Surveillance of Occupational Fatality Trends Among Vulnerable Populations in North Carolina

SURVEILLANCE OF OCCUPATIONAL FATALITY TRENDS AMONG VULNERABLE POPULATIONS IN NORTH CAROLINA

Morgan Miller Richey

A dissertation submitted to the faculty at the University of North Carolina at Chapel Hill in partial fulfillment of the requirement for the degree of Doctor of Philosophy in the Department of Epidemiology in the Gillings School of Global Public Health.

Chapel Hill 2021

Approved by:

David B Richardson

Yvonne M Golightly

Alexander P Keil

Stephen W Marshall

Wendy M Novicoff

ABSTRACT

Morgan Miller Richey: Surveillance of Occupational Fatality Trends Among Vulnerable Populations in North Carolina (Under the direction of David Richardson)

North Carolina is one of the fastest growing states in the nation, experiencing rapid population and economic growth as it transitions from an economy dominated by agriculture and manufacturing to a post-industrial technology and service economy. North Carolina is highly diverse, including a growing elderly and Latinx population. Sustaining economic growth requires developing, attracting, and retaining workers. Of great import to employees, employers, and public health, is the occurrence of occupational injury. Reduction of occupational injury and fatality rates in the 20th century has been hailed as one of the major achievements in public health. In this dissertation I investigate trends of fatal occupational injury in North Carolina during the period between 2000 and 2017.

The first aim of this dissertation examines rates of fatal occupational injury in different racial and ethnic groups. Latinx workers, the fastest growing ethnic group in North Carolina, also had the highest rate of fatal occupational injuries during the study period. Standardization, employed to adjust for age, sex, industry and occupational differences between groups, suggest that there were 134 more deaths among Latinx workers during the study period than would be expected if they experienced the fatal occupational injury rates of White non-Latinx workers.

iii

In the second aim, rates of fatal occupational injury among older workers, defined as 55 years of age or greater, are estimated before and after the economic recession of 2008. In the pre-recession period between 2000 and 2008, the rate of fatal occupational injury among workers over 55 years of age was in decline. In the post-recession period this trend reversed, and fatality rates began to increase.

In the final aim, rates of fatal occupational injury due to homicide, suicide and overdose are examined. Contrary to population trends, we found that fatal occupational injury due to homicide and suicide declined during the study period, while rates of overdose increased.

Surveillance of fatal occupational injury is an important component of injury prevention and a sustainable, equitable economy. Further investment in surveillance, as well as investigation into the etiology and prevention of fatal occupational injury is necessary to continue to reduce occupational fatality.

To my parents, who never gave up. To my friends, who never gave me a break.

To those who are no longer with us.

ACKNOWLEDGEMENTS

First, I must thank my parents, who have always had my back during my winding path through life. Thank you to David Richardson, my chair, for his unflagging guidance, mentorship, and inspiration during this work. I feel privileged to have such a supportive, insightful, and enjoyable mentor and to have been able to contribute to valuable and informative work. A special thank you to Wendy Novicoff, if not for her belief and continuous investment in my potential I would not be here. I would also like to thank Maryalice Nocera, without question the most competent project manager I have ever had the pleasure to work with. I would like to thank Samantha Wooten, with her incisive eye for detail and an indefatigable work ethic. I would like to extend a special thanks to the OINC study team for their dedication, grit and commitment to ensuring that no stone was left unturned. Finally, a special thank you to my partner, Jessica, who has supported me throughout this process, and has shown me what real grit is.

TABLE OF CONTENTS

LIST OF TABLES ...... xi

LIST OF FIGURES ...... xii

SPECIFIC AIMS ...... 1

CHAPTER 1: BACKGROUND AND SIGNIFICANCE ...... 3

1.1 History of Occupational Fatality Surveillance in the United States...... 6

1.1.1 Energy Transfer and the Epidemiologic Triangle of Occupational Fatality ...... 8

1.1.2 Conceptual Underpinnings of Workplace Safety, Culture, and Climate ...... 9

1.1.3 The Recurring Costs of Workplace Safety ...... 10

1.2 Trends in Occupational fatality ...... 12

1.2.1 National Trend Data ...... 12

1.2.2 Opiate Availability ...... 13

1.2.3 Demographic Change ...... 14

1.2.4 High Risk Industries and Occupations ...... 15

1.3 A Shift Towards Occupational Fatality Investigation in North Carolina...... 15

1.3.1 Relevance to the US South ...... 16

1.3.2 The Rise and Fall of Industries...... 17

vii

CHAPTER 2: MATERIALS AND METHODS ...... 21

2.1 Case Ascertainment ...... 22

2.1.1 Office of the Chief Medical Examiner (OCME) and State Center for Health Statistics (SCHS) ...... 23

2.2 Case definition...... 25

2.2.1 Study Variables Defined Based on OCME Fatality Data ...... 27

2.3 Outcome Definition and Variable Validation ...... 29

2.4 Rate Denominator Data Based on US Census Bureau Microdata ...... 32

2.5: Analytic Methods ...... 32

Aim 1 Approach ...... 32

Aim 2 Approach ...... 33

Aim 3 Approach ...... 34

Human Subjects and IRB Approval ...... 34

CHAPTER 3: RACIAL AND ETHNIC DISPARITIES IN OCCUPATIONAL FATAL INJURY TRENDS IN NORTH CAROLINA ...... 35

3.1 Introduction ...... 35

3.2 Methods ...... 36

3.2.1 - Study Setting ...... 36

3.2.2 - Fatal Occupational Injuries ...... 37

3.2.3 - Population at Risk ...... 38

viii

3.2.4 - Latinx Ethnicity ...... 38

3.3 Results ...... 40

3.4 Discussion ...... 42

3.5 Tables and Figures ...... 45

CHAPTER 4: FATAL OCCUPATIONAL INJURY RATES AMONG OLDER WORKERS BEFORE AND AFTER THE GREAT RECESSION OF 2008 ...... 49

4.1 Introduction ...... 49

4.2 Methods ...... 50

4.2.1 Estimates of the Population of Older Workers...... 51

4.2.2 Statistical Methods ...... 52

4.3 Results ...... 53

4.4 Discussion ...... 56

4.5 Tables and Figures ...... 60

CHAPTER 5: FATAL OCCUPATIONAL INJURY DUE TO HOMICIDE, SUICIDE, ALCOHOL POISONING OR OPIATE OVERDOSE ...... 67

5.1 Introduction ...... 67

5.2 Methods ...... 68

5.2.1 Population at Risk ...... 69

5.3 Results ...... 70

5.4 Discussion ...... 71

5.5 Tables and Figures ...... 75

CHAPTER 6: DISCUSSION ...... 78

6.1 Study Strengths ...... 78

6.1.1 Exposure Classification ...... 79

6.1.2 Outcome Classification...... 79

6.1.3 Population at Risk ...... 80

6.2 Study Limitations ...... 80

6.2.1 Potential for Misclassification of Outcome ...... 81

6.2.3 Potential for Misclassification of Ethnicity in Cases: ...... 82

6.2.3 Potential for Misclassification of Population at Risk ...... 83

6.3 Funding...... 85

6.4 Conclusions ...... 85

REFERENCES ...... 88

LIST OF TABLES

Table 2.1 - Study Variables ...... 28

Table 3.1 – Change in North Carolina Workforce: 2000-2017 ...... 45

Table 3.2 - Occupational Fatality Rate (per 100,000 worker-years) During Two Calendar Periods and Estimated Annual Average Change in Fatality Rate, by Race and Latinx Ethnicity. North Carolina, 2000-2017...... 48

Table 4.1 - Change in North Carolina Workforce Composition During Study Period by Age Group ...... 60

Table 4.2 - Change in Distribution of Workforce by Industry: Pre-Recession (2000-2007) Vs Post-Recession (2009-2017), N.C...... 61

Table 4.3 - Leading Industries by Count of Intentional and Unintentional Injuries. N.C. 2000-2017 ...... 62

Table 4.4 - Leading Industries by Rate of Fatal Occupational Injury and Intentionality ...... 63

Table 4.5 - Rates of Fatal Occupational Injury Among Age Groups by Study Period, N.C., 2000-2017 ...... 64

Table 5.1 - Occupational Fatality Change During the Study Period: 2000-2017, N.C...... 75

Table 5.2 - Leading Industries by Count of Homicide, Suicide and Overdose Deaths. N.C. 2000-2017 ...... 75

Table 5.3 - Leading Industries by Rate of Fatal Occupational Injury Due to Homicide, Suicide and Overdose ...... 76

LIST OF FIGURES

Figure 2.1 - Flow Chart of Data Abstraction Process ...... 22

Figure 3.1 - Fatal Occupational Injury Trends by Race and Latinx Ethnicity (adjusted) ...... 48

Figure 4.1 - Annual Fatal Occupational Injury Rate Among Older Workers (55 Years of Age or Older): 2000-2017, N.C...... 65

Figure 4.2 - Annual Fatal Occupational Injury Rate among Workers 65 Years of Age and Older: 2000-2017, N.C...... 66

Figure 4.3 - Years of Potential Life Lost Among Older Workers: 2000-2017, N.C...... 66

Figure 5.1 - Yearly Rate of Fatal Occupational Injuries: Homicide, 2000-2017, N.C...... 76

Figure 5.2 – Yearly Rate of Fatal Occupational Injuries: Suicide, 2000-2017, N.C...... 77

Figure 5.3 – Yearly rate of Fatal Occupational Injuries: Overdose, 2000-2017, N.C ...... 77

xii

SPECIFIC AIMS

The frequency and distribution of occupational fatalities in North Carolina has changed dramatically in the last 40 years as the occupational landscape has evolved from an agricultural and manufacturing dominant economy to post-industrial sectors such as service and construction.

Identifying trends in fatal occupational injury, including disparities experienced by groups of workers, is critical to informing state and national efforts to identify and reduce occupational fatalities. While economic and employment trends in North Carolina are unique, North Carolina has been considered representative of labor conditions in the South. North Carolina also has a rapidly growing population of Latinx workers, a group that has been shown to be more likely to be employed in high-risk occupations as well as experience a high rate of fatal occupational injury. In addition, North Carolina has a substantial aging workforce, especially in agriculture, whose members may be at high risk for occupational death as mechanized farm equipment becomes more common.

Economic and cultural transitional forces in the United States continue to reshape the occupational landscape, and racial and ethnic disparities in occupational fatality have persisted in many industries. Vulnerable groups continue to shoulder a disproportionate burden of workplace fatalities, through (a) being more likely to work in a high-risk profession and (b) experiencing higher rates of fatal workplace injury than their colleagues. Understanding the distribution and determinants of occupational fatality during the economic transition from the early 2000’s to present is important to inform efforts to improve occupational safety.

In the service of identifying industry-specific trends in occupational fatality among vulnerable workers in North Carolina, I aim to:

Aim 1: Estimate long term trends in occupational fatal injuries in North Carolina among Latinx workers.

I leverage occupational death records abstracted from the Office of Chief Medical

Examiner (OCME) for the state of North Carolina, and US census estimates of the population at risk to estimate rates of workplace fatality among North Carolina workers who are determined to be Latinx. As North Carolina continues to benefit from an influx of Latinx workers, understanding the distribution of these workers in high-risk professions, as well as the rate of fatality within these professions, is important for prevention efforts. Recent research suggests that Latinx workers experience a higher rate of workplace fatality than their Black and White colleagues.

Aim 2: Estimate rates of fatal occupational injury among older workers during the study period, contrast the rates prior to the great recession of 2008 to those in the period following the recession.

The dramatic contraction of the economy as a consequence of the great recession of 2008 may have had disparate effects depending on class of worker, sex, age, and race, and contributed to overall trends in workplace death. I aim to estimate the rate of occupational fatality during the pre-crash and post-crash period, as well as the years of potential life lost among selected groups of workers.

Aim 3: Describe trends of suicide, homicide, drug or alcohol overdose, that have risen rapidly beginning in the year 2000 and have continued unabated.

Identification and surveillance of new sources of fatal occupational injury is important to inform directed research towards changes in policy and intervention. I estimate trends in

2 occupational fatalities caused by suicide, homicide, overdose, and alcohol poisoning, within selected groups of workers, during the period between 2000 and 2017.

CHAPTER 1: BACKGROUND AND SIGNIFICANCE

Previous research has identified groups at uniquely high risk of fatal occupational injury in North Carolina. Latinx workers may have higher rates than other ethnic groups and are the fastest growing ethnic group in North Carolina. Older workers, especially those over 65, have the highest rate of occupational fatality nationwide, and their participation in the workforce has never been higher.1 Older workers are also particularly vulnerable to economic shocks that affect retirement income,2,3 and the great recession of 2008 has been followed by historically high rates of workforce participation among these workers. While the rate of fatalities by suicide, homicide and overdose of opiates and alcohol have been rising,4–6 the rate of these death types in the workplace has only recently begun to be investigated.7–9 In this dissertation I (i) estimate and compare trends in occupational fatality among and between White non-Latinx workers, Black non-Latinx workers and Latinx workers, (ii) estimate occupational fatality rates and trends among older workers before and after the economic recession of 2008, and (iii) estimate rates and trends in occupational fatality due to suicide, homicide, and overdose of drugs or alcohol.

Fatal occupational injuries are overwhelmingly preventable events, and the surveillance and investigation of deaths at work is a bedrock component of an equitable, sustainable, and profitable occupational climate. While the rate of fatal occupational injury has declined significantly since the early 1900s, high rates persist in numerous industries. Within and between industries, disparities in fatality have been identified in race, sex, ethnicity, and age

4 groups. Within industries with nominally low rates of occupational injury there may be occupations or types of employees at unusually high risk. Technology has evolved considerably, resulting in an expansion of mechanized equipment in most industries, especially manufacturing, farming, and transportation. Economic transitions have also occurred, as more of the riskiest manufacturing, mining, and processing jobs are relocated overseas. In previous work on the topic, shrinking industries have been found to experience rising deaths at work, possibly due to decreases in investment in training and equipment.10 In contrast, industries that are experiencing rapid growth have been found to experience decreasing occupational fatalities. The rapidly evolving labor market, advances in equipment capability and affordability, and changes in workplace composition and culture make surveillance of occupational fatality an interesting and important challenge.

According to national statistics compiled by the Bureau of Labor Statistics (BLS), there were 5,250 fatal occupational injuries in 2018. According to the BLS, preventable deaths in the workplace are defined as all deaths not occurring due to natural causes, homicide, or suicide.

Occupational fatalities are unique in that they are overwhelmingly, if not completely preventable, and the measurable of improvement in occupational fatality rates in the last 100 years can be attributed, in part, to the imposition of regulation and standards by state and federal agencies. In the past 25 years, there have also been significant improvements in training methodology, advancements in the design of machinery, and organization-led efforts to improve workplace safety. Despite these improvements, annual occupational fatality rates in some industries and groups have plateaued, and in others increased.11 From an economic and a human perspective, the cost of each occupational fatality is immense. The National Safety Council estimates that each workplace fatality has a total economic cost of 1.2 million dollars, along with the loss of

5 years of life to the decedent and the immeasurable effect on their families and loved ones.12

1.1 History of Occupational Fatality Surveillance in the United States

The phrase “safety regulations are written in blood” is heard in a variety of industries and serves as a grim reminder of the reactive nature of workplace safety regulation in the United

States. Occupational fatalities have been a major driving force of workplace safety laws in the

United States. For example, in the Triangle Shirtwaist Factory fire in New York in 1911, 146 workers perished as they were unable to escape their locked factory, a gruesome incident that came only a day after progressive workplace safety laws were repealed in New York superior court.13 While the incidence of occupational fatality has decreased considerably since the 146 lives lost in this event, an average of 14 Americans died at work every day during 2015.14 The deaths of the Triangle Shirtwaist workers, most of them young women, shocked the United

States into action on workplace safety, while the 14 workers who perished each day in 2015 died in isolated events less visible to the public and generate less outrage. The effect of the fire was to highlight the consequences of poor occupational safety, and significant improvement has been made to the safety environment of the American workplace since. The National Safety Council reports that the national occupational fatality rate has decreased from over 37 deaths per 100,000 workers in 1933 to 4 per 100,000 in 1997.15,16 While great strides have demonstratively been made, there is still much work to be done.

Riding the wave of public outcry against the death of workers in the Triangle Shirtwaist fire, the flurry of research that followed focused on the motivations and behaviors of individual employees, with an early focus on identifying accident-prone employees, were conducted by

6 applied and industrial psychologists.17 This focus on the individual employee as source of workplace injury risk dominated the early 20thcentury approach to workplace safety, based on the concept that hiring “fit” employees was the primary driver of both productivity and safety.18

This approach spawned increasingly involved evaluation of individual employees, including batteries of tests for cognitive ability, reaction time, tolerance for repetitive actions, and increasingly in-depth simulation of daily duties.

After World War II, the lessons learned during the massive mobilization of an untrained workforce to wartime production yielded an awareness of the importance of training, and greater focus on the motivations of workers. Dissecting workplace injuries and fatalities after the fact became common, notably using Critical Incident technique, a precursor of the epidemiological triangle.19 This period of occupational safety research was characterized by a focus on iterative improvements in the training and management of workers in high-risk occupations. It is worth noting that epidemiologists were involved in both evaluating the correlates of occupational fatalities, and were instrumental in creating the theory of the epidemiologic triangle and the development of injury research as a scientific discipline.20 As the field of injury research continued to evolve, the focus on individual behavior and characteristics gave way to multilevel thinking, as the characteristics, culture and policies of organizations were identified as primary determinants of occupational fatality. The era of placing the onus of safety on the individual was over, and the focus shifted to intervening at the organizational level to improve workplace safety and reduce occupational injury and fatality.

1.1.1 Energy Transfer and the Epidemiologic Triangle of Occupational Fatality

The beginning of modern occupational injury research and practice was, fittingly, an occupational injury that could have been a fatality. After surviving what by all accounts should have been a fatal fall from an airplane during World War I, an engineer named Hugh DeHaven became convinced that the survivability of force that was directed at the human body was susceptible to statistical analysis.21,22 His initial studies in of the capabilities of the human body laid the groundwork for acceleration seats, flight suits and crash dummies. This work was incorporated into the work of Gibson, a psychologist, who defined injury as resulting in the transfer of energy to the human body above a threshold of tolerance.23 However, it wasn’t until a collaboration with Haddon, a physician and epidemiologist, that these observations about the relationship between force absorption and injury were rearranged into a unifying theory: prevention of injury depends on removing or reducing the transfer of energy to the human body.24,25

This framework allows injury epidemiologists to evaluate the forces that cause fatal trauma with a focus on how to remove or minimize them. In a machine shop, for example, the force transmission potential of a bench grinder is massive but can be minimized if the operator employs a face shield, heavy leather clothing and the wheel is regularly inspected and the grinding surface groomed. The power of the energy transfer theory is in its application to injury prevention: a well-designed question of how to prevent injury, when posed in the framework of energy transfer, contains its own answer.

