If you want to keep all your digits and limbs, you probably want to avoid working at Anheuser-Busch’s Metal Container Corp., in Arnold, Missouri. That worksite was recently cited by OSHA for hazards related to incidents last fall in which one worker lost fingers in machinery, and another worker had a foot amputated because of a forklift incident. The 13 serious and one willful violation come with a proposed $107,200 penalty. Those two work-related amputations are just two of the estimated 3.9 million cases of injuries and illness that occur each year in U.S. workplaces. About a third of those cases are serious enough that the worker has to miss at least one day of work.

These figures come, for the most part, not from an actual count of injury cases, but from a survey conducted by the US Department of Labor’s (DOL) Bureau of Labor Statistics (BLS) of about 200,000 workplaces. The annual Survey of Occupational Injuries and Illnesses (SOII) is the most often cited source of occupational injury and illness data for U.S. workers. The SOII depends primarily on input from two DOL agencies: an Occupational Safety and Health Administration (OSHA) regulation on injury recordkeeping, and a survey of those records conducted by BLS. SOII also relies on data directly from the Mine Safety and Health Administration (MSHA) for injury cases among mine workers and from the Federal Railroad Administration for cases among rail workers. Employers in these industries are not subject to OSHA regulations, but follow injury reporting requirements mandated by these other safety agencies.

The SOII relies largely on an OSHA regulation (29 CFR 1904) which requires some employers to keep records of work-related injuries and illnesses suffered by their employees. These employers use one form to make an initial record of the injury or illness (OSHA Form 301 or a comparable incident report), another to keep a running tally of incidents (OSHA Form 300 Log), and an annual summary of work-related injuries and illnesses.

OSHA’s recordkeeping requirements, as currently written (and if employers complied fully with them) should capture all work-related illnesses or injuries that involve lost worktime, medical treatment (other than first-aid), restriction of work, loss of consciousness, or transfer to another job. Under these parameters, recordable cases cover a broad range of events—from extremely serious events that require surgery, hospitalization or other intensive medical care—to less severe cuts or sprains that may require a few work shifts of lighter-duty tasks. A “recordable case” is, by intent, broadly defined in order to capture the fullest scope of harm-inducing events related to work.

There are about 9 million workplaces in the U.S., employing 130.3 million workers. More than 85%, or 111.5 million, are employed in private industry. OSHA’s injury/illness recordkeeping rule applies to about 1.36 million worksites, or 18 percent of all U.S. private sector workplaces. OSHA does not require employers with 10 or fewer employees, or those involved in many retail, financial, and service industries to follow these recordkeeping requirements. Businesses that are exempt from this OSHA rule include car dealers, gas stations, restaurants and bars, retail stores, hair salons and barber shops, funeral parlors, furniture repair shops, child care centers, schools and colleges, medical and dental labs and offices, retail bakeries, meat and fish markets, and hardware stores. A vast majority, 82 percent of private sector workplaces, are not required by OSHA to keep injury/illness records. Some may do so because their insurer recommends it, or to assess the effectiveness of their management systems, or to continuously improve their operation.

You can consider step one of this national surveillance system the OSHA regulation outlining how to record injuries and illnesses. Step two is an annual survey by BLS of these injury records to estimate the incidence of work-related injuries and illnesses for the nation as a whole. BLS uses a probability sample of about 200,000 worksites (i.e., 2 percent of U.S. total) to create a representative mix of industries and workplace sizes. Some employers in the BLS sample would not normally keep the OSHA injury records because they are in industries considered “low hazard” and therefore exempt from the OSHA regulation. These employers are asked by BLS to keep a record of injuries for one year following the OSHA rules, for the sole purpose of the survey. The SOII statistics are wholly dependent on the accuracy and completeness of the data provided by the selected employers.

Policy makers, the press and others use statistics from BLS’ SOII to characterize the state-of-workplace-safety in the U.S. Occupational injury rates have decline substantially over the last two decades. The rate of recordable injuries was about 9 per 100 full-time workers in 1980 and about 8 in 1995. It declined rapidly after that, from 5.0 per 100 full-time employees in 2003, to 3.6 in 2009. For more serious injuries, such as those requiring at least one day away from work, the rate dropped from about 4 per 100 full-time employees in 1980, to 1.5 in 2003, to 1.1 per 100 full-time employees in 2009.

