Unintentional Fentanyl Overdoses in New Hampshire: An NDEWS HotSpot Analysis

Marcella H. Sorg, Ph.D.
Jamie A. Wren, MPH
Margaret Chase Smith Policy Center
University of Maine
Kathleen Stewart, Ph.D.
Yanjia Cao
Center for Geospatial Information Science
University of Maryland

To download this report, scroll to Table of Contents.


Through HotSpot studies, National Drug Early Warning System (NDEWS) Coordinating Center staff can launch rapid, limited onsite investigations of an important drug outbreak in a specific locality. Each HotSpot study includes a 3–5-day site visit by NDEWS scientists to gather additional data and descriptive information that can be used to help interpret the information collected and analyzed by NDEWS staff prior to the site visit. NDEWS also convenes a Planning Committee composed of NDEWS staff and local experts who help to plan the site visit and arrange meetings with persons who can provide the most useful information. The study team is multidisciplinary and may be composed of scientists, public health practitioners, and law enforcement personnel, as the situation requires. 
NDEWS identified New Hampshire as a HotSpot location because of the rapid increase in both drug overdose deaths and opioid-related emergency department (ED) visits. Under commission from NDEWS, the University of Maine Rural Drug and Alcohol Research Program completed a rapid report on mortality patterns in mid-2016. Later that same year, we embarked on the current, expanded study. In the preliminary 2016 report, we showed that drug overdose deaths in New Hampshire increased by 1,629% between 2010 and 2015, largely as a result of fentanyl. The county most impacted by these overdose deaths is Hillsborough County in southern New Hampshire, home to 30.6% of the state’s population, where 43.6% of the fentanyl deaths occurred. 
The Office of the Chief Medical Examiner (OCME) is responsible for determining the cause and manner of death for all drug overdoses statewide. The Chief Medical Examiner (CME) and Deputy Chief Medical Examiners (DCME) perform or oversee all death investigations for the OCME as part of a centralized system of state medical examiners. Assistant Deputy Medical Examiners (ADMEs) are trained death investigators for the OCME system. They respond to death scenes, collect information about the circumstances of death, perform an external examination of the deceased, and document their findings. The system has struggled to address the dramatic increase in case load without sacrificing the quality of toxicological analysis and overall death investigation. 
In this analysis, we expanded our preliminary 2016 study to focus on deaths caused by fentanyl and fentanyl analogs. We used medical examiner files, including the death investigation report, the medical examiner report and death certificate information, the autopsy report, and the toxicology findings. By combining these data sources, this report looks beyond the cause and manner of death to the details of these death events, at the decedents’ interrelated medical history and social circumstances, as well as at their toxicology and autopsy findings. 
In this study, 541 individuals were identified who died of fentanyl-induced overdose in New Hampshire from January 1, 2015 to September 30, 2016. This dataset was used for the geographic information systems (GIS) location analysis. Cases were included in the final analytical sample if fentanyl, a fentanyl metabolite, or a fentanyl analog were mentioned as a cause of death, but only if the cases were complete (included toxicology and complete medical examiner and death investigator report) and the manner of death was accidental. A sample comprising 505 decedents resulted. These are “occurrent” rather than “resident” deaths, i.e., just those who died in the state of New Hampshire, even if they were legal residents elsewhere. Of the 505 cases, 189 had an autopsy. We sent 136 urine samples originally obtained at autopsy to NDEWS Coordinating Center staff who asked the toxicology lab of the Armed Forces Medical Examiner System (AFMES) to conduct comprehensive drug screening tests.
We also worked with NDEWS collaborator Dr. Kathleen Stewart from the University of Maryland Center for Geospatial Information Science (CGIS) to analyze GIS-based maps exploring the spatial relationships between the location of residence, injury, death, and other sociodemographic characteristics. 
Spatial Relationships (n = 541)
New Hampshire is a small, largely rural state with a population of approximately 1.3 million. A substantial concentration of this population is clustered in the southeastern urban hub of Manchester/Nashua in Hillsborough County. Fentanyl deaths are also concentrated in Hillsborough County, which includes 30.6% of the state population. This county was the residence of 39.6% of the decedents who died in New Hampshire from fentanyl, a fentanyl metabolite, or a fentanyl analog in 2015–2016 and the location of where 43.6% of the deaths in the state occurred. Nearly all (96%) of the decedents were New Hampshire residents. Overall, deaths were more likely to occur in an urban rather than in a rural, small town or micropolitan township, and they were significantly more likely to occur in townships with greater population density. Furthermore, 70.3% of fatal overdoses occurred within 5 miles of an interstate highway, and the fatal drug use [1] occurred at the decedent’s residence in 68.6% of deaths.[2]  Relationships between the distribution of decedents’ residences and sociodemographic factors of employment, median income, and percent with college education were not statistically significant.
Decedent Characteristics (n = 505)
The population of decedents is younger and consists of proportionally more males than the census population of the state. The mean age of decedents is 36.2, with males slightly younger than females, and outnumbering them 2 to 1. Mirroring the census population, 95.0% of the study population is White. Compared with the state census population, fewer in this population had a high school diploma or general equivalency diploma (GED; 60.8% vs. 92.3%), and fewer were veterans (3.8% vs. 10.3%). Among the 490 decedents for whom household composition was known, 20.0% lived alone and 4.5% were homeless.