1.1.2 Conceptual Underpinnings of Workplace Safety, Culture, and Climate

Formally, workplace safety is an amalgamation of two concepts, safety culture and safety climate.26 Culture is often comprised of top-down management policy, often dictated by insurance requirements, and crafted with the input of administrators, risk management teams and supervisory personnel. Culture describes the micro-level environment surrounding a worker, which is mediated by their direct supervisor, colleagues, and the physical environment in which they operate.27,28 In an ideal state, culture and climate will be well aligned, however in practice substantial variation between culture and climate in different departments of a workplace is often observed.29,30

Individual behavior is affected by climate and culture, in that workers seek to maximize their opportunity for advancement (being recognized by managers as worthy of promotion), while minimizing their exposure to disciplinary action (avoiding being seen by managers as a troublemaker or failure-prone). In this context, game theory has been used to analyze the micro- decisions that employees must regularly make as they go about their daily duties. Employees are sensitive to tradeoffs between safety and productivity, especially in high-throughput environments where local managers are held responsible for productivity goals. Employees are also keenly observant of the behavior of management when a safety issue comes to their attention. For example, if a critical piece of equipment necessary to a manufacturing operation has malfunctioned in a way that could cause a safety concern, but is still operable, employees can infer much about the alignment between culture and climate by observing the actions of management. The decision to immediately remove the equipment from service for repair will eliminate a threat to employee safety at the cost of productivity, in that one or more production lines will halt until the repair is complete. In companies with nominally strong workplace safety

9 standards, the expectation that the equipment would be taken down for repair soon after the problem is identified seems reasonable. When this does occur, employees see convincing evidence of alignment between workplace safety policies (culture) and the action of managers

(climate). In some establishments, however, management may decline to remove the piece of equipment from service, delay repair, or depend on employees to manage their risk while using the machine (perhaps with a sign on the machine describing the failure). In this environment, employees can come to the (perhaps correct) conclusion that the hierarchy of priorities in their workplace places productivity at the top, with safety below. Thus, safety culture begins to diverge from the local, proximal safety climate. While these decisions are nominally the mandate of local managers, supervisors, and foremen, in some organizations they may lack sufficient administrative authority to remove a machine from service or remediate a potentially dangerous situation. It is an unfortunately regular occurrence in many industries, that workers observe that upper management’s commitment to safety is in name only, and that policy prioritizes productivity, even in the face of strenuous objection by supervisors. This realization can create an environment where the worker believes, often correctly, that the organization is willing to overlook risks to workers if productivity can be maintained or improved.

1.1.3 The Recurring Costs of Workplace Safety

Initial investment in protective gear by safety-conscious companies is sometimes seen as a one-time investment, especially in companies where administrative personnel are less familiar with the work environment on the production floor or in the field. In these situations, rigid enforcement of policy (such as mandating protective eyewear) without the accompanying regular investment to replace worn, damaged, or missing safety equipment can lead to non-compliance

10 by employees who compare the risk of being caught in a safety violation with the real risk of injury due to using defective or substandard safety equipment. Disconnects between policymakers, supervisors and those charged with operational budgets are often a source of conflict in the world of workplace safety. Regular inspection of equipment, repair, and maintenance, purchasing new safety equipment, and investing in training are significant monetary outlays for organizations. As injuries that do not occur cannot be measured, companies that are successfully preventing workplace death can fail to see the results from regular investment in workplace safety. Similar to the success of vaccination against disease in modern medicine, organization can struggle to recognize the absence of events as a real measure of thoughtful policy, effective supervisors, and diligent employees. Those who wish to evaluate effective safety policies and performance are often limited to measurements such as days since last event – a common feature at construction sites. As companies face financial shortfalls, this disconnect may contribute to cutting costs that affect safety, including the delaying the purchase of new equipment, training sessions and maintenance on critical machinery. A useful example from the author’s experience involves the safety/budget tradeoffs unique to the transportation industry, which spends a large proportion of their maintenance budget on tires. Studies have consistently shown that the last 4 millimeters of tread life in a truck or car tire has rapidly decreasing grip, and low tire tread is associated with higher rates of crash and fatality.31,32

Failure to replace these tires is false economy, as the expenses of even infrequent crashes and occupational injuries or fatalities are likely to outweigh the cost of tires. While it is an accepted concept that replacing tires before grip thresholds are reached is a worthwhile investment in reducing crashes, for a company with 1,000 semi-trucks, extending this range down to the last remaining millimeter could result in significant cost savings. These savings may come at the

11 expense of human life: The National Highway Traffic Safety administration reports that in 2018,

738 road fatalities were tire-related.33

Accurately evaluating these tradeoffs can be difficult for managers with limited experience in risk management as well as the day-to-day duties of employees they employ.

Surveillance of occupational fatality trends can contribute to understanding of the distribution of fatal occupational injury to inform better regulatory and risk management strategies. For example, in response to reports on tire-related deaths by the NHTSA, in the year 2000 the U.S. congress passed the TREAD act, which mandated the construction of a system by which a vehicle operator would be alerted to an underinflated tire.34 By September 1, 2007 all vehicles sold in the United States were equipped with automatic tire pressure monitoring systems.

1.2 Trends in Occupational fatality

1.2.1 National Trend Data

Based on national BLM data, the number of occupational fatalities observed per year was trending downwards between 1990 and 2008, when the trend reversed, and fatalities began to increase. The rate of occupational fatality, defined as the number of deaths per 100,000 workers, followed this pattern until 2008, then plateaued. In 2008, the BLS changed their calculations of rates from employment-based rates to hours-based, and now report fatalities per 100,000 full- time equivalent (FTE) work hours instead of workers.35 This change in measure may explain the discontinuity between a steadily increasing count of deaths nationwide and a stable death rate.

This plateauing of the occupational fatality rate may be associated with the economic crash of

2008, if the contractions of the economy forced the same number of workers to work higher-risk overtime hours, or a bolus of newer and less experienced workers entered the labor force.

1.2.2 Opiate Availability

Cultural changes are also at play in the safety environment of the workplace, in this work the focus will be on an unprecedented increase in availability of opiates now widely acknowledged as “the opioid epidemic,” which is of particular concern for occupational safety.

Exposure to opiates is not evenly distributed throughout the workforce. Workers at the highest risk of musculoskeletal injury are often also employed in industries or occupations that are known for high risk of occupational fatality, as work with heavy and dangerous equipment, animals or in extreme environmental conditions confers increased risk of both injury and death.

Clinical treatment of musculoskeletal injury during the period between 2000 and 2017 was characterized by a new focus on pain control, with widespread use of opiates for pain control and benzodiazepines for muscle relaxation. Both drugs are respiratory depressants with high addictive potential. In 2017, the age-adjusted rate of overdose was 3.6 times that of 1999, and increased in all age groups.6 This rapid increase in prescription of opiates and benzodiazepines is highly relevant to workplace fatality, as treatment for work-related musculoskeletal injury has been shown to be common among opiate overdose deaths,36 and may represent the index exposure of many Americans to these addictive and psychoactive drugs. Polydrug use is also common among opiate overdoses, twenty-three percent of opiate deaths in 2015 tested positive for benzodiazepines, highlighting musculoskeletal injury treatment as a potential precursor to and multiplier of opiate death risk.37 Adding evidence to these suspicions are studies evaluating risk factors for opioid-related deaths, in which work-related injuries have been found to be strong predictors of opioid death.36 The effects of opiates in this context can be twofold: opiates are

13 known to cause issues with proprioception, balance and perception of risk as well as cause death when taken in moderately large amounts, especially with alcohol.38–40 Alcohol use is also common among workers in industries with some of the highest rates of occupational injury,41–44 potentially further increasing risk, as workers in industries at the highest baseline risk of fatality are prescribed medications that may facilitate further injury from drug interactions, fatal overdose or external causes. This simultaneous transition from blue-collar to a whiter-collar economy, a rapid cadence of economic shocks starting in 2001, including a recession in 2008, as well as the massive increase in prescribing of opiates have yet to be fully investigated in the occupational safety literature.

1.2.3 Demographic Change

The demographic composition of the United States’ labor force is also changing. This change is most noticeable in the older age groups, as older workers continue to participate in the workforce at a historically high rate. This older group of workers, especially in agriculture, have been shown to experience the highest risk of occupational fatality, potentially due to changes in proprioception, balance and the accumulation of musculoskeletal injury.45,46 This disproportionate burden of occupational fatality born by older adults has continued to increase, in

2016 the rate of fatal injuries in the 65 and older age group was 9.6 per 100,000 – more than 2.5 times the average rate of 3.6 per 100,000, and more than double the rate in 55–64-year-old workers.47

Economic stressors, including the recession of 2008 that will be examined in this dissertation, may have contributed to older adults remaining in the workforce. Employees who would normally have retired in the years proximal to 2008 may have been forced to delay

14 retirement due to loss of retirement savings. Older workers may also face ageism in their job search, including unfamiliarity with newer technology, potentially relegating them to higher-risk industries and occupations. Older workers who are forced to delay retirement may also face job loss and a difficult re-entry into the labor market, where they may take a position in which the risks are unfamiliar, and thus even further increase the rate of occupational injury in this age group. These factors, among others, may contribute to the observed historically high rates of workforce participation among older workers and the hypothesized increase in fatalities in the period following the 2008 economic crash.

1.2.4 High Risk Industries and Occupations

Traditionally agriculture, mining, transportation, and construction have been high-risk industries for occupational fatalities. Between 2012 and 2018, the industries with the highest rate of occupational fatality at the national level were agriculture (including hunting, fishing, and forestry), mining, transportation, construction, and wholesale trade. These 3 high-rate industries are characterized by interaction with heavy equipment, objects, and dangerous animals in remote environments with long distances to medical care.

1.3 A Shift Towards Occupational Fatality Investigation in North Carolina

North Carolina is a unique venue in which to evaluate trends in the occupational fatality landscape. North Carolina has experienced extremely rapid growth in the last 25 years, including a 50% increase in population accompanied by relatively high rates of employment.

North Carolina moved from the 10th to the 9th most populous state during the study period, at 10

15 million residents, and has enjoyed a massive influx of diverse new residents, including one of the fastest-growing Latinx populations in the United States.48 North Carolina also has a substantial aging population, a group that has been recently observed to delay retirement and stay employed longer than historically observed.49,50 Representative of the “new south,” North Carolina has experienced a rapid transition from agriculture and manufacturing to a more diversified and service-dominated economy, with concomitant changes in worker training, experience, and injury risk. This rapid economic transition is a result of advanced planning by a relatively far- seeing state legislature that saw the economic climate changing and saw the need to invest in retraining and reshaping the economic climate in NC.51 Investment in the Research Triangle has been one of the more successful arms of this investment and has helped to transform the North

Carolina piedmont from a primarily agricultural and manufacturing economy to include growing biotechnology, information technology and medical research sectors.52 This transition away from industries traditionally known to experience higher rates of occupational fatality would be expected to be associated with a reduction in the overall rate of fatal workplace injury, however evidence exists that these industries may experience increasing rates of fatality as they decline.10

Much work remains to be done as the occupational, social, and economic landscape continues to shift, and inference from analyses in North Carolina may be generalizable to other southern states as nationwide economic forces continue to shape the occupational landscape.

1.3.1 Relevance to the US South

Examination of occupational fatality trends in North Carolina may be applicable to other southern states. North Carolina has a distinctly southern occupational climate, characterized by low union membership, relatively deregulated industries and a legacy of disparities in occupational fatality compared to northern states.53 North Carolina residents also continue to

16 shoulder the enduring legacy components of slavery, as Black workers have been found to be at the highest risk of occupational fatality in the south,53 as well as North Carolina.54 This emerging signal that Latinx workers are experiencing rates of occupational fatality significantly higher than other groups is of great concern. Some of this increase may be explained by increased participation in industries with high risk of fatality, however early analyses have suggested that even within these industries, Latinx-Americans experience anomalously high rates of fatal occupational injury.53

1.3.2 The Rise and Fall of Industries

Recent investigation of workplace fatality in North Carolina have revealed other problematic trends, including increasing rates of fatality in industries that are experiencing a shrinking workforce.10 An opposite trend has also been observed, where industries experiencing rapid growth have workplace deaths rates that are trending downwards. The legacy of North

Carolina as a prior manufacturing and agricultural powerhouse is also important in this context, as the transition from dominance in these sectors has left a large cohort of workers to retrain.

1.3.3 Vulnerable Populations

Racial and Ethnic Groups

Occupational fatality is almost always preventable. In North Carolina, this is particularly problematic given the fact that occupational fatality disproportionately affects workers with lower income, less formal education, and members of minority groups.53–55 Historically, Black

Americans experienced the highest rates of workplace fatality, and there appears to be little evidence in the literature of concerted efforts to intervene on these high rates. Of equal concern

17 is recent data suggesting that the rate of occupational death among Latinx workers is equal or greater than Black workers.53 This is plausible as immigrants, especially those for whom English is not their first language, have historically been given the riskiest and lowest paying jobs, and census data has identified a significant increase in the Latinx population in North Carolina.48 As with other states, especially in the South, North Carolina depends on immigrant labor for some of the highest risk jobs. This dependence has resulted in disproportionate representation of

Black and Latinx workers in industries with the highest baseline rates of occupational fatality, especially agriculture, construction, and meat processing. Even within these high-risk industries, preliminary data suggests that the occupational fatality rates among Latinx workers is higher than White and Black workers.

Physical Occupations, Musculoskeletal Injury and Substance Use

Occupational fatalities include deaths that occur at work caused by both licit and illicit substances. Alcohol consumption within 12 hours of a work shift has been associated with increases in the rate and severity of workplace injury, especially in occupations that require operation of heavy machinery and manipulation of heavy loads.44,56 Homicide, suicide and overdose on drugs or alcohol have been observed to be increasing in the general population.57–60

This suite of outcomes could be sentinel events, as increases in these outcomes in the population may be a precursor to increases in the occupational environment. While the rates of these outcomes have been trending upwards, their distributions and determinates have only recently been investigated in the workplace.7,8,61 Of special concern is the association between difficult occupations, especially those that involve grueling physical labor, and resulting non-fatal workplace musculoskeletal injuries.45,62 The treatment of these injuries in the United States often

18 involve benzodiazepines as muscle-relaxants and opiates as pain killers, as discussed previously.

For many workers, their index exposure to opiates may be due to an injury or disease associated with their work.36 This relationship may be more common in some groups, for example the observed increase in the rates of legally prescribed coincided with the “greying” of the population of American workers, who may have an accumulated burden of musculoskeletal injuries from years of work. We hypothesize that this relationship between nonfatal work injury, increases in hazardous medications to treat injuries, and an older workforce, may be associated with upward trends in overdose among workers. Opiate overdose has a complex pathophysiology, as a dose taken hours before work, or even the night before, can cause respiratory depression and death hours later, after the worker has arrived at work. For this reason, onset of injury and time of death are critical factors in classifying an overdose as a workplace fatality.

Older workers

Older Americans are participating in the workforce at historically high rates, workers 55 and older comprise more than 20.3% of the workforce, and projections indicate that older workers will continue to increase their participation rate.63 This older workforce, especially in agriculture, has been shown to experience the highest rates of fatality, potentially due to changes in proprioception, balance and the accumulation of musculoskeletal injury.45,46,63 Given the composition of the workforce in 2018, the 4.8% of workers 65 or older account for 14.8% of the occupational fatalities.64 This is even more dramatic in agriculture, forestry, fishing and hunting where workers 65 and older account for 32.9% of all fatalities, more than any other age group.

When workers 55 and older are considered, they experience 55.1% of all fatalities in agriculture,

19 forestry fishing and hunting – supporting the hypothesis that age-related changes to proprioception and situational awareness are associated with fatality in industries requiring high levels of caution, coordination and attention.14

This disproportionate burden of occupational fatality born by older adults has continued to increase, in 2017 the rate of fatal injuries among workers 65 and older was estimated at 10.3 per 100,000 worker-years, – more than 2.5 times the average rate of 3.5 and more than double the rate in 55–64-year-old workers.47

CHAPTER 2: MATERIALS AND METHODS

Research Setting

North Carolina is the ninth largest state in the nation, with over 10.5 million residents.

During the period between 2000 and 2017, the population of North Carolina grew by over 3 million residents, most of whom migrated from other states.65,66 The economic climate of North

Carolina has been in transition, as ongoing industrial restructuring from a primarily agricultural and manufacturing economy toward technology, research and service has provided significant job opportunities. Considered representative of the “new south,” North Carolina is demographically diverse, with a large African-American population, a sizeable elderly population, and a rapidly growing Latinx population.48,67 During the study period, significant change in the economic, demographic and social environment occurred, including aforementioned industrial restructuring, the aging of the workforce, increases in Latinx immigration and the onset of the opiate epidemic. It is in this labile setting that we have investigated trends in fatal occupational injury among groups we hypothesized would be most vulnerable.

2.1 Case Ascertainment

Accurate identification and Figure 2. 1 - Flow Chart of Data Abstraction Process verification of individual cases of Deaths classified as Deaths classified as “Occupational” by the “Occupational” by the occupational fatality are of high Office of the Chief Medical State Center for Health Examiner Statistics importance when conducting analyses Study Population of rates. One of the major challenges

Probabilistic faced by researchers who investigate matching of death certificates rates of fatal occupational injury is that

Abstraction of death of verifying whether the injury that certificate and associated evidence. caused death occurred while working for pay. Researchers regularly depend Final adjudication of cases on accurate classification of fatalities

Final study in electronic databases of state population agencies charged with tracking fatalities and determining cause and manner of death, which often includes an occupational activity flag for decedents working at the time of death. This determination is often challenging, complicated by self-employed workers, workers for whom commuting is part of their job, and professions who may be considered “on-duty” when not officially at work – such as police officers, nurses, and doctors. Our case ascertainment process takes a highly sensitive approach, first informed by determination of occupational fatality by two state agencies, the State Center for Health Statistics (SCHS) and the Office of the Chief Medical Examiner (OCME). Each case contained in either of these databases is validated with a comprehensive, case-by case review by trained investigators who abstract pertinent information from each death record. Records

22 reviewed by abstractors include the official death certificate, coroner report, autopsy results, witness and family narratives and the final certified medical examiners report. Narratives recorded by police officers, coworkers and medical professionals who interviewed or treated mortally injured workers are often crucial pieces of evidence, rarely captured in electronic databases, to assist in verifying that a decedent was engaged in work activity at the time that the fatal injury occurred. I considered all civilian deaths on North Carolina soil that were determined to be at work by either agency as eligible for review if the time between injury onset and death was 30 days or less.

