Occupational Patterns in Unintentional and Undetermined Drug-Involved and Opioid-Involved Overdose Deaths — United States, 2007–2012

The opioid epidemic affects multiple segments of the U.S. population (1). Occupational patterns might be critical to understanding the epidemic. Opioids are often prescribed for specific types of work-related injuries, which vary by occupation* (2). CDC used mortality data from the National Occupational Mortality Surveillance (NOMS) system to examine unintentional or undetermined drug overdose mortality within 26 occupation groups. This study included data from the 21 U.S. states participating in NOMS during 2007–2012.† Drug overdose mortality was compared with total mortality using proportional mortality ratios (PMRs) indirectly standardized for age, sex, race, year, and state. Mortality patterns specific to opioid-related overdose deaths were also assessed. Construction occupations had the highest PMRs for drug overdose deaths and for both heroin-related and prescription opioid–related overdose deaths. The occupation groups with the highest PMRs from methadone, natural and semisynthetic opioids, and synthetic opioids other than methadone were construction, extraction (e.g., mining, oil and gas extraction), and health care practitioners. The workplace is an integral part of life for the majority of the adult U.S. population; incorporating workplace research and interventions likely will benefit the opioid epidemic response.

Usual occupation, recorded as free-text on the death certificate, was coded to 1990 or 2000 U.S. Census occupation codes** by NIOSH or by the state. A crosswalk based on U.S. Census data was used to convert the 1990 U.S. Census occupation codes to the 2000 U.S. Census occupation codes. † † Occupation codes were binned into 26 groups based on job duties. § § For each outcome, the proportion of deaths among each occupation group was compared with the proportion of deaths among all occupations combined using PMRs indirectly standardized by age, sex, race, calendar year, and state of occurrence. A PMR >1.00 indicated that the proportion of deaths within that occupation group is higher than the proportion of deaths among all occupation groups combined. Corresponding 95% confidence intervals were calculated. § 2012 is the most recent year for which NOMS data are available. NOMS is the largest source of U.S. population-level mortality data that contains occupation and industry information. Only 21 states contributed data to NOMS during the study period. ¶ From 2013 to 2014, a large increase in illicitly manufactured fentanyl occurred. Within the study period, this category is mostly recording information about pharmaceutical synthetic opioids. ** https://www2.census.gov/programs-surveys/demo/guidance/industryoccupation/occ2000t.pdf. The 1990 census occupation codes are available upon request from the NOMS program (https://www.cdc.gov/niosh/ topics/noms/). † † The crosswalk is based on data in Table 2  The analysis identified 57,810 drug overdose deaths within the study population (1.4% of the 4,024,086 deaths). The majority of drug overdose deaths were among persons who were male (61.8%), white (89.8%), and aged 45-54 years (30.1%) or 35-44 years (24.1%). ¶ ¶ PMRs from drug overdose were significantly above 1.00 for the following six occupation groups: 1) construction (1.25); 2) extraction (1.16); 3) food preparation and serving (1.11); 4) health care practitioners and technical (1.16); 5) health care support (1.18); and 6) personal care and service (1.10) ( Table 1). PMRs from drug overdose were also significantly elevated among deaths where the usual occupation was unpaid/unemployed (1.10)*** or unknown (1.31). † † † For each specific opioid type, significantly elevated PMRs were generally observed only for those occupation groups that also had a significantly elevated PMR for drug overdose overall (Table 1) ( Table 2). The only two exceptions were the arts, design, entertainment, sports, and media occupation group and the building and grounds cleaning and maintenance occupation group. For these groups, the proportion of drug overdose deaths among the two occupation groups was similar to the proportion of drug overdose deaths overall (i.e., PMR approximately = 1.00), whereas the proportional distribution of specific drugs involved in an overdose was different, § § § with heroin-involved overdose deaths higher than expected ( Table 1). The highest PMRs for methadone, natural and semisynthetic opioids, and synthetic opioids were in the construction (1.34), extraction (1.39), and healthcare practitioner (1.81) occupation groups, respectively.
Because the PMRs for all opioid types within the construction occupation group were elevated, a subanalysis further examined opioid-related deaths in this group. The analysis identified 7,402 drug overdose deaths among persons aged ≥18 years within the construction occupation group. The majority of decedents were male (96.7%), white (92.6%), and aged 45-54 years (30.4%) or 35-44 years (26.9%). ¶ ¶ ¶ ¶ ¶ Opioid overdose decedent median age = 43 years. *** Homemaker (not working on a farm), volunteer, or student. † † † This category includes deaths with insufficient information available on the death certificate to apply a U.S. Census occupation code or for which the usual occupation field was left blank. § § § The proportional distribution of drugs (i.e., the proportion of total drug overdoses deaths for each drug type) involved in drug overdose deaths include opioids and nonopioids (e.g., cocaine). PMRs are mutually dependent and a higher proportion for one cause (e.g., a specific drug) results in a lower proportion for another cause. In this analysis, cause-specific outcomes (e.g., heroin-related overdose or prescription opioid-related overdose) are not independent and are partially overlapping. Decedents might have multiple drug types within their system at time of death and therefore counted in more than one cause-specific outcome category. ¶ ¶ ¶ Construction occupation group. These deaths were examined by the following occupation subgroups****: first-line supervisors and managers, † † † † construction trade workers (e.g., carpenters, electricians, painters, iron and steel workers, operating engineers, and construction equipment operators), construction trade helpers, and other construction and related workers (e.g., building inspectors, hazardous waste workers, and highway maintenance workers). PMRs were significantly elevated for all types of opioids within the occupation subgroup construction trade workers (Table 3).