According to death investigator reports, the study population includes a majority (63.4%) with a history of opioid abuse, 9.5% with a history of chronic pain, 12.7% with a history of previous overdose, and 14.9% with an opioid prescription during the previous 12 months. Of those with an opioid abuse history, 89.7% were reported as injection drug users, 81.6% had been using heroin, 7.2% had used a combination of heroin and pills, and only 4.7% had used pills alone.
Female decedents were much more likely to use medical services. They are significantly more likely to have had a prescription for an opioid in the previous 12 months (25.2% vs. 11.9% for males). They are also significantly more likely to have an EMS response (71.2% vs. 59.4%) and significantly more likely to have naloxone present in their toxicology findings (25.2% vs. 15.2%) compared with males.
Decedents 40 years old or older are less likely to be male than those younger than 40 (74.1% vs. 80.1%). They are significantly more likely than those younger than 40 to be living alone (28.7% vs. 14.5%) and significantly more likely to live in a rural area (32.8% vs. 22.7%). Older decedents are significantly less likely to have a history of opioid abuse (50.6% vs. 70.1%) but significantly more likely to have a history of chronic pain (70.8% vs. 29.2%) and significantly more likely to have pharmaceutical opioids present in their toxicology findings (32.2% vs. 17.2%). Older decedents are significantly less likely to have received an EMS response (55.7% versus 65.3%).
Decedents with a known opioid prescription during the previous 12 months numbered 75, 14.9% of the primary study population. Of these, 40.0% had pharmaceutical opioids in their toxicology, which is significantly higher than those with no opioid prescription history, 19.3%. Approximately one third (32.0%) of those with an opioid prescription also had a history of chronic pain, which is significantly more than those lacking an opioid prescription, 5.6%. Having a “rapid” overdose was significantly less common among decedents with a recent opioid prescription, 6.7% compared with 19.8%, but they were no more likely to have naloxone in their toxicology findings.
There are 48 decedents with a reported history of chronic pain. This small subpopulation was significantly different from others only in terms of whether they also had an opioid prescribed within the previous 12 months (50.0% vs. 11.2%) and whether they had a pharmaceutical opioid present in their toxicology findings (41.7% vs. 20.4%).
Most decedents, 63.4%, had a history of opioid abuse. Within this subpopulation, only 8.8% also had a history of chronic pain, 15.9% had a recent prescription for an opioid, and 22.8% had pharmaceutical opioids present in their toxicology. Those with an opioid abuse history were significantly less likely to be alone when they died (53.8% vs. 46.3%), and significantly less likely to have an EMS response (65.6 vs. 34.4%), but they were no more likely to have naloxone present in their toxicology (16.9%). They were, however, significantly more likely to have a history of previous overdose (17.5% vs. 4.3%) and significantly more likely to have been recently released (past two weeks) from jail or substance abuse treatment (6.6% vs. 1.6%). There was no significant difference between those with an opioid abuse history and others in the study population in terms of the percent unemployed, disabled, living alone, or homeless. 
Death Event Characteristics (n = 505)
Most (62.6%) victims were reported to be alone when they took the fatal overdose, slightly fewer when they died (57.6%). In 13.9% of cases, there was a witness who was aware of the ingestion. The decedent was found decomposed in 6.3% of cases.
Most victims ingested the drug(s) at their own residence (69.9%). Many others were at another person’s residence when they took the drug(s) (17.6%) and died there (13.3%). Seventeen percent of victims died in the emergency department or hospital, regardless of where they ingested the drugs.
911 was called in 92.5% of cases, and EMS responded in 62.0%. In 11.7% of cases, the ADME reported that naloxone was administered. EMS administered 79.7% of the ADME-reported naloxone administrations. A friend, relative, or girlfriend/boyfriend administered it in 0.8% of cases. Within the toxicology data, there were 17.4% cases positive for naloxone, 27 more than reported by the ADME, which means they were probably administered in the emergency department.
Autopsies are routinely done for cases where the likely cause of death is unclear or for potential overdoses where other causes of death cannot be ruled out. Autopsies are also done in cases that may result in a prosecution, which is a decision made by the Attorney General’s Overdose Task Force. Within the autopsy study population (n = 189), 37.4% of decedents received an autopsy. Surprisingly, autopsies were significantly more likely to be done on cases with female decedents and on decedents who are 40 years old and older. Decedents who received an autopsy were significantly less likely to have a history of opioid abuse. 
Toxicology Findings
Polydrug complexity is an important feature of toxicology findings for this population. Within the medical examiner’s toxicology findings (most are blood specimens), there are 114 drugs and metabolites identified. Individual toxicology tests ranged widely from just one substance reported (9.7%) to 19 substances in one decedent. Fentanyl was found in 98.4% and fentanyl analogs in 11.7% of cases; these cases overlap. The mean number of parent [3] drugs was 6.23. Key co-intoxicant and potentially synergistic drugs present include heroin/morphine (20.6%), non-fentanyl opioids (34.5%), benzodiazepines (27.5%), cocaine (31.1%), and alcohol (32.9%).
Postmortem levels of fentanyl confirmed in our sample range widely from 0.75 to 113.00 ng/mL, with a mean of 9.96. We compared the distributions of fentanyl levels for cases where fentanyl was the only drug found with cases with key co-intoxicants (opioids, benzodiazepines, or alcohol); the distributions were not statistically different. The distribution of fentanyl levels among the subgroup of decedents who reportedly had a “rapid” overdose was not statistically different from other decedents.
We collaborated with NDEWS to submit 136 urine specimens from the autopsy subpopulation to the Armed Forces Medical Examiner System (AFMES) to screen quickly for a wide range of drugs. Their analysis also reflected polydrug complexity, with an average of 2.88 drug categories present per specimen. Among all urine tests, 98% tested positive for any form of fentanyl, 52% for non-fentanyl opioids, 28% for benzodiazepines, and 37% for cocaine.
Discussion and Conclusions

