We examined the prevalence and correlates of NFOD among PWUD across MA between 2017 and 2019. We observed a high prevalence of lifetime (62.5%) and past year (37%) overdose in the study population. Nonprescription buprenorphine use and fentanyl were significantly, positively, and independently associated with higher odds of NFOD, whereas marijuana and methadone use were associated with lower odds of NFOD.

The estimates of NFOD prevalence observed in our study are significantly higher than recently published estimates of the history of lifetime NFOD in PWUD, which range from 15 to 58% [8, 31–34]. The rise of fentanyl in the illicit drug supply across the USA has resulted in a high prevalence of NFOD and a sharp rise in fatal OD [35–37]. Forty-five percent of our study participants also reported using fentanyl and among those 76% reported any overdose. Our findings are consistent with previous studies in MA, which have linked 75% to 94% of opioid-related overdoses to fentanyl [35, 38]. Findings from our multivariable regression models also strongly support the association between specific substance use (e.g., heroin use, fentanyl) and use patterns (i.e., injection drug use) with NFOD. These findings are consistent with prior research linking overdose risk to heroin and injection drug use [34, 39]. Criminalization of illicit substances can create an environment where people who use drugs opt toward high-risk behavior such as sharing and reusing needles. Previous studies indicate that such practices can also predispose individuals to the risk of transmitting infectious diseases such as HIV and HCV [12, 40, 41]. In light of these risks and the current HIV outbreak in the state, findings from the present study underscore the urgency of efforts to address the health of people using heroin and fentanyl in MA.

Prior research attributes fatal overdose rates to the high prevalence of fentanyl in the drug market [6, 35, 42]. Access to medication for opioid use disorders (MOUD) and harm reduction services—two evidence-based overdose prevention measures [43–45], have been shown to reduce fatal overdoses in many regions of the USA [9, 43, 46]. In examining access to harm reduction services among PWUD in MA, we found that a high proportion of participants reported carrying naloxone (62.7%) compared to estimates (17%-48%) from studies in other locations [34, 37, 47]. This higher prevalence of naloxone access among participants in our sample is likely due to the fact that MA has been at the forefront of public health responses to the opioid overdose epidemic [46, 48, 49]. Since 2006, MA has provided access to health insurance for all residents. Additionally, MA has implemented state-funded harm reduction services such as overdose education and naloxone distribution since 2005, with additional access to naloxone through pharmacies since 2018 [46, 48–50]. The observation that difficulty in accessing naloxone was associated with NFOD experience suggests there is still much room for improvement in community naloxone provision and equity in who can easily access this lifesaving medication. Further, the state's well-established 9-1-1 Good Samaritan Law, which provides limited immunity for drug-related charges when responding to a suspected overdose emergency, may have contributed to increased overdose rescues and thus improved survival in at-risk individuals, thereby leading to a higher prevalence of having experienced NFOD [46, 48, 49, 51, 52]. However, it is important to note that our study did not specifically examine the details surrounding the participants overdose to conclusively estimate the effect of GSL on the number of overdose rescues.

Participants who reported using non-prescribed buprenorphine in the 30 days before the study interview had approximately twofold increased odds of having experienced NFOD during the past year and in their lifetime, whereas current use of prescribed buprenorphine was not associated with NFOD. Buprenorphine is a key medication for the effective treatment of OUD [53]. But like all medication treatments for chronic conditions, it is most effective with prolonged and consistent therapeutic use. It is generally understood that the use of nonprescription buprenorphine is an attempt to self-treat withdrawal symptoms associated with opioid dependence, which is facilitated by buprenorphine's low risk of adverse events, a safety profile as a partial agonist, and greater availability [54–57]. Prior studies have reported a lower risk of NFOD with non-prescribed buprenorphine use [34, 56], though neither were conducted in places where or during periods when fentanyl was the dominant opioid being consumed. As we are unable to establish temporality due to the cross-sectional nature of the study design, it is difficult to disentangle whether the withdrawal symptoms may have played in mediating NFOD risk. It is possible that individuals who have had prior overdoses may be motivated to utilize non-prescribed buprenorphine as a means to reduce their risk of overdose and also serve as the first step to initiation of formal treatment [54–56]. The utilization of nonprescription of buprenorphine represents a complex dynamic of interaction between people with OUD and their substance use behaviors, substance use treatment providers, substance use treatment policy, and rurality. In states where prescribed buprenorphine access is limited due to stricter regulation of MOUD treatment, the presence of higher substance use-related stigma, or rurality, use of nonprescription buprenorphine may confer added protection to an individual with OUD [58]. In MA, where MOUD is more easily accessible, the use of nonprescription buprenorphine could also be indicative of higher risk-taking behavior and or mistrust of substance use treatment providers [59]. The findings highlight the importance of ensuring access to buprenorphine as a harm reduction measure, decriminalization of buprenorphine diversion, and improved availability through pharmacies [60, 61].

