Epidemiology of Musculoskeletal Disorders and Workplace Factors

Abstract

The title of this chapter contains three key elements: (1) epidemiology, (2) musculoskeletal disorders (MSDs), and (3) workplace factors. In order to understand how these three elements are linked, it is necessary to separately define each element.

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Overview

The title of this chapter contains three key elements: (1) epidemiology, (2) musculoskeletal disorders (MSDs), and (3) workplace factors. In order to understand how these three elements are linked, it is necessary to separately define each element.

Epidemiology

Epidemiology is the study of the distribution and determinants of diseases and injuries in human populations (American Heritage Dictionary Editors, 2011). Disease does not develop randomly, and all individuals are not equally likely to develop a specific disease at a given time. Therefore, the risk of developing a disease is different for each individual, and it is a function of his or her personal characteristics (inheritance) and his/her surroundings (environment). In the occupational setting, epidemiology is often used to determine association or causation (and, hence, financial responsibility). Epidemiology can also help direct prevention programs by reduction of risk (Melhorn, 1996, 1999a; Melhorn, Wilkinson, Gardner, Horst, & Silkey, 1999; Melhorn, Wilkinson, & O’Malley, 2001; Melhorn, Wilkinson, & Riggs, 2001). The epidemiological literature on occupational disorders is often confusing because of conflicting evidence on the importance of various potential risk factors. This chapter describes basic epidemiologic methods so the reader can learn to evaluate and critically analyze the published literature on occupational disorders. Epidemiology requires a methodology for testing scientific hypotheses in groups of individuals (Melhorn & Hegmann, 2011). By understanding the fundamental strengths and limitations of the study design, combined with the implementation of published studies, it is possible to evaluate the strength of the evidence derived from these studies and even to make sense of conflicting results from different studies on the same issue. In this chapter, we will present an overview of the basic terminology used in epidemiology and their characteristics. Additional information regarding strengths and limitations of analytic (hypothesis testing) study designs, with an emphasis on observational study designs, can be found in this chapter(Melhorn, 2012a; Melhorn, Brooks, & Seaman, 2013).

Musculoskeletal Disorders

MSDs are not a specific medical diagnoses but are labels or descriptive terms for aches and pains that can affect the body’s muscles, joints, tendons, ligaments, and nerves. Unlike a specific medical diagnosis that requires precise criteria for a diagnosis (such as an appropriate subjective history, unique physical examination findings, and exact supporting studies), MSDs are basically “I hurt and I hurt at work or with physical activities” (Melhorn, 1994, 2012b). Musculoskeletal pain with physical activity is a normal physiological process. Energy is required to perform work. Work requires muscles to move. By-products, such as lactic acid, are created as potential energy and are turned into kinetic energy to move the muscle and complete the physical activity. Increasing functional capacity (the ability to do more work) is the key to physical conditioning that all athletes understand. This understanding gives rise to the common adage of coaches’ “no pain, no gain.” However, there are occasions when the work activities can contribute to the MSD or pain, and therefore, the musculoskeletal pain or disorder is considered juristically as work-compensable (Melhorn, 1997). This does not mean that the job caused the MSD, but it implies that the job activities may have contributed to the disorder. This determination requires an understanding of the legal threshold that is established by each jurisdiction as to what is considered work-compensable (Hegmann, Thiese, Oostema, & Melhorn, 2011; Melhorn, Ackerman, Talmage, & Hyman, 2011).

Workplace Factors

The World Health Organization (WHO) has characterized “work-related” diseases as multifactorial and considers the following list to illustrate a number of risk factors (e.g., physical, work organizational, psychosocial, individual, and sociocultural) that can contribute to causing these diseases. WHO also acknowledges that much of the controversy surrounding work-related MSDs is a result of their multifactorial nature. Commonly described workplace factors are included in Table 10.1 and commonly described individual risk factors are included in Table 10.2 (Melhorn, 1999b, 2000a; Melhorn, Wilkinson, & O’Malley, 2001; Melhorn, Wilkinson, & Riggs, 2001).

Table 10.1 Common list of possible workplace risk factors

Table 10.2 Common list of possible individual risk factors

Although there have been concerns expressed regarding the inclusion criteria and methodology, an additional reference source is Musculoskeletal Disorders and Workplace Factors—A Critical Review of Epidemiologic Evidence for Work-Related Musculoskeletal Disorders of the Neck, Upper Extremity, and Low Back by the US Department of Health and Human Services, Public Health Service, Centers for Disease Control and Prevention, and National Institute for Occupational Safety and Health (NIOSH), July 1997 (public domain) at www.cdc.gov/niosh/docs/97-141/pdfs/97-141.pdf.

Linking the Key Elements

Why is this explanation important? The answer is that because most of the “science” we have regarding MSDs comes from epidemiological studies which often use the inclusion criteria of a “survey” to establish a “diagnosis” of MSD, which is then used in determining an association (risk) between a specific activity and the onset of the MSD in question. Therefore, occupational exposures and their association with, or causation of, injuries and illnesses are often debated. Because a determination for association or causation is required to determine eligibility for compensation and, therefore, financial responsibility for workers’ compensation or tort cases, debates and disputed legal cases often ensue (Melhorn & Ackerman, 2008). The significance of such disputes is underscored by the reported 1997 data listing direct health-care costs for the nation’s work forces of more than $418 billion and indirect costs of more than $837 billion (Brady et al., 1997).

An example of how the science may differ from public opinion would be helpful. Carpal tunnel syndrome was linked to keyboard activities. Because this proposed linkage is appealing and pervasive and seems to make sense, the lay press has advanced this association despite quality scientific investigations that found little or no relationship between carpal tunnel syndrome and occupation or hand use (Andersen et al., 2003; Brenner, Bal, & Brenner, 2007; Clarke Stevens, Witt, Smith, & Weaver, 2001; Egilman, Punnett, Hjelm, & Welch, 1996; Fisher & Gorsche, 2004; Garland et al., 1996; Hadler, 1999; Lo, Raskin, Lester, & Lester, 2002; Lozano Calderon, Anthony, & Ring, 2008; Melhorn, Martin, Brooks, & Seaman, 2008, 2011, Nathan, Keinston, & Meadows, 1993, Nathan & Keniston, 1993; Nathan, Keniston, Myers, & Meadows, 1992; Nathan, Meadows, & Istvan, 2002; Nordstrom, Vierkant, DeStefano, & Layde, 1997; Ring, 2007). Using Bradford Hill causation criteria, if an activity is the cause, removal or modification of the activity (the keyboard) should result in a reduction of the incidence. Two studies found that keyboard modification did not change the incidence of carpal tunnel syndrome (Lincoln et al., 2000; Rempel, Tittiranonda, Burastero, Hudes, & So, 1999). So how do we know what we know? This chapter will discuss what we know and how we know the Epidemiology of Musculoskeletal Disorders and Workplace Factors. The terms musculoskeletal disorders (MSDs), musculoskeletal condition, and musculoskeletal pain will be used interchangeably.

Introduction

The huge costs of work-related musculoskeletal pain and its associated disability are not new or unique to the population of the United States, but are a worldwide problem. Many historical manifestations of workplace pain have been related to innovation and changing technology. Some examples include miners’ nystagmus (change from candle to battery-powered head lamps), train dispatchers’ nystagmus (due to watching fast moving trains pass by the station), telegraphists’ cramp (1900s due to tapping on key), and watchmakers’ cramp (spasm of the finger) (Culpin, 1933). A list of other historical conditions is provided in Table 10.3 (Zeppieri & Melhorn, 2000).

Table 10.3 History of workplace diseases

Musculoskeletal pain is often separated into two categories: occupational and nonoccupational. This distinction is often considered when reviewing the outcome of treatment but is commonly overlooked during treatment. This legal distinction is not required by the physician for treatment of the condition, but it has great importance for the patient. Injuries or illnesses can cause musculoskeletal pain in the workplace. An occupational injury by definition is one that results from a work-related event or from a single instantaneous exposure in the work environment. Injuries are reportable by the employer on the Occupational Safety and Health Administration (OSHA) 300 log if they result in lost work time, require medical treatment (other than first aid), or the worker experiences loss of consciousness, restriction of work activities or motion, or is transferred to another job (United States Bureau of Labor Statistics, 1997). An occupational illness is any abnormal condition or disorder (other than one resulting from an occupational injury) caused by exposure to a factor(s) associated with employment. Included in this category are acute and chronic illnesses or diseases that may be caused by inhalation, absorption, ingestion, or direct contact (United States Bureau of Labor Statistics, 1997). Musculoskeletal injuries are often defined as traditional traumatic injuries such as fractures, sprains, strains, dislocations, or lacerations, while musculoskeletal illnesses are commonly called cumulative trauma disorders (CTD), repetitive motion injuries (RMI), or musculoskeletal disorder (MSD).

