Worldwide Prevalence and Trends in Unintentional Drug Overdose: A Systematic Review of the Literature

METHODS

The potential scope of literature that may be considered of interest to the topic is vast. We conducted a systematic review of publications available in the US National Library of Medicine’s PubMed electronic database³¹ in July 2013 by using different combinations of Medical Subject Headings (MeSH) search terms. We restricted the search to English-language articles published between 1980 and July 2013 that deal with unintentional overdose, focusing on 1 or more of the following key constructs: prevalence, time trends, mortality rates, and correlates. The term “overdose” as a MeSH Major Topic yielded 1076 publications. In addition, we searched the following combinations of nonmajor MeSH terms: “street drugs” and “overdose” yielded 180, “death” and “overdose” yielded 114, and “poisoning” and “drug users” yielded 17. There was some overlap among the searches. We did not use “unintentional” as a search term so as not to leave out studies that dealt with both intentional and accidental overdoses. There were no limitations in this original search on type of sample, or location or geography, but we excluded studies dealing with children, because, although this group is important, we aimed to focus only on studies that reported on unintentional drug overdoses among adults. We included commentaries because of their potential to shed light on potential interventions or trends. In addition, we reviewed reference lists of chosen articles to select additional studies or commentaries.

With these search and inclusion and exclusion criteria, we selected a total of 169 relevant articles for this study through a close review of abstracts (Figure 1). We were interested in articles that reported

1. population-based prevalence measures (46 articles; reported in Table 1),

2. studies that reported percentage of changes over time in trends of overdose (34 articles; Table 2),

3. mortality rates (53 articles; Table 3), and

4. correlates of overdose (summarized in the Results section, but not included in tables because of the lack of common quantitative measures to record these topics).

RESULTS

Table 1 shows prevalence of overdose found in relevant articles, organized by type of drug, sample, and type of prevalence (witnessed vs experienced, fatal vs nonfatal, and time point). Lifetime prevalence of witnessed overdose among drug users (n = 17 samples from 18 articles) ranged from 50% to 96%,^{15–17,32,33,37,39,40,43–45,53,54,57–60,64} with a mean of 73.3%, a median of 70%, and a standard deviation of 14.1%. Although samples differed in demographics, length of drug use, and type of drug involved (e.g., prescription opioids, cocaine), the lifetime prevalence of witnessing an overdose did not differ greatly between studies conducted in different settings and countries.

Two studies reported lifetime prevalence of witnessing a fatal overdose specifically: 1 found that 4 of 24 drug users attending a general medical practice in inner-city Dublin, Ireland, in 1998 to 1999 had witnessed a fatal overdose⁴⁰; the other reported a 15% prevalence among 763 injection drug users from 16 different cities of the Russian Federation.⁵⁹ Two studies reported past-year prevalence of witnessed fatal or nonfatal overdose: 50% among 973 users aged younger than 30 years who had injected drugs in the previous month in 4 neighborhoods of San Francisco, California,⁵⁸ and 58% of 212 users in Dorset, United Kingdom, who had injected in the previous 2 months.⁶² One study of drug users aged 18 years or older in New York City (n = 1184) reported past-6-month prevalence of witnessing overdose at 35.2%¹⁶ and 2 samples of injection drug users reported past-3-month prevalence of witnessing overdose at 48.5% and 50%.^32,55

Lifetime prevalence of drug users personally experiencing a nonfatal overdose had a considerably wider range across studies (n = 29 samples), varying from 16.6% to 68.0%,^{9,13–16,32,37-44,46–48,50,53,56,58-61,63,64,67–69} with a mean of 45.4%, a median of 47%, and a standard deviation of 14.4%. Past-year prevalence of drug users personally experiencing a nonfatal overdose (n = 14 samples) ranged from 4.0% to 38.0%,^{9,35,36,42,46–50,52,58,62,63,67} with a mean of 18.8%, a median of 16.8%, and a standard deviation of 8.9%. Past-6-month prevalence of drug users personally experiencing a nonfatal overdose ranged from 6.7% to 33% (n = 7 samples),^{19,34,51,61,65,66,70} with a mean of 18.5%, a median of 17%, and a standard deviation of 10.4%. Two studies reported past-3-month prevalence of overdose: one found that 11.9% of drug users had overdosed out of 644 users younger than 30 years in San Francisco who had injected in the past 30 days⁵⁶ and the other found a prevalence of 15.2% among injection drug users, most of whom were homeless or in temporary housing, in Los Angeles, California, in 2006.⁵⁵

DISCUSSION

We systematically synthesized the peer-reviewed literature to document the global epidemiological profile of unintentional drug overdoses, and the prevalence, time trends, mortality rates, and correlates of drug overdoses. We found, first, wide variability in the lifetime prevalence of experiencing a nonfatal overdose and in mortality rates attributable to overdoses, depending on the study setting. Second, the majority of studies on longitudinal trends of overdose death rates or overdose-related hospitalization rates showed increases in overdose death rates and in overdose-related hospitalization rates across time, which have led to peaks in these rates now. In particular, an overall trend of increasing deaths from prescription opioid use and decreasing deaths from illicit drug use in the past several years has been noted across most of the literature.^{80–82,89,90,96,97,99,100} Third, with the increase in prescription opioid overdose deaths, drug overdose is no longer just an urban problem: rural areas have seen an important increase in overdose deaths. Fourth, cocaine, prescription opioids, and heroin are the drugs most commonly associated with unintentional drug overdoses worldwide and the demographic and psychiatric correlates associated with unintentional drug overdoses are similar worldwide. However, more studies examining these correlates are needed from non–English-speaking countries.