2.1.1 Office of the Chief Medical Examiner (OCME) and State Center for Health Statistics (SCHS)

The responsibility for investigating fatalities in North Carolina falls to county medical examiners, appointed, and supervised by the Chief Medical Examiner (CME). The chief medical examiner is appointed by the state secretary is a board-certified forensic pathologist who is licensed to practice medicine. Article 16 Statutes § 130A-383 of the NC general statutes describes the jurisdiction of medical examiners:

“Upon the death of any person resulting from violence, poisoning, accident, suicide, or

homicide; occurring suddenly when the deceased had been in apparent good health or

when unattended by a physician; occurring in a jail, prison, correctional institute, or in

police custody; or occurring under any suspicious, unusual or unnatural circumstance, the

medical examiner of the county in which the deceased is found shall be notified…”

- NC General Statutes § 130A-383

Occupational fatalities are given special attention by medical examiners. During an investigation of an occupational fatality the medical examiner considers a variety of information from several sources. Medical examiners are empowered to seek administrative search warrants, request medical records, perform autopsies and toxicology screenings, and review the testimony of witnesses and family members. A certificate of death is completed by the medical examiner that explicitly states the disease or means of death, the manner of death, age, sex, place of birth, and whether the death occurred at work. Death certificates are reviewed by the chief medical examiner, who has the right to conduct additional procedures including transport of the body to the primary facility for additional autopsy, tissue sampling and other investigative efforts. In deaths determined to be homicides, the final copy of the death certificate is forwarded to the district attorney for use in criminal and civil cases. Occupational fatalities are classified based on a combination of witness and police testimony, location of death, hospital records and coworker and family interviews. Age, sex, race, and ethnicity of decedents is determined through a similar process and are coded by both the OCME and SCHS. Final determination of an occupational fatality triggers the forwarding of a copy of the certificate to the Commission of

Labor for review.68

Death certificate data is digitized by the OCME and transmitted to the State Center for

Health Statistics, a state agency charged with health surveillance. Analysts at the SCHS link

OCME death records with demographic information derived from vital statistics databases, code manner of death using International Classification of Disease (ICD) coding and classify industry and occupation using U.S. Census categories. This demographic linkage compliments the detailed death information, yielding a richer and more comprehensive source of information than the OCME alone. This additional information is critical to analysis as they provide further

24 context on deaths, provide additional information for the adjudication process, and serve as a validation of the electronic records provided by the OCME.

2.2 Case definition

All civilian, noninstitutionalized persons engaged in work for pay who are above 18 years of age are eligible for inclusion in this study. I restrict my analysis to decedents who were physically in North Carolina when they sustained a workplace injury that resulted in death within

30 days of injury, regardless of documentation or citizenship status. Two state databases, provided by the OCME and SCHS, provide an initial list of all decedents classified as occupational fatalities between 2000-2017, and are the sample frame from which cases are drawn. Trained investigators abstract all available information on each occupational fatality from OCME case files, including autopsy, toxicology screening, description of personal effects, interviews with family members, first responders and law enforcement. Cases that do not fully meet criteria for inclusion, or cases in which the determination is questionable, are presented to the primary investigator team in bi-weekly meetings and adjudicated on a case-by-case basis.

2.2.1 Analytical Challenges in Case Acquisition and Enumeration

Several challenges confront researchers who conduct analyses using administrative data, including the rarity of complete data in each dataset and the concomitant necessity of integrating multiple data sources to achieve completeness. The office of the Chief Medical Examiner

(OCME) holds physical copies of all materials pertaining to deaths investigated by the OCME in

North Carolina; these files are the focus of our abstraction efforts. As discussed previously,

25 occupational fatality flags from either the SCHS or the OCME were used to compile a list of potential occupational fatalities eligible for abstraction. The first challenge was encountered after both datasets were cleaned and formatted, a mismatch between death certificate numbers was observed: no unique key linking the SCHS and OCME database was present.

The OCME uses an in-house death certificate numbering system that does not correlate to the official death certificate number, and a case folder identifier that corresponds only to the physical location of the death certificate, autopsy results and associated records. In contrast, the

SCHS assigns a non-unique death certificate number to decedents; the first death for every year during the study period was assigned the death certificate number 000001. This lack of a unique key to link each database necessitated a probabilistic matching operation to link decedents identified by the SCHS as occupational fatalities with the physical location of the death file at the

OCME archives.

The fidelity of a probabilistic matching is dependent on the level of congruence between potential matching variables. Maximum probability of accurate matching is obtained when matching variables are recorded in the exact same format in both databases. Unfortunately, the

OCME and SCHS had significantly different structure for critical matching variables. Variations in naming convention, including middle names as initials compare to full spelling, accommodating multiple middle names (especially important for the Latinx cohort), and considering suffixes (Jr., Sr., III, etc.) as part of the last name or distinct variables, are a short list of the differences between the databases. Additional variation between age, birth, and death date, likely due to errors in character recognition software, clerical errors and differences between interview-derived age and birth certificate derived age were also present. The misalignment of these two databases in terms of typical matching variables necessitated a finely

26 tuned probabilistic linkage operation to match SCHS-identified occupational fatalities with their

OCME records.

The CDC-developed linkage tool Link Plus (Version 2.0, Center for Disease Control,

Atlanta GA) was employed to probabilistically match cases identified as on-the-job in the SCHS database with the physical location (case folder number) at the OCME archives. Matching variables (also known as blocking variables) included death date (within 3 days), sex, last name, first and middle name. Fuzzy matching methods were employed on name variables to account for differences in spelling, length, suffixes, and inversions of matching variables. Match results were checked by hand, and mismatches resolved through database search and retrieve operations.

Final match results included a 100% match of SCHS cases to their respective OCME case folder, except for 1 ME case verified as truly missing from the SCHS database. Newspaper articles,

OSHA reports and obituaries were obtained to verify the name, sex, ethnicity, date of death, age, and occupational activity at time of death.

A study team was formed, and investigators trained to interpret and abstract both hand- written and print death records and related files in the case folders at the OCME. These include the official death certificate, medical examiner report including toxicology and autopsy results, narratives from witnesses, family, and police as well as industry and occupation at the time of death. Data were coded into a Microsoft Access database by abstractors and imported into SAS software (SAS 9.4, Cary N.C) for cleaning, formatting, and analysis.

2.2.1 Study Variables Defined Based on OCME Fatality Data

Trained investigators abstracted three types of data from case folders at the Office of the

Chief Medical Examiner: demographic, situational and occupational, summarized in Table 2.1.

Table 2.1 - Study Variables

Demographic OCME SCHS Study Team

Age (in years), X X

Sex (coded as a binary variable), X X X

Race (coded as White, Black, Other) X X X

Ethnicity (Hispanic Yes/No) X X X

Situational OCME SCHS Study Team

Immediate cause of death (free text) X

Contributing cause of death (free text) X

Underlying cause of death X X X

(coded to ICD9),

Description of injury leading to death (free X X

text)

Manner/Means of death X X

(coded as accident, homicide, suicide)

Manner/means of death (free text) X X

Narrative of death (free text) X

Occupational OCME SCHS Study Team

Usual occupation/industry X X

Occupation/industry at time of death X X

2.3 Outcome Definition and Variable Validation

Aim 1 and Aim 2

Any fatal injury determined to have been sustained at work, with death occurring within

30 days, was counted towards the outcome.

Aim 3

Homicide cases were identified using the final manner of death from the medical examiner and were each validated by the study team.

Suicide cases were also identified using the final manner of death (“Suicide”) from the medical examiner’s office.

Overdose cases were identified using the final means of death (“Poisoning”) from the medical examiner’s report. Toxicology screening results were also utilized to verify the cause of poisoning deaths (e.g., not due to carbon monoxide poisoning), as well as identify potentially misclassified overdose cases. For example, three decedents who were classified as accidental deaths by the medical examiner had toxicology results that, in the opinion of the medical examiner, identified an amount of drug or alcohol in their systems to be sufficient to cause death independently, and thus were classified as overdose cases.

2.3.1 Validation of Study Variables

Two electronic databases provide validation of variables abstracted for analysis. The disparate nature of the data sources used allowed for comparison of the determination of each death by data source. Database operations conducted in each aim were distributed to the subsequent (and previous) aims.

Aim 1: Latinx Ethnicity

Validation of ethnicity was required due to 89 hand-written ethnicity entries, entered in the section for “Race” on the official death certificate. Review of these certificates revealed that country of origin was often entered in place of race. Decedents whose race was entered as

“Hispanic,” “Latino,” or a country in Latin America (e.g., Guatemala, Mexico) were classified as

Latinx. Race was considered missing.

Aim 2: Age

Age was determined using the age at death derived from the death certificate. Missing ages were calculated using birth dates and death dates. Implausible birth dates, primarily due to storing of year as a 2-digit number, were identified and compared to calculation of birth and death dates. Recorded ages that differed by birth/death-date calculations were evaluated; two decedents had discrepancies that would potentially put them in a different age category The correct ages for these decedents were established using supplemental materials, including online obituaries and news articles.

Aim 3: Intentionality and Means of death

The final determination of the intentionality and means of death was determined by abstractors, and the circumstances surrounding the fatal injury recorded.

Homicide cases were all classified as homicides by the OCME, however there was one suicide case that was considered undetermined. The circumstances surrounding this case were verified as suicide, involving deliberate running of a car exhaust hose into the decedents office, resulting in carbon monoxide poisoning.

Overdose cases were classified by a manner of death of “poisoning” that included alcohol overdose, prescription, or illegal drugs, as determined by the medical examiner, validated with toxicology screening data provided by the OCME. Among the 15 overdose cases, 4 were initially classified as accidents – perhaps intending to indicate accidental overdose as opposed to deliberate overdose. In three of these cases, the amount of drug detected in the toxicology screen was considered sufficient to cause death independently by the medical examiner, with no evidence of suicidal intent. Based on this information, we classified these 3 decedents as overdose deaths. One decedent either did not receive a toxicology screen, or it was negative (the

OCME database does not distinguish between null and absent toxicology screen results). For this decedent, the medical examiner determined the cause of death to be due to poisoning due to probable ephedrine toxicity. This cause of death meet our criteria for overdose, however further investigation was performed. A review of toxicology and pharmacokinetics of ephedrine, a common stimulant used illicitly by athletes, suggests that ephedrine has a short half -life (3-6 hours) and is rapidly eliminated in urine, suggesting that a delayed screening is a plausible explanation for the lack of positivity on toxicology screening.69–71 Based on these data, we classified this decedent as an overdose death.

2.4 Rate Denominator Data Based on US Census Bureau Microdata

The population at risk in each stratum was estimated for each year of the study using linear interpolation between decennial census population estimates in 2000 and 2010, as well as with American Community Survey (ACS) estimates. Estimates for the population at risk for each year was cross classified by age, sex, race, ethnicity, industry, and occupation.

2.5: Analytic Methods

For each aim, unadjusted rates of occupational fatality will be modeled with Poisson regression using yearly counts of deaths derived from adjudicated case reports and a census- derived, denominator of the population at risk cross classified by age, sex, race, ethnicity, industry, and occupation included as a regression model offset. For aim 1 and 2, comparisons between groups and time periods, respectively, will be conducted using a standardized mortality ratio.

Aim 1 Approach

I aim to estimate crude rates of occupational fatality among Latinx workers during the study period. I construct a standardized mortality ratio (SMR) contrasting the expected number of deaths among Latinx workers by transporting the cross-classified age, sex, occupation, and industry rates of White workers to Latinx workers.

Poisson regression models with a log link were employed to estimate rates of fatal occupational injury in three groups of workers: White non-Latinx, Black non-Latinx and Latinx workers. I will derive unadjusted yearly rates of occupational fatality using the equation:

log(퐹푎푡푎푙푖푡푦푅푎푡푒퐴푛푛푢푎푙) = 훽0 + 훽1(푌푒푎푟) + 훽2(퐿푎푡푖푛푋) and report the annual percentage change in rate by exponentiating the coefficient of β1 , subtracting 1 and multiplying by 100. The log-transformed estimate of the population at risk in each calendar year, cross-classified by age, sex, industry, and occupation, will be used as the offset. Industry specific rate estimates for the study period will be estimated using the formula log(퐹푎푡푎푙푖푡푦푅푎푡푒 ) = 훽0 + 훽1(퐼푛푑푢푠푡푟푦)

Estimates of the age, sex, industry and occupation-specific rates of fatal occupational injury in

White, non-Latinx workers will be applied to Latinx workers, the SMR calculated, and the count of observed vs expected fatalities compared.

Aim 2 Approach

I aim to estimate unadjusted rates of occupational fatality among groups of older workers

(over 55 years of age), contrasting the years prior to the 2008 crash to the years following it, along with calculations comparing the years of potential life lost (YPLL) in the two periods. I will calculate the SMR for this group by standardizing the age, sex and employment structure of the second period (2009-2017) to that of the first period to control for changes in demographics, and industry structure.

Generalized linear models utilizing a Poisson regression link will be employed to estimate rates of fatal occupational injury before (2000-2007) and after (2009-2017) the great recession of 2008 among workers over 55 years of age. Estimates of fatality rates by calendar year will be calculated, and direct standardization approaches employed to estimate the effect of employment structure and increased participation in the workforce on annual rates. I will estimate the crude yearly rate of occupational fatality among older workers using the equation: log(퐹푎푡푎푙푖푡푦푅푎푡푒퐴푛푛푢푎푙) = 훽0 + 훽1(푌푒푎푟) + 훽2(퐴푔푒퐺푟표푢푝)

33 and report the annual percentage change in rate by exponentiating the coefficient of β1 and multiplying by 100. Years of potential life lost will be calculated using the difference between a decedents age at death and 75. Decedents older than 75 will contribute zero years to YPLL estimates.

Aim 3 Approach

I aim to estimate unadjusted rates of occupational fatality due to homicide, suicide and overdose by opioids or alcohol during the study period using the equation: log(퐹푎푡푎푙푖푡푦푅푎푡푒퐴푛푛푢푎푙) = 훽0 + 훽1(푌푒푎푟) where the outcome, FatalityRate, here denotes the fatal occupational injury due to homicide, suicide, drug overdose, or alcohol poisoning. As with the other aims, I will report the annual percentage change in rate by exponentiating the coefficient of β1 and multiplying by 100.

Human Subjects and IRB Approval

This work was approved by the IRB of the University of North Carolina, Chapel Hill

(IRB# 20-3674).

CHAPTER 3: RACIAL AND ETHNIC DISPARITIES IN OCCUPATIONAL FATAL INJURY TRENDS IN NORTH CAROLINA

3.1 Introduction

High rates of occupational fatality have historically been observed among Black non-

Latinx and Latinx workers, especially in the South.54,72,73 Latinx is a gender-neutral term that imperfectly groups peoples from Latin America and the Caribbean, a highly heterogenous group with a diverse set of cultural practices and language. North Carolina has seen an increase in the

Latinx population, increasing from 380,000 residents in 2000 to 997,000 residents in 2017.48

Prior studies of occupational fatality rates in the South have identified disparities in fatal occupational injury rates across racial and ethnic groups.53,55 These disparities are also observed at the national level,53,74–76 and persist despite an observed decline in employment in industries and occupations with high rates of occupational fatality, such as agriculture and manufacturing, and an expansion in the comparatively lower-rate service, retail, and information technology sectors.

Rates of fatal occupational injury have been steadily decreasing since the 1970s.11,77–79

Interpretation of this decrease is complicated by shifts in demographics, employment patterns, and macroeconomic trends towards globalization. For example, in the South, there has been decline in employment in industries such as agriculture and manufacturing, with concomitant rapid growth in the service and biotechnology sectors. Prior work identified high rates of fatal occupational injury among workers employed in industries where the size of the labor force was

35 declining , but concluded that less than 10% of the decrease in occupational fatality rates can be attributed to the decline of high-rate industries.80

The Latinx population is a steadily growing proportion of the overall workforce, in part due to migration. Recently arrived migrant and other under-resourced populations are often employed in occupations that are hazardous and/or may be assigned more hazardous tasks within an occupation/industry group;73 and, they tend to have high rates of fatal and nonfatal occupational injury.73,75,81,82 The purpose of this paper was to examine recent trends in fatal occupational injury among Latinx workers in North Carolina, relative to non-Latinx workers, during the period between 2000-2017 that was characterized by rapid growth of the Latinx population. We used standardization to estimate the expected mortality among Latinx workers had they experienced the age, sex, industry, and occupation specific fatality rates of White non-

Latinx workers.

3.2 Methods

3.2.1 - Study Setting

North Carolina is located in the Southeastern United States and is the ninth most populous state in the US. While historically the state’s economy was largely based on textile production, furniture manufacturing, and tobacco and other agricultural crops, since the 1980s the state’s urban population has rapidly grown, and the growth of the state’s economy has substantially shifted towards service sector and technology industries.

3.2.2 - Fatal Occupational Injuries

The North Carolina Medical Examiner system is a well-established statewide reporting, coding and data retrieval system maintained by the Office of the Chief Medical Examiner

[OCME]. The OCME is part of the Division of Public Health in the North Carolina Department of Health and Human Services. Investigating fatal injuries is the jurisdiction of local medical examiners in each of the 100 counties in North Carolina. Each medical examiner determines the cause of death and circumstances surrounding it, including whether the injury that led to death occurred at work.68 Individual medical examiners’ findings are reported to the OCME whose staff includes forensic pathologists and data coders.

We defined an occupational fatality as any injury leading to death within 30 days, both intentional and unintentional, sustained by a worker in North Carolina engaged in legal work for pay. Occupational fatality cases identified for review were drawn from two data sources; the

OCME data system and North Carolina State Center for Health Statistics death certificate data system. Deaths during the study period [2000-2017] that were flagged as “at work” in the State

Center for Health Statistics data system or “on the job” in the OCME data system were eligible for study inclusion.

Detailed information pertaining to occupational fatalities between January 2000 and

December 2017 was abstracted from the records of the Office of the Chief Medical Examiner.

These files contain the Medical Examiner’s report, the official death certificate, autopsy results, toxicology results and, if available, family interviews, witness, and police statements. A trained team of investigators abstracted study data from the physical death records, as well as from supplemental sources such as news clippings, court transcripts and crime scene reports. A

37 review of the circumstances surrounding each death was conducted, and complex cases adjudicated by experienced investigators to make a final determination of “at work” status for purposes of inclusion.

Information on the age, sex, race, and ethnicity of decedents was abstracted from the death certificate and the medical examiner report. Information on usual occupation and industry of the decedent was abstracted from the death certificate; and information on the occupation, and industry at time of fatal injury was abstracted from the medical examiner report. The current report concerns only the decedent’s occupation and industry at the time of fatal injury.

Occupations and industries were coded to the U.S. Census year 2000 guidelines.

