Clarifying CDC’s Efforts to Quantify Overdose Deaths

In 2020, nearly 92 000 drug overdose deaths occurred in the United States. ¹ Provisional estimates from the Centers for Disease Control and Prevention (CDC) indicate that overdose deaths continued to rise in 2021. ² The rise in opioid overdose deaths can be described in the following 3 waves ³ : (1) increase in deaths involving prescription opioids starting in the late 1990s, ⁴ (2) increase in deaths involving heroin starting in 2010, ⁵ and (3) increase in deaths involving synthetic opioids, predominantly illicitly manufactured fentanyl and fentanyl analogs (IMFs), starting in 2013.^6,7 Synthetic opioids remain the primary driver of overdose deaths.^2,8

To inform public health policy and practice, it is vital to understand the changing drug overdose landscape and distinguish among the contributions of specific opioid types to the ongoing crisis. Under the International Classification of Diseases, Tenth Revision (ICD-10) coding system,^9,10 opioid-involved overdose deaths are classified as involving opium (T40.0), heroin (T40.1), natural and semisynthetic opioids (T40.2), methadone (T40.3), synthetic opioids other than methadone (T40.4), and other and unspecified narcotics (T40.6). Historically, CDC categorized prescription opioid–involved overdose deaths as those involving natural and semisynthetic opioids (commonly prescribed opioids, such as morphine, oxycodone, and hydrocodone), methadone, and synthetic opioids (eg, tramadol, fentanyl). ¹¹

In April 2018, Seth et al published an editorial describing CDC’s updated method to calculate the number of deaths involving prescription opioids because of the changing illicit opioid drug supply. ¹¹ CDC updated estimates because new data indicated that the increase in deaths involving synthetic opioids starting in 2013 was driven primarily by IMFs, not by prescription synthetic opioids. ³ Therefore, starting with 2014 data, CDC adopted a more conservative approach and removed synthetic opioids when approximating deaths involving prescription opioids. Seth et al outlined the rationale for this approach and showcased other CDC investments, such as the State Unintentional Drug Overdose Reporting System (SUDORS), that provide data from death certificates and medical examiner/coroner (ME/C) reports, including full postmortem toxicology results.

A 2021 editorial by Peppin and Coleman criticized CDC for this approach and incorrectly stated that the agency was admitting an error in its previous calculation of prescription opioid–involved overdose deaths and overrepresenting the role of prescription opioids in overdose deaths. ¹² The authors make other inaccurate assertions about CDC’s approach to quantifying overdose deaths that warrant further discussion. The purpose of our commentary is to present the fact-based calculation of overdose deaths based on the best scientific methods available, correct misstatements, and systematically refute the arguments made by Peppin and Coleman. We also outline CDC’s continuing efforts to modernize its ability to track drug overdose deaths with greater precision by investing in new surveillance systems (eg, SUDORS).

Peppin and Coleman ¹² misconstrue the intended use of drug overdose data captured from death certificates, overlooking improvements made in the last 15 years. ICD-10 cause-of-death coding captures drug involvement in a death using an internationally recognized standard coding system. ⁹ ICD coding is not intended to differentiate between prescription and illicit drugs, to capture whether drugs are prescribed or diverted, or to delineate whether the drug was prescribed for pain or used as treatment for opioid use disorder. ⁹ CDC uses these codes to identify the type of substances involved in the death, not to differentiate how the substance may have been obtained.

Additionally, CDC does not control the ICD coding system or the level of granularity in coding drug overdose deaths. ICD codes are published by the World Health Organization to promote international comparability in the collection, processing, classification, and presentation of mortality data (eg, format for reporting cause[s] of death on a death certificate). The coding rules improve the usefulness of mortality data by prioritizing certain categories, by consolidating conditions, and by systematically selecting a single cause of death. ¹⁰ Thus, CDC has been reporting on data collected within the constraints of this international coding system.

Appreciating the limitations of the ICD-10 system and how medicolegal death investigations vary across jurisdictions, CDC has worked closely with the ME/C community to improve drug overdose death reporting. For example, in 2012, CDC worked with the National Association of Medical Examiners to develop a position paper to provide evidence-based recommendations for death investigations and certifications involving opioids and other substances to improve mortality data for surveillance. ¹³ More recently, recognizing the overdose crisis and advances in forensic science, CDC worked with the ME/C community on an updated position paper. ¹⁴ These efforts have strengthened documentation of drug overdose deaths based on improvements in the percentage of drug overdose deaths that list at least 1 specific drug as a contributing cause of death (76% in 2012, 94% in 2020). ¹⁵