Occupational medicine presents a number of challenges to the physician. Management of work-related musculoskeletal pain is often frustrating. Patients may have more complaints and longer recovery times, require longer and more frequent office visits, and may be accompanied by the employer or nurse case manager during the office visit (Black & Frost, 2011; Daniell, Fulton-Kehoe, Chiou, & Franklin, 2005). They frequently have more questions about work status, require more phone calls, and have more paper work requirements. Many have attorneys, and they commonly require a permanent physical impairment rating with subsequent depositions or mandatory court appearances. NCCI (National Council on Compensation Insurance, Inc.) data suggest that the average treatment duration is four times greater in workers’ compensation (WC) cases than in non-WC cases—206.6 versus 51.9 days, respectively (https://www.ncci.com/NCCIMain/Pages/Default.aspx). Treatment outcomes often shift from good to poor (Kasdan, Vender, Lewis, Stallings, & Melhorn, 1996). The negative shift in outcome indicates that WC involvement introduces additional factors that influence patients and complicate treatment efforts (Berecki-Gisolf, Clay, Collie, & McClure, 2012). Traditional Western medical education is heavily weighted in the scientific study of the biologic systems of health and disease, often to the exclusion of biopsychosocial factors (Zeppieri, 1999). Physicians who provide care to those with work-related injuries are often inadequately prepared to deal with the biosocial (also labeled as psychosocial or biopsychosocial) issues—including motivation, social factors, psychological overlays, economic incentives, and legal complications—that influence the outcomes of treatment (Marchand & Durand, 2011; Melhorn, 1998a). Those physicians who are adequately prepared are often faced with the difficult task of separating fact from fiction. Occasionally, the patient’s symptoms can be disproportional to the clinic examination. Because an occupationally related OSHA event requires only a complaint of pain, multiple subjective issues must be reviewed. This can make the clinical picture confusing and require more tests and studies to be used to arrive at the appropriate medical diagnosis, relative to a similar nonoccupational patient. Other factors impacting treatment costs might include somatization behavior among patients and medicalization among physicians (Barsky & Borus, 1995; Gross & Battie, 2005), cost shifting from commercial insurance to WC insurance (Butler, 1996), and removing disincentives for early return to work (National Practitioner Data Bank, 1994).

According to a 2011 survey conducted for the Center for Disease Control (http://www.cdc.gov/Workplace/), 51.5 % of adults reported a chronic musculoskeletal condition in 2009, twice the rate of chronic heart or respiratory conditions. Musculoskeletal conditions are so ubiquitous that they have become the third most common reason that Americans seek medical attention. A US Department of Health study showed that, from 1996 to 2004, managing musculoskeletal conditions, including lost wages, costs an average $850 billion annually (compared to the 1997 data above), making it the largest WC expense (http://www.hhs.gov/news/). For employers paying WC claims, the economic strain has reached a breaking point. How significant is the category of musculoskeletal conditions? Consider the following data:

• 80 % of all claims under WC are musculoskeletal sprain/strain injuries, with lower back injury consuming more than 33 % of every WC dollar.

• Back pain causes more than 314 million bed days and 187 million lost work days yearly (data from the US Department of Labor, 1998–2005).

• Employers lose 5.9 h of productivity per week from those suffering from musculoskeletal pain who continue to be on the job (referred to as “presenteeism”).

It should be noted that the exact prevalence rates/figures for occupational injuries and illnesses are not available. The best data for the United States are provided by the Annual Survey of Occupational Injuries and Illnesses by the Bureau of Labor Statistics (BLS), US Department of Labor. The annual BLS data are obtained by having employers complete their data entry at http://www.bls.gov/respondents/iif/. The website states “Welcome to the Survey of Occupational Injuries and Illnesses respondent’s website. This website is your source for information that will help you to complete and submit your response to the Survey of Occupational Injuries and Illnesses. You have been selected to participate in this survey to help us to obtain a complete and accurate representation of work-related injuries and illnesses in America’s work places.”

In order to understand the data, it is important to know the definitions for injuries and illnesses. According to OSHA, an occupational injury is any injury such as a cut, fracture, sprain, or amputation that results from a work accident or from a single instantaneous exposure in the work environment. Minor injuries are defined as injuries requiring only first aid treatment (e.g., not involving medical treatment, loss of consciousness, restricted work, or transfer to another job) and are not recorded in the logs. An occupational illness is any abnormal condition or disorder, other than one resulting from an occupational injury, caused by exposure to environmental factors associated with employment. Occupational illnesses include acute and chronic illnesses or diseases that may be caused by inhalation, absorption, ingestion, or direct contact. All occupational illnesses are recordable. However, there are known limitations of the BLS data (Melhorn & Ackerman, 2008). The survey estimates of occupational injuries and illnesses are based on a selected probability sample, rather than a census of the entire population. Because the data are based on a sample survey, the injury and illness counts are helpful estimates but are not accrued values. Underreporting, along with selection bias, can occur. Additionally, the survey measures only the number of new work-related injury and illness cases that are recognized, diagnosed, and reported during the year.

In September 2010, the BLS completed a major revision to the Occupational Injury and Illness Classification System (OIICS). The OIICS is used in the Census of Fatal Occupational Injuries (CFOI) and the Survey of Occupational Injuries and Illnesses (SOII) to code various circumstances of the individual injury or illness reported. OIICS provides a structure to classify the nature of the injury and part of the body affected, source and secondary source of the injury, and event or exposure that precipitated the injury. Data for 2010 reported 3,063,400 cases involving days away from work. Sprains, strains, and tears were 370,130, back injuries were 185,270, and falls were 208,470. The total recordable cases of nonfatal occupational injury and illness incidence rates among private industry employers declined in 2010 to 3.5 cases per 100 workers, from 3.6 in 2009 (http://www.bls.gov/news.release/osh.toc.htm).

Interesting facts include:

• Manufacturing was the sole private industry sector to experience an increase in the incidence rate of injuries and illnesses in 2010—rising to 4.4 cases per 100 full-time workers, from 4.3 cases the year earlier. The increased rate resulted from a larger decline in hours worked than the decline in the number of reported cases in the industry sector.

• Health care and social assistance experienced an incidence rate of injuries and illnesses of 5.2 cases per 100 full-time workers—down from 5.4 cases in 2009—and was the lone industry sector in which both reported employment and hours worked increased in 2010.

• National public sector estimates, covering more than 18.4 million state and local government workers, are available for the third consecutive year, with an incidence rate of 5.7 cases per 100 full-time workers in 2010; this was relatively unchanged from 2009 (Fig. 10.1).

  Fig. 10.1

  

  Nonfatal occupational injury and illness incidence rates by case type and ownership, 2010

• Approximately 2.9 million (94.9 %) of the 3.1 million nonfatal occupational injuries and illnesses in 2010 were injuries. Of these, 2.2 million (75.8 %) occurred in service-providing industries, which employed 82.4 % of the private industry workforce covered by this survey. The remaining 0.7 million injuries (24.2 %) occurred in goods-producing industries, which accounted for 17.6 % of private industry employment in 2010, while workplace illnesses accounted for 5.1 % of the 3.1 million injury and illness cases in 2010. The rate of workplace illnesses in 2010 (18.1 per 10,000 full-time workers) was not statistically different from the 2009 incidence rate (18.3 cases).

• Goods-producing industries, as a whole, accounted for 36.3 % of all occupational illness cases in 2010, resulting in an incident rate of 31.8 per 10,000 full-time workers—up from 29.1 cases in 2009. The manufacturing industry sector accounted for over 30 % of all private industry occupational illness cases, resulting in the highest illness incidence rate among all industry sectors of 41.9 cases per 10,000 full-time workers in 2010—an increase from 39.0 cases in 2009. Service-providing industries accounted for the remaining 63.7 % of private industry illness cases and experienced a rate of 14.6 cases per 10,000 full-time workers in 2010—statistically unchanged from the prior year. Among service-providing industry sectors, health care and social assistance contributed 24.2 % of all private industry illness cases and experienced an incidence rate of 30.2 cases per 10,000 full-time workers in 2010—down from 34.8 cases in 2009.

• Review of injury case type and the employer type is suggesting and interesting pattern to nonfatal injury and illness. Efforts by private industry to reduce “risk factors” in the workplace appears to be having some impact, while state and local government efforts have been less successful.

• Another source for data is the NCCI at www.ncci.com/. Their Workers Compensation Temporary Total Disability Indemnity Benefit Duration 2012 Update (https://www.ncci.com/nccimain/IndustryInformation/ResearchOutlook/Pages/WC-Temp-Benefit-2012-Upate.aspx) found that the average duration of temporary total disability (TTD) indemnity benefits began to increase at the onset of the recent recession and that the rate of increase had moderated for injuries occurring during the first 6 months of 2010. Using an additional 12 months of reported data, they find that this more moderate rate of increase continues for injuries occurring through the first 6 months of 2011.

• NCCI estimated that the ultimate mean duration of TTD indemnity benefits rose from 130 days for Accident Year 2005 to 147 days for Accident Year 2009 and rose again to 149 days for claims in the first half of Accident Year 2011. The national unemployment rate deteriorated from 4.6 % in December 2007 to 8.9 % in December 2011.