3.2.3 - Population at Risk

Annual estimates of the North Carolina work force, cross-classified by age, sex, race

(classified a White, Black, or other), ethnicity (Hispanic or non-Hispanic), occupation (in 33 groups defined by census codes), and industry (in 51 groups defined by census codes),80 were derived from the 2000 decennial US Census and the 2004-2017 American Community Survey

(ACS). We use annual census estimates of the working population as an estimate of the number of person-years at risk in that stratum in that calendar year. The American Community Survey replaces the long-form decennial census and provides yearly estimates of the population of the

United States, including demographic and occupational information, beginning in 2004.

3.2.4 - Latinx Ethnicity

Classification of decedents according to race and ethnicity was based on information recorded on the NC state death certificate. This included information identifying Latinx

38 decedents (categorized as Hispanic on the death certificate) and detailing the decedent’s country of origin. While information on race of the decedent was relatively complete overall, when

Latinx ethnicity was identified on the death certificate, most decedents (80%, n=189) were missing information on race. In most cases, country of origin was hand-written on the death certificate. Rather than attempt to cross-classify Latinx decedents by race, we classified decedents as either Latinx, White non-Latinx, or Black non-Latinx for this analysis. We acknowledge the likelihood of significant unmeasured heterogeneity by country of origin, language spoken, race and other factors.

Statistical Methods

To examine fatal occupational injury rates, and trends in rates over the study period, we modeled the annual rate of occupational fatality using Poisson regression, where the number of deaths due to fatal occupational injury in a given calendar year, cross-classified by categories of age, sex, race, ethnicity, occupation, and industry was modeled as a Poisson distributed variable, and the natural logarithm of the annual estimate of the North Carolina work force in that group served as a model offset. The unadjusted model for a crude linear trend in rates was:

퐸[퐿푛(퐴푛푛푢푎푙 푅푎푡푒)] = β0 + β1 (year − 2000). Calendar year was included as a continuous variable to estimate the average change in rate per year. The variable was centered by subtracting

2000 so that the regression model intercept, β0, can be interpreted as the log of the baseline fatality rate in the year 2000. Estimated trends in the rate of fatal injury can be reported as the annual percent change in rate, obtained from the model as 100[exp(β1) − 1]. Age and sex- adjusted fatal occupational injury rate estimates were calculated by including a binary indicator

39 variable for sex (1 if male, else 0) in the regression model, and including indicator variables for age group (10-year age groups beginning at 18) in the Poisson regression model.

To assess whether differences in annual rates and rate trends between Latinx and non-

Latinx workers were due to differences in demographic and employment characteristics, standardized mortality ratios (SMR) were calculated to evaluate differences between expected and observed counts of occupational fatality in Latinx workers.

3.3 Results

Between 2000 and 2017 a total of 1,775 occupational fatalities were observed. Of these,

237 (13.4%) decedents were identified as Latinx. Of the Latinx deaths, 230 (97%) were male.

During the years under study, the NC labor force increased by 28%, from 3.8 million to 4.8 million. During this same period, the Latinx labor force in NC increased by an estimated 226%, from 183,800 workers to 414,800 (Table 3.1). The proportion of Latinx workers in the workforce increased from 4.8% of the workforce in 2000 to 8.7% of the workforce in 2017.

A substantial number of fatalities among Latinx workers occurred in utility and repair services, while proportionately fewer White non-Latinx or Black non-Latinx workers died in these industries (Table 3.2). Black non-Latinx workers experienced a larger proportion of fatalities in food and related manufacturing operations, and in water and sanitation, compared to

White non-Latinx and Latinx workers. White non-Latinx workers saw higher numbers of fatalities in the professional services than Black non-Latinx and Latinx workers. Industry- specific rates of occupational fatality tended to be similar between groups (Table 3.3), although

Black non-Latinx and Latinx workers had higher annual fatal occupational injury rates in forestry, hunting/fishing/trapping, and transportation than White non-Latinx workers.

The unadjusted rate of occupational fatality among Latinx workers declined during the study period, from a high of 5.7 per 100,000 workers in the first half of the study (2000-2008) to

3.4 during the second half (2009-2017) (Table 3.4). Black non-Latinx workers also saw significant decreases in the unadjusted rate of occupational fatalities, decreasing from 2.7 deaths per 100,000 workers in the first half of the study to 1.8 in the second half of the study period.

White non-Latinx workers experienced a similar reduction in unadjusted occupational fatality rates over the study period, decreasing from 2.7 deaths per 100,000 workers to 2.0. Age and sex-adjustment reduced the rate estimates in White non-Latinx workers but increased the rate in

Black non-Latinx and Latinx workers (Table 3.4).

To account for differences in the age, sex, and employment distribution of Latinx workers relative to White non-Latinx workers, standardization was used to calculate the expected number of deaths among Latinx workers if they had experienced the age, sex, industry, and occupation specific rates of White non-Latinx workers, expressed as a SMR. The SMR comparing the observed to expected deaths among Latinx workers was 2.31 (95% CI: 2.0, 2.6), indicating that fatal occupational injury rates for Latinx workers was more than twice as high as

White non-Latinx workers after accounting for different distributions of age, sex, industry, and occupation between these groups.

3.4 Discussion

The rate of fatal occupational injury declined in North Carolina between 2000-2017, especially among Black non-Latinx workers. Latinx workers also experienced a decline in occupational fatal injury rates, but Latinx workers had the highest rate of fatal occupational injury of the three groups. Standardization approaches suggest higher observed than expected deaths among Latinx workers, where expected numbers were based on fatal occupational injury rates among White non-Latinx workers; we estimate that the fatal occupational injury rate was more than twice as high among Latinx workers during the study period as among White non-

Latinx workers.

We found evidence of disparities in occupational fatality rates, with persistently high fatal occupational injury rates among Latinx workers compared to White non-Latinx workers.

Historical forces underpin differential occupational fatality rates in southern states, including structural segregation of employment in which racial or ethnic groups are overrepresented in occupations with the highest rates of fatality. However, standardizing the Latinx population to the age, sex, and employment patterns of White non-Latinx workers revealed that segregation in employment does not fully explain the differences in fatal occupational injury rates. We acknowledge that our measure of employment does not account for all types of segregation, and that differences in our standardized measure may be due to different distribution of hazardous tasks, personal protective equipment, and other factors. We estimated the standardized mortality ratio (SMR) to be 2.31, suggesting that the number of occupational fatalities among Latinx workers is higher than we would expect if the rage, sex, and employment specific rates of Latinx workers were similar to White non-Latinx workers. While full explanation for these trends is

42 beyond the scope of this work, persistently high rates of occupational fatality concurrent with a period of population growth is of significant concern. Our results are consistent with research showing Latinx workers as experiencing the highest rate of occupational fatality nationwide.74,76,82,83

Early work on construction safety noted characteristics of the worker, including English- language proficiency, documentation status, cultural factors, and the desire to appear industrious, as contributors to risk disparities between groups of workers. 84,85 Recent work has increasingly focused on the occupational environment surrounding the workers, suggesting that understanding of labor laws and protections available to workers, perceptions by employers that Latinx workers will accept risky assignments, and poor communication (including by supervisors and coworkers) also are likely to contribute to the disproportionate burden of occupational fatality.86,87 For example, workers unfamiliar with American labor laws may not be provided personal protective equipment, be threatened with termination for reporting injuries, and experience higher levels of job insecurity.81,88,89 Workers may also be less likely to decline overtime, not be provided training for unfamiliar equipment or be made aware of the availability of PPE, or feel empowered to decline a hazardous task or report an unsafe work environment.82,90

Immigrant workers, especially those in temporary positions, have been shown to experience additional pressure to meet deadlines driven by fear of termination, and are more vulnerable to employer abuse that contributes to occupational health disparities.91,92 Declining to perform a dangerous component of a job due to inclement weather or other safety concern may also result in job loss, which may affect their ability to get another job through “blacklisting,” or by affecting immigration status.81

Our study has several strengths, including a robust system of medical examiners in North

Carolina who are required to investigate accidental and intentional deaths, and are empowered to seek administrative search warrants as part of their investigations. We employed a case ascertainment process that leveraged multiple state databases to identify potential occupational fatalities.

Some limitations of our analysis include our focus only on fatal occupational injuries that result in death within 30 days. Workers who sustain cumulative injuries, such as silicosis or asbestosis, would not be included in our estimates. We note that 80% of Latinx decedents were missing data on race, limiting our ability to distinguish between groups of Latinx workers whose workplace experience was likely to have differed. Recent immigrants also bring diverse work culture and practices to the U.S. workplace. The term “Latinx” encompasses workers from many countries, with different mother tongues, English competency, and varying cultural expectations around work, a concept that may not be well understood by many Americans. Bias may be present in our denominator, since prior research suggests that undercounting of Latinx workers may be present in Census data, especially in areas of larger Latinx populations.93–95 Systematic undercounting of the Latinx workforce could yield an overestimate of the yearly rate, however, we note that estimates of trend are robust to misclassification of the population at risk, provided that misspecification of the workforce is consistent over time.

Given persistent historical inequities between racial and ethnic groups, the overall decrease in fatality rates among all groups is encouraging. Black non-Latinx workers in North

Carolina, historically a group with high levels of occupation fatality, have rates similar or lower to White non-Latinx workers. We note a concerning increase in rate during the final years of the study, suggesting the need for continued surveillance. While this general decline in fatality rates

44 represents measurable progress by regulatory agencies, stakeholder groups, managers and individual workers, the fatality rate among Latinx workers has remained significantly higher than their White non-Latinx counterparts. This and other recent work suggest that continued surveillance coupled with state-level investigation of the etiology of high rates of occupational fatality is warranted.

3.5 Tables and Figures

Table 3.1 – Change in North Carolina Workforce: 2000-2017

2000 2017 2000 2017 Change Change Workforce Workforce Percentage Percentage Group (2000- (2000- size (in size (in of the of the 2017) 2017) thousands) thousands) workforce workforce

Latinx 183 415 232 4.8% 8.7% 81.3% Black 665 973 308 17.7% 20.5% 15.8% Non-Latinx White 2,783 3,153 370 74.0% 65.2% -11.9% Non-Latinx Other 129 269 140 3.4% 5.6% 64.7% Non-Latinx Total* 3760 4750 990 100% 100% --

Table 3.2 - Leading Industries by Number of Fatal Occupational Injuries by Race and Latinx Ethnicity: NC 2000-2017

Latinx Fatal Black Fatal White Fatal injuries non-Latinx injuries non-Latinx injuries

Transport Construction 122 59 Construction 251 Bus/Truck (51.5%) (19.5%) (20.5%)

Agriculture 42 Construction 54 Agriculture 125 (17.7%) (17.8%) (10.2%) Transport Forestry/Logging 6 Agriculture 15 160 Bus/Truck (2.5%) (5.0%) (13.1%) Justice, Sawmills/Wood 6 Forestry 13 Public 64 Products (2.5%) (4.3%) Order/Safety (5.2%) Justice, Wholesale 6 Public 12 Retail 49 (2.5%) Order/Safety (4.0%) (4.0%)

Retail 6 Wholesale 11 Wholesale 44 (2.5%) (3.6%) (3.6%) Business/Repair Food and 5 10 Forestry 43 Svc Assoc. Mfg (2.1%) (3.3%) (3.5%) Water Transport Professional 4 Supply/ 10 42 Bus/Truck Services (1.7%) Sanitation (3.3%) (3.4%) Automotive Utilities 4 Retail 10 Sales/Service 29 Electric/Gas (1.7%) (3.3%) Repair (2.4%) Automotive Warehouse Transport / Sales/Service 4 9 Storage/ 26 Taxi Repair (1.7%) (3.0%) Transport (2.1%)

Table 3.3 - Leading Industries by Fatal Occupational Injury Rate, by Race and Latinx Ethnicity. NC 2000-2017

Black Non- White Non- Latinx Rate/105 Rate/105 Rate/105 Latinx Latinx Fishing / 103.6 314.5 42.3 Forestry Hunting / Forestry (46.6, 230.7) (79.1, 1,258.6) (31.4, 57.0) Trapping Fishing / Transport/ 54.0 72.8 41.9 Forestry Hunting / Taxi (17.4, 167.5) (43.9, 120.9) (24.8, 70.7) Trapping 23.4 27.6 Transportatio 17.0 Utilities Transport/Taxi (8.8, 62.3) (14.3, 53.0) n/Taxi (7.1, 40.9) Detective / 12.4 14.2 15.3 Protective Agriculture Agriculture (1.8, 88.1) (9.4, 21.3) (13.0, 17.8) Services Constructio 10.0 Transportation 14.0 Transportatio 13.5 n (8.4, 12.0) Bus / Truck (10.8, 18.1) n Bus/Truck (11.3, 16.1)

9.9 11.5 Gas Service 12.6 Agriculture Mining (7.3, 13.4) (1.6, 81.5) Stations (7.9, 20.0)

9.3 11.2 Stone / Glass / 9.0 Sawmills Construction (4.2, 20.8) (8.6, 14.7) Concrete Mfg (5.8, 14.1)

Justice/Pub 8.4 Water Supply 10.9 Warehouse 7.1 lic Safety (2.1, 33.6) Sanitation/Utili (5.9, 20.2) Storage/ (4.9, 10.5) ty Transport Transportat 6.6 Gas Service 9.4 Justice/Public 6.6 ion (2.5, 17.7) Stations (3.5, 24.9) Safety (5.1, 8.4) Bus/Truck Stone / Glass/ Transportation 6.5 7.2 6.4 Concrete Rail / Water / Construction (1.6, 26.0) (2.7, 19.2) (5.7, 7.3) Manufactur Air ing

Annual 95% Confidence Group 2000-2008 2009-2017 change Interval Unadjusted Latinx 5.7 (125) 3.4 (112) -5.2% (-7.7%, -2.7%) Black 2.7 (179) 1.8 (137) -3.6% (-5.8%, -1.4%) White 2.7 (683) 2.0 (541) -3.4% (-4.6%, -2.2%) Age and Latinx 6.0 (125) 3.8 (112) -5.0% (-7.7%, -2.3%) sex Black 3.2 (179) 1.3 (137) -3.7% (-5.9% -1.5%) adjusted White 2.5 (683) 1.5 (541) -3.8% (-5.0%, -2.6%)

Figure 3. 1 - Fatal Occupational Injury Trends by Race and Latinx Ethnicity (adjusted)

Age/Sex Adjusted Fatal Occupational Injury Trends by Race and Ethnicity

8.0

6.0 4.0 2.0 0.0

Deaths Deaths per worker 105 Years CalendarYear

White Black LatinX

CHAPTER 4: FATAL OCCUPATIONAL INJURY RATES AMONG OLDER WORKERS BEFORE AND AFTER THE GREAT RECESSION OF 2008

4.1 Introduction

Older workers historically have been identified as experiencing the highest rate of fatal occupational injury among all age groups.96–100 Changes in proprioception, balance and visual acuity, that are normal parts of the aging process, may predispose older workers to injury, while an injury that would have been severe but survivable for a younger worker may be fatal to an older worker.46,101 Technology has also altered the work environment, with increasing use of heavy equipment, computerized systems and power tools that may be unfamiliar to workers trained prior to the advent of these newer technologies.

Research on older workers has identified high levels of skill, consistent delivery of high- quality work, and less absenteeism compared to younger workers.102,103 These experienced employees may identify workplace hazards more rapidly, mediate interpersonal conflict in the work environment, as well as discourage unsafe behaviors in younger workers, although they are often victims of bullying and prejudice themselves.104,105 High rates of fatal occupational injury in this group is of particular concern not only from an occupational safety perspective but also because of the social role that older adults play; older workers make valuable contributions to the economy and society, and are often deeply embedded with their community as well as their extended family.

We examine fatal occupational injury rates among older workers in North Carolina during an 18-year period, centered on the great recession of 2008, during which many workers experienced employment changes. The great recession affected different age groups of workers differently; and, we hypothesize that older workers experienced different stresses that affected their employment patterns than younger workers (e.g., deterioration of retirement savings).3

Older workers also may find it more challenging to retrain or move locations to find work, and have been shown to experience employment discrimination due to their age.106–108 In this work, we estimate trends in fatal occupational injury rates among older workers, as well as years of potential life lost, during the years surrounding the recession of 2008.

4.2 Methods

Fatal occupational injuries of all types are required to be investigated in the state of North

Carolina by local medical examiners, overseen by the chief medical examiner in the state capital of Raleigh. Medical examiners are charged with determining the venue, cause of death, and circumstances that lead to each fatal injury.68 For the purposes of this study, a fatal occupational injury is defined as any injury resulting in death within 30 days, whether intentional or unintentional, sustained by any person in North Carolina who is working for legal pay. Trained investigators abstracted demographic and occupational data from the records of the Office of the

Chief Medical Examiner, including the official death certificate, autopsy results, family interviews, and witness and police statements, of all fatal occupational injuries between January

1, 2000, and January 31, 2017.

All decedents 55 years of age or older (at death) are eligible for inclusion, regardless of immigration status. Information on age, sex and ethnicity was used to categorize decedents, in 6 age groups (ages 25-34, 35-44, 45-54, 55- 64, 65-74, 75+ years), 3 race groups (White, Black,

Other), manner of death (accident, suicide, homicide) and ethnicity (Latinx or non-Latinx) derived from the death certificate and medical examiner report. Occupation and industry at time of fatal injury were abstracted from the medical examiner report and coded to the U.S. Census year 2000 guidelines. The recorded age at death was compared to the difference between the decedent’s recorded birth date and death date; decedents for whom this derived age at death differed by more than 1 year from the recorded age at death, or with discrepancies in age sufficient to move them from one age group to another, were adjudicated using supplemental materials, including news articles, obituaries, and memorials.

4.2.1 Estimates of the Population of Older Workers.

The decennial US censuses and the American Community Survey were used to generate cross-classified estimates of the yearly population of workers in North Carolina in age, sex, race, ethnicity, occupation (in 33 groups defined by census codes), and industry (in 51 groups defined by census codes) categories. Estimates in each stratum include workers who are classified as currently participating in the labor force, by answering “yes” to questions regarding labor force participation and “work for pay,” on either the 2000 decennial census or the American

Community Survey.

4.2.2 Statistical Methods

To evaluate yearly rates of occupational fatality, and rate trends in the pre-recession and post-recession period, we employed Poisson regression modeling techniques. The overall unadjusted model for a crude trend in rates is expressed as:

log(annual rate) = β0 + β1 (Year − 2000) . Calendar year was entered as a continuous variable to estimate the average change in rate per year. Sex-adjusted estimates of fatal occupational injury rates were calculated by including a binary indicator variable for sex in the regression model. To estimate the trend in rates in the pre-recession period, this model was fitted to the study data restricted to the period 2000-2007. To estimate the trend in rates in the post- recession period this model was fitted to the study data restricted to the period 2009-2017. We report estimates of the yearly percent change in fatality rate in each period derived from the fitted

Poison regression model as 100[exp(β1)−1]. To evaluate our hypothesis that the rates in the two periods would be different, we included a binary variable for the post-recession period (1 if post, else 0) as well as an interaction term between this variable and calendar year. The contrast in rate between the two periods is estimated using the Wald test statistic for the interaction term between the post-period variable and calendar year, evaluated at the α=.05 level. To evaluate the difference in trends of occupational fatality between the two study periods, a model was fit to the data for the entire period 2000-2017 of the form log(퐴푛푛푢푎푙푅푎푡푒) = β0 + β1(Year − 2000) +

β2(Y2008) + β3(PostRecession) + β4(Y2009), where the variable Y2008 is a binary variable that takes a value of 1 in 2008, else 0, the variable ‘post-recession’ is a binary variable that takes a value of 1 if year>=2009, else 0; and, the variable Y2009 is a variable that takes a value of

(year-2008) if year>=2009, else 0. The contrast in rate trend between the pre and post period is

52 estimated by exponentiating the β4 coefficient, and a test of the null hypothesis that the trends are equal was evaluated using a Wald test statistic at the α=.05 level.