Peppin and Coleman also focus on the ICD-10 coding for benzodiazepines because, as with opioids, it is not possible to decipher an illicit versus licit drug or to determine whether someone was prescribed the drug. ¹² Additionally, the authors question the role of methadone and present a rather simple perspective: because methadone administration from opioid treatment programs has been increasing and dispensing of methadone from pharmacies (where it is used only for pain) has been decreasing, overdose deaths must be attributable to methadone for opioid use disorder treatment. ¹² Thus, they claim that CDC should have never included methadone in calculating prescription opioid–involved overdose deaths. As mentioned, ICD-10 is intended only to identify the type of opioid, not the therapeutic intent. Contrary to the view of Peppin and Coleman, ¹² prior investigations have identified methadone used for pain treatment as the primary contributor to overdose deaths involving methadone.^16,17 Also, a recent study found that increases in overdose deaths involving methadone during the COVID-19 pandemic were likely because of IMF, rather than take-home policies from opioid treatment programs. ¹⁷ Research has shown a strong relationship between trends in methadone dispensing in retail pharmacies for pain, methadone diversion, and methadone-involved overdose deaths. ¹⁸

Moreover, Peppin and Coleman imply that because of the ICD-10 coding scheme, CDC did not capture polysubstance involvement in overdose deaths and, thus, failed to acknowledge the role of other substances in opioid-involved overdose deaths. ¹² Even within the ICD-10, CDC has long focused on the role of multiple substances involved in overdose deaths. In 2013, CDC documented other prescription drugs involved in prescription opioid–involved overdose deaths. Specifically, findings highlighted the frequent involvement of drugs such as benzodiazepines, antidepressants, and antipsychotics in overdose deaths. ¹⁹ Also, researchers from the US Department of Health and Human Services published 2 reports that examined the role of benzodiazepines in prescription opioid–involved overdose deaths and emergency department visits and the combined prescribing of these 2 drug classes during the same period.^20,21 In 2014, CDC and the US Food and Drug Administration documented the role of alcohol in prescription opioid– and benzodiazepine-involved emergency department visits and overdose deaths. ²²

Most concerning, Peppin and Coleman assert that because the origins of prescription opioids or other drugs cannot be distinguished, policy has been misguided and erroneously focused on reducing opioid prescribing and related harms. ¹² This view overlooks the overwhelming data that corroborate the role of prescription opioids and prescribing of these medications in the first wave of the epidemic. ⁴ Multiple data sources that do not rely on ICD-10 coding have clearly documented that prescription opioids, rather than illicit opioids, were the drugs driving increases in opioid-involved morbidity and mortality in the late 1990s through the early 2010s. Data from the Treatment Episode Data Set, ²³ the Drug Abuse Warning Network, ²⁴ and the National Survey on Drug Use and Health^25,26 all pointed to large increases in high-frequency prescription opioid misuse and opioid use disorder from 2002 to 2013. These increases paralleled the rise in prescription opioid–involved overdose deaths and the prescribing and distribution of prescription opioids. In 2011, CDC documented that prescription opioid sales, treatment admissions, and deaths rose >300% from 1999 to 2008. ⁴ During this same time, other studies documented high doses of opioid prescriptions and longer durations of treatment as further connections among prescription opioids, certain prescribing behaviors, and rising morbidity and mortality.^27-29 Additional studies found declines in prescription opioid–related harms when policies were enacted to curb inappropriate prescribing of prescription opioids,^30,31 without a shift to illicit opioid–related harms. To overlook these documented connections would have been a shortsighted perspective and would have had grave implications for public health.

Peppin and Coleman incorrectly suggest that CDC misled policy makers for years based on how the agency calculated prescription opioid–involved overdose deaths. ¹² Prior to 2016, CDC’s approximation of prescription opioid–involved overdose deaths included deaths coded as T40.2 (natural and semisynthetic opioids), T40.3 (methadone), and T40.4 (synthetic opioids). ¹¹ The authors ¹² inaccurately state that CDC admitted, in a 2018 editorial, ¹¹ to an error in how these deaths were previously calculated, which included synthetic opioids, to misrepresent the role of prescription opioids in overdose deaths. Instead, CDC was being responsive to new data documenting that the observed increases in synthetic opioid–involved overdose deaths were primarily due to IMFs rather than prescription synthetic opioids. Therefore, CDC updated its calculation by removing synthetic opioids (T40.4) when calculating deaths involving prescription opioids to provide the most precise number of deaths within the parameters of the ICD-10 coding scheme. Any concern about “overestimation” would have been limited primarily to 2013 data, when rapid increases in deaths involving IMFs were observed. This issue was quickly addressed with analysis modifications implemented with the 2014 data, which were available December 2015 and published in 2016.^32,33