Therefore, the statistics surrounding musculoskeletal conditions clearly define them as the primary threat to employers’ WC programs. The magnitude of this problem is related to the three principle issues related to the delivery of efficient and effective care: (1) The condition often lacks a reliable or precise diagnosis. (2) This can lead to the use of ineffective treatment methods. (3) And there has been limited application or emphasis on self-care and preventive strategies (http://www.ctdmap.com/downloadsinfo/1887.aspx). Thus, the occupational physician must recognize, understand, and address these multiple factors to achieve the more favorable outcomes to treatment that are seen in non-WC injuries and illnesses (Melhorn & Talmage, 2011).

Definitions

In order to provide a consistent approach to definitions and terms, this section has been provided with permission from the American Medical Association’s Press Guides to the Evaluation of Disease and Injury Causation (editors J. Mark Melhorn and William E. Ackerman, Chapter 1 Introduction) (Melhorn & Ackerman, 2008).

Evidence-Based Literature

Evidence-based medicine has become the standard for determining appropriate medical care. The most common definition was provided by Dr. David Sackett: “Evidence-based medicine is the conscientious, explicit, and judicious use of current best evidence in making decisions about the care of individual patients … [which] means integrating individual clinical expertise with the best available external clinical evidence from systematic research” (Sackett, Rosenberg, Gray, Haynes, & Richardson, 1996). Unfortunately, randomized controlled clinical studies are difficult to perform in the workplace and, hence, are uncommon. Therefore, most of the information available is from epidemiologic studies that can disprove, but not prove, an association (Hadler, 1999).

Epidemiology

As noted earlier, epidemiology focuses on the distribution and determinants of disease in groups of individuals who happen to have some characteristics, exposures, or diseases in common. Viewed as the study of the distribution and societal determinants of the health status of populations, epidemiology is the basic science foundation of public health (Melhorn, 1999c). The goal of epidemiologic studies is to identify factors associated (positively or negatively) with the development or recurrence of adverse medical conditions. A search strategy of bibliographic databases was used to identify epidemiologic literature that addresses causation of specific medical conditions, as outlined in Guides to the Evaluation of Disease and Injury Causation (editors Melhorn and Ackerman, Chapter 4 Methodology) (Melhorn & Hegmann, 2008). Although the referenced Chapter 4 is copyrighted, Drs. Melhorn and Hegmann have decided to offer the materials therein as “in the public domain and may be freely copied or reprinted” if appropriate acknowledgment of the reference source is used.

Specific Definitions

• Medical conditions are defined as an injury or illness that meets the standard criteria for an ICD-10 diagnosis (Melhorn & Ackerman, 2008).

• Disability refers to an alteration of an individual’s capacity to meet personal, social, or occupational demands or statutory or regulatory requirements because of impairment. Disability is a relational outcome, contingent on the environmental conditions in which activities are performed (AMA, 2001).

• Impairment refers to a loss, loss of use, or derangement of any body part, organ system, or organ function (AMA, 2001).

• Occupational exposures and physical factors at work are defined as identifiable occupational exposures to possible exacerbating or aggravating agents. For the musculoskeletal system, physical factors are often described in terms of repetition, force, posture, vibration, temperature, contact stress, and unaccustomed activities (CtdMAP, 2006; Melhorn, 1998b). For hearing, sound levels are measured in decibels. Radiation exposure is measured in millirads, and chemical exposure in milligram per cubic meter or parts per million.

• Nonoccupational exposures are defined as individual risk characteristics such as age, gender, hand preference, comorbid medical conditions such as diabetes, body mass index (BMI), depression, and hobbies.

• Under paragraph 1904.5(b)(1), the Occupa-tional Safety and Health Act (OSHA) defines the work environment as the establishment and other locations where one or more employees are working or are present as a condition of their employment. The work environment includes not only physical locations but also the equipment or materials used by the employee during the course of his or her work (U.S. Department of Labor, 2006a).

• Aggravation refers to a preexisting injury or illness that has been significantly aggravated, for purposes of OSHA injury and illness record keeping, when an event or exposure in the work environment results in any of the following:

  • Death, provided that the preexisting injury or illness would likely not have resulted in death but for the occupational event or exposure

  • Loss of consciousness, provided that the preexisting injury or illness would likely not have resulted in loss of consciousness but for the occupational event or exposure

  • One or more days away from work or days of restricted work or days of job transfer that otherwise would not have occurred but for the occupational event or exposure

  • Medical treatment in a case where no medical treatment was needed for the injury or illness before the workplace event or exposure or a change in medical treatment was necessitated by the workplace event or exposure (U.S. Department of Labor, 2006a)

The above are similar to aggravation as defined by the AMA Guides to the Evaluation of Permanent Impairment, fifth Edition: a factor(s) (e.g., physical, chemical, biological, or medical condition) that adversely alters the course or progression of the medical impairment or worsening of a preexisting medical condition or impairment (AMA, 2001).

• Exacerbation is defined as a transient worsening of a prior condition by an injury or illness, with the expectation that the situation will eventually return to baseline or pre-worsening level (Talmage & Melhorn, 2005). Some take issue with this definition because the signs or symptoms of a preexisting injury or illness may be temporarily worsened by something (i.e., activity, exposure, weather, reinjury), but the “something” is not an injury or illness. For example, a set of tennis will temporarily worsen the symptoms of degenerative arthritis in the serving shoulder, but tennis is neither an injury nor illness. This concept is clarified by the following.

Exacerbation is defined in the AMA Guides to the Evaluation of Permanent Impairment, sixth Edition, as temporary worsening of a preexisting condition. Following a transient increase in symptoms, signs, disability, and/or impairment, the person recovers to his or her baseline status or what it would have been had the exacerbation never occurred. Given a condition whose natural history is one of progressive worsening, following a prolonged but still temporary worsening, return to pre-exacerbation status would not be expected, despite the absence of permanent residuals from the new cause (Oakley, 2011, p. 611).

• Recurrence is defined as the reappearance of signs or symptoms of a prior injury or illness with minimal or no provocation and not necessarily related to work activities (Talmage & Melhorn, 2005).

• Apportionment is defined as a distribution or allocation of causation among multiple factors that caused or significantly contributed to the injury or disease and resulting impairment. The factor could be a preexisting injury, illness, or impairment (AMA, 2001).

For purposes of this present chapter, the words diagnosis, disorder, condition, injury, or illness are essentially considered the same.

Why Epidemiology?

Epidemiology, the science, is used to determine association or causation between MSDs and risk factors (individual and workplace). Understanding association or causation allows for intervention and treatment (medical) and the determination of compensability (legal and financial responsibility). Medical and legal “causation” are not the same. This concept of medical and legal causation has been discussed in detail in the following publications: Chapter 1 Introduction, Chapter 2 Understanding Work-Relatedness, and Chapter 3 Causal Associations and Determination of Work-Relatedness in Guides to the Evaluation of Disease and Injury Causation, Second Edition, editors Melhorn, Ackerman, Talmage, and Hyman, AMA Press (2011) granted. Medical intervention and treatment can include prevention. Prevention comes in three forms (jmm, 1999). Primary prevention keeps disorders from occurring. It is focused on the universal application of safety and health and, when successful, reduces the risk and obviates the need for secondary or tertiary prevention. Once a disorder has emerged (become detectable), primary prevention is not possible; secondary prevention must be designed to keep the disorder from increasing in severity. The goal of secondary prevention is to arrest the growth of the disorder and, if possible, reverse or correct it. This process is the traditional health-care model. Tertiary prevention is designed for disorders that have reached advanced stages of development and threaten to produce significant side effects or complications. The goal of tertiary prevention is to keep the disorder from overwhelming the individual, leading to long-term disability. Unfortunately, prevention of musculoskeletal conditions has also been limited by legislated mandates such as ADA and Equal Employment Opportunity Commission (EEOC). ADA is the Americans with Disabilities Act of 1990, including changes made by the ADA Amendments Act of 2008 (Pub L 110-325), which became effective on January 1, 2009. The ADA was originally enacted in public law format and later rearranged and published in the US Code. The EEOC is the agency that watches for discriminatory practices that are prohibited under Title VII, the ADA, GINA, and the ADEA, for any aspect of employment, including testing, training and apprenticeship programs, and other terms and conditions of employment. GINA is the Genetic Information Nondiscrimination Act of 2008 (Pub L 110-233, 122 Stat. 881, enacted May 21, 2008). ADEA is the Age Discrimination in Employment Act of 1967. Although the goals of such legislation are socially appropriate, the application of the law has been detrimental to the prevention of musculoskeletal workplace injuries and illnesses. Additionally, the above legislation can be in conflict with the Occupational Safety and Health Act of 1970 (OSH Act) which requires the employer to provide a safe workplace. OSH Act is Pub L 91-596, 84 STAT. 1590, 91st Congress, S.2193, December 29, 1970, as amended through January 1, 2004, “To assure safe and healthful working conditions for working men and women; by authorizing enforcement of the standards developed under the Act; by assisting and encouraging the States in their efforts to assure safe and healthful working conditions; by providing for research, information, education, and training in the field of occupational safety and health; and for other purposes.” For example, if medical screening or testing has determined an increased risk for a musculoskeletal condition in an individual, an attempt by the employer to reduce the workplace risk could be interpreted as “a discriminatory action and therefore punishable” even though the effort on the employer’s part is to reduce risk of the musculoskeletal condition for the individual in compliance with the OSH Act. This obvious confusion has led many employers to elect not to precede with appropriate prevention programs to the detriment of the worker (Melhorn et al., 1999, Melhorn, Kennedy, & Wilkinson, 2002; Melhorn, Wilkinson, & Riggs, 2001; Melhorn, Wilkinson, & O’Malley, 2001; jmm, 1999).