During a September 17, 2008 congressional hearing, one Labor Department official used the SOII data to illustrate the effectiveness of the GW Bush Administration’s worker safety policies. Assistant secretary of labor for policy Leon Sequeira said:

“…not only are injury and illness rates declining, they are at the lowest level in recorded history, under this Administration.”

The Chamber of Commerce and National Association of Manufacturer’s “Coalition for Workplace Safety” use the SOII data to argue that cooperative programs, not enforcement, are more effective in reducing worker injury rates.

But others question the veracity of the SOII data. They point to both structural limitations and empirical evidence that the BLS survey substantially understates the true magnitude of work-related injury and illness. Even the current Labor Secretary, Hilda Solis, showed reticence when announcing the BLS numbers last year:

“OSHA is aggressively working to ensure the completeness and accuracy of injury data compiled by the nation’s employers.”

The Secretary knows that when OSHA looks carefully at employers’ records, it finds some workplaces have suspicious or even gross errors on their injury logs. Here are just a couple of recent examples:

• OSHA issued 83 will violations against Houston-based Goodman Manufacturing Co. in September 2010 for blatant injury recording discrepancies. As many as 72 percent of the entries on their injury logs from January 2008 to March 15, 2010 were incorrect. The company contested OSHA’s citations and the proposed $1.2 million penalty, saying they were just clerical errors. The case is still open.

• OSHA found 38 violations of injury recordkeeping requirements at Louisiana-based Superior Energy Services and five subsidiary companies. The February 2011 citations came with a proposed penalty of $337,500. The company is contesting the OSHA citations and the case is still open.

• OSHA cited Lowe’s Home Centers’ regional distribution center in Rockford, IL for four willful violations and proposed a $182,000 penalty for failing to comply with injury recordkeeping requirement. Lowe’s contested the citations and the case is still open.

The logs examined by OSHA inspectors with these serious errors, are the same logs collected by BLS for its annual SOII. Not every employer’s logs are inaccurate, but it’s fair to suspect that if some portion of employers have incomplete or inaccurate injury logs, the representative sample collected for the SOII will also contain errors. But what is the magnitude of the under- or over-reporting? A few public health researchers have tried to answer that question.

Leigh and colleagues (2004) used data from eight different studies to calculate a range and median percentage of cases missed by the SOII. They estimated that more than half of the cases, 54.9 percent, that should have been recorded on the employers’ log and subsequent submission to the SOII were not reported.

Rosenman and colleagues (2006) studied injury and illness reporting of particularly serious cases–those in which the worker had to miss 7 or more days of work. They examined cases from the State of Michigan and used data from five different data sources. They applied capture-recapture techniques to match injury cases from 1999 through 2001 and found that only one-third of all work-related injuries and illnesses were captured by the SOII.

Boden and Ozonoff (2008) used a similar methodology to analyze data from six States, assembling cases from 1998-2001 in which a worker missed 3 or more days of work due to their injury. The authors reported that one-quarter to one-half of the injury cases were not captured by the Labor Department’s SOII. Others report less disparity between workers compensation records and SOII cases among more serious injuries, such as those requiring the worker to miss enough days of work to be eligible for wage replacement benefits.

These studies and others suggest that when researchers apply cross-verification methods to the Labor Department’s SOII data, such as with workers’ compensation records, state-based physician or employer reporting systems, deficiencies in SOII are identified. The assistant commissioner for BLS, John Ruser, PhD, who is responsible for the SOII program, described in 2008 ways in which they were addressing structural deficiencies in the survey. He did not agree however with assertions of a serious undercount by SOII of injuries. One major improvement came with the 2008 survey, when BLS included injury data for workers employed by state and local governments (e.g., police officers, school nurses and teachers, bus drivers.) Limitations to completeness of the SOII remain, including complications related to occupational illness reporting, and exemptions from the survey for agricultural employers with 10 or fewer employees, independent contractors and other self-employed individuals.

In installment one of this series, I examined the purpose and authority for an occupational injury and illness surveillance system; and in part two, I described the annual census of work-related fatal njuries and discussed the challenges related to occupational illness surveillance. This third installment described the annual SOII, and in the final piece, I’ll examine the system for ensuring the accuracy of employer injury reports, and ask whether we should rely solely on a survey of employers to provide the national statistics for work-related injuries and illnesses.

Read Part 1 and Part 2 of this series