Discussion
In this study, unintentional and undetermined overdose deaths varied by occupation group, with the construction group having elevated PMRs for all drug types. Although few related studies have been conducted, similar results have been observed. In Kentucky (2011) (3) and Ohio (2016) (4), for example, overdose deaths varied by industry and occupation and were highest among construction workers. Multiyear studies conducted in two Massachusetts jurisdictions (Barnstable County and Mystic Valley Public Health Coalition communities) found trade workers (e.g., construction, building/grounds maintenance, and mechanics) had the largest proportion of opioid overdose deaths (37% and 42%, respectively) (5,6). Variation was expected because work-related injuries and illnesses vary by occupation and industry. In addition, other factors that might affect opioid use, such as psychosocial workrelated stress (e.g., job insecurity or high demand/low control jobs), socioeconomic standing, and education level, also vary by occupation and industry (7)(8)(9).
The specific drugs influencing higher than expected proportions of overdose deaths also varied by occupation group. In this study, heroin PMRs were highest for the construction; food preparation and serving; and arts, design, entertainment, sports, and media occupation groups. Among the drug types evaluated, heroin is illicit, whereas among the other types, usage is both licit (i.e., prescribed and used as directed) and illicit. Data from the National Survey on Drug Use and Health illustrate that self-reported illicit drug use varies by  were accommodations and food services; arts, entertainment, and recreation; and construction (10). The variation by occupation group in this study leads to speculation about opioid initiation or use and the work environment. A single on-the-job injury (e.g., fracture or dislocation) or chronic work-related pain (e.g., caused by repetitive motion or lifting) might result in a prescription for pain medication (2,8). Workers' compensation data from  The findings in this report are subject to at least six limitations. First, data were analyzed in aggregate, but occupational patterns for each drug type might have differed by year. Second, NOMS has limited information on the specific circumstances of death. It is not known, for example, whether the death occurred at work. Death certificates do not state whether decedents were employed at their usual job (listed on the death certificate), another job, or unemployed at the time of death; if the drug use was legal or illegal; or if drug use was initiated while decedents were employed at their usual job, another job, or before employment. Third, the specific drug involved in the drug overdose death might have been misclassified (e.g., heroin deaths misclassified as morphine deaths because of similar metabolites) or given nonspecific codes (1). Within this study, the only drug code listed for one fourth of overdose deaths was "other and unspecified drugs" (T50.9 excluding T36-T50.8). Outcome misclassification might vary by state and year. Fourth, intentional overdose deaths were excluded; however, an unknown proportion of undetermined deaths might have included homicides or suicides and might therefore have resulted in overestimates. In this study, 9.6% of drug overdose deaths were of undetermined intent. The distribution of overdose deaths by intent and occupation group need to be explored. Fifth, PMRs are mutually dependent and cannot distinguish whether occupation was associated with increasing a specific cause of death, preventing the occurrence of other causes of death, or some combination of these effects. Finally, only 21 states participated in NOMS during the study period, which limits generalizability of the findings.
This study identified occupation groups with a higher proportion of drug and opioid-specific overdose mortality but was unable to identify specific factors that might have led to the observed results. The surveillance data presented in this study generated many questions; future studies are needed to identify potential work-related factors along the causal pathway from drug initiation to overdose mortality and to investigate ways of tailoring prevention measures to specific occupations. Workplace-specific programs and policies to reduce the impact of the opioid epidemic can be implemented. Since 2009, a decline in opioid use among nonsurgical workers' compensation claims in 26 states has occurred, which is associated with changes to workers' compensation laws and regulations regarding pain management and the prescribing and distribution of opioids, in addition to corresponding national and state-level legislative and regulatory changes (2). Examples of programs † † † † † that might address both licit and illicit opioids include comprehensive drug-free workplace programs, employee assistance programs, peer-support networks, and education targeted to employees † † † † † The Substance Abuse and Mental Health Services Administration (SAMHSA) provides detailed information on drug-free workplace programs, related laws and regulations, and a toolkit for employers (https:// www.samhsa.gov/workplace).

Summary
What is already known about this topic?
A majority of the U.S. population participates in the workforce. A person's job affects both physical and psychological wellbeing. The opioid epidemic negatively affects workers, workplaces, and employers.
What is added by this report?
During 2007-2012 proportional mortality ratios (PMR) for heroin-related overdose deaths (1.46) and methadone-related overdose deaths (1.34) were highest for the construction occupation group. PMRs for natural and semisynthetic opioids were highest for the extraction (1.39) and health care practitioner (1.81) occupation groups.
What are the implications for public health practice?
Identification of occupations associated with drug overdose deaths further characterizes the opioid epidemic. Incorporating workplace research and targeted interventions might benefit the opioid epidemic response. and employers (3,5,6). Continued evaluation of the effectiveness and impact of these programs and interventions are needed to prevent opioid misuse and abuse and to reduce opioid-related morbidity and mortality.