Because the distribution of age differed between the pre and post period (Table 4.1), standardized mortality ratios (SMR) were calculated to evaluate differences between expected and observed counts of occupational fatality in older workers. We used the age, sex, occupation, and industry-specific rates observed in the pre-recession period as the reference rates; we applied these to the person-time in the post-recession period to derive an expected number of occupational fatal injuries if stratum-specific rates in the post-recession period had been identical to that of the same stratum in the pre-period. The ratio of the observed number of occupational fatal injuries to this expected number is reported as an SMR, along with its associated 95% confidence interval. Years of potential life lost (YPLL) were calculated, defined as the difference between age of death and 75 years of age, decedents 75 years and older contribute zero years to

YPLL calculation. Adjusted estimates of years of potential life lost are calculated by standardizing the age distribution and workforce participation proportion between the two periods.

4.3 Results

Between 2000 and 2017, 1,775 occupational fatalities were observed in North Carolina, of which 469 (26.4%) were identified as 55 years of age or older. Four hundred and thirty decedents (95.7%) were male, 396 (84.4%) were White, 73 (15.6%) were Black, and 19 of

(4.1%) were identified as Latinx and White. During the years between 2000-2017, the NC labor force increased by an estimated 28%, from 3.8 million in 2000 to 4.8 million in 2017. Half of

53 this growth was due to a doubling of the number of older workers from 506,000 workers to

1,055,000 during the same period. The proportion of older workers in the workforce increased from 13.5% of the workforce in 2000 to 22.0% of the workforce in 2017. All age groups in

North Carolina increased in size during the study period, with the largest percentage increase among older workers (Table 4.1). The change in the workforce among selected industries are detailed in Table 4.2, stratified by greatest absolute and percentage change, as well as greatest decline.

The industries with the highest counts of death are detailed in Table 4.3, along with the distribution of manner of death in each industry. Older workers employed in construction, agriculture and transportation died in the largest numbers, while the rates of occupational fatality were highest in forestry, hunting fishing and trapping, and transportation (Table 4.4).

Disaggregating fatal occupational injury counts and rates into intentional and unintentional injuries reveals overlap between the highest number and highest rate industries. For example, grocery stores and gas stations were among the highest 3 industries for both rate and number of intentional occupational fatality.

During the entire study period, the 18 years between 2000 and 2017, the estimated annual rate of fatal occupational injury declined among older workers by an estimated 2.8 percentage points per year (95% confidence interval: -4.5%, -1.0%) compared to an estimated 4.1% (-5.2, -

3.0) per year among younger workers (Table 4.5, Figure 4.1). Comparing the average rate of fatal occupational injury in the pre-recession period to the post-recession period, a decline of approximately 1.0 deaths per 100,000 workers (4.0 vs 3.0, Wald χ2 = 5.1, p=.02) was observed.

After estimating the rate during the entire study period and comparing the average rate in the pre- recession and post-recession period, we examined whether annual trends in fatal occupational

54 injury rate among older workers increased in the post-recession period. In the pre-recession period, spanning the years 2000-2007, the fatal occupational injury rate among older workers decreased by an estimated 7.1 percentage points per year (-11.7%, -2.2%), while younger workers experienced an estimated decline of 9.6 percentage points per year (-12.7, -6.4). In the post-recession period, fatal occupational injury rates among older workers increased by an estimated 0.9 percentage points (-3.8%, 5.9%) per year, and younger workers experienced an increase of 2.6 percentage points per year (-1.0, 6.2). Workers 65 years and older experienced annual rate increase in both periods, increasing by 1.4% (-9.0%, 12.9%) per year in the pre- period and 4.4% (-3.9%, 13.5%) per year in the post-period (Table 4.4, Figure 4.2).

Years of potential life lost, defined as the difference between age of death and 75 years, increased, from an average annual loss of 295 years in the pre period to 335 years in the post period (Figure 4.3). We calculated an adjusted estimate of the relative years of life lost between the two periods to account for the age distribution and workforce size differences between the two periods. We estimate we would have observed more than 105 additional annual years of life lost in the post period if the age and workforce participation were identical to the pre-recession period.

Standardization of the post-recession rates of fatal occupational injury to the age, sex, industry, and occupational rates of the pre-recession period produced an SMR of 0.82 (95% CI:

0.72, 0.92), consistent with the overall decline in rates of fatal occupational injury among older workers.

4.4 Discussion

The proportion of the workforce composed of older workers increased steadily during the two study periods (Table 4.1). During the years immediately following 2008, the workforce participation among older workers shrunk, (Figure 4.4) likely due to a variety of recession- related outcomes including job loss, acceptance of early retirement offers, and inability to find work. By 2011, workforce participation among older workers recovered, and continued to increase throughout the post-recession period. A strong economic climate in the first half of the study period may have fueled the increase in pre-recession employment of older workers, as previous studies have suggested an association between economic improvement and increased employment among older workers.109 We hypothesized that the great recession of 2008 would further increase this participation by delaying the retirement of the oldest workers via loss of value of retirement investments.2,3 Distance to retirement has been identified as a significant driver of employment patterns among older workers, especially in competitive labor markets.110–

112 Increased participation rates during economic downturn can relate to fatal occupational injury rates via a variety of mechanisms, including (but not limited to): (a) older workers entering dangerous or unfamiliar professions after loss of their primary job, (b) decreased spending by organizations on new machinery, equipment, and safety training and (c) a decrease in the number of available workers for dangerous tasks, potentially increasing the requirements for overtime and probability of fatal incident.

We hypothesized that the rate of occupational injury in the period following the recession of 2008 would increase. We found evidence that, contrary to our hypothesis, the rate of fatal occupational injury among older workers declined from a mean 4.0 deaths per 100,000 worker-

56 years in the pre-recession period to 3.0 deaths in the post-recession period (Wald χ2 = 5.1 p=.02).

While the rate of fatal occupational injury declined, the pre-recession trend of yearly declines in fatal occupational injury rate reversed, from a decline of 7.1% per year in the pre-recession period to an annual increase of 0.9% in the post-recession period (Table 4.4). We estimate that the annual rate of fatal occupational injury among workers 65 and older increased in both the pre and post period, with a greater estimated yearly increase in the post period. Total years of potential life lost increased among all groups of older workers, while standardization of the age and workforce participation of the post-recession period to that of the pre-recession period suggests that this increase is largely explained by greater participation in the workforce by older workers.

The aging of the baby boomer generation is receiving greater attention as a significant public health concern. Older workers have historically experienced high rates of fatal occupational injury, and our results suggest that while the recession of 2008 was associated with a rate decrease, there is evidence that rates may be increasing in the post-recession period.

Changes in the economic landscape may have, and continue to, affect trends in fatal occupational injuries among older workers through a variety of mechanisms, including rapid technological advances in many industries that have increased the amount of heavy machinery present on job sites. If these trends continue, we can expect further increases in the fatal occupational injury rate as well as years of potential life lost as workforce participation continues to increase, and workers continue to age into higher-rate strata.

Older workers may be uniquely vulnerable to occupational fatality as rapid technological advances remodel the workplace.113 Workers expert in methods in use prior to technological change may find it harder to adapt to due to unfamiliarity with computerized systems, frustration

57 with having to retrain, or general resistance to change.114 Training or retraining may also be especially challenging for older workers, especially in unfamiliar, fast-paced or technology- dense professions.115 Age-related changes may also contribute to the relationship between age and fatal occupational injury, as proprioceptive decline, including balance, vision, and hearing, may increase the likelihood of falls or collisions, while decreasing a worker’s ability to see a hazard to hear a warning signal.46,101,116 Older workers may also experience elevated levels of financial uncertainty, especially during the period surrounding the great recession, that combined with ageism in hiring may make changing jobs more difficult than for their younger colleagues.

117,118 Physical limitations are also a factor in employment transition for older workers, as the accumulated burden of musculoskeletal injuries,45 the potential necessity for ergonomic accommodations,119 and reductions in a workers ability to endure extreme environmental or strenuous labor conditions may reduce competitiveness in the job market.120,121 Companies may also perceive, sometimes correctly, that older workers may have higher health insurance costs, greater rates of absenteeism, and take longer to recover from injuries.122

The loss of older workers is of concern to their families, public health as well as the companies that employ them. Older workers are often charged with training employees and may have extensive experience in safely completing hazardous tasks. The death of these employees also represents a loss of institutional memory functions, including those that identify policies that will increase hazards. Older workers may also play a mediating role as organizations seek to improve their profitability (or stem their losses) during a downturn, by pushing back against austerity initiatives that could increase hazards, such as reducing training/onboarding time, decreasing the frequency of maintenance, or assigning inexperienced employees to hazardous jobs. The presence of older workers may also exert a stabilizing effect in some industries,

58 decreasing bullying, horseplay, risk-taking and posturing among younger workers. While older workers themselves may have high rates of occupational fatality, the loss of older workers in the workplace may have unexpected effects on the occupational fatality rates of their colleagues.

As demographics continue to shift in the United States, and the largest generation in

American history continues to age, we present evidence of a post-recession trend towards increasing fatality rates of older workers as a significant public health concern. While fatality rates declined from their highest levels in the year 2000, this trend stalls or reverses after the recession of 2008. Among workers 65 years of age or greater, yearly rates of fatal occupational injury increased during the pre-recession period, with a larger annual increase observed during the post-recession period. Increases in years of potential life lost among older workers is also concerning, as these lost years represent significant earnings, time to enjoy retirement with family and friends, and continued contribution to society. Ongoing surveillance of occupational fatality trends focusing on populations at high risk of occupational fatality is a useful tool to identify novel and track persistent patterns of fatality. Further investigation into the etiology of fatal occupational injury among older workers, as well as applied research towards methods of remediating them are important aspects of public health inquiry.

4.5 Tables and Figures

Table 4.1 - Change in North Carolina Workforce Composition During Study Period by Age Group

2000-2007 2009-2017 Age Change Percent Number of Number of Group (/1000) Change workers (/1000) workers (/1000) 25-34 927 1075 148 16.0% 35-44 1013 1146 133 13.1% 45-54 879 1132 253 28.8% 55-65 476 789 313 65.8% 65-75 109 200 91 83.5% 75+ 24 37 13 54.2%

Table 4.2 - Change in Distribution of Workforce by Industry: Pre-Recession (2000-2007) Vs Post-Recession (2009-2017), N.C.

Largest Percentage Decline Largest Absolute Increase Largest Percentage Increase / Smallest Percentage Increase

Change in Change in Change in Industry Industry workforce Industry workforce workforce (%) (%) (%) Paper and 2,411 -11,380 Professional 1,313,138 Allied Product Furniture and Services (108%) (1,512%) Fixtures Mfg (-883%) Mfg Wood Finance, 224,537 12,562 Buildings and -542 Insurance, Agriculture (149%) (703%) Mobile (-675%) Real Estate Homes Mfg

221,769 Sawmills and 6,957 Textile Mill -34,405 Retail (186%) Wood Mfg (406%) Product Mfg (-373%)

Business and Printing and 12,397 Apparel and -14,185 167,363 Repair Finished (90%) Publishing (394%) (-347%) Services Textiles Mfg Fishing / Food and 152,444 707 58,396 Construction Hunting / Associated (204%) (318%) (67%) Trapping Product Mfg

Chemicals Misc 16,462 65,104 Public 130,485 and Administration (137%) Manufacturing (302%) Petroleum (71%) Mfg Machinery 3,881 5,784 and Transport 107,556 Forestry Transport /

Equipment (89%) (255%) Taxi (73%) Mfg Personnel 78,940 78,940 27,155 Wholesale Wholesale Supply (227%) (227%) (77%) Services Detective / Eating and Electric Gas / 78,796 15,224 36,451 Drinking Protective Pipeline / (127%) (204%) (84%) Places Services Other Utility

152,444 5,168 Transport 72,291 Construction Mining and Bus/Truck (150%) (204%) oil (85%)

Data are presented as difference in number of employees (% difference) between the pre-recession and post-recession periods

Table 4.3 - Leading Industries by Count of Intentional and Unintentional Injuries. N.C. 2000-2017

All Cause Unintentional Intentional

Rate/105 Rate/105 Rate/105 Industry Deaths Industry Deaths Person Industry Deaths Person Person Years Years Years 10.3 10.3 Constructio 10.3 Construction 85 Construction 85 85 (8.3, 12.8) (8.3, 12.8) n (8.3, 12.8)

24.0 24.0 24.0 Agriculture 73 Agriculture 73 Agriculture 73 ( 19.1, 30.2) ( 19.1, 30.2) ( 19.1, 30.2)

Transport Transport 22.4 22.4 Transport 22.4 69 69 69 Bus/Truck (17.7, 28.3) Bus/Truck (17.7, 28.3) Bus/Truck (17.7, 28.3)

2.4 2.4 2.4 Retail 27 Retail 27 Retail 27 (1.7, 3.2) (1.7, 3.2) (1.7, 3.2)

3.9 3.9 3.9 Wholesale 18 Wholesale 18 Wholesale 18 (2.4, 6.1) (2.4, 6.1) (2.4, 6.1)

65.6 65.6 65.6 Forestry 17 Forestry 17 Forestry 17 (40.8, 105.6) (40.8, 105.6) (40.8, 105.6)

Professional Professional 0.3 0.3 Professiona 0.3 14 14 14 Services (0.2, 0.5) Services (0.2, 0.5) l Services (0.2, 0.5) Automotive Automotive Automotive Sales / 4.9 4.9 Sales / 4.9 Sales/Service 13 13 13 (2.8, 8.4) Service (2.8, 8.4) Service (2.8, 8.4) Repair Repair Repair Justice / Justice / Justice / 6.8 6.8 6.8 13 Public 13 Public 13 Public Safety (3.9, 11.7) (3.9, 11.7) (3.9, 11.7) Safety Safety Data are presented as number of deaths, fatal occupational injury rate, (95% confidence interval)

Table 4.4 - Leading Industries by Rate of Fatal Occupational Injury and Intentionality

All Cause Intentional Unintentional

Rate/105 Rate/105 Rate/105 Industry Industry Industry Person Years Person Years Person Years 65.6 65.6 Transport/ 20.1 Forestry Forestry (40.8, 105.6) (40.8, 105.6) Taxi (6.5, 62.3) Fishing / Fishing / 55.8 55.8 Gas Service 18.5 Hunting / Hunting / (18.0, 172.8) (18.0, 172.8) Stations (8.3, 41.2) Trapping Trapping Wood 3.3 Transport / 33.5 Buildings / 26.6 Grocery / Food (1.6, 6.5) Taxi ( 13.9, 80.4) Mobile Home ( 6.7, 106.2) Stores Manufacturing

Wood 2.1 Buildings / 26.6 23.7 Justice / Agriculture (0.8, 5.6) Mobile Home ( 6.7, 106.2) (18.8, 29.8) Public Safety

Manufacturing Automotive 1.9 24.0 Transport/ 21.4 Agriculture Sales/Service (0.8, 4.5) ( 19.1, 30.2) Bus/Truck (16.8, 27.2) Repair Warehouse 1.9 Transport/ 22.4 13.4 Transport/Taxi Storage / (0.5, 7.5) Bus/Truck (17.7, 28.3) (3.4, 53.5) Transport Stone/Clay/ Stone / Clay / 1.7 Gas Service 18.5 10.3 Glass/Concrete Glass/ Concrete (0.2, 12.2) Stations (8.3, 41.2) (4.6, 23.0) Manufacturing Manufacturing Stone/Glass/ 9.8 1.4 12.1 Eating / Concrete Construction (7.9, 12.2) (0.5, 3.6) (5.8, 21.3) Drinking Places Manufacturing Warehouse 7.5 1.1 10.3 Construction Storage / (3.8, 15.0) Lodging (0.2, 7.8) (8.3, 12.8) Transport

Warehouse 6.5 1.0 9.4 Transport/ Storage/ Mining and Oil (0.9, 46.0) (0.3, 3.0) (5.0, 17.4) Bus/Truck Transport Data are presented as rates per 100,000 worker-years (95% confidence interval)

Table 4.5 - Rates of Fatal Occupational Injury Among Age Groups, by Study Period, N.C., 2000-2017

Full Study Period 2000-2007 2009-2017 2000-2017 Rate /105 Annual Rate Rate /105 Annual Rate Annual Rate Age Rate /105 (95% Change (95% Change Change Group (95% CI)1 CI)1 (95% CI)2 CI)1 (95% CI)2 (95% CI)2

2.2 -4.1 2.7 -9.6 1.8 2.6 (25-54) (2.1, 2.4) (-5.2, -3.0) (2.5, 3.0) (-12.7, -6.4) (1.7, 2.0) (-1.0, 6.2) 1.9 -5.1% 2.4 -9.8% 1.5 -2.5% 25-34 (1.7, 2.1) (-7.0, 3.0) (2.1, 2.8) (-15.5, -3.7) (1.3, 1.8) (-8.7, 4.1) 2.2 -3.7% 2.8 -11.1% 1.8 10.4% 35-44 (2.0, 2.5) (-5.6, 1.9) (2.4, 3.2) (-16.1, -5.8) (1.6, 2.1) (4.0, 17.3)

Younger workers Younger 2.5 -3.8% 3.0 -8.1% 2.1 0.0 45-54 (2.3, 2.7) (-5.6, -2.0) (2.6, 3.5) (-13.3, -2.5) (1.8, 2.4) (-5.3, 5.7)

-2.8% -7.7% 3.4 4.0 3.0 1.4% (55+) (-4.5%, - (-11.7%, - (3.1, 3.7) (3.5, 4.6) (2.7, 3.4) (-3.8%, 5.9%) 1.0%) 2.2%) 2.9 -3.5% 3.4 -11.5% 2.6 -1.3% 55-64 (2.6, 3.2) (-5.6, -1.4) (2.9, 4.1) (-17.9, -4.6) (2.2, 3.1) (-7.0, 4.8) 1.4%

Olderworkers 6.1 -1.7% 6.0 4.5 4.4% 65+ (-9.0%, (4.7, 7.7) (-4.6, 1.5) (4.8, 7.6) (3.7, 5.6) (-3.9%, 13.5%) 12.9%) Data are presented as: 1 Average rate of fatal occupational injuries per 100,000 worker-years during time period indicated (95% confidence interval) 2 Yearly change in rate during period indicated (95% confidence interval)

Figure 4.1 - Annual Fatal Occupational Injury Rate Among Older Workers (55 Years of Age or Older): 2000-2017, N.C.