What We Know and How We Know It

Epistemology (E·pis·te·mol·o·gy) [ih-pis-tuh-mol-uh-jee] is a branch of philosophy that investigates the origin, nature, methods, and limits of human knowledge. In other words “What We Know and How We Know It” or “What There Is to Know About Knowing” (Melhorn, 2008). Therefore, as is often the case, a decision on association or causation may be difficult because the determination is based on imperfect or inadequate information (the science). To understand epidemiology it is important to acknowledge these intrinsic limitations. Limiting causal conclusions to proven and established facts does not guarantee that future studies will not prove the current data wrong. Conversely, shunning everything unproven will result in rejection of many statements that are true but just not proven. The best illustration of this concept is in Table 10.4.

Table 10.4 Medical knowledge

The Science

Health-care providers are often asked whether a condition is work-related or not (i.e., if it is causally related to a specific occupational injury or exposure). It is incumbent upon the clinician to give an opinion based on a careful review and analysis of the individual’s clinical findings and his or her workplace exposures and the literature linking (or not) the injury or exposure of concern and the condition in question (Melhorn & Ackerman, 2008). In contrast to a witnessed occupational injury causing immediate symptoms and corroborated by objective physical and diagnostic test findings, a cause-and-effect relationship between a disease (nontraumatic injuries are classified by OSHA as illnesses) and an agent or condition in the workplace may be unclear. Occupational diseases may develop slowly, with months or years between exposure and onset of symptoms and/or signs. Disease manifestations may be confused with changes due to normal aging. Information on past work exposure is often unavailable, inadequate, or incomplete. In addition, not all individuals react or respond in the same way to similar exposures to disease-producing agents. In some cases, there is a clearly identifiable single cause for the condition, whether work-related or nonoccupational. More often, causation is multifactorial, with one or more nonoccupational causes (e.g., age-related degeneration, smoking, or obesity), in addition to varying contribution from the workplace.

Causality determination may be difficult and result in contested claims. Honest differences of opinion are common when the facts are subject to different interpretations. Therefore, considerable judgment is necessary when data are lacking or incomplete. With occupational diseases, what appears obvious to some may nevertheless still be controversial, and it is important to assemble a complete database (history including occupational and nonoccupational exposures, physical and test findings, health-care records, etc.), be familiar with the relevant medical literature, and then review and analyze the data in a logical and unbiased manner to ensure a correct and equitable decision on causation. In 1976, the NIOSH created A Guide to the Work-Relatedness of Disease (Publication No. 79-116) to assist clinicians and, therein, provided a six-step method to assist in this decision-making process (Hegmann & Oostema, 2008; NIOSH, 1979). These six steps are listed in Table 10.5.

Table 10.5 NIOSH causation decision-making process

Consideration of Evidence

The first step in determining the probability of a cause-and-effect relationship, between an exposure in the workplace and the subject illness, is to establish that a disease does in fact exist and the disease and its manifestations appear to be the result of exposure to a specific harmful agent. Evidence elicited in the course of a medical evaluation should address these questions and specifically include the following:

• Complete medical, personal, family, military, and occupational histories from the employee

• A thorough physical examination and acquisition or review of appropriate radiographic, laboratory, or other diagnostic tests

• Analysis and reporting of these clinical data

The occupational history should include, but is not limited to:

• Job titles

• Type of work performed (complete listing of actual duties)

• Duration of each type of activity

• Dates of employment and worker’s age for each job activity

• Geographical and physical location of employment

• Product or service produced

• Condition of personal protective equipment used (if any) and frequency and duration of periods of use

• Nature of agents or substances to which worker is, or has been, exposed, if known (including frequency and average duration of each exposure situation)

The resultant report should include a complete list of all diagnoses, with an opinion, whenever possible, as to which diagnoses are occupationally related and which are not.

Consideration of Epidemiological Data

The essential approach of epidemiology is the investigation of relative and absolute measures of frequency while comparing the characteristics of individuals with and without the condition. The most obvious measures of frequency are case counts and their variations, which are often referred to as numerator data. This number (the numerator) describes the frequency of the disorder, without reference to the underlying population at risk (the dominator data). The US Congress recognized that statistics on workplace injuries and diseases were essential to an effective national program of occupational disease prevention (Melhorn & Ackerman, 2008). Therefore, when the OSHA was passed in 1970, employers were required to maintain records on workplace injuries and illnesses (commonly labeled as OSHA 300 logs). The act delegated the responsibility for collecting statistics on these occupational injuries and illnesses to the BLS. To comply with the OSHA, the BLS conducts an annual survey of the occupational injuries and illnesses in the United States (U.S. Department of Labor, 2006b). The survey compiles the OSHA 300 logs from over 200,000 establishments, grouped together by industry codes established by BLS as the North American Industry Classification System (NAICS) (http://www.bls.gov/bls/naics.htm). The frequency of the particular disorder can also be expressed as a proportionate ratio (the number of cases of the particular disorder, compared to cases of all disorders, in the study population). By itself, numerator data cannot provide useful information regarding the risk or probability of acquiring the disorder. The case frequency has to be related to the underlying population that could have potentially developed the disorder (the denominator). Without the denominator (the number of people at risk), it is not possible to estimate the risk of a specific condition in the population or to test hypotheses regarding risk factors for a specific condition.

There are, though, known limitations of the BLS data. The survey estimates of occupational injuries and illnesses are based on a scientifically selected probability sample rather than a census of the entire population. Because the data are based on a sample survey, the injury and illness estimates probably differ from the figures that would be obtained from all units covered by the survey. Also, the survey measures the number of new work-related illness cases that are recognized, diagnosed, and reported during the year. Some conditions (e.g., long-term latent illnesses caused by exposure to carcinogens) often are difficult to relate to the workplace and are not adequately recognized and reported. These long-term latent illnesses are believed to be understated in the survey’s illness measures. In contrast, the overwhelming majority of the reported new illnesses are those that are easier to track (e.g., contact dermatitis) (Melhorn & Ackerman, 2008). Furthermore, employer bias in selecting which conditions to report may result in underreporting. Additionally, the OSHA definition for work-relatedness is more inclusive than most. Injuries and illnesses that occur at work may not have a clear connection to an occupational activity or substance peculiar to the work environment. For example, an employee may trip for no apparent reason while walking across a level factory floor, be sexually assaulted by a co-worker, or be injured accidentally as a result of an act of violence perpetrated by one co-worker against a third party. For this reason, rates are often used when the objective is to assess the risk of the disorder or determinants of disorders or their outcomes.

Rates

Rates describe the frequency of a disorder (or disorder per unit size of the population per unit time of observation). The most common rates are incidence and prevalence. The incidence rate is based on new cases of a disorder, whereas the prevalence rate reflects existing cases. Because they are based on new versus existing cases, incidence and prevalence rates have different uses and limitations. Therefore, the incidence rate is a rate of change, often described as the frequency with which people change from healthy to injured, sick, or disabled. Thus, the appropriate denominator is the population at risk of acquiring the disorder (i.e., those who are free of the disorder at the start of the time interval). The incidence rate may be quantified in a number of ways when the population is stable and the number of new events is counted each year. This is often expressed as the number of new events per 1,000 persons per year. Alternatively, incidence rate may be quantified as the number of new events per 1,000 person-years, as is done in prospective studies where a fixed population is followed until the end of the study. In practice, although the best denominator for incidence rates is the number of people free of the disorder at the start of the time interval, surveillance incidence rates (and prevalence rates) that are based on case reports often use the total population derived from estimates or census data.