Compared with the New Hampshire census population, the fentanyl overdose study population proportionally over-represents males approximately 2:1 and over-represents males and females in their 20s and 30s as well as, to a slightly lesser extent, those in their 40s and early 50s. Compared with the population of living fentanyl users, it is likely that the study population of those who died contains older, more physically and/or more medically vulnerable individuals. Thus, our findings should not be generalized to this population without elucidating the potential differences.
The GIS analytical support provided by NDEWS helped to elucidate spatial distribution relationships. Although fentanyl-associated mortality has reached most communities in the state, it is disproportionately located in urban centers in the southeast quadrant of the state, especially Hillsborough County. We ruled out statistically significant relationships between decedent residence location and township-level socioeconomic factors of employment, income, and educational level. The concentration of deaths is, however, related to population density, urban status of the area, and proximity to major highways. 
Despite the ubiquitous presence of multiple drugs in these decedents, the effects of fentanyl were evidently so strong that there were no statistical differences in the fentanyl level (mean and standard deviation) with or without the presence of these co-intoxicants. The range of fentanyl levels was wide, from 0.75 to 113.00 ng/mL, with an average of 9.96 ng/mL. Nevertheless, the mean and range of fentanyl levels when fentanyl was the only drug found in toxicology were statistically the same as the mean and range for the cases where fentanyl was only one of several synergistic co-intoxicants. This suggests that fentanyl presence alone seems to be sufficient to cause death. 
Most decedents in the study population received some type of medical intervention: EMS response, naloxone administration, emergency room visit, or hospital admission. Certain subpopulations were, however, significantly less likely to receive any of these interventions: decedents 40 years or older; decedents we categorized as “opioid-naïve” (those who lacked any history of opioid abuse, recent opioid prescription, or previous overdose); and decedents residing in rural or micropolitan townships. Age plays an important role in these patterns because older decedents are significantly more likely to be living alone, to reside in a rural township, or to be categorized as “opioid naïve.”
In conclusion, the findings of our research shed new light on the population of decedents who died in 2015 and 2016 as a result of unintentional fentanyl poisoning. They focus attention on a mixture of rural and urban decedents from a primarily White population who died in New Hampshire. Victims who were older, lived in more rural areas, and lacked an opioid-related history had significantly less access to care. Although most decedents were found to use multiple drugs, fentanyl levels ranged broadly among them, with no significant relationship to the presence of other co-intoxicants, or to decedents’ opioid-related history.