7.0

6.0 Years

- 5.0

4.0

Worker 5 3.0

2.0

Rate per 10 per Rate 1.0

0.0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Calendar Year

Yearly Rate Model-Predicted Rate

Figure 4.2 - Annual Fatal Occupational Injury Rate among Worker 65 Years of Age and Older: 2000-2017, N.C

9.0

8.0

7.0

Years - 6.0

5.0 Worker

5 4.0

3.0

2.0 Rate per 10 per Rate 1.0

0.0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Calendar Year

Yearly Rate Model-Predicted Rate

Figure 4.3 - Years of Potential Life Lost Among Older Workers: 2000-2017, N.C.

500 450 400 350 300 250 200 150

100 Years Years of Potential Life Lost 50 0

Calendar Year

CHAPTER 5: FATAL OCCUPATIONAL INJURY DUE TO HOMICIDE, SUICIDE, ALCOHOL POISONING OR OPIATE OVERDOSE

5.1 Introduction

Drug overdose and suicide in the United States have been rising since the early 2000s, while homicides have been rising since 2014.4,5 Drug overdose, especially those due to opiate- induced respiratory depression, increased dramatically since 2000.123 The rate of suicide increased by an estimated 35% in the period between 1999 and 2018, with the highest rates, and increases in rates, observed among 45-64 year-olds.124 Homicide rates, after years of decline, began to increase in 2014, with an estimated increase of 15% between 2014-2018.125

The occupational context for these fatal events is important to consider. Suicide and homicide at work have long been targets of occupational injury prevention research, more recent work has included drug overdose at the workplace.7 Drug use at work is an important avenue of investigation, as narcotic use on the job is both associated with fatal overdose due to the effects of the drug, as well as with experiencing a fatal injury from external causes.126,127 Alcohol and pharmaceuticals, especially opiates, have been identified as co-abused substance, the combination of which significantly increases the risk of respiratory depression and death.128

Furthermore, the etiology of overdose for many Americans may involve their occupation, as research has suggested that the first encounter with an opiate for many people is not a street drug, but is prescribed by a physician for legitimate pain from a work injury.36

Places of work are a unique venue, in that several people observe the behavior of employees on a daily basis, and changes in behavior or affect are readily observed. The potential of the workplace as a venue for identification of mental health and substance abuse issues, is important to consider. Workplace interventions may have merit, as resources may be more easily assigned to those needing them when attached to employment, for example with a guarantee of not losing your job conditional on acceptance of treatment. Understanding the distribution of these outcomes, and trends in these outcomes over time, is critical to informing efforts to reduce them. In this work, we estimate the yearly rate and trends of fatal occupational injury due to homicide, suicide, and drug or alcohol overdose among North Carolina workers between the years of 2000-2017.

5.2 Methods

In North Carolina, injuries that result in a fatality are the jurisdiction of medical examiners, who are empowered to conduct autopsies, toxicology screenings and request administrative search warrants. Two medical examiners are appointed in each of the 100 counties in North Carolina and are overseen by the chief medical examiner in the state capital of

Raleigh. Medical examiners determine the location at which a deadly injury occurred, the cause, means and manner of death, as well as record demographic information about decedents, including industry and occupation at time of death.68 We define a fatal occupational injury as any injury leading to death within 30 days sustained on North Carolina soil while engaged in legal work for pay. All information pertaining to each fatal occupational injury occurring between

January 1, 2000, and January 31, 2017, including toxicology screening results and determination of means and manner of death, were abstracted by trained investigators for analysis.

The activity at the time of fatal injury was used to classify decedents’ industry and occupation, coded to U.S. Census guidelines for the year 2000. Toxicology results, including drug name, concentration, and determination of lethality, were abstracted, and categorized into 3 types (sufficient to cause death independently, capable of causing death in combination with other drugs, indirectly contributed to death).

In the current analysis, we focus on decedents whose cause of death was determined to be suicide, homicide, or poisoning (due to alcohol or drug overdose). Poisoning deaths with positive toxicology screen were categorized by type of substance detected on toxicology screen, and decedents with alcohol, prescription or recreational drugs detected included in our analysis.

Decedents whose deaths were caused by exposure to industrial chemicals, such as carbon monoxide or toluene, were not included in the overdose group. Deaths who were determined to be intentional suicide involving lethal doses of drugs were classified as suicides and not included in the overdose group.

5.2.1 Population at Risk

Yearly employment estimates for North Carolina workers during the study period, cross- classified by industry and occupation, were extracted from the decennial US censuses and the

American Community Survey. Industry was grouped into 33 categories defined by census code, while occupational was grouped in 51 census-derived categories.129

5.2.2 Statistical Methods

To estimate yearly rates and trends in fatal occupational injury due to suicide, homicide and overdose, Poisson regression models were constructed for each outcome, using the count of fatal injuries of each type per year, and an offset defined as the log-transformed population of employed North Carolinians. An unadjusted model for the crude trend in rates during the study period is expressed for each outcome separately as: log(annual rate) = β0 + β1 (Year-2000).

Calendar year was centered on the year 2000 and entered as a continuous variable to estimate the average change in rate per year. The fit of the Poisson regression model was used to estimate the average change in rate per year for each outcome, expressed with the equation 100[exp(β1)−1].

Estimates were adjusted for sex by the inclusion of a binary indicator for sex in the regression model.

5.3 Results

Between 2000 and 2017, 1,775 occupational fatalities were observed, 275 (15.5%) of which were identified as having been caused by homicide, suicide, or drug/alcohol overdose.

Two hundred and thirty-five decedents (85.4%) were male, 218 (79.3%) were white, 57 (20.7%) were black, and 25 (9.1%) were identified as Latinx and White. Homicide was the most common cause of death, with 200 deaths (72.7%) during the study period, followed by 55 suicide deaths (20.0%), 15 overdose deaths (5.5%) and 5 alcohol poisoning deaths (1.8%). (Table 5.1)

The rate of occupational fatalities due to homicide and suicide decreased during the study period, while occupational fatalities due to suicide increased. We estimate that the rate of

70 occupational homicide declined an average of 3.1% per year (-5.7%, -0.5%) from a high of 4.3 deaths per million workers in the year 2000 to 3.1 deaths in 2013. (Figure 5.1) A similar pattern was observed for occupational suicide rates, declining by an average of 3.3% per year (-8.1%,

1.7%). (Figure 5.2) Rates of overdose due to alcohol, or prescription, or recreational drugs increased by an estimated 5.5% per year (-3.9%, 15.9%) during the study period, from 0 deaths per million workers in 2000 to 0.8 deaths in 2017. (Figure 5.3)

Industries with the highest counts of death, including eating and drinking places, public safety, and grocery stores, are detailed in Table 5.2, along with the distribution of manner of death in each industry. Table 5.3 details leading industries, including transportation and taxi, gas stations, and grocery stores.

5.4 Discussion

Detailed death records derived from the files of the Office of the Chief Medical Examiner

(OCME) were abstracted, and the cause of death, toxicology reports, and employment information used to categorize decedents. Estimates of the working population of North

Carolina, cross-classified by industry, occupation and sex were used to define the population at risk. Poisson regression modeling approaches were utilized to estimate the yearly rate of fatal occupational injury due to homicide, suicide, and overdose in the population of North Carolina as well as within industries.

During the study period, we found evidence of an increases in the rate of occupational drug overdose, with decreases observed in the rate of occupational homicide and suicide. (Table

5.1) We estimate that the rate of fatal occupational injury due to homicide or suicide decreased

71 by approximately 3% per year, (Table 5.1, Figure 5.1, 5.2) while the rate of fatal occupational injury due to overdose increased by an estimated 5.5% per year (Figure 5.3).

Reducing the rate of workplace homicide has been a priority for injury prevention research. While a variety of methods to decrease workplace homicide have been proven effective in pilot studies, such as cash control devices for retail establishments, uptake of these findings has been shown to be low.130 We note that of the five highest homicide-rate industries, three (taxi, gas stations, grocery stores) are industries that require taking payments from customers. While paying with cash is becoming less common, these businesses are legally required to accept cash, which requires having sufficient cash to make change, and the accumulation of cash at busy periods during the workday. Cash control strategies, identified as one effective method to reduce robbery and homicide, require significant investment, as the safes required must be robust, be anchored to solid objects, and staff trained to deposit cash when it reaches a defined amount. In some industries, profit margins are slim, and turnover is high, making investment in cash control devices challenging, and increasing the likelihood that untrained staff neglect to follow deposit procedures. Suicide in the occupational setting has recently begun to receive increased attention from injury prevention research. Suicide rates in the general population have been rising, and recent research suggests that national rates of occupational suicide are also increasing.8 We identified industries with the highest rates of occupational suicide in Table 5.3 and note alignment with other studies that have found high rates among gas station workers, automotive sales and repair, and truck transportation.59 We note an overlap in both the highest rates and highest counts of homicide and suicide among the gas station, and truck transportation industries, suggesting a potential association between violence risk from homicide and self-directed violence in the form of suicide. Another

72 explanation recently forwarded by researchers has been that adverse conditions in the workplace are associated with suicide, though suicides are not always explicitly identified as occurring at the workplace.131 This association is increasingly viewed as plausible, as a growing body of literature has found associations between adverse work conditions and major depressive disorder, a known risk factor for suicide.132–134 As the study period encompasses the economic crisis of

2008, we also note a larger evidence base that associates economic downturns and increases in suicide rates.135

The dramatic increase in opiate overdose in the general population, colloquially known as the opiate epidemic, has also been identified in the workplace.7 The onset of the opiate crisis was sudden, and the source of a large proportion of opiates are legal prescriptions, making intervention challenging. For many illicit users of opiates, their first encounter with these highly addictive drugs was through a prescription.136 Workplace injury is also associated with opiate prescription, and many workers who’s injuries may not have prompted a physician to write an opiate prescription prior to the JACO guidance on pain management, may have received their first dose of these highly addictive drugs.36 These results suggest that the need for effective non- opiate treatments for injury as well as investigation and evaluation of methods of overdose prevention is high.

This study has several strengths, including trained investigators abstracting medical examiner files, official determinations of cause of death, and toxicology screenings with determinations of lethality. Limitations include the pathophysiology of overdose, where workers who take a lethal quantity of a drug at work may not develop symptoms for hours and may not succumb to the effects of the drug until after they get home from work. Similarly, workers who have taken a potentially fatal dose the night before a workday may not show up to work, and thus

73 would not be counted in our estimates. Workers who have become addicted to opiates may also lose their jobs, and thus cannot die at work. Associations between mental health issues and addiction may also limit our ability to estimate these rates effectively, as workers who are depressed may drop out of the workforce and begin to use drugs, or get fired due to drugs.137

We found that rates of fatal occupational injury due to homicide or suicide are decreasing during the study period, while rates of overdose increased. These results suggest that applied occupational injury research efforts focusing on homicide and suicide may have been successful in reducing the rate of these events at work, even in the setting of increases in homicide and suicide in the general population. To address the rapid increase in overdose identified here, similar levels of investment in research and evaluation of intervention methodologies are needed.

5.5 Tables and Figures

Table 5.1 - Occupational Fatality Change During the Study Period: 2000-2017, N.C.

Type Occupational Fatalities

Average Rate Number Rate Change / 106 Worker- of Deaths per Year years

3.5 -3.1% Homicide 200 (2.7, .5) (-5.7%, -0.5%)

1.0 -3.3% Suicide 55 (0.6, 1.6) (-8.1%, 1.7%)

0.2 5.5% Overdose 20 (0.1, 0.8) (-3.9%, 15.9%)

Table 5.2 - Leading Industries by Count of Homicide, Suicide and Overdose Deaths. N.C. 2000-2017

Homicide Suicide Overdose

Industry Deaths Rate/106 Industry Deaths Rate/106 Industry Deaths Rate/106

Justice / 19.8 Constructio 1.2 0.7 Public 27 7 Construction 5 Safety (13.6, 28.9) n (0.7, 2.5) (0.3, 1.8)

Auto Sales Grocery 16.6 5.1 Professional 0.1 26 / Service / 7 3 Stores (11.3, 24.3) (2.5, 10.7) services (0.0, 0.4) Repair Eating Eating and and 5.9 1.2 2.4 25 Drinking 5 Agriculture 2 Drinking (4.0, 8.8) (0.5, 2.8) (0.6, 9.6) Places Places

Finance / Gas 80.8 1.1 Transport 0.7 18 Insurance / 5 1 Stations (50.9, 128.2) (.4, 2.6) Bus / Truck (0.1, 5.0) Real Estate Business 2.8 Transport / 3.5 0.3 Retail 16 5 and Repair 1 (1.7, 4.6) Bus / Truck (1.4, 8.3) (0.0, 0.2) Services Data are presented as rates per 1,000,000 worker-years (95% confidence interval)

Table 5.3 - Leading Industries by Rate of Fatal Occupational Injury Due to Homicide, Suicide and Overdose

Homicide Rate/106 Suicide Rate/106 Overdose Rate/106 Paper / Transport / 183.3 9.0 2.7 Gas Stations Allied Taxi (95.1, 353.4) (2.2, 35.9) (0.4, 19.3) Products 90.2 Auto Sales / 5.1 2.4 Gas Stations Agriculture (51.0, 159.4) Repair (2.4, 10.7) (0.6, 9.6)

Justice / 19.8 Transport / 3.5 0.7 Public Safety (13.6, 28.9) Bus Construction (1.4, 8.3) (0.3, 1.8) Truck

Grocery / 16.6 2.4 Transport / 0.7 Agriculture Food Stores (11.3, 24.3) (0.6, 9.6) Bus / Truck (0.1, 5.0)

15.1 1.9 Grocery 0.6 Forestry (3.8, 60.3) Utilities (0.3, 13.8) Stores (0.1, 5.0)

Data are presented as rate per 1,000,000 worker-years (95% confidence interval)

Figure 5.1 - Yearly Rate of Fatal Occupational Injuries: Homicide, 2000-2017, N.C.

6.0

5.0

Years - 4.0

Worker 3.0 6

2.0

1.0 Rate per Rate 10 0.0

Calendar Year

Yearly Rate Average Rate

Figure 5.2 – Yearly Rate of Fatal Occupational Injuries: Suicide, 2000-2017, N.C.

1.6

1.4 Years

- 1.2 1.0

Worker 0.8 6 0.6 0.4

0.2 Rate per Rate 10 0.0

Calendar Year

Yearly Rate Average Rate

Figure 5. 3 – Yearly rate of Fatal Occupational Injuries: Overdose, 2000-2017, N.C

0.9 0.8

0.7

Years - 0.6

0.5 Worker

6 0.4 0.3 0.2

Rate per Rate 10 0.1 0.0

Calendar Year

Yearly Rate Average Rate

CHAPTER 6: DISCUSSION

Significant reductions in the rates of fatal occupational injury were observed among

North Carolina workers, although rates in the last 5 years of the study period began to trend upwards. Black non-Latinx workers, a group historically with the highest rate of fatal occupational injury, experienced rates statistically similar to that of White workers for most of the study period, in the final 5 years of the study fatality rates began to rise. Latinx workers consistently experienced the highest rates of fatal occupational injury, and also experienced a rate increase in the last 5 years of the study. Older workers experienced relatively high rates of fatal occupational injury but declined steadily during the study period.

6.1 Study Strengths

This study has several strengths, including a robust case identification process, multiple sources of data for sensitivity analyses and an experienced team of investigators with in-depth knowledge of the legal, occupational, and regulatory environment in North Carolina. Advanced techniques in modeling of rates will be employed utilizing the finer resolution provided by the

ACS 3 and 5-year intercensal estimates of the population at risk.

6.1.1 Exposure Classification

Much of the work in the field of occupational fatality depends on the accurate coding of industry and occupation by state personnel in administrative databases. I utilize state-coded occupation and industry data in the initial survey of occupational fatality, then build on this approach via hand-abstraction of all records pertaining to decedents and consideration of alternate sources of data provided by the State Center for Health Statics. OCME records include testimony from witnesses as well as family regarding the usual industry and occupation, as well as the characteristics of the activities engaged in by decedents preceding death. At approximately the halfway mark of abstraction, agreement between state coded data and the study team is good.

It is worth noting that in several cases, the usual industry/occupation of decedents coded by the state databases has in some cases conflicted with the activities the decedent was engaged with at the time of death. A significant strength in this study is the improved classification of exposure via detailed abstractions by trained investigators of the activities of the deceased prior to death. I propose that this methodology decreases the probability of exposure misclassification, and the clear definition of analytic methods and inclusive case definition improves the interpretability of calculations of stratum-specific rates of occupational fatality.

6.1.2 Outcome Classification

Standard methodology of occupational fatality surveillance involve analyses using data provided by state agencies. I improve on this approach using three approaches, hands-on abstraction of physical paperwork contained at the OCME, regular adjudication of cases using flexible but clear guidelines regarding case definition, and sensitivity analyses using a 1% sample of non-cases. The outcome measure should also enjoy excellent sensitivity, as deaths at

79 work are carefully screened for at hospitals and by medical examiners. Within the study protocols I improve on this sensitivity by consulting two state databases to identify cases, and trained investigators utilize multiple sources of information to verify whether occupational activities were ongoing at the time of death. To estimate the completeness of the population and to assess specificity of the outcome measure, a 5% sample of cases deemed not at work by both the SCHS and OCME was conducted. The rate of misclassification of workplace fatality was

3.8% in this sample, suggesting that misclassification of outcome is uncommon.

6.1.3 Population at Risk

Prior to 2005, estimation of a population at risk using census data required linear interpolation of the populations within groups of interest, including ethnicity, industry, occupation, and age. Beginning in 2005, the American Community Survey (ACS) began to release 1-year, 3-year and 5-year estimates of the US population. These intercensal estimates can be used to improve the validity of estimates between the 10-year census counts. This is especially important for the period surrounding the great recession, where sudden job loss in

2008 may not be captured well by linear interpolation of between the 2000 and 2010 census.

The utilization of intercensal estimates and advanced Poisson modeling techniques for estimating rates, will allow for a more flexible approach to the population at risk.