The prevalence rate is the number of existing cases of a disorder in a given population in a given time period, while point prevalence is the number of cases per unit population at one moment. For example, point prevalence would be the number of railroad employees receiving disability because of a medical condition on a specific day such as January 1, 2010. Therefore, the unit of time is not expressed. A period prevalence would be the number of cases existing at one time during a definable time interval such as 1 year. Lifetime prevalence (which is a form of period prevalence) is defined as the number of individuals in a population that, at some point in their lives (which could be several to more than 100 years), have the condition in question, compared to the total number of persons. Prevalence is sometimes not defined as a rate because, in practice, data are often derived from surveys that are difficult to assign to a specific time interval. A number of variables, other than the risk factor under study, may affect the incidence and prevalence rates. Examples include demographic characteristics of the underlying population. The most common variable is age distribution because aging is associated with the onset of most disorders. Gender and ethnicity distributions must also be taken into account. Other confounders that can distort the incidence rate include company policies, WC claims, and health-care system influences that affect the likelihood of seeking medical attention, of being diagnosed with a given disorder, or of having the disorder reported. These variables must be considered when measures of disorder frequency are evaluated, particularly when changes are assessed over time or when different populations are compared. In order to eliminate the effects of differences in these variables, the rates may be adjusted or standardized algebraically. The adjusted rates express the risk of acquiring the disorder in the populations being compared as if they had the same age, sex, and ethnicity distributions. In other words, the “variables” have been accounted for. Sometimes, it is appropriate to not account for these variables (e.g., the morbidity rates within population strata defined by age, sex, and ethnicity). Remember, the number of existing cases of a disorder at any time is a function of both the rate of new cases (incidence) and the duration of that disorder. Accordingly, a change in prevalence may reflect changes in the incidence rate, duration, or both. Consequently, when a population is stable and the duration of a disorder is also stable, it is possible to estimate prevalence from incidence and vice versa, according to the following approximation:

$$ \mathrm{Prevalence}\approx \mathrm{incidence}\times \mathrm{duration} $$

Therefore, rates become the first step in considering causality and lead to further epidemiological studies.

Epidemiological Study Design

Epidemiological studies are of two major types which can be subdivided. The first is the descriptive epidemiology study, which drives the need to explain variation and formulate causal hypotheses that draw on current available information. However, while it supports the development of causal hypotheses, descriptive epidemiology does not itself support conclusions about disorder causality or any hypotheses. In descriptive epidemiology, the frequency of a disorder in the population is characterized in terms of person (e.g., individual risk factors—age, gender, ethnicity-specific incidence rates, economic, behavioral, occupational, and other factors), place (country, rural, urban, type of industry, job requirements), and time (day, week, month, year, lifetime), as seen in Table 10.5. Each epidemiologic study also has certain basic elements: occurrence relation, outcome, determinant(s), study population, and domain. Determinants are defined as the risk factors related to the diagnosis. The study population must be well defined in order to allow the data obtained to be applied, or theoretically generalized, to a larger population called the domain. This requirement is often described as external validity (Table 10.6).

Table 10.6 Descriptive epidemiology study design

Specific hypotheses are developed by inductive reasoning to explain observed patterns of variation and then evaluated using a study designed to test them. Studies that test specific hypotheses are analytic epidemiologic studies (the second major type of epidemiologic study). As the results of hypothesis testing (analytic), studies are accrued, and their data added to the basis for causal inference, depending on their strengths and generalizability. Hypotheses can then be supported, modified, or negated. Analytic (hypothesis testing) epidemiology relies on two types of study designs: observational and experimental. In observational studies, exposure to the hypothesized causal factor and the subsequent development of the selected disorder in the population under study occur in the natural course of events; in other words, the investigator does not cause the exposure to the causal factor. The study is designed and implemented to maximize the extent to which it is a natural experiment. Extraneous sources of variation are eliminated, and only exposure to the alleged cause and the frequency of the selected disorder vary between populations being compared. Once substantial observational evidence has accrued, causality is often widely accepted. However, only prospective randomized interventional or experimental studies can prove causation; and these are unlikely to be performed in the workplace as it would require exposing individuals to known or suspected risk factors and, thereby, potential harm.

Literature Review Summaries

Epidemiological versus individual causal assessment requires determination that a “risk factor” is truly a disease determinant, rather than merely an associated factor (Hegmann & Oostema, 2008). If the risk factor is causal, then elimination of the risk factor must result in fewer cases of the particular disease. Literature review summaries require five steps, as listed in Table 10.7. To summarize, one must try to avoid omitting articles in the review of the literature. Of course, the purpose of a well-designed study is to provide insight into the “truth” regarding causation. The ability to determine the truth, or to infer from a limited sample to the whole, is compromised by a systematic or study design flaw in the form of bias and/or confounding. Alternatively, chance or random occurrence may influence whether the results of a study accurately reflect the truth. Etiologic epidemiology tests whether a hypothesized factor is a determinant or cause of a disorder in previously healthy population, whereas, in clinical epidemiology, one tests whether risk factors are determinants of the specific disease. The classic observational analytic study designs are the cohort study, the case–control study, and the cross-sectional study.

Table 10.7 Literature review

Table 10.8 summarizes the various types of study designs, based on their strength.

Table 10.8 Study design pyramid

Because it is not possible to study the entire universe of potentially eligible subjects (workers), epidemiologic studies are conducted on samples of the population of interest. Even a study of an entire city’s work force constitutes a sample. The method of sampling should not introduce selection biases, but epidemiological studies are commonly affected by them. For example, no characteristics of the individuals should affect the likelihood of selection for the study. However, a volunteer study is potentially susceptible to selection bias because the health behavior and health status of people who volunteer for research are known to be better than those who refuse. Relatedly, internal validity refers to both how well a scientific study was conducted and how confidently one can conclude there is a cause-and-effect relationship. Research design, definitions used, what variables were and were not measured, how accurately they were measured, completeness of data collection, and other factors all influence validity. This applies to both descriptive and experimental studies. However, in experimental studies, one also wants to know how certain it is that the effect was caused by the independent variable rather than extraneous ones. For example, did the treatment really cause or contribute to the difference observed between subjects in the control and experimental groups? If there is inaccuracy (measurement error) in the information collected, the ability to detect the association of interest is reduced. If the accuracy of information is worse for one exposure group than another, the effect on the study results may not be predictable. Hence, evaluation of the accuracy (or validity) of measurements is necessary for any study. Research reports should describe the validity of the sources of information. For example, questionnaires or reporting methods that have been validated in the study population, or in similar populations or circumstances, should be used. Finally, the strength of evidence regarding etiology varies depending on the type of study. Prospective cohort studies are best, while retrospective cohort studies are of low to medium strength, and case–control and cross-sectional are of low strength. Frequently used analysis tools are listed in Table 10.9.

Table 10.9 Common statistical tests

Cohort Study

Cohort studies can be prospective or retrospective. The direction of data collection or investigation can be seen in Table 10.10. When well designed and executed, a cohort study produces the soundest results for incidence rates, disorder etiology, and/or prognostic determinants of all the observational study designs. The hallmark of a cohort study is that a population is initially free of the disease of interest. Potential confounders and important covariates are identified and characterized with respect to the hypothesized risk factor. The population (cohort) is observed for a period of time adequate for development of the disorder (exposure), and the new (incident) cases are recorded (outcomes). Rates of disorder development are compared between those who are and are not exposed to the hypothesized risk factor. Loss to follow-up, though, is a potential problem. If a number of individuals are lost to follow-up, the observed relative risk underestimates the true relative risk. Selective survival or selective attrition bias can occur. Long latency periods increase the cost to continue these studies. In retrospective studies, the exposure occurs first and then the outcome is observed, which allows for the individual to be included in the “cases.”

Table 10.10 Direction of investigation

Case–Control Study

The essential feature of the case–control study that differentiates it from other observational study types is that individuals are selected for the study based on the presence of the disorder in question (cases) and then compared with others who do not have the disorder (control subjects). The presence or absence of the hypothesized cause is then ascertained in both case and control subjects. Although this appears to be a simple undertaking, case–control studies present a number of methodological challenges that must be solved for the study results to be valid. Case–control studies frequently suffer from information biases and unbiased recall failure.

Cross-Sectional Study

Cross-sectional studies simultaneously ascertain exposure to risk factors and the presence of the disorder in question in a population sampled, without regard to the presence of either. This type of sampling is sometimes called naturalistic sampling. In contrast to a cohort study, which follows subjects over time and ascertains incidence, a cross-sectional study ascertains conditions present at the moment of study, that is, prevalence of the disorder. The estimates of relative risk derived from cross-sectional studies are therefore estimates of prevalence relative risk. Cross-sectional or survey studies are often undertaken because, unlike case–control studies, they require few a priori decisions regarding subject selection and, unlike cohort studies, it is not necessary to wait for the study outcome. These advantages are offset by their susceptibility to some of the problems of both cohort and case–control studies listed above and resultant decreased strength of evidence.

Assess the Methods of Each Study

Strength and weakness of the data should be assessed. Another way of looking at this is to consider “threats to validity.” There are three general reasons why the results of a study may not be valid: chance, bias, and confounding.