Drug overdose is an important, yet inadequately understood, public health problem. A review of existing studies points to several key methodological issues that future studies need to address and limitations in extant studies. Methodological issues include enhancing data collection methods on unintentional fatal and nonfatal overdoses, and collecting more detailed information on drug use history, source of drug use (for prescription drugs), and demographic and psychiatric history characteristics of the individual who overdosed. For example, future studies could employ consistency check strategies to compare medical examiner information on fatal overdoses with available medical records and information from the deceased individuals’ family members.

Future data collection efforts could also include more detailed questions when interviewing individuals who report nonfatal overdoses, and check consistency of such information with hospital discharge data and information from family or other third parties (e.g., friends who witnessed the overdose episode). The field still also lacks studies in rural areas as well as longitudinal studies of population-based drug users (i.e., non–treatment samples) in multiple urban and rural areas that would provide us with comparable prevalence estimates across different contexts.

In addition, several of the selected studies did not disclose how they defined nonfatal overdoses to study respondents,^33,34,37 simply mentioning that they asked them if they had ever experienced or witnessed an overdose, which could be somewhat misleading and biased. On the other hand, it should be noted that, among the studies that provided information on the definition of nonfatal overdoses, the definition of nonfatal overdoses was overall extremely consistent and standardized across the reviewed studies, similar to the definition we present in the introduction of this review.^{12,14–17,36,39,40}

The marked increase in prescription opioid overdoses described in the studies reviewed is likely attributable to a rapid increase in sales of prescription opioids during the past decade; the quantity of prescription opioids sold to pharmacies, hospitals, and doctors’ offices in the United States was 4 times larger in 2010 than in 1999.⁸ The increase in overdoses caused by prescription opioid use in the United States parallels the increase in the availability of prescription opioids since the early to mid-1990s in the United States.⁹⁰ As described in one of Nora Volkow’s recent presentations to the US Senate:

The number of prescriptions for opioids (like hydrocodone and oxycodone products) have escalated from around 76 million in 1991 to nearly 207 million in 2013, with the United States their biggest consumer globally, accounting for almost 100 percent of the world total for hydrocodone (i.e., Vicodin) and 81 percent for oxycodone (i.e., Percocet).^161(p3)

There is a need to invest in research to understand the distinct determinants of this type of overdose worldwide. Several other countries need to collect in a systematic and continuous fashion such data on sales of prescription opioids and other prescription drugs, nonmedical use of prescription drugs, and hospitalization secondary to overdoses on prescription drugs. Moreover, there is the need for future studies to further investigate whether those who overdose while using prescription medication had been prescribed these drugs recently or in the past because only 1 of the reviewed studies included this information.⁶⁹

Finally, the sparse evidence on the environmental determinants of overdose suggest a need for research that will inform the types of environmental interventions we can use to prevent drug overdose. A combination of studies on individuals who overdose and the settings where overdoses occur will help us better tailor interventions to the types of strategies that are most likely to have a major impact on this epidemic.

Limitations of the reviewed studies are also noted. First, most studies did not collect information on the sources of drug used, which can provide insights into the ways individuals learn about and acquire drugs, and can thus inform prevention opportunities. Drugs can be acquired from family, friends, and drug dealers, but they can also be prescribed for legitimate medical reasons (in the case of prescription drugs). More studies that investigate prescribing patterns, such as the article by Logan et al. on potentially inappropriate prescriptions given in the emergency department,¹⁶² are needed. Moreover, only about a third of the mortality rate studies, half of trends studies, and one sixth of the prevalence studies stratify results by specific types of prescription drugs.

Second, it is already well established that there might be an underreporting of intentional overdoses, because when substance users are asked whether their overdose was or was not intentional, most report it as unintentional.^29,163,164 In our review, we focused on unintentional overdoses, but there is the possibility that some of these overdoses were truly intentional.

Third, because substance users who experience an overdose are usually polydrug users, it is very difficult to distinguish which and how many drugs were used before an overdose episode.^165–168 Because of the high prevalence of multiple drugs being present in overdoses,^11,139 it is hard to tease out the effect of one drug over another. Overall, there is a mix of how well studies differentiate, but many aim to show an overall measure of burden of this issue, not truly trying to make a causal association with any 1 type of drug.

Fourth, studies have shown that forensic toxicology laboratories are limited in the number of drugs for which they can screen and that drug levels can change postmortem.^165–168 Further complicating this issue is the fact that data collected by medical examiners and coroners are usually inadequate, incomplete, and inconsistent with regard to polydrug use, estimated drug dosage (when, for instance, the individual had a legitimate prescription for at least 1 of the drugs involved in the fatal overdose episode—e.g., a legitimate opioid prescription), and patient characteristics.¹⁶⁹

Fifth, if we are to identify potential avenues to prevent drug overdose, more concerted investment needs to be made to understand the environmental drivers of overdose risk, and the individual and interpersonal mechanisms whereby environmental characteristics can increase the risk for overdose. This is particularly important for overdose risk in rural areas, as the bulk of the literature on environmental determinants of overdose has focused on urban areas. This also has important implications for prevention and intervention strategies that need to focus on harm reduction strategies such as the availability of safe injection facilities or mobile safe injection facilities where individuals can inject preobtained drugs under the supervision of a medical doctor, such as facilities that exist in Canada and in some European countries,¹⁹ while also offering them treatment for their drug-using behavior.