Trends in Substances Involved in Polysubstance Overdose Fatalities in Maryland, USA 2003-2019

Abstract

Background.

The substances driving the overdose epidemic in the United States have changed over time. Since 2013, fentanyl-analogues have become the primary opioids driving the epidemic. Recently, polysubstance related deaths have come to the forefront of the epidemic. Therefore, we explored trends in polysubstance involvement in overdose fatalities in Maryland between 2003 and 2019.

Methods.

We used records for the Maryland Office of the Chief Medical Examiner between 2003 and 2019. First, we assessed trends in the number of drugs (1, 2, 3+) involved in overdose fatalities over time. Then, we assessed linear and quadratic trends in nine substance categories using logistic regression.

Results.

Overtime, the proportion of overdose deaths involving one (ß=−0.28, p<0.001) or two (ß=−0.14, p<0.001) drugs decreased, while deaths involving three or more drugs increased (ß=0.31, p<0.001). The involvement of most drugs decreased over the study period. Only cocaine (linear ß=−0.08, p<0.001; quadratic ß=0.29, p<0.001) and fentanyl (linear ß=1.67, p<0.001; quadratic ß=0.75, p<0.0001) showed significant increases over time. Post hoc analyses showed that cocaine involvement only increased in the presence of fentanyl (linear ß=1.41, p<0.001; quadratic ß=0.70, p<0.001) and decreased when fentanyl was not present (linear ß=−0.81, p<0.001; quadratic ß=−0.31, p<0.001).

Conclusions.

Polysubstance involvement in overdose fatalities has become more common over time in Maryland. Most individual substances became less common over time, while cocaine and fentanyl involvement increased. Unintentional fentanyl consumption through contaminated cocaine may be an increasingly important driver of deaths in Maryland, indicating the importance of implementing widespread harm reduction programs, especially drug checking.

Introduction

As the overdose epidemic has progressed in the United States (US), the primary substances driving the epidemic have changed. Beginning in the late 1990’s, the overdose epidemic was largely driven by prescription opioids, driven by increased chronic pain, heightened expectations for pain treatment, and systemic factors favoring pharmaceutical pain treatments (Dasgupta, Beletsky, & Ciccarone, 2018). Around 2010, the epidemic shifted away from prescription opioid overdose towards heroin as a result of a broader susceptible market affected by a restricted prescription opioid supply (Compton, Jones, & Baldwin, 2016; Mars et al., 2014; Rudd et al., 2014; Unick et al., 2013). In 2013, the epidemic shifted again, as illicitly manufactured fentanyl was introduced into the drug supply, resulting in a dramatic spike in overdose mortality (Ciccarone, 2017; Prekupec, Mansky, & Baumann, 2017).

Research suggests that we are entering a new wave of the opioid crisis related to stimulant and polysubstance use (Jones et al., 2020). Deaths associated with cocaine and other psychostimulants have been increasing since 2015 (Jones, Einstein, & Compton, 2018; Kariisa et al., 2019). Polysubstance involvement has also been increasing; for example, two-thirds of opioid-related deaths from 25 states involved cocaine, methamphetamine, or a benzodiazepine (Gladden, O’Donnell, Mattson, & Seth, 2019). Polysubstance use is a prevalent and growing aspect of the opioid crisis than needs to be well-understood in order to curb mortality.

The overdose crisis in Maryland has followed the broader pattern of the opioid crisis, recently experiencing increases in fentanyl and stimulant related overdoses (Vital Statistics Administration, 2018). However, changes in polysubstance involvement have not yet been explored. Accordingly, we explore how the number of drugs involved in overdose deaths in Maryland have changed since 2003 and if the involvement of specific substances in polysubstance overdose deaths have changed over time.

Methods

Data Source.