• Chance

Chance is defined as the absence of any cause of events that can be predicted, understood, or controlled (American Heritage Dictionary Editors, 2011). Measurements made during research are nearly always subject to random variation. Determining whether findings are due to chance is a key feature of statistical analysis. The best way to avoid error due to random variation is to ensure that the sample size is adequate (O’Keefe, Huffman, & Bukata, 2008). The confidence interval is a plus-or-minus figure and is often reported as the margin of error. For example, with a confidence interval of 5, if 40 % percent of a sample picks an answer, one can be reasonably certain that between 35 and 45 % (40 ± 5) of the entire study population would have selected the same answer if asked that question. The confidence level is the statistical likelihood that a variable lies within the confidence interval, expressed as a percentage, such as, 50, 95, or 99 %. It measures the reliability of a statistical result and indicates the probability of the result is correct. The 95 % confidence level is most commonly used. The sample size required depends on the confidence and confidence level (Creative Research Systems, 2011). For a given confidence level, the larger the sample size, the smaller your confidence interval and the more certain one can be that the results truly reflect the entire population.

• Bias

While chance is due to random variation, bias is caused by systematic variation. A systematic error in the selection of study subjects, disease or condition, outcome measures, or data analysis will lead to inaccurate results. The numerous types of bias can be broadly divided into three categories delineated below:

• Selection bias: The selection of individuals for a sample or their allocation to groups may produce a sample not representative of the entire population. Random selection and allocation prevent this type of bias.

• Measurement bias: Measurement of a condition or outcomes may be inaccurate due to inaccuracy in criteria for diagnosis of the disease or a measurement instrument or bias in the expectations of study participants or researchers. The latter may be addressed by blinding both the subjects and investigators to the real purpose of the study.

• Analysis bias: The protection against bias afforded by randomization will be maintained only if subjects remain in the study group to which they were allocated and then complete follow-up. Participants who change groups, withdraw from the study, or are lost to follow-up may be systematically different from those who complete the study. Analysis bias can be reduced by maximizing follow-up.

There are also other factors to be considered in determining the strength studies, as delineated below.

• Accuracy and precision

Random variation (chance) leads to imprecise results, while systematic variation (bias) leads to inaccurate results. For example, a large observational study involving thousands of individuals may produce results that are precise (specific), but not accurate. A small, high-quality randomized controlled trial may produce results that are accurate but not precise.

• Confounding

This is similar to bias and is often confused. Whereas bias involves error in the measurement of a variable, confounding involves error in the interpretation of what may be an accurate measurement. A classic example of confounding is to interpret the finding that people who carry matches are more likely to develop lung cancer as a cause-and-effect relationship. Smoking is the confounder. Smokers are more likely to carry matches and also more likely to develop lung cancer. Confounding occurs when the study results can be explained by a factor unnecessary to the hypothesis being tested. A potential confounding factor must be associated with both the disorder in question and the hypothesized cause. That is to say, the study group with the disorder having the confounding exposure must be different from the study group without the disorder who also have the confounding exposure. Additionally, it is necessary that the study group of those with the hypothesized cause and confounding exposure are different from the group not exposed to the hypothesized cause who have the confounding factor. For example, a study finding an association between low job satisfaction and occupational carpal tunnel syndrome could be confounded by the physical requirements of work. Specifically, those individuals whose work involves repetitive high force activities in a cold environment are at greater risk of developing carpal tunnel syndrome, but this group also has a lower job satisfaction than individuals employed in less physically demanding occupations. So, which factor is actually responsible for the risk of occupational carpal tunnel syndrome? Potential confounding factors can be eliminated in the design of the study by restricted or matched sampling or, in the data analysis phase, by stratified or multivariate analysis, for example. In the study just described, if statistical analyses controlled for the physical requirements of work, or if the researchers conducted an exploratory analysis and found no association between job satisfaction and the physical requirements of work, the confounding could be reduced or eliminated. In experimental studies, potential confounding should be eliminated by truly random assignment of individuals to the treatment and control groups. Comparability of the groups should be confirmed by presentation of the baseline characteristics of each group upon entry to the study. Thus, confounding invalidates a study as a test of the null hypothesis, and its results cannot be taken as evidence of causality. Lack of generalizability, unlike confounding, does not invalidate a study’s results, but merely restricts inference to populations similar to those under study.

Because of its importance, we will summarize the concept of confounders.

• What is a confounder?

A confounder is any factor that is prognostically linked to the disease of interest and unevenly distributed between the study groups. A factor is NOT a confounder if it lies on the causal pathway between the elements of interest. For example, the relationship between diet and coronary heart disease may be explained by serum cholesterol level. Elevated cholesterol is not a confounder because it may be the causal link between diet and coronary heart disease.

• Known confounders

Dealing with confounding is relatively easy if the likely confounders are known. The data could be stratified. For example, in the study on diet and coronary heart disease, smokers and nonsmokers could be analyzed separately, or one could use statistical techniques to adjust for confounding.

• Unknown confounders

Allocating for unknown confounders is much more difficult. There is always a risk that an apparent association between a risk factor and the disease is being mediated by an unknown confounder. This is particularly true of observational studies where selection is not randomized. Again, randomization suggests that both known and unknown confounders will be approximately evenly distributed between two study groups.

Ascertain Statistical Significance

Research is typically conducted on a sample of individuals (the study group) from a target population. Therefore, the results of such studies are estimates of the true means, proportions, relative risks, etc. of the populations from which the sample groups were selected. The precision of a study estimate of the population value is described by the standard error of estimate. The standard error (SE) is the square root of the ratio of the variance (s ²), or variability of the measurement in the sample, to the number of subjects (N) in the study, as expressed by the formula

$$ S{E}_{mean}=\sqrt{\frac{s^2}{N}}. $$

The statistical hypothesis test evaluates the null hypothesis that the study results observed occurred because of sampling error when there was no true association in the population from which the sample of study subjects was derived. If the observed association is large enough that this kind of error is improbable, the null hypothesis is rejected. The investigators then accept the alternative hypothesis that the observed estimates of relative risk or association reflect the true situation in the sampled population. By convention, the cutoff for rejecting the null hypothesis is usually set at 0.05. Thus, if the probability (p value) that the observed results are due to sampling error is less than 0.05, the null hypothesis is rejected. The results are declared statistically significant because, within an acceptable margin of error, they probably did not occur by chance. The larger the observed association, relative to the underlying variability of the outcome being measured, the more likely it will be statistically significant.

A Type I error, also known as a false positive, occurs when a statistical hypothesis test rejects the null hypothesis, even though it is true. For example, the null hypothesis states a new treatment is no better than an older, less expensive one. A Type I error would occur if researchers concluded the new treatment produced outcomes when in reality there was no difference. The rate of Type I errors is represented by the Greek letter alpha (α) and usually equals the significance level of a test. Relatedly, a Type II error, also known as a false negative, occurs when a statistical hypothesis test fails to reject a false null hypothesis. Continuing the prior example, a Type II error would occur if researchers concluded there was no difference in outcomes between the new and old treatments when, in fact, the new treatment was more effective. False-negative results are often due to too small sample sizes. The rate of Type II error is represented by the Greek letter beta (β). Finally, the probability that a study will be able to correctly reject the null hypothesis when it is false, that is, correctly detect an association when there is one in the population, is called statistical power (1 − β). Table 10.11 illustrates the different conditions and possible results of a statistical hypothesis test.

Table 10.11 Testing hypothesis

In the planning phase of research, investigators should determine how strong an association (how large an estimated relative risk or how big a difference between treatments) would be statistically significant. Because a valid study requires that it be a true test of the research hypothesis, it is important to design it so the study has sufficient statistical power to detect a statistically significant association. The larger the sample size, the more power the statistical test has to detect associations. In other words, as expected differences or relative risks get smaller, the number of subjects studied must increase to have adequate statistical power to test the hypothesis. With very large sample sizes, it is possible to declare trivial associations statistically significant. When studies with small sample sizes report results that are not statistically significant, they should also report how large an association would have been required to detect it. One should also evaluate whether the observed difference and its upper confidence limit, although not statistically significant, are clinically significant. Conversely, when studies with large numbers of subjects report statistically significant results, one needs to determine if the differences are significant clinically or not.

Other Important Terms

An independent variable is one whose value determines that of other variables. A dependent variable is that which is observed in a study and whose changes are determined by the presence and extent of one or more independent variables. A continuous variable describes numerical information that can be any value within a range. Continuous data may be parametric or nonparametric. Parametric data may be represented in a distribution explained by a single mathematical equation. Nonparametric data are not represented by a single mathematical equation and do not belong to any particular distribution. Relative risk (RR) estimates the magnitude of the association between the exposure and disease of interest. RR equals the incidence of disease in exposed subjects divided by that in unexposed individuals. A RR of 1.0 means the disease incidence rates are identical in the exposed and unexposed study groups. A RR greater than 1.0 suggests a positive association (increased incidence in exposed group), while a RR of less than 1.0 suggests a negative or inverse association (decreased incidence in the exposed study group). Finally an odds ratio (OR) is used in retrospective case–control studies where incidence cannot be determined. OR equals the probability (odds) of being exposed in the group with the disease divided by the probability of exposure those without the disease. Statistical tests of inference require assumptions about the data type and distribution. Examples of statistical tests were previously listed in Table 10.9.