Peppin and Coleman ¹² do not mention that CDC was one of the first to document the emergence of IMFs in overdose deaths in 2013 and early 2014.^34,35 CDC conducted 2 in-depth investigations to further understand whether pharmaceutical fentanyl or IMFs were driving increases in overdose deaths.^36,37 Additionally, CDC found that among drug overdose deaths in 2014, the rate of overdose deaths involving synthetic opioids increased by 80% during 2013-2014, and it stated that this increase was likely because of IMFs. ³³ Changes to analyses were implemented as soon as sufficient data confirmed that the increase in overdose deaths involving synthetic opioids (T40.4) was not an isolated incident and was due to IMFs, not pharmaceutical fentanyl. Although toxicology tests are unable to distinguish between prescription and illicit fentanyl, these deaths coincided with increased availability of IMFs based on drug seizure data from the US Drug Enforcement Administration. ³⁸ While overdose deaths involving synthetic opioids were associated with the number of drug products obtained by law enforcement testing positive for fentanyl, they were not associated with fentanyl prescribing rates. ⁷ Additionally, in its report ³³ on drug overdose deaths in 2014 and a Health Advisory Network release, ³⁸ CDC noted as early as 2015 that IMFs were often being combined with heroin or sold as heroin and may have been contributing to recent increases in overdose deaths involving heroin. These findings suggested that IMFs posed an emerging threat. With the 2015 mortality data, ³⁹ CDC noted that from 2014 to 2015, the overdose death rate involving synthetic opioids and heroin increased by 72.2% and 20.6%, respectively. These actions run counter to the assertions by Peppin and Coleman ¹² that CDC was attempting to misrepresent the role of prescription opioids in the overdose crisis and misguide policy.

Finally, Peppin and Coleman ¹² imply that CDC should have been separating out synthetic opioids from the definition of prescription opioid–involved overdose deaths starting in 2007, when the number of overdose deaths spiked, because of heroin being tainted with IMFs. They suggest that CDC inappropriately included several thousand deaths each year for nearly a decade in overdose deaths involving prescription opioids. Again, the authors fail to acknowledge the data showing that the spike in overdose deaths that occurred in 2007 was due to fentanyl-tainted heroin from a single clandestine laboratory in Mexico that was subsequently decommissioned ⁴⁰ ; afterward, the occurrence of fentanyl in the illicit drug supply declined, and overdose deaths involving synthetic opioids decreased. Most of these deaths occurred in Chicago, Detroit, and Philadelphia, where IMF was mixed with heroin and found in counterfeit pills. These data are corroborated by the US Drug Enforcement Administration’s National Forensic Laboratory Information System, which shows that drug products seized and tested for fentanyl spiked in 2006, rapidly subsided from 2007 to 2013, and then increased in 2014, coinciding with increases in overdose deaths involving synthetic opioids.^41,42 Thus, contrary to a statement by Peppin and Coleman, ¹² CDC did not overestimate prescription opioid–involved overdose deaths for more than a decade, with the timing of CDC’s decision on calculation modifications being accurate, data informed, and justified.

The evolving drug overdose landscape adds to the complexities of categorizing drug overdose deaths, especially with an overburdened ME/C system. Death certificates have been used to identify drug overdose deaths and have been categorized by using ICD-10 codes since 1999.^9,10 Peppin and Coleman ¹² inaccurately assert that CDC failed to respond to the limitations of ICD-10 categorization and the modernization of overdose mortality surveillance. To address these limitations, CDC expanded its overdose mortality surveillance efforts. In 2016, SUDORS began with CDC’s Enhanced State Opioid Overdose Surveillance (ESOOS) program. SUDORS includes information from ME/C reports, including detailed postmortem toxicology findings, in addition to death certificate data (Table). The additional data required for SUDORS provide a comprehensive assessment of the cause of death, the specific substances involved in, and the circumstances surrounding drug overdose deaths of unintentional and undetermined intent. Also, drug overdose deaths are identified in SUDORS by a combination of ICD-10 codes and ME/C cause-of-death determinations based on postmortem toxicology results. The ESOOS program initially funded 12 states to collect data on all opioid overdose deaths of unintentional and undetermined intent, and in 2017 CDC expanded to 32 states and the District of Columbia (Figure). In 2019, with CDC’s Overdose Data to Action (OD2A) program, SUDORS expanded to collect information on all drug overdose deaths of unintentional and undetermined intent and now includes 47 states and the District of Columbia.

Table.

Examples of data elements collected in the State Unintentional Drug Overdose Reporting System, by data source

Data source	Data element examples
Death certificate	• Demographic characteristics • County and state where overdose occurred • Cause and manner of death • Other substantial conditions contributing to death • How overdose occurred • Place of death (eg, hospital, home) • Date of death
Medical examiner/coroner report	• History of overdoses • Treatment for substance use disorder • Prescription drug misuse or illicit drug use history • Routes of drug administration (eg, injection, smoking) • Presence of potential bystanders • Naloxone administration
Postmortem toxicology report	• All drugs detected • Drugs contributing to death • Date specimens collected

Figure.