For this analysis, we used records from Maryland Office of the Chief Medical Examiner for all investigated deaths between the years 2003 and 2019. We identified records for deaths where the cause of death was identified as an overdose (N=20,538). For suspected overdoses, the medical examiner conducts a toxicology panel to identify involved substances. A small number of cases were missing toxicology information (n=535), so these cases were excluded from our analysis. We also removed 287 cases where no involved substances were identified by the toxicology panel, yielding an analytic sample of 19,716 overdose deaths.

Measures.

The medical examiner tests for a broad range of individual substances, many of which belong to similar classes of drugs. The toxicology procedure will detect any substances whose metabolites are present at the time of death, though this does not necessarily mean that a particular substance contributed to the death. The toxicology panel was reasonably standard over time, with no changes to testing for major substance of abuse. Only one antidepressant (venlafaxine) was added to the toxicology panel, as it was approved for use in the US during the study period. Testing was done with aqueous humor, when possible, as it remains sterile for the longest period after death allowing for the most accurate results (especially in the case of alcohol which is the most affected by body decomposition). The toxicology screening included indicators for the presence of alcohol, fentanyl (including 9 common analogues), benzodiazepines, cocaine, methadone, and morphine. The presence of morphine could indicate either prescription medications or the presence of heroin, as heroin metabolizes to morphine and the more specific molecular indicators for heroin were not systematically tested for across time. The toxicology screening measures individual substances that belong to broader classes of drugs, particularly for prescription drugs, so we created combined indicators for antidepressants (e.g., bupropion, fluoxetine, venlafaxine, amitriptyline, nortriptyline, doxepin, nordoxepin, mirtazapine, desmethylsertraline, sertraline, citalopram, paroxetine, trazodone) and prescription opioids (e.g., meperidine, tramadol, propoxyphene, oxycodone). We created an indicator for the presence of other substances which were too rare to include separately, including neuroleptics, anticonvulsants, amphetamines, sedative/tranquilizers, and a variety of other drugs that did not fit into any of these categories. We generate a binary indicator to indicate the presence of each of these substance categories. We also generated a categorical variable for the number of substances involved each death, indicating whether one, two, or three or more substances were involved. When considering the number of drugs involved in a given case, we counted the number of drug categories (e.g. two antidepressants would be counted as one drug).

Analysis.

We used binomial logistic regression models to assess trends in the number of substances and individual substances involved in overdose fatalities (i.e. a series of models where each substance is individually regressed on time). Time was treated as a discrete. We then used a post-estimation orthogonal polynomial contrast of marginal linear trends to obtain an overall estimate for trends over time. This approach allows for the modeling of both linear and quadratic trends over time. Quadratic trends allow for the assessment of non-linear changes, with positive coefficients indicating a convex trends and negative coefficients indicating concave ones. We report linear and quadratic trends for the number of involved drugs and each substance. When interpreting the trends observed in this study, both linear and quadratic trends should be considered together, as interpretation in isolation may result in misleading conclusions. For example, a situation where there are positive linear and negative quadratic trends reflects a situation with an initial increase followed by a later decline (producing a concave shape). If only the linear trend was considered, one may mistakenly draw the conclusion that there was a consistent increase across the study period. As trends among all deaths and polysubstance deaths were near identical for all drugs, only trends among polysubstance deaths are presented in the results (Appendix 1). Analyses were conducted using Stata 14 (StataCorp, 2015).

Results

The number of deaths increased over the study period, starting at 768 deaths in 2003 and ending at 2,323 deaths in 2019. Across years, morphine was the most common drug (51.3% of deaths), followed by fentanyl (38.5%), alcohol (37.3%), and cocaine (32.6%). Antidepressants (24.4%), prescription opioids (19.7%), benzodiazepines (16.6%), methadone (16.5%), and other drugs (24.9%) were less common.

Over time, the number of drugs involved in fatalities increased (Figure 1). The proportion of overdoses that involved a single substance decreased linearly over time (ß=−0.28, p<0.001). The proportion of deaths involving two substances also decreased (ß=−0.14, p<0.001). Deaths involving three or more substances increased linearly over time (ß=0.31, p<0.001). There were no significant quadratic trends for the number of drugs involved over time.

