed. [Google Scholar]
- 30.Liao C-Y, Garcia G-G, DiGennaro C, et al. Racial Disparities in Opioid Overdose Deaths in Massachusetts. JAMA Netw Open. 2022;5:e229081. doi: 10.1001/jamanetworkopen.2022.9081. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Kennedy-Hendricks A, Ettman CK, Gollust SE, et al. Experience of Personal Loss Due to Drug Overdose Among US Adults. JAMA Health Forum. 2024;5:e241262. doi: 10.1001/jamahealthforum.2024.1262. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Pustz J, Srinivasan S, Larochelle MR, et al. Relationships between places of residence, injury, and death: Spatial and statistical analysis of fatal opioid overdoses across Massachusetts. Spat Spatiotemporal Epidemiol. 2022;43:100541. doi: 10.1016/j.sste.2022.100541. [DOI] [PubMed] [Google Scholar]
- 33.Data brief: opioid-related overdose deaths among massachusetts residents. 2024. [05-Feb-2025]. https://www.mass.gov/doc/opioid-related-overdose-deaths-among-ma-residents-june-2024-0 Available. Accessed.
- 34.U.S. Department of Agriculture, Economic Research Service Rural-urban continuum codes. [26-Aug-2024];2024 https://www.ers.usda.gov/data-products/rural-urban-continuum-codes/documentation/ Available. accessed.
- 35.Spencer MR, Miniño AM, Garnett MF. Co-involvement of Opioids in Drug Overdose Deaths Involving Cocaine and Psychostimulants, 2011-2021. NCHS Data Brief. 2023:1–8. [PubMed] [Google Scholar]
- 36.Ciccarone D. The rise of illicit fentanyls, stimulants and the fourth wave of the opioid overdose crisis. Curr Opin Psychiatry. 2021;34:344–50. doi: 10.1097/YCO.0000000000000717. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.DEA . Drug Enforcement Administration; 2022. Drug fact sheet: fentanyl. [Google Scholar]
- 38.The rising menace: street drugs laced with fentanyl. [3-Sep-2024]. https://texaslawhelp.org/article/the-rising-menace-street-drugs-laced-with-fentanyl Available. Accessed.
- 39.DEA The growing threat of xylazine and its mixture with illicit drugs. 2022. [23-Feb-2025]. https://www.dea.gov/sites/default/files/2022-12/The%20Growing%20Threat%20of%20Xylazine%20and%20its%20Mixture%20with%20Illicit%20Drugs.pdf Available. Accessed.
- 40.CDC: Public Health Law . CDC; 2024. [27-Feb-2025]. Massachusetts coroner/medical examiner laws.https://www.cdc.gov/phlp/php/coroner/massachusetts.html Available. Accessed. [Google Scholar]
- 41.Gutkind S, Marziali ME, Bruzelius E, et al. Misclassification of opioid-involvement in drug-related overdose deaths in the United States: A scoping review. Ann Epidemiol. 2025;102:8–22. doi: 10.1016/j.annepidem.2024.12.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data may be obtained from a third party and are not publicly available.