6.2 Study Limitations

This study has several limitations, including potentially missing cases, misclassifying exposures or outcomes, and under or overestimating the population at risk. More generally, the

80 outcomes under study are a tiny subset of occupational injuries, those acute enough to cause death within 30 days of injury onset. While occupational fatality can be seen as the ultimate workplace injury, occupational fatality is a small component of occupational injury. Recent work by the AFL-CIO suggest that the rate of workplace death that occurs in the longer term

(and therefore outside of the sample) dwarfs acute occupational injury leading to fatality.138

6.2.1 Potential for Misclassification of Outcome

I consider that there may be true occupational fatalities that are not reported as such to hospitals, law enforcement and thus be missing from this study. This is plausible in that some employers may seek to reduce their exposure to lawsuits, increases in their workman’s compensation insurance premiums and investigation by OSHA or other state agencies. While I acknowledge that this outcome, while rare, may be more common among vulnerable groups of workers, I am unaware of any evidence that this represents a significant threat to inference.

My inclusion criteria, restricting to cases in which injury is followed by death within 30 days may also allow for some fatality misclassification. Some mortally injured workers may survive for more than 30 days before succumbing to their wounds. The number of individuals who do not meet study criteria due to delays between injury and outcome will be evaluated as a group, with the potential for sensitivity analysis if warranted.

6.2.2 Potential of Misclassification of Occupation

Occupational misclassification in this study is possible but unlikely, as a substantial amount of information in a death file would have to be misleading. One plausible mechanism of

81 exposure misclassification among cases could involve self-employed workers. Sole proprietors of companies with a diverse work portfolio, such as handymen, mechanics and migrant workers may experience a workplace fatality in their garage, driveway or at the home of a friend. This residential location of work is not uncommon but may complicate investigation by law enforcement and medical examiners, as it may not be clear that the decedent was working for pay. Other factors can affect the motivations of witnesses and informants, as many self- employed persons operate on a cash basis. Hesitance to provide detailed information to law enforcement may result in hesitance to provide details to investigators regarding occupational activities due to perceived legal consequences.

Misclassification of the industry and occupation of census data is also possible, as the census asks participants to self-report their industry and occupation. Persons who believe that census operations may result in investigation by local authorities may neglect to report engaging in side-work or other cash jobs.

6.2.3 Potential for Misclassification of Ethnicity in Cases:

Ethnicity is a concept that presents special challenges for investigators, especially those who rely on secondary data. Persons perceived as being members of an ethnic group may in fact not be, as well as vice versa. In this study, ethnicity is derived from death certificates, autopsies, and medical examiner reports. These reports are filled out by medical professionals based on the appearance of a decedents body after death, on interviews with family and first responders, and on examinations conducted during investigation of the death. Our determination of ethnicity is dependent, in the absence of family, coworkers or first responders’ narratives, on the judgement of the medical examiner. As the ability to determine ethnicity is likely to vary among medical

82 examiners, we acknowledge the possibility of misclassification of ethnicity by the medical examiner. Misclassification is also possible during data analysis, for example in Aim 2, where determination of ethnicity is required to group decedents into Latinx and non-Latinx categories.

During chart review, abstractors encountered death certificates in which ethnicity was hand- written in a free text section. In many cases, this text entry appears to be country of origin (e.g.

El Salvador, Mexican). For some decedents with countries of origin that included central and

South America, the birthplace recorded on their death certificate was in the United States of

America. Final classification of ethnicity for aim 2 encompassed any ethnicity determination that included a country in central or South America, and the phrase “Hispanic” or “Latino”.

6.2.3 Potential for Misclassification of Population at Risk

Census data was chosen to estimate the population at risk as it represents the highest fidelity source of longitudinal, group-stratified (age, sex, race, ethnicity, industry/occupation) estimates of the population of the United States. Challenges in the utilization of census-derived denominator data are of particular concern in this analysis, as in the setting of limited cases, a large misspecification of the population at risk will bias calculations of rates. In terms of aim 1,

Latinx residents have been shown to be more likely to be undercounted by the census than other groups,93 and some estimates suggest that more than 5% of the true population could be missed.72 While this phenomenon has been documented primarily in the southwest part of the

United States, we feel it is important to acknowledge the possibility of systematic undercounting of Latinx workers in North Carolina. We note that while yearly estimates would be affected by this misspecification, to affect estimates of trends in rates of occupational fatality, the proportion of undercount would have to change during the study period. It is also important to acknowledge

83 that the inclusion criteria allow all workers, regardless of documentation status to be counted as cases, and thus it is possible to have people counted in the numerator but absent from the denominator. I hypothesize that undocumented workers are both more likely to be absent from the denominator and may be more likely to experience an occupational fatality (and thus be present in the numerator), a relationship that could bias estimated rates away from the null.

While it is important to acknowledge the possibility of these concerns, Poisson regression approaches are relatively insensitive to misspecification of the denominator in the setting of a small numerator and large denominator.

Overestimation of the population at risk is of greatest concern in aim 2, where I consider the great recession of 2008 as a natural experiment and hypothesize that the distribution of occupational fatality will change in the years following compared to the years prior. Beginning in September 2008 Americans found themselves unemployed at a historically rapid rate, with 1.8 million jobs lost between September and December 2008, and employment over 10% until 2010.

139 While this is a concern, the timing of the 2010 census is very fortunate. At the height of unemployment, a ten-year census was performed, and thus the rapid job loss experienced after

2008 will be better captured by interpolated estimates of the population at risk. While this proximity to the decennial census may improve the estimate of the denominator, I consider the possibilities that the population at risk in industries that were the first to experience rapid job loss may be overestimated. Again, I rely on the robust estimating characteristics of Poisson regression, and the approach of evaluating trends prior to and following the recession of 2008, to minimize the effect of this concern on our estimates.

6.3 Funding

Data collection, abstraction, linkage, and analysis are all supported by NIOSH R01

OH011256-01A1. Morgan Richey is supported by a research assistantship through the UNC

Injury Prevention Research Center.

6.4 Conclusions

In an encouraging sign of improvement in occupational safety, workers in North Carolina experienced decreases in the rate of fatal occupational injury during the period between 2000-

2017, albeit with concerning increases towards the end of the study period. By the end of the study period, Black workers, who have historically experienced high rates of fatal occupational injury, had similar rates to that of White workers. Older workers also experienced lower rates of fatal occupational injury during the study period, though an increase in rates in the years following the great recession was observed. Fatal occupational injury due to homicide and suicide declined during the study period, even as the rates in the general population increased.

While these improvements are important to note, there is still much work to be done. We identified high rates of fatal occupational injury among Latinx workers, a rapidly growing group in North Carolina. Older workers, who’s rate of occupational fatality declined during the period preceding the recession of 2008, saw a reversal of this trend and rising rates in the post-recession period. This relationship was more pronounced among workers 65 years of age or greater.

Finally, fatal occupational injury due to overdose on prescription or illicit drugs or alcohol increased throughout the study period.

We feel it is important to highlight these improvements and credit the contributions of occupational safety initiatives on the national, state, and local level, as well as efforts by individual managers, supervisors, and workers to improve workplace safety. The success of injury prevention efforts is the culmination of rigorous research into the etiology of occupational injury, as well as research and refinement of both established and novel methodologies.

Continued investment and innovation in injury prevention research and practice is necessary to continue to decrease the rate of fatal occupational injury among workers in the United States.

Surveillance of these events among vulnerable groups is a challenging but critical component of injury prevention. These results suggest that even in the setting of declining average rates of fatal occupational injury, groups of workers may be experiencing rate increases. Continuing to improve the systems that support surveillance, including funding of injury prevention training and research is critical. Harmonizing federal, state, and local data collection, coding and reporting is an important aspect of injury prevention reporting systems, as standardized and automated reporting mechanisms have the potential to shorten the duration between the emergence of a threat to occupational safety and its resolution

The occupational sphere is an important venue for addressing disparities by race, ethnicity, and age, among others. In this work we identified high rates of fatal occupational injury among Latinx workers, which remained even after standardizing the age, sex, occupation, and industry fatality rates to that of white workers. While the etiology of these rates is beyond the scope of this work, the persistence of rate disparities as estimated by the standardized mortality ratio suggests that even within the same age, sex, industry, and occupational strata – the rate of fatal occupational injury among Latinx workers was higher than an otherwise identical white worker. Older workers, a group in which membership is inevitable for most people, have

86 high rates of fatal occupational injury, and these rates may be more vulnerable to economic uncertainty. Increases in drug and alcohol overdose, first observed in the general population, is increasing at the work. Injury prevention research approaches are well positioned to identify and address the etiology of these disparities, test interventions and disseminate the effective prevention strategies that will be necessary to reduce the overall rate of fatal occupational injury and disparities in those injuries.

REFERENCES

1. Civilian Labor Force Participation Rate by Age, Sex, Race, and Ethnicity : U.S. Bureau of Labor Statistics.; 2020. Accessed June 25, 2021. https://www.bls.gov/emp/tables/civilian- labor-force-participation-rate.htm

2. Szinovacz ME, Davey A, Martin L. Did the Great Recession influence retirement plans? Res Aging. 2015;37(3):275-305. doi:10.1177/0164027514530171

3. Ondrich J. The Great Recession and the Retirement Decisions of Older Workers | Center for Retirement Research. Center for Retirement Research, Boston College; 2013. Accessed November 28, 2020. https://crr.bc.edu/working-papers/the-great-recession-and- the-retirement-decisions-of-older-workers/

4. Rudd RA, Aleshire N, Zibbell JE, Gladden RM. Increases in Drug and Opioid Overdose Deaths — United States, 2000–2014. Morb Mortal Wkly Rep. 2016;64(50 & 51):1378- 1382.

5. Curtin SC, Warner M, Hedegaard H. Increase in Suicide in the United States, 1999-2014. U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Health Statistics; 2016.

6. Hedegaard H, Miniño AM, Warner M. Drug Overdose Deaths in the United States, 1999- 2017. NCHS Data Brief. 2018;(329):1-8.

7. Tiesman HM, Konda S, Cimineri L, Castillo DN. Drug overdose deaths at work, 2011– 2016. Inj Prev. 2019;25(6):577-580. doi:10.1136/injuryprev-2018-043104

8. Tiesman HM, Konda S, Hartley D, Menéndez CC, Ridenour M, Hendricks S. Suicide in U.S. Workplaces, 2003–2010: A Comparison With Non-Workplace Suicides. Am J Prev Med. 2015;48(6):674-682. doi:10.1016/j.amepre.2014.12.011

9. Moracco KE, Runyan CW, Loomis DP, Wolf SH, Napp D, Butts JD. Killed on the clock: A population-based study of workplace homicide, 1977–1991. Am J Ind Med. 2000;37(6):629-636. doi:10.1002/(SICI)1097-0274(200006)37:6<629::AID- AJIM7>3.0.CO;2-7

10. Richardson D, Loomis D. Trends in fatal occupational injuries and industrial restructuring in North Carolina in the 1980s. Am J Public Health. 1997;87(6):1041-1043. doi:10.2105/AJPH.87.6.1041

11. Stout NA. Occupational injury prevention research: progress and priorities. Inj Prev. 2002;8(90004):9iv-914. doi:10.1136/ip.8.suppl_4.iv9

12. Work Injury Costs.; 2019. Accessed November 3, 2020. https://injuryfacts.nsc.org/work/costs/work-injury-costs/

13. Hofmann DA, Burke MJ, Zohar D. 100 years of occupational safety research: From basic protections and work analysis to a multilevel view of workplace safety and risk. J Appl Psychol. 2017;102(3):375-388. doi:10.1037/apl0000114

14. Census of Fatal Occupational Injuries (CFOI) - Current and Revised Data. Published 2016. Accessed November 3, 2020. https://www.bls.gov/iif/oshcfoi1.htm#2015

15. Corn M, Corn JK. The United States experience with regulation of health and safety: 1970-1989. In: Health Care in the Changing Economic and Social Environment. 7th ed. ; 1993:225-238.

16. Work-Related Fatality Trends.; 2018. Accessed November 3, 2020. https://injuryfacts.nsc.org/work/work-overview/work-related-fatality-trends/

17. Burtt HE. Psychology and Industrial Efficiency. Appleton; 1929:xviii, 395.

18. Burnham J. Accident Prone. University of Chicago Press; 2009.

19. Flanagan JC. The critical incident technique. Psychol Bull. 1954;51(4):327-358. doi:10.1037/h0061470

20. Conroy C, Fowler J. The Haddon Matrix: Applying an Epidemiologic Research Tool as a Framework for Death Investigation. Am J Forensic Med Pathol. 2000;21(4):339-342.

21. De Haven H. Mechanical analysis of survival in falls from heights of fifty to one hundred and fifty feet*. Inj Prev. 1942;6(1):62-68. doi:10.1136/ip.6.1.62-b

22. Guarnieri M. Landmarks in the history of safety. J Safety Res. 1992;23(3):151-158. doi:10.1016/0022-4375(92)90018-5

23. Gibson J. The contribution of experimental psychology to the formulation of the problem of safety: A brief for basic research. Behav Approaches Accid Res. 1961;1:77-89.

24. Haddon W. A Note Concerning Accident Theory and Research with Special Reference to Motor Vehicle Accidents. Ann N Y Acad Sci. 1963;107(2):635-646. doi:10.1111/j.1749- 6632.1963.tb13307.x

25. Haddon W. THE CHANGING APPROACH TO THE EPIDEMIOLOGY. PREVENTION, AND AMELIORATION OF TRAUMA: THE TRANSITION TO APPROACHES ETIOLOGICALLY RATHER THAN DESCRIPTIVELY BASED. Published online 1968:8.

26. Guldenmund FW. The nature of safety culture: a review of theory and research. Saf Sci. 2000;34(1):215-257. doi:10.1016/S0925-7535(00)00014-X

27. Ashkanasy NM, Wilderom CPM, Peterson MF. Handbook of Organizational Culture and Climate. SAGE; 2004.

28. Ostroff C, Kinicki AJ, Muhammad RS. Organizational Culture and Climate. In: Handbook of Psychology, Second Edition. American Cancer Society; 2012. doi:10.1002/9781118133880.hop212024

29. Mearns KJ, Flin R. Assessing the state of organizational safety—culture or climate? Curr Psychol. 1999;18(1):5-17. doi:10.1007/s12144-999-1013-3

30. Zohar D. A group-level model of safety climate: Testing the effect of group climate on microaccidents in manufacturing jobs. J Appl Psychol. 2000;85(4):587-596. doi:10.1037/0021-9010.85.4.587

31. Krishnappa V, Matthews HS, Liu Y. Data-Driven Analysis to Support Revised Tire Tread Inspection Standards. Transp Res Rec. 2019;2673(11):517-528. doi:10.1177/0361198119851456

32. Royal Society for the Prevention of Accidents. Tyre Tread Depth and Stopping Distances. Road Saf Inf UK. Published online July 2005:2.

33. Tires | NHTSA. Accessed June 2, 2021. https://www.nhtsa.gov/equipment/tires

34. Federal Motor Vehicle Safety Standards; Tire Pressure Monitoring Systems. Federal Register. Published July 12, 2007. Accessed June 2, 2021. https://www.federalregister.gov/documents/2007/07/12/07-3382/federal-motor-vehicle- safety-standards-tire-pressure-monitoring-systems

35. Census of Fatal Occupational Injuries - Hours-Based Rates. Published 2010. Accessed November 3, 2020. https://www.bls.gov/iif/oshnotice10.htm

36. Cheng M, Sauer B, Johnson E, Porucznik C, Hegmann K. Comparison of opioid-related deaths by work-related injury. Am J Ind Med. 2013;56(3):308-316. doi:10.1002/ajim.22138

37. Abuse NI on D. Benzodiazepines and Opioids. National Institute on Drug Abuse. Published March 15, 2018. Accessed November 2, 2020. https://www.drugabuse.gov/drug-topics/opioids/benzodiazepines-opioids

38. Buckeridge D, Huang A, Hanley J, et al. Risk of Injury Associated with Opioid Use in Older Adults. J Am Geriatr Soc. 2010;58(9):1664-1670. doi:10.1111/j.1532- 5415.2010.03015.x

39. Krebs EE, Paudel M, Taylor BC, et al. Association of Opioids with Falls, Fractures, and Physical Performance among Older Men with Persistent Musculoskeletal Pain. J Gen Intern Med. 2016;31(5):463-469. doi:10.1007/s11606-015-3579-9

40. Shaw WS, Roelofs C, Punnett L. Work Environment Factors and Prevention of Opioid- Related Deaths. Am J Public Health. 2020;110(8):1235-1241. doi:10.2105/AJPH.2020.305716

41. Grzywacz JG, Quandt SA, Isom S, Arcury TA. Alcohol use among immigrant Latino farmworkers in North Carolina. Am J Ind Med. 2007;50(8):617-625. doi:10.1002/ajim.20482

42. Gutin I, Hummer RA. Occupation, employment status, and “despair”-associated mortality risk among working-aged U.S. adults, 1997–2015. Prev Med. 2020;137:106129. doi:10.1016/j.ypmed.2020.106129

43. Hawkins D, Davis L, Punnett L, Kriebel D. Disparities in the Deaths of Despair by Occupation, Massachusetts, 2000 to 2015. J Occup Environ Med. 2020;62(7):484-492. doi:10.1097/JOM.0000000000001870

44. Stallones L, Kraus JF. The occurrence and epidemiologic features of alcohol-related occupational injuries. Addiction. 1993;88(7):945-951. doi:10.1111/j.1360- 0443.1993.tb02112.x

45. Algarni FS, Gross DP, Senthilselvan A, Battié MC. Ageing workers with work-related musculoskeletal injuries. Occup Med. 2015;65(3):229-237. doi:10.1093/occmed/kqu213

46. Thomas EA, Cadogan MP. Work-Related Injury and the Older Adult. J Gerontol Nurs. 2012;38(9):9-12. doi:10.3928/00989134-20120807-02

47. Smith S, Pegula S. Fatal occupational injuries to older workers. Mon Labor Rev. Published online January 30, 2020. doi:10.21916/mlr.2020.2

48. North Carolina’s Hispanic Community: 2019 Snapshot. Carolina Demography. Published September 26, 2019. Accessed November 3, 2020. https://www.ncdemography.org/2019/09/26/north-carolinas-hispanic-community-2019- snapshot/

49. Ellis CD, Munnell AH, Eschtruth AD. Falling Short: The Coming Retirement Crisis and What to Do About It. Oxford University Press; 2014. doi:10.1093/acprof:oso/9780190218898.001.0001

50. Wainwright D, Crawford J, Loretto W, et al. Extending working life and the management of change. Is the workplace ready for the ageing worker? Ageing Soc. 2019;39(11):2397- 2419. doi:10.1017/S0144686X18000569

51. Morgan JQ. Governance, Policy Innovation, and Local Economic Development in North Carolina. Policy Stud J. 2010;38(4):679-702. doi:https://doi.org/10.1111/j.1541- 0072.2010.00379.x

52. Link AN, Scott JT. The Growth of Research Triangle Park. Small Bus Econ. 2003;20(2):167-175. doi:10.1023/A:1022216116063

53. Richardson DB, Loomis D, Bena J, Bailer AJ. Fatal Occupational Injury Rates in Southern and Non-Southern States, by Race and Hispanic Ethnicity. Am J Public Health. 2004;94(10):1756-1761. doi:10.2105/AJPH.94.10.1756

54. Loomis D, Richardson D. Race and the risk of fatal injury at work. Am J Public Health. 1998;88(1):40-44. doi:10.2105/AJPH.88.1.40

55. Richardson D, Loomis D, Wolf SH, Gregory E. Fatal agricultural injuries in North Carolina by race and occupation, 1977–1991. Am J Ind Med. 1997;31(4):452-458. doi:10.1002/(SICI)1097-0274(199704)31:4<452::AID-AJIM12>3.0.CO;2-Z

56. Bernhardt JH, Langley RL. Analysis of Tractor-related Deaths in North Carolina From 1979 to 1988. J Rural Health. 1999;15(3):285-295. doi:10.1111/j.1748- 0361.1999.tb00750.x

57. Berg MT. Trends in the Lethality of American Violence. Homicide Stud. 2019;23(3):262- 284. doi:10.1177/1088767919849643

58. Hockenberry JM, Joski P, Yarbrough C, Druss BG. Trends in Treatment and Spending for Patients Receiving Outpatient Treatment of Depression in the United States, 1998-2015. JAMA Psychiatry. 2019;76(8):810. doi:10.1001/jamapsychiatry.2019.0633

59. McIntosh WL, Spies E, Stone DM, Lokey CN, Trudeau A-RT, Bartholow B. Suicide Rates by Occupational Group — 17 States, 2012. Morb Mortal Wkly Rep. 2016;65(25):641-645.