Figure 1.

Percentage of Maryland Overdose Fatalities Involving One, Two, or Three or More Substances.

The trends for individual drugs involve in polysubstance fatalities showed that the involvement of most drugs declined or remained stable over time, with a few notable exceptions (Figure 2). Among opioids, involvement of methadone (linear ß=−0.39, p<0.001; quadratic ß=−0.21, p<0.001) and prescription opioids (linear ß=−0.10, p<0.001; quadratic ß=−0.28, p<0.001) both showed overall downwards trends that accelerated over the study period. Morphine involvement also demonstrated a slight downward trend (linear ß=−0.06, p<0.001; quadratic ß=−0.02, p=0.35), with a notable uptick between 2013 and 2016. Fentanyl involvement was uncommon at the beginning of the study period, then increased dramatically beginning around 2014 (linear ß=1.67, p<0.001; quadratic ß=0.75, p<0.0001).

Figure 2.

Trends in Substances Involved in Polysubstance Overdose Fatalities in Maryland, USA.

Among non-opioids, alcohol involvement declined slightly over the study period (linear ß=−0.08, p<0.001; quadratic ß=−0.03, p=0.094). Benzodiazepine (linear ß=0.28, p<0.001; quadratic ß=−0.46, p<0.001) and other drug (linear ß=0.07, p<0.001; quadratic ß=−0.23, p<0.001) involvement both initially increased until 2011 and then declined for the rest of the study period. Antidepressant involvement declined over the study period (linear ß=−0.09, p<0.001; quadratic ß=−0.13, p<0.001). Cocaine involvement, however, initially declined and then increased from 2015 onwards (linear ß=−0.08, p<0.001; quadratic ß=0.29, p<0.001).

Based on the trends observed for individual substances, we conducted a post hoc analysis to assess if the simultaneous increases in cocaine and fentanyl were related to deaths that involved both cocaine and fentanyl. Using the same method as for single substances, we assess trends in deaths that involved cocaine and fentanyl and deaths that involved cocaine without fentanyl among polysubstance overdose deaths. For cocaine deaths that also involved fentanyl, there was a positive linear trend over time (ß=1.41, p<0.001) and a positive quadratic trend (ß=0.70, p<0.001), indicating that the increase accelerated over the study period. For cocaine deaths without fentanyl, we observed negative linear (ß=−0.81, p<0.001) and quadratic (ß=−0.31, p<0.001) trends, indicting an accelerating decrease over time.

Discussion

Overdose deaths in Maryland have increasingly involved multiple substances. The proportion deaths involving three or more substances increased across the entire study period. This finding further reflects an absolute increase in polysubstance involved deaths, as the number of overdoses by year increased across the study period as well. The involvement of most individual drugs in polysubstance deaths declined over time, except for fentanyl and cocaine. Our findings align with some, but not all, previous research on changes in polysubstance involvement in overdose fatalities. The declines we observed for substances like alcohol and benzodiazepines were in contrast to some previous research in other states, which suggested these drugs are on the rise (Gladden et al., 2019; Tori et al., 2020).