[1] The “location of fatal drug use” will also be referred to in this report as the “location of injury.” The latter phrase is used on the death certificate. In overdose cases, it is taken to mean where the fatal drug use occurred.
[2] By using the broader death certificate dataset of 541, which was used for GIS mapping, the percentage of decedents for whom the location of fatal drug use was the same as the residence location was 68.6%. When the smaller analytical dataset of 505 was used, that percentage was 69.9%.
[3] “Parent” drugs refer to the original drug taken by the decedent before the body metabolizes it and changes its chemistry slightly. The term for drugs that have been metabolized is “metabolite.” 

Download: Unintentional Fentanyl Overdoses in New Hampshire: An NDEWS HotSpot Analysis (58 pages / 1.4MB)

Table of Contents

Executive Summary_________________________________4
  • Spatial Relationships (n = 541)
  • Decedent Characteristics (n = 505)
  • Death Event Characteristics (n = 505)
  • Toxicology Findings
Discussion and Conclusions_________________________________8
Study Populations and Data Sources_________________________________13
  • Overview of Sample Selection
  • Spatial Analysis (n = 541)
  • Occurrent Ratios (n = 773)
  • Medical Examiner Report (n = 505)
  • Autopsy Report (n = 189)
  • Death Investigation Report (n = 505)
  • Toxicology Report (n = 505)
  • Expanded Toxicology Testing by AFMES (n = 136)
  • Education and Veteran Status Derived Provided by Vital Records (n = 556)
Data Collection Process_________________________________16
Overview and Occurrent Ratios_________________________________17
Spatial Analysis_________________________________17
  • Population Density
  • Urbanity
  • Proximity to Interstate Highways
  • Sociodemographics
  • Distance Between Place of Residence, Fatal Drug Use, and Death
Decedent Characteristics (n = 505) _________________________________19
  • Demographics
  • Medical History
  • Literal Cause of Death Drug Frequencies
  • Toxicology Findings: Polydrug Complexity
  • Analysis of Fentanyl Levels
  • Expanded Toxicology of Urine Samples (n = 136)
Death Event Characteristics (n = 505) _________________________________22
Comparison of Decedent Subpopulations_________________________________23
  • Autopsy or No Autopsy Subpopulations
  • Age Group Subpopulations
  • Male and Female Subpopulations
  • Subpopulation with History of Chronic Pain
  • Subpopulation with Opioid Prescription in Previous 12 Months
  • Subpopulation with History of Opioid Abuse
  • Subpopulation with History of Previous Overdose
  • Homeless Subpopulation
Discussion and Conclusions_________________________________28