It should also be noted that disease detection and correct diagnosis depend on the sensitivity and specificity of tests. A test that yields a positive result when the disease is present is called a true positive. A positive test result when the disease is not present is a false positive. A negative result when the disease is not present is a true negative, whereas a negative result when the disease is present is a false negative. The positive predictive value (PPV) of a test is the probability that the patient has the disease when the test result is positive, specifically the number of true positives (TP) divided by the sum of true positives and false positives (FP). So, PPV = TP/(TP + FP). It can also be described as PPV = [(prevalence)(sensitivity)]/[(prevalence)(sensitivity) + (1-prevalence)(1-specificity)]. Negative predictive value (NPV) of a test is the probability that the patient does not have the disease when the test result is negative, specifically the number of true negatives (TN) divided by the sum of true negatives and false negatives (FN). NPV = TN/(TN + FN). It can also be calculated as NPV = [(specificity)(1 − prevalence)]/[(specificity)(1 − prevalence) + (1 − sensitivity)(prevalence)]. Disease prevalence affects the PPV. Also note that PPV is not intrinsic to the test. Table 10.12 summarizes some of the above review.

Table 10.12 Sensitivity and specificity

Conclusions About the Degree of Causal Association

Strength of evidence of causation in epidemiological studies can be combined using a point value to suggest an association between a risk factor and a disease or condition as very strong evidence, strong evidence, some evidence, or insufficient evidence (Melhorn & Hegmann, 2008). Additional details on this methodology are available in Guides to the Evaluation of Disease and Injury Causation (editors Melhorn and Ackerman), Chapter 4 Methodology, or can be downloaded from www.ctdmap.com/downloadsinfo/29101.aspx. This method is in the public domain and may be copied and used if appropriate acknowledgment is used. For additional information, the reader is referred to Users’ Guide to Medical Literature by Gordon Guyatt and Drummond Rennie (editors), AMA Publication ISBN 1-57947-191-9, and Guides to the Evaluation of Disease and Injury Causation by J. Mark Melhorn and William E. Ackerman (editors), AMA Press ISBN 978-1-57947-945-9.

Consideration of Evidence of Exposure

How does the evaluator take the general epidemiological data and apply this information to the specifics of the individual in question regarding causation? Occasionally, occupational data will be presented for each relevant job or duty. The following information would be helpful: the identification of risk factor(s) and data from industrial hygiene studies, especially any that indicates the magnitude of worker exposure. With regard to occupational disease, there is no generally accepted medical definition of aggravation. However, for WC, aggravation of a disease or impairment may be defined as any occupational occurrence, act, or exposure that permanently worsens, intensifies, or increases the severity of any preexisting physical or mental problem. The existence of a condition before exposure does not necessarily mean before employment. Furthermore, an individual may experience multiple exposures while working for different employers having different WC insurance carriers. An example may be helpful.

A 35-year-old man has worked as a chain saw logger for the past 15 years and complains of a 10-year history of numbness in both hands and digits. History, physical examination, and nerve conduction study by a physician reveal bilateral carpal tunnel syndrome. A judge decided his condition was compensable under the current state’s workers’ compensation system and asked for an apportionment of the medical condition. During his 15 years of employment, he has worked for three different companies. The last employer changed insurance carriers 1 month before he filed the workers’ compensation claim.

Appointment

WC boards in all jurisdictions are faced with an expanding challenge in the management of occupational disease claims. In some cases, there are multiple risk factors, with multifactorial causation, and the resultant needs to clarify the contribution of each risk factor to the condition in question in order to apportion liability and financial costs. The process of adjudicating WC claims depends on the applicable statute but may involve differentiation between occupational and nonoccupational causes of disease and injury. Although, in practice, this can be exceedingly difficult and, in some cases, impossible, establishing causation and apportionment are integral parts of the philosophy of workers’ compensation. Apportionment by cause is the estimation in an individual case of the relative contribution of several risk factors or potential causal exposures to the disease. In the tort system, the equivalent concept is apportionment of harm (meaning responsibility for causing harm). However, because WC is a no-fault insurance system, assignment of blame or responsibility is generally irrelevant.

Apportionment by cause must be performed on the individual case, which may vary from the population as a whole. However, often apportionment cannot be determined with certainty, and epidemiologic data may then be used to derive an estimate of the relative contribution of risk factors in an individual claim. The estimate for apportionment of causation derived should not to be confused with the apportionment of its social derivative, disability. The benefits of fair and accurate apportionment, when it can be done, are obvious. Adjudication may be simpler, quicker, cheaper, and fairer to injured workers, employers, and insurers. Workers might be encouraged to take responsibility for their own health. The financial resources conserved could be used to increase benefits and/or decrease premium costs. Fiscal exposure for health care, disability, and impairment would be more fairly divided among payers, such as provincial or private health-care plans, Social Security, or WC.

Although apportionment is an attractive option for adjudication in workers’ compensation, it has many drawbacks and uncertainties (Melhorn, Andersson, & Mandell, 2001). The single greatest obstacle to apportionment is the availability of data and limitations on the methodology of assessment of relative contribution to the disease. Therefore, apportionment is often consensus- or expert-derived. WC carriers are generally required to accept medical claims in their totality if a component of the disease is work-related. However, there is wide variation between jurisdictions regarding how big the occupational component must be before the condition is accepted as work-related. A minimal contribution from work, even one iota, is sufficient to render a disease compensable in some states, whereas others require that work have been the substantial factor, the major contributing cause, or a significant contributor. Furthermore, defining what constitutes a substantial, significant, or even minimal component is often difficult. Apportionment is more often applied to the permanent impairment rating, which is often used to determine a financial settlement in workers’ compensation claims.

A special case of apportionment is presumption, of which there are two types: A rebuttable presumption shifts the burden of proof to the part against which the presumption applies (Melhorn, Ackerman, Glass, & Deitz, 2008). In law, it is an assumption by a court that is considered true until a preponderance of evidence disproves (rebuts) the presumption. For example, in criminal law, a defendant is presumed innocent until proven guilty. On the other hand, an irrebuttable presumption establishes a legal conclusion which may be based not on scientific or other evidence but the desire for social justice and fair play (Melhorn, Ackerman, et al., 2008). Judges and legislatures have the power to substitute convenience for science. For example, an irrebuttable presumption may be made based on information (not necessarily facts) that because most cases of a disease in persons with a given job can be attributed to an occupational risk factor, any such case for which a claim is filed will be accepted as work-related. Many presumptions are written into law, often without good evidence, such as the presumption in California that heart disease among firefighters and police officers is work-related (Brooks & Melhorn, 2008). Others are “scheduled” or designated on lists. Presumption logically requires both strong evidence of an association and a risk that is at least doubled. A simple association can be accepted at a greater than 50 % level of certainty for the occupational group overall at whatever degree of association, but a simple association is not the same as presumption. A presumption involves the same degree of certainty but an actual proportion of the disease (attributable fraction) compatible with at least 50 % in the occupation or population overall. At such a high frequency, it is statistically likely that for any one individual drawn from that population (and submitting a claim) who presents with the disease or outcome in question, the occupational cause would be the risk factor. For example, firefighters have a much higher risk of kidney cancer than the general population, but their risk of lung cancer is only elevated by about 50 % (Guidotti, 2006). One can justify a presumption for kidney cancer, but not for lung cancer. Any firefighter with kidney cancer probably would not have been at risk if he or she were not in that occupation. Presumptions can also be legislated in the opposite fashion. For example, in 1996, the Virginia Supreme Court said that carpal tunnel syndrome does not make a person eligible for WC benefits even if a doctor insists the problem is work-related (Carrico, 1996).

The effects of increasing requests for apportionment are reflected in the definition used in the AMA Guides to the Evaluation of Permanent Impairment. The fourth edition, published in 1995, states apportionment is an estimate of the degree to which each of the various occupational or nonoccupational factors may have caused or contributed to a particular impairment. For each alleged factor, two criteria must be met: The alleged factor could have caused or contributed to the impairment, which is a medical determination, and in the case in question, the factor did cause or contribute to the impairment, which usually is a nonmedical determination. The physician’s analysis and explanation of causation is significant. The fifth edition, published in 2001, defines apportionment as the distribution or allocation of causation among multiple factors that caused or significantly contributed to the injury or disease and existing impairment. The sixth edition, published in 2008, states apportionment is the extent to which each of two or more probable causes are found responsible for an effect (injury, disease, impairment, etc.). Only probable causes (at least more probable than not) are included. Hence, the first step in apportionment is scientifically based causation analysis. Second, one must allocate responsibility among the probable causes and select apportionment percentages consistent with the medical literature and facts of the case in question. Arbitrary, merely opinion-based unscientific apportionment estimates which are nothing more than speculations must be avoided. When appropriate, the current impairment can also be apportioned to more than one cause.