We also observed that the odds of lifetime NFOD were lower in Hispanic and NH Black people compared to NH white people. However, a different pattern of past year NFOD emerged, suggesting similarities with respect to risk, which is in congruence with previously reported findings [10, 62]. The opioid overdose epidemic primarily affected NH white people in the early and mid-2000s owing to the disparities in opioid prescribing [4]. However, as the epidemic evolved, a much broader population is being impacted. A recently published study indicated that the greatest increase in opioid-related mortality was seen in NH Black men, indicating the need to provide focused interventions for a minority population that was previously thought to have lower risks [15, 63]. The odds of past year NFOD were no different between NH white and Blacks; thus, our sample corroborates the recent and concerning fatal overdose trends nationally by race and, for the first time in the literature, suggests that the alarming increase appears to extend to NFOD events as well.

The findings from our study also indicate that methadone and marijuana use is associated lower odds of having a history of NFOD, even while using other illicit drugs in the past month. Methadone is commonly used in the treatment of OUD. Given that methadone maintenance reduces opioid cravings and that there is relatively limited diversion risk, the association between methadone and lower odds of past year NFOD suggests that methadone has been effective in reducing behaviors (e.g., fentanyl use, injection of drugs) and circumstances (e.g., incarceration, loss of tolerance) that increase the risk for fatal overdose. More research, however, is needed to understand the motivation for taking methadone among those with and without a history of NFOD (both lifetime and last year). Additionally, we found a strong inverse association between marijuana use and lifetime NFOD (aOR: 0.6). The legalization of medical marijuana has been shown to reduce the use of prescription drugs in Medicare part D and Medicaid enrollees in some studies [64, 65]. Segura et al. showed no effect of marijuana laws on the non-medical use of opioids [66], while other studies have found that marijuana use is related to an increased risk of having OUD [67–69]. Additional studies show the reduced use of illicit opioids in people who used marijuana frequently [70–73]. The association between marijuana and reduced odds of NFOD could be channeled through several pathways. Marijuana could possibly be used to manage underlying pain, stave off drug cravings, or intentionally limit the use of opioids by active users. Recent legal access to marijuana, for recreational or medical purposes, may encourage reduced use of illicit substances by creating safer, more consistent channels to a mind-altering substance. It is also likely that we observed these associations (reduced risk) because marijuana is more frequently being used by people who do not intentionally use opioids and, therefore, would be at low risk of NFOD.

Our findings should be considered in light of several limitations. We focused on high-overdose-risk communities in MA; hence, our results may not be generalizable to other regions of the state. Additionally, the data we collected were based on self-report, which is prone to recall and reporting bias. The lack of substance use testing can also affect the reported rates of heroin and fentanyl use as many participants could have unknowingly used heroin contaminated with fentanyl. The high rate of naloxone possession observed in our study could also have been due to the oversampling of participants from substance use treatment and harm reduction centers. To protect the confidentiality of participants and encourage participation, exact dates of care and health events were not obtained for this rapid assessment. Month and year of events were self-reported by participants, however, and allowed broadly for exploring the order of exposure if not the precise timing of exposure, which may be better achieved in a different study design. Furthermore, we carried out the rapid assessments over an extended period between August 2017 and November 2019, during which time substance use patterns could have changed. Notably, however, we used a random effects model to account for any variability associated with recruitment location, method, and time. The study also lacked an assessment of mental health issues in the population, which could have a significant association with a history of both lifetime and recent NFOD. Future research with this population could explore behavioral health considerations more broadly as well as motivations for using prescribed, non-prescribed, and illicit substances in relation to overdose risk.