Our data support research indicating that cocaine-related overdose fatalities are rising as a result of co-use or contamination with fentanyl. We found that cocaine has become more prevalent among polysubstance overdose fatalities when in combination with fentanyl, but when fentanyl is absent cocaine involvement has declined. This suggests that in Maryland co-exposure to fentanyl and cocaine is a primary driver of the increase in cocaine-related fatalities we observed, possibly due in part to unintentional consumption by individuals who primarily use stimulants and do not typically use opioids. It should be noted that a substantial portion of deaths involving cocaine and fentanyl are likely due to speedball use (intentional co-consumption of heroin and cocaine) where the heroin was cut with or replaced by fentanyl. Research from other settings has also suggested that there has been a surge in deaths related to fentanyl and cocaine (Gladden et al., 2019; McCall Jones, Baldwin, & Compton, 2017; Nolan et al., 2019). This highlights the importance of broadly implementing drug checking technologies, including fentanyl test strips, and emphasizing the dangers associated with co-use to the community. Health departments and other public health organizations should prioritize fentanyl test strip distribution to people who use drugs and education for this population about the potential benefits of being able to check their drugs for fentanyl. Previous work has shown that people who use drugs will engage in risk reduction strategies, like using tester shots, snorting instead of injecting, or using less than they would if they knew the drug did not contain fentanyl, when using drugs if they are aware that fentanyl is present in their drugs (Park et al., 2020; Peiper et al., 2019). Test strip distribution should be paired with the implementation of comprehensive drug checking programs in communities with high overdose burdens, including drug checking devices. Once installed, drug checking devices can serve as fixed location testing resources, where individuals can bring their drugs to test for the presence of fentanyl (Laing, Tupper, & Fairbairn, 2018).

Drug checking programs should be further complemented by an array of harm reduction programs to prevent overdose fatalities. Public awareness campaigns around the potential for fentanyl to be present in non-opioid drugs are needed to raise the awareness of overdose risk among individuals who primarily use stimulants. Naloxone distribution programs should also actively seek to reach individuals who primarily use stimulants to help address the deaths due to unintentional fentanyl consumption. Individuals who use stimulants like cocaine but do not intentionally use opioids may not perceive a need to carry naloxone regularly, as they may believe that they are not at risk for an overdose. Substance use treatment services should also be expanded for stimulants, using the 2020 federal funding provisions that allow for opioid related funding to also be used to address stimulants ("Consolidated Appropriations Act," 2020). By raising risk awareness through media campaigns while simultaneously expanding risk reduction tools, like drug checking resources, treatment, and naloxone for overdose reversals, we can meaningfully turn the trend of overdose fatalities in the state of Maryland.

This study does have some limitations to consider. Most importantly, we were unable to distinguish between prescription morphine and heroin based on the toxicology results. The toxicology panel only tested for common fentanyl analogs, so it is possible that there were fentanyl-involved deaths that were not detected if a rarer analog was involved. We are also unable to conclusively determine if deaths involving multiple substances were the result of intentional polysubstance use or unintentional use due to drug contamination. We also combined drugs of the same class together, therefore reducing the number of deaths that were considered polysubstance involved. Our results likely only apply to Maryland and other states with similar drug markets, as there are large regional differences in substance use across the US.

Polysubstance involvement in overdose fatalities has become increasingly common in Maryland since the early 2000s. Most individual substances became less common among polysubstance deaths with the exception of cocaine and fentanyl deaths, which increased over time. Increases in the presence of cocaine were due to the co-presence of fentanyl, highlighting the importance of making drug checking technologies available to people who use drugs to avoid unintentional fentanyl consumption and resulting overdoses.

Highlights.

• Polysubstance involvement in overdose deaths increased from 2003-2019 in Maryland.

• Most substances decreased over time, while fentanyl and cocaine increased.

• Cocaine involvement only increased while fentanyl was also present.

Appendix 1.