Expansion of the State Unintentional Drug Overdose Reporting System. As of 2019, all funded states reported data on drug overdose deaths of unintentional and undetermined intent, regardless of when initially funded.

The detailed data collected in SUDORS from death scene investigations and toxicology reports allow for better differentiation between prescription and illicit substances that are involved in drug overdose deaths of unintentional and undetermined intent. ⁴³ Since 2017, CDC has been reporting on SUDORS data, and findings have highlighted the involvement of IMFs in drug overdose deaths—something not possible with the ICD-10 system.^44,45 SUDORS data have reported on the continued and growing threat of IMFs and provided precise data on substances causing and contributing to drug overdose deaths, including polysubstance-related overdose deaths. Despite citing a CDC Vital Signs report that used SUDORS data to detail the involvement of IMFs in drug overdose deaths, ⁴⁶ Peppin and Coleman ¹² did not acknowledge SUDORS as an enhanced CDC overdose mortality surveillance system, and they make an inappropriate comparison between SUDORS and ICD-10 coding. To implement a new, enhanced surveillance system nationally takes a considerable amount of time. They also stated that SUDORS received data from fewer than half of US states, ¹² not accounting for the rapid growth from 12 states in 2016 to 47 states and the District of Columbia in 2019. The implementation and expansion of SUDORS has allowed for improvements in understanding substances involved in overdose deaths. Using SUDORS, CDC reported that >80% of overdose deaths during the first half of 2019 were attributed to IMFs, heroin, cocaine, or methamphetamine—either alone or in combination (ie, polysubstance). ⁴⁶ Alarmingly, >60% of drug overdose deaths during that time had at least one opportunity for intervention (eg, linkage to care). Additionally, through OD2A, CDC supports data linkage activities. For example, several states are linking their prescription drug monitoring program data with SUDORS data. The comprehensive data from SUDORS and data linkage efforts allow for more data-driven, actionable prevention efforts. CDC also encourages sharing of these data with ME/Cs to help improve drug overdose death investigations.

CDC has invested in critical and continued improvements in timeliness and quality of mortality data through standards-based interoperability between ME/C case management systems and state electronic death reporting systems. Improvements build on existing efforts to increase the speed of cause-of-death coding, expand the details of coding available for drugs involved in the death, and enhance capacity for rapid release of provisional estimates.

CDC also has dedicated resources to enhance forensic toxicology testing of suspected overdose deaths. Starting in 2017, states funded through the ESOOS and OD2A programs were required to spend a portion of funding on comprehensive toxicologic testing of suspected opioid overdose deaths (ESOOS and first 2 years of OD2A) and suspected stimulant overdose deaths (beginning in 2021 for OD2A). Additionally, to address a critical gap in the ability of US laboratories to characterize drugs involved in opioid-involved overdoses, CDC developed Traceable Opioid Material (TOM) Kits, which are collections of reference materials that can be used to develop methods for detecting most opioids identified by the US Drug Enforcement Administration drug product submissions for testing. These kits provide reference materials for 250 synthetic opioids, with 210 being fentanyl analogs, and are freely available to any laboratory in the United States. Finally, CDC has made substantial investments to build capacity in the ME/C community on drug overdose death investigation and improve drug specificity included on death certificates. These efforts led to the following: an updated position paper by the National Association of Medical Examiners detailing recommendations for the investigation, diagnosis, and certification of drug overdose deaths ¹⁴ ; a reference guide for completing the death certificate for drug toxicity deaths ⁴⁷ ; a free online training on drug overdose death investigation ⁴⁸ ; a drug overdose death investigation workshop provided at the International Association of Coroners and Medical Examiners annual training symposium; and a free 3-part basic-to-advanced forensic toxicology of drug overdose deaths training and accompanying Grand Rounds series provided by the Center for Forensic Science Research and Education and the National Association of Medical Examiners.

CDC has made substantial efforts and investments to increase the timely reporting of drug overdose death data; to improve the quality, comprehensiveness, and specificity of overdose mortality data; to modernize its surveillance systems; to enhance forensic toxicology testing; and to support the ME/C community. While limitations to the ICD-10 coding system exist, the categorization has allowed for timely reporting of data during an escalating overdose crisis to inform data-driven prevention efforts. CDC continues to make critical investments to improve understanding of overdose deaths and associated risk factors.

The overdose crisis is one of the historic public health challenges of our lifetime. Urgent data-driven action is critical to inform public health prevention and response efforts. These actions include improvements in linking to and retaining people in treatment (ie, medications for opioid use disorder); expanding cognitive behavioral, contingency management, and harm reduction interventions (eg, access to syringe service programs, fentanyl test strips, naloxone); increasing provision of recovery support services; and scaling up primary prevention to reduce substance use initiation and the development of substance use disorders. Timely surveillance data are the foundation of this comprehensive, multipronged public health approach.