The Changing Threshold for WC Compensability

Kansas House Bill No. 2134, An Act Concerning Workers’ Compensation, passed May 15, 2011, as Law 04-18-2011 (Kansas House, 2011) defines “Prevailing Factor Test” in which, under this test, the employee’s work must be the “prevailing factor” for the injury to be considered work-compensable. If it is not the “prevailing factor,” the injury is not compensable. It is believed that, under this standard, employers will have a defense against preexisting degenerative conditions not “caused” by work. In other words, employees will no longer have a compensable claim for an aggravation, acceleration, or intensification of a preexisting condition. Unfortunately, the legislature did not define “prevailing factor” and did not reference the Missouri law which could then have been used as “case law” to assist in the definition. Consensus opinion is that the need to define “prevailing factor” will increase ligation and the associated costs. The Missouri Workers’ Compensation Law was passed and is in Chapter 287 of the Revised Statutes of Missouri (www.moga.mo.gov/statutes/statutes.htm).The WC statute is the law that controls the rights and obligations of employees and employers when employees are injured at work. It outlines that the work injury must be the “prevailing factor” in causing the level of disability and medical condition for it to be compensable. Medical causation is needed in cases where a traumatic incident occurred or where an employee is injured by an occupational exposure or repetitive motion that is part of their employment. The prevailing factor is the primary factor in comparison to any other possible factor resulting in the employee’s injury. It should also be pointed out that Oklahoma SB878 (Oklahoma, 2011) passed May 2011 now requires “major cause provision” (which has yet to be defined by case law but will probably “contribute more than half”) and the requirement that objective medical evidence be determined by the Daubert criteria (Fed Rule of Evidence 702) for the opinion of “major cause.”

Consideration of Validity of Testimony

Nonprofessional persons cannot be expected to collect and evaluate all the information needed in causation analysis. In most cases, physicians will provide testimony on test results, medical conditions, and causation, using information from industrial hygienists regarding exposure and epidemiologists regarding epidemiologic data. These professionals must consider all pertinent facts and literature in their area of expertise to present an accurate and meaningful evaluation of the available data. The judicial entity requesting the information or determination may have additional expert witness criteria, such as the Daubert Standard, to be discussed next. The law of evidence governs whether testimony (e.g., oral or written statements, such as an affidavit), exhibits (e.g., physical objects), and other documentary material are admissible (i.e., allowed to be considered by the trier of fact, whether a judge or jury) in a judicial or administrative proceeding (e.g., a court of law) and how they are used (Wikipedia, 2011). In 1993, the US Supreme Court established the current standards for the admissibility of scientific evidence in Daubert v. Merrell Dow Pharmaceuticals, Inc. This decision abolished the old “general acceptance” test and set forth a new standard which focuses on the reliability and relevance of scientific testimony. In evaluating scientific testimony, trial courts will consider the following factors: whether the research was conducted prior to the litigation, whether it has been tested, whether it had been subjected to peer review and publication, whether there is a known or potential rate of error, whether the research was conducted according to fixed standards, and whether the technique is generally accepted in the scientific community (Klimek, 2001).

The Daubert Standard is a rule of evidence regarding the admissibility of expert witness testimony during US federal legal proceedings (Wikipedia, 2011). Pursuant to this standard, a party may raise a Daubert motion, which is a special case of motion in limine raised before or during trial to exclude the presentation of unqualified evidence to the jury. A motion in limine (Latin, “at the threshold”) is one made before the start of a trial requesting that the judge rule that certain evidence may, or may not, be introduced at trial. This is done in judge’s chambers or in open court with the jury absent. Usually, it is used to shield the jury from evidence that may be inadmissible and/or unfairly prejudicial. Nonexpert witnesses are permitted to testify only about facts they observed and not their opinions about these facts. An expert (professional) witness is one who, by virtue of education, training, skill, or experience, has knowledge and expertise in a scientific, technical, or other subject beyond that of a layperson, sufficient that others may rely upon his or her opinion on evidence and facts within the expert’s area of expertise, even though he or she was not present at the time of the injury, exposure, or other event. In law and religion, testimony is a solemn attestation as to the truth of a matter, and the expert opinion is intended to educate the judge and/or jury on a specialized subject matter, thereby assisting the trier of fact.

Consideration of Other Relevant Factors

Medical causation is drawn from science while, in law, causation is the connection between a wrongful act and harm (Melhorn, Ackerman, et al., 2008). Work-relatedness, in the context of occupational injury or illness, involves concepts of both medical and legal causation. The two may be mutually exclusive. Definitions of medical causation and legal causation arise from different sources—one from science and the other from desire for social justice (with origins in religion). This has been described by Melhorn as “the difference between why things are as they are, and how things ought to be” (Melhorn, 2008). For physicians treating injured or ill workers, understanding the differences between the two concepts is essential. Legal causation requires two components: cause in fact and proximate (or “legal”) cause (Melhorn, Ackerman, et al., 2008). Both must be present. If the occurrence of one event brings about another, the former can be considered the cause in fact of the latter. This is true regardless of the number of events involved. Causal fallacies exist, and at least one of them requires attention here. The post hoc ergo propter hoc fallacy occurs when “after this, therefore because of this” reasoning leads to the assertion of a causal relationship. It is a fallacy to conclude that the occurrence of one event followed by a second necessarily demonstrates a causal relationship between the two. The second part of the legal analysis of causation seeks to determine whether two events that are linked in fact should also be linked in law. This second test, proximate or legal cause, is whether the two events are so closely linked that liability should be attached or assigned to the first event that produced the harm, the second event. The most common legal threshold is “that the injury arises, in whole or in significant part, out of or in the course and scope of employment” (Melhorn, 2000b). The most common level of legal certainty needed to establish medical causation in the legal system is “more likely than not” or “more probable than not” (Melhorn, 2007).

Evaluation and Conclusions

Seven questions that can be used to address the question of causation based on the best available science are as follows:

• 1. Has a disease condition been clearly established?

• 2. Has it been shown that the disease can result from the suspected agent(s)?

• 3. Has exposure to the agent been demonstrated (by work history, sampling data, expert opinion)?

• 4. Has exposure to the agent been shown to be of sufficient degree and/or duration to result in the disease condition (by scientific literature, epidemiologic studies, special sampling, and replication of work conditions)?

• 5. Has nonoccupational exposure to the agent been ruled out as a causative factor (or a contributory factor—suggesting apportionment)?

• 6. Have all special circumstances been considered?

• 7. Has the burden of proof been met—did the evidence prove that the disease resulted from, or was aggravated by, conditions at work?

Summary and Conclusions

What is cause? An event, condition or characteristic that plays an essential role in producing an occurrence of a disease.

What is causation? The act of causing.

What is causality? The relationship of cause to effect.

What is risk? Risk is the probability that an event will occur if exposed to the risk factor.

What is a causal relationship? Inferring a causal relationship requires an understanding of epidemiology.

Epidemiology is a science and, as such, adopts the scientific standard of proof, generally greater than or equal to 95 % probability. However, civil litigation and adjudication hold to a different standard. How does one apply epidemiology when the standard is “more likely than not?” In general, this requires a relative risk odds ratio of greater than 2.0. Unfortunately, there is often insufficient data to establish a relative risk. Furthermore, conventional statistics for risk derived from epidemiologic research are generalized and may be difficult to apply to an individual case because the individual’s experience in the future may not be similar to the group studied. Health and medical knowledge are essential to the resolution of disputes in legal and administrative applications (such as WC), and it provides essential input into public policy decisions. There are no socially agreed-upon rules for the application of this knowledge except in the law. On a practical level, the legal system lacks the capacity to evaluate the validity of knowledge as evidence and therefore relies heavily on expert opinion.

The determination of causation may be extremely difficult in contested claims. Honest differences of opinion are common when the “facts” may be subject to different interpretations. The practitioner faced with the above questions should know the legal definitions that determine whether the condition is considered work-related. Although the condition may not meet the medical criteria of causation, the condition may meet the legal threshold and therefore be considered work-compensable. Considerable judgment is necessary when data are lacking or incomplete. It is important to assemble complete information in a logical and orderly sequence wherever possible to ensure a correct and an equitable decision regarding causation.

Future Directions

Epidemiology will play a major role in future research into the ever broadening field of health-care and public health concerns. This is evidenced by the exponential growth of research studies claiming to be “evidence-based.” This growth will not occur without challenges and opportunities, examples listed below.

The challenges include:

• 1. The growing threats to data access. Unfortunately, many recent attempts to place limits on the collection and storage of personal health data have completely ignored the potential impact of the proposed legislation on epidemiological medical research. This has been less of a problem for Europe which has “universal health care” and government data available for research. For perspective, remember, this limitation of health-care data access is occurring at a time when “social data” is commonly shared by individuals on the internet on multiple websites.

• 2. The challenge of communicating epidemiologic data to the public. Often, the science is not intuitive and therefore can be difficult to accept.

• 3. The intensifying interface between epidemiologic data and the legal and legislative system.

The opportunities include:

• 1. Scientific answers or insight to specific questions

• 2. The ability to convert global data to the individual and thereby reduce individual risk factors

• 3. The potential to maintain and improve the public and thereby reduce impairment and disability while improving the quality of life for the individual