ALL OVERDOSE DEATHS (% of deaths drug was involved in)
Year	Alcohol	Anti- depressants	Fentanyl	Benzo- diazepines	Cocaine	Methadone	Morphine	Pres. Opioids	Other
2003	39.11	27.47	4.04	7.94	36.59	20.18	61.20	17.19	18.75
2004	38.99	22.45	1.44	5.90	37.27	19.86	54.68	17.55	17.84
2005	33.48	22.70	2.30	5.46	34.91	26.44	50.57	17.53	17.96
2006	40.35	23.16	4.98	6.97	40.10	25.53	49.19	19.30	19.93
2007	39.00	22.71	3.45	12.60	36.98	27.82	52.44	17.84	21.88
2008	39.94	30.10	3.88	19.14	28.29	24.80	47.57	25.38	23.58
2009	39.27	26.44	4.19	20.55	27.88	19.63	54.58	24.35	26.31
2010	37.50	27.06	6.03	21.91	24.56	27.50	41.18	28.53	26.18
2011	37.22	30.34	5.34	31.46	24.30	25.98	42.13	33.15	37.22
2012	37.83	29.24	4.30	29.95	24.11	22.20	53.46	28.16	35.8
2013	40.25	30.16	7.85	27.80	22.76	17.04	63.68	24.55	29.71
2014	39.05	28.42	19.01	25.16	23.21	16.59	62.63	25.44	30.20
2015	36.45	26.02	28.93	16.65	21.11	15.58	63.09	21.26	26.78
2016	41.29	21.89	57.18	14.64	29.00	10.75	64.52	16.95	24.63
2017	37.75	22.62	72.21	14.49	35.92	11.56	51.05	15.77	23.22
2018	34.16	21.66	82.06	14.55	42.25	9.55	39.98	16.70	24.19
2019	31.87	20.40	84.20	15.28	40.94	9.08	36.12	13.22	22.77
Linear Trend	−0.04, 0.014	−0.03, 0.046	1.67, <0.001	0.33, <0.001	−0.05, 0.003	−0.36, <0.001	−0.05, <0.001	−0.04, 0.027	0.13, <0.001
Quadratic Trend	−0.05, 0.003	−0.12, <0.001	0.70, <0.001	−0.44, <0.001	0.27, <0.001	−0.22, <0.001	−0.06, <0.001	−0.27, <0.001	−0.20, <0.001
POLYSUBSTANCE OVERDOSE DEATHS (% of deaths drug was involved in)
Year	Alcohol	Anti- depressants	Fentanyl	Benzo- diazepines	Cocaine	Methadone	Morphine	Pres. Opioids	Other
2003	46.36	33.95	4.90	10.30	43.24	22.80	65.37	21.45	23.48
2004	46.18	29.12	1.81	8.23	43.98	24.30	63.45	22.89	21.89
2005	39.40	29.20	3.20	7.60	42.00	31.00	57.60	21.60	23.40
2006	45.71	29.87	5.94	8.58	46.86	30.03	55.94	24.09	24.59
2007	44.51	28.21	4.23	16.46	44.36	31.35	57.37	22.73	28.37
2008	45.41	38.20	4.86	24.32	32.79	27.75	51.53	31.53	30.27
2009	45.54	32.18	4.46	25.58	30.86	21.62	58.91	29.37	32.01
2010	41.65	33.02	6.94	27.77	28.71	30.21	43.34	35.08	32.46
2011	40.31	36.02	5.83	38.25	27.44	28.99	45.28	37.91	44.08
2012	41.36	34.54	5.08	36.14	27.72	24.09	56.60	32.66	42.53
2013	44.31	35.91	8.81	33.47	26.15	18.97	66.94	28.59	34.82
2014	41.79	33.41	20.56	29.83	27.15	18.55	66.93	30.17	35.08
2015	42.06	31.10	32.51	20.51	24.76	17.39	67.77	24.57	31.29
2016	45.06	24.52	62.40	16.73	32.52	11.54	69.46	18.66	27.11
2017	41.50	25.77	75.77	16.63	40.38	12.49	57.81	18.01	26.25
2018	37.05	24.63	85.43	16.63	47.11	10.60	45.39	18.74	26.59
2019	35.63	23.82	87.46	17.91	46.33	10.37	42.08	15.07	25.64
Linear Trend	−0.08, <0.001	−0.09, <0.001	1.67, <0.001	0.28, <0.001	−0.08, <0.001	−0.39, <0.001	−0.06, <0.001	−0.10, <0.001	0.07, <0.001
Quadratic Trend	−0.03, 0.094	−0.13, <0.001	0.75, <0.001	−0.46, <0.001	0.29, <0.001	−0.21, <0.001	−0.02, 0.35	−0.28, <0.001	−0.23, <0.001