60. Wang J, Sumner SA, Holland KM, Halpin J, Ivey-Stephenson A, Crosby AE. National Trends in hospitalizations for self-directed violence related to opioids and/or depression — United States, 2000–2015. Prev Med. 2020;134:106051. doi:10.1016/j.ypmed.2020.106051

61. Tiesman HM, Konda S, Bell JL. The Epidemiology of Fatal Occupational Traumatic Brain Injury in the U.S. Am J Prev Med. 2011;41(1):61-67. doi:10.1016/j.amepre.2011.03.007

62. Mora DC, Miles CM, Chen H, Quandt SA, Summers P, Arcury TA. Prevalence of musculoskeletal disorders among immigrant Latino farmworkers and non-farmworkers in North Carolina. Arch Environ Occup Health. 2016;71(3):136-143. doi:10.1080/19338244.2014.988676

63. Data and Statistics - Productive Aging and Work | NIOSH | CDC. Published February 21, 2020. Accessed June 24, 2021. https://www.cdc.gov/niosh/topics/productiveaging/dataandstatistics.html

64. Work Injuries and Illnesses by Age.; 2020. Accessed November 3, 2020. https://injuryfacts.nsc.org/work/industry-incidence-rates/work-injuries-and-illnesses-by- age/

65. NC Population Reaches 10.3 Million in 2017.; 2017. Accessed June 24, 2021. https://www.ncdemography.org/2017/12/20/nc-population-reaches-10-3-million-in-2017/

66. North Carolina on Track to Surpass 10.6 Million by 2020, Gain in the House.; 2020. Accessed June 24, 2021. https://www.ncdemography.org/2020/01/16/north-carolina-on- track-to-surpass-10-6-million-by-2020-gain-in-the-house/

67. NC in Focus: Black Population in North Carolina, 2016.; 2018. Accessed June 24, 2021. https://www.ncdemography.org/2018/02/08/nc-in-focus-black-population-in-north- carolina-2016/

68. 2018 North Carolina General Statutes :: Chapter 130A - Public Health :: Article 16 - Postmortem Investigation and Disposition. :: § 130A-385 - Duties of medical examiner upon receipt of notice; reports; copies. Justia Law. Published 2018. Accessed November 3, 2020. https://law.justia.com/codes/north-carolina/2018/chapter-130a/article-16/section- 130a-385/

69. Chester N, Mottram DR, Reilly T, Powell M. Elimination of ephedrines in urine following multiple dosing: the consequences for athletes, in relation to doping control. Br J Clin Pharmacol. 2004;57(1):62-67. doi:10.1046/j.1365-2125.2003.01948.x

70. Gosselin RE. Clinical Toxicology of Commercial Products, 5th Ed.; 1985. Accessed June 24, 2021. https://onlinelibrary.wiley.com/doi/abs/10.1002/jps.2600741037

71. Csajka C, Haller CA, Benowitz NL, Verotta D. Mechanistic pharmacokinetic modelling of ephedrine, norephedrine and caffeine in healthy subjects. Br J Clin Pharmacol. 2005;59(3):335-345. doi:10.1111/j.1365-2125.2005.02254.x

72. Statistics (U.S.) NC for H. Quality of Death Rates by Race and Hispanic-Origin: A Summary of Current Research, 1999. U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Health Statistics; 1999.

73. Orrenius PM, Zavodny M. Do immigrants work in riskier jobs? Demography. 2009;46(3):535-551. doi:10.1353/dem.0.0064

74. Richardson S. Fatal work injuries among foreign-born Hispanic workers : Monthly Labor Review: U.S. Bureau of Labor Statistics. Published 2005. Accessed March 3, 2021. https://www.bls.gov/opub/mlr/2005/article/fatal-work-injuries-among-foreign-born- hispanic-workers.htm

75. Dong X, Platner JW. Occupational fatalities of Hispanic construction workers from 1992 to 2000. Am J Ind Med. 2004;45(1):45-54. doi:https://doi.org/10.1002/ajim.10322

76. Centers for Disease Control and Prevention, USA. Work-related injury deaths among Hispanics--United States, 1992-2006. MMWR Morb Mortal Wkly Rep. 2008;57(22):597- 600.

77. Bailer AJ, Stayner LT, Stout NA, Reed LD, Gilbert SJ. Trends in rates of occupational fatal injuries in the United States (1983-92). Occup Environ Med. 1998;55(7):485-489. doi:10.1136/oem.55.7.485

78. Smith GS. Public health approaches to occupational injury prevention: do they work? Inj Prev. 2001;7(suppl 1):i3-i10. doi:10.1136/ip.7.suppl_1.i3

79. Centers for Disease Control and Prevention, USA. Improvements in workplace safety-- United States, 1900-1999. MMWR Morb Mortal Wkly Rep. 1999;48(22):461-469.

80. Richardson D, Loomis D. Trends in fatal occupational injuries and industrial restructuring in North Carolina in the 1980s. Am J Public Health. 1997;87(6):1041-1043. doi:10.2105/AJPH.87.6.1041

81. Smith JDR. Immigrant workers and worker’s compensation: The need for reform. Am J Ind Med. 2012;55(6):537-544. doi:10.1002/ajim.22033

82. Byler CG, Robinson WC. Differences in Patterns of Mortality Between Foreign-Born and Native-Born Workers Due to Fatal Occupational Injury in the USA from 2003 to 2010. J Immigr Minor Health. 2018;20(1):26-32. doi:10.1007/s10903-016-0503-2

83. Byler CG. Hispanic/Latino Fatal Occupational Injury Rates. Mon Labor Rev. 2013;136:14.

84. Sinyai C, Choi S. Fifteen years of American construction occupational safety and health research. Saf Sci. 2020;131:104915. doi:10.1016/j.ssci.2020.104915

85. Al-Bayati AJ, Abudayyeh O, Fredericks T, Butt SE. Reducing Fatality Rates of the Hispanic Workforce in the U.S. Construction Industry: Challenges and Strategies. J Constr Eng Manag. 2017;143(3):04016105. doi:10.1061/(ASCE)CO.1943-7862.0001269

86. "Safety Is Seguridad: A Workshop Summary by the National Academy of Sciences. National Academies Press (US); 2003. doi:10.17226/10641

87. Flynn MA. Safety & the Diverse Workforce. Prof Saf. 2014;59(6):52-57.

88. Fuentes CMD, Pantoja LM, Tarver M, Geschwind SA, Lara M. Latino immigrant day laborer perceptions of occupational safety and health information preferences. Am J Ind Med. 2016;59(6):476-485. doi:10.1002/ajim.22575

89. Landsbergis PA, Grzywacz JG, LaMontagne AD. Work organization, job insecurity, and occupational health disparities. Am J Ind Med. 2014;57(5):495-515. doi:10.1002/ajim.22126

90. Gany F, Novo P, Dobslaw R, Leng J. Urban Occupational Health in the Mexican and Latino/Latina Immigrant Population: A Literature Review. J Immigr Minor Health. 2014;16(5):846-855. doi:10.1007/s10903-013-9806-8

91. Martínez AD, Piedramartel A, Agnew J. Going Beyond the Injury: Regulatory Conditions Contributing to Latina/o Immigrants’ Occupational Psychosocial Stressors. Front Public Health. 2015;3. doi:10.3389/fpubh.2015.00240

92. Haro AY, Kuhn R, Rodriguez MA, Theodore N, Melendez E, Valenzuela A. Beyond Occupational Hazards: Abuse of Day Laborers and Health. J Immigr Minor Health. 2020;22(6):1172-1183. doi:10.1007/s10903-020-01094-3

93. Visser MA. Two Plus Two Equals Three: Classification Error and the Hispanic Undercount in United States Census Surveys. Am Rev Public Adm. 2014;44(2):233-251. doi:10.1177/0275074012461296

94. Kissam E. Differential undercount of Mexican immigrant families in the U.S. Census. Stat J IAOS. 2017;33(3):797-816. doi:10.3233/SJI-170388

95. O’Hare WP. Differential Undercounts in the U.S. Census : Who Is Missed? Springer Nature; 2019. Accessed June 30, 2021. https://library.oapen.org/handle/20.500.12657/23087

96. Chi C-F, Wu M-L. Fatal occupational injuries in Taiwan — Relationship between fatality rate and age. Saf Sci. 1997;27(1):1-17. doi:10.1016/S0925-7535(97)00046-5

97. Janicak CA. Occupational fatalities due to electrocutions in the construction industry. J Safety Res. 2008;39(6):617-621. doi:10.1016/j.jsr.2008.10.007

98. Jeong BY. Occupational deaths and injuries in the construction industry. Appl Ergon. 1998;29(5):355-360. doi:10.1016/S0003-6870(97)00077-X

99. Lin Y-H, Chen C-Y, Luo J-L. Gender and age distribution of occupational fatalities in Taiwan. Accid Anal Prev. 2008;40(4):1604-1610. doi:10.1016/j.aap.2008.04.008

100. Sorock GS, Smith EO, Goldoft M. Fatal Occupational Injuries in the New Jersey Construction Industry, 1983 to 1989. J Occup Environ Med. 1993;35(9):916-921.

101. Farrow A, Reynolds F. Health and safety of the older worker. Occup Med. 2012;62(1):4- 11. doi:10.1093/occmed/kqr148

102. Findley ME, Bennett JO. Safety & the silver collar worker. Prof Saf. 2002;47(5):34-38.

103. Hobson J, Smedley J. Fitness for Work: The Medical Aspects. Oxford University Press; 2019.

104. Blackstone A, Houle J, Uggen C. “I Didn’t Recognize It as a Bad Experience until I was Much Older”: Age, Experience, and Workers’ Perceptions of Sexual Harassment. Sociol Spectr. 2014;34(4):314-337. doi:10.1080/02732173.2014.917247

105. Einarsen S, Skogstad A. Bullying at work: Epidemiological findings in public and private organizations. Eur J Work Organ Psychol. 1996;5(2):185-201. doi:10.1080/13594329608414854

106. Harris K, Krygsman S, Waschenko J, Laliberte Rudman D. Ageism and the Older Worker: A Scoping Review. The Gerontologist. 2018;58(2):e1-e14. doi:10.1093/geront/gnw194

107. Liu S-F, Courtenay BC, Valentine T. Managing Older Worker Training: A Literature Review and Conceptual Framework. Educ Gerontol. 2011;37(12):1040-1062. doi:10.1080/03601277.2010.500576

108. McVittie C, McKinlay A, Widdicombe S. Committed to (un)equal opportunities?: ‘New ageism’ and the older worker. Br J Soc Psychol. 2003;42(4):595-612. doi:https://doi.org/10.1348/014466603322595293

109. Harootyan R. A RISING TIDE LIFTS ALL BOATS: OLDER WORKER EMPLOYMENT RATES. Innov Aging. 2019;3(Supplement_1):S646-S646. doi:10.1093/geroni/igz038.2398

110. Hairault J-O, Sopraseuth T, Langot F. Distance to Retirement and Older Workers‘ Employment: The Case for Delaying the Retirement Age. J Eur Econ Assoc. 2010;8(5):1034-1076. doi:10.1111/j.1542-4774.2010.tb00547.x

111. Kanfer R, Beier ME, Ackerman PL. Goals and motivation related to work in later adulthood: An organizing framework. Eur J Work Organ Psychol. 2013;22(3):253-264. doi:10.1080/1359432X.2012.734298

112. Desmette D, Gaillard M. When a “worker” becomes an “older worker”: The effects of age‐related social identity on attitudes towards retirement and work. van der Heijden B, Schalk R, J.P.M. van Veldhoven M, eds. Career Dev Int. 2008;13(2):168-185. doi:10.1108/13620430810860567

113. Moyers PA, Coleman SD. Adaptation of the older worker to occupational challenges. Work. 2004;22(2):71-78.

114. Czaja SJ. Technological change and the older worker. In: Handbook of the Psychology of Aging, 5th Ed. Academic Press; 2001:547-568.

115. Lain D, Airey L, Loretto W, Vickerstaff S. Understanding older worker precarity: the intersecting domains of jobs, households and the welfare state. Ageing Soc. 2019;39(10):2219-2241. doi:10.1017/S0144686X18001253

116. Shephard RJ. Human rights and the older worker: changes in work capacity with age. - Abstract - Europe PMC. Accessed June 7, 2021. https://europepmc.org/article/med/3574049

117. Lain D, Phillipson C. Extended Work Lives and the Rediscovery of the “Disadvantaged” Older Worker. Generations. 2019;43(3):71-77.

118. Sweet S. The Older Worker, Job Insecurity, and the New Economy. Generations. 2007;31(1):45-49.

119. Garg A. Ergonomics and the Older Worker: An Overview. Exp Aging Res. 1991;17(3):143-155. doi:10.1080/03610739108253894

120. Baugher D. Is the Older Worker Inherently Incompetent? Aging Work J Age Work Retire. 1978;1(4):243-250.

121. Charness N. A Modest Proposal for Solving the Older Worker Productivity Problem. Work Aging Retire. 2019;5(2):204-206. doi:10.1093/workar/waz001

122. Wiatrowski WJ. Older Workers and Severity of Occupational Injuries and Illnesses Involving Days Away from Work. Mon Labor Rev. 2005;2005. Accessed June 7, 2021. https://fraser.stlouisfed.org/title/monthly-labor-review-6130/july-2005-598218/olderworkers-severity-occupational-injuries-illnesses-involving-days-away-work-566460

123. Phillips DM. JCAHO Pain Management Standards Are Unveiled. JAMA. 2000;284(4):428-429. doi:10.1001/jama.284.4.423b

124. Hedegaard H. Drug Overdose Deaths in the United States, 1999–2017. 2018;(329):8.

125. Curtin, Sally. Death Rates Due to Suicide and Homicide among Persons Aged 10–24 : United States, 2000–2017. Department of Health and Human Services; 2019. Accessed June 19, 2021. https://stacks.cdc.gov/view/cdc/81944

126. Abd-Elsayed A. Prescription Drugs and the US Workforce:Results from a National Safety Council Survey. Pain Physician. 2020;1;23(1;1):1-16. doi:10.36076/ppj.2020/23/1

127. Kowalski-McGraw M, Green-McKenzie J, Pandalai SP, Schulte PA. Characterizing the Interrelationships of Prescription Opioid and Benzodiazepine Drugs With Worker Health and Workplace Hazards. J Occup Environ Med. 2017;59(11):1114-1126. doi:10.1097/JOM.0000000000001154

128. Jones CM, Paulozzi LJ, Mack KA. Alcohol Involvement in Opioid Pain Reliever and Benzodiazepine Drug Abuse–Related Emergency Department Visits and Drug-Related Deaths — United States, 2010. MMWR Morb Mortal Wkly Rep. 2014;63(40):881-885.

129. Loomis DP, Richardson DB, Wolf SH, Runyan CW, Butts JD. Fatal Occupational Injuries in a Southern State. Am J Epidemiol. 1997;145(12):1089-1099. doi:10.1093/oxfordjournals.aje.a009071

130. Peek-Asa C, Casteel CH. Documenting the need for translational research: an example from workplace violence prevention. Inj Prev. 2010;16(1):50-52. doi:10.1136/ip.2009.022756

131. Baumert J, Schneider B, Lukaschek K, et al. Adverse conditions at the workplace are associated with increased suicide risk. J Psychiatr Res. 2014;57:90-95. doi:10.1016/j.jpsychires.2014.06.007

132. Blackmore ER, Stansfeld SA, Weller I, Munce S, Zagorski BM, Stewart DE. Major Depressive Episodes and Work Stress: Results From a National Population Survey. Am J Public Health. 2007;97(11):2088-2093. doi:10.2105/AJPH.2006.104406

133. Stansfeld SA, Fuhrer R, Shipley MJ, Marmot MG. Work characteristics predict psychiatric disorder: prospective results from the Whitehall II Study. Occup Environ Med. 1999;56(5):302-307. doi:10.1136/oem.56.5.302

134. Wang J, Patten SB, Currie S, Sareen J, Schmitz N. A Population-based Longitudinal Study on Work Environmental Factors and the Risk of Major Depressive Disorder. Am J Epidemiol. 2012;176(1):52-59. doi:10.1093/aje/kwr473

135. Barr B, Taylor-Robinson D, Scott-Samuel A, McKee M, Stuckler D. Suicides associated with the 2008-10 economic recession in England: time trend analysis. BMJ. 2012;345(aug13 2):e5142-e5142. doi:10.1136/bmj.e5142

136. Cicero TJ, Ellis MS, Surratt HL, Kurtz SP. The changing face of heroin use in the United States: a retrospective analysis of the past 50 years. JAMA Psychiatry. 2014;71(7):821- 826. doi:10.1001/jamapsychiatry.2014.366

137. Eisen EA, Chen KT, Elser H, et al. Suicide, overdose and worker exit in a cohort of Michigan autoworkers. J Epidemiol Community Health. Published online July 8, 2020:jech-2020-214117. doi:10.1136/jech-2020-214117

138. Death on the Job: The Toll of Neglect, 2019 | AFL-CIO. Accessed November 3, 2020. https://aflcio.org/reports/death-job-toll-neglect-2019

139. Farber HS. Job Loss in the Great Recession: Historical Perspective from the Displaced Workers Survey, 1984-2010. National Bureau of Economic Research; 2011. doi:10.3386/w17040