Data are collected using a variety of source documents (See data source section for more information.) and entered manually by state partners once a fatal occupational injury has been identified. As more information is gathered, the data are revised until the case has been fully coded.
The system uses a matching scheme to keep duplicate entries for the same fatal injury from being entered. After data entry but before tabulation, a final manual check and final computer-assisted check ensure there are no duplicates. The CFOI program is a true census in that few, if any, cases are missing from the data. This thoroughness is due to the use of multiple source documents rather than relying on a single data collection tool. Outside researchers have reached a consensus confirming the completeness of the data.
To ensure an accurate count of fatal occupational injuries, the CFOI requires that, for each case, the work relationship (that is, whether a fatality is work-related) be substantiated by two or more independent source documents or a source document and a followup questionnaire. Followup questionnaires are sent either to the employer or to another contact with knowledge of the incident. The followup questionnaire is also used to collect information that may be missing from the source documents. At the end of the collection period, fatal injuries for which the state has only one source document are reviewed by BLS. The case is included in the national database only if the state and BLS agree that there is sufficient information on the sole source document to determine that it is indeed work-related.
Data validation
There are many levels of data validation to verify the completeness and correctness of the coding of a case itself. Within the system, there are edits that might indicate missing information or a mismatch in codes according to various coding rules and structures. (See concepts section for more information.)
The CFOI program uses a three-tiered system of manual reviews of both the case and source documents to ensure correct information is being transcribed and coded, which is initiated by the regional BLS offices. Every case is subsequently also reviewed by the national BLS staff to ensure consistency throughout the nation and accuracy of the coding. During the national office review, cases with unusual circumstances may be further vetted by senior staff to ensure those cases meet the scope criteria for the CFOI, according to the definitions found on the CFOI definitions page.
Because the CFOI uses multiple documents, key variables are rarely missing or coded as unknown. The timetable for collecting and coding the CFOI starts at the beginning of the calendar year and runs until late fall of the following year (almost 2 full years for a January death and about 10 months for a December death) to help ensure that the CFOI data are complete and accurate. Ninety-nine to 100 percent of the data are known and coded for selected key data elements that are required to be entered, which excludes the general unknown code.
Data compilation
Data are collected (See data sources section.) and validated. Once data validation checks have been completed, data are compiled by simple aggregation. For a more complete listing of how the data are presented, see the presentation section.
Fatal injury rate methodology and calculation
Fatal injury rates depict the risk of incurring a fatal occupational injury faced by all workers or a subgroup of workers and are used to compare risk over time or with other worker groups. Workers can be grouped for comparison by a number of variables, including industry, worker’s age, or gender. Because employment data are not collected by the CFOI, fatal injury rates are calculated using estimates of hours worked from the Current Population Survey (CPS).[1]
In 2008, the CFOI adopted hours-based employment as the denominator of fatal injury rates to measure fatal injury risk (rates were calculated retroactively using this new method for 2006 and 2007 data as well). This methodology is generally considered to be more accurate than employment-based rates per the standardized length of exposure to risk of occupational injury. It is defined as the average number of persons at work over the year multiplied by the average hours each employee works over the year. More information on the change from employment to hours-based fatal injury rates is available online. The article “Change to Hours-Based Fatality Rates in the Census of Fatal Occupational Injuries” details the impact of the methodology change.
National rates
To accurately describe fatal injury risk for a worker group, we must refer the numerator (fatal injuries) and denominator (total hours worked) of the rate to the same group of workers. The hours-worked data from the CPS used in the rate calculations do not include workers under the age of 16, volunteers, and members of the resident military. Therefore, fatal injuries occurring to these workers are also excluded from the numerator. Industry data for national rates are broken out by ownership: private, overall government, federal government, state government, and local government.
National fatal injury rates use data from the CPS. As opposed to the employment number, data on people at work exclude those who were temporarily absent from a job (classified in the zero-hours-worked category, with a job but not at work). Those not at work were absent from their jobs for the entire week for such reasons as bad weather, vacation, illness, or involvement in a labor dispute. Two estimates, at work and average hours, are combined to create the denominator, annual total hours worked:
where:
• EH = total hours worked by all employees in a group during the calendar year
• AW = at work (number of employees working in a group)
• H = average hours (average annual hours worked by an employee in that group)
The hours-based rate (expressed per 100,000 workers) is:
where:
• N = number of fatal injuries in a group
• EH = total hours worked by all employees in a group during the calendar year
• 200,000,000 = base for 100,000 equivalent full-time workers (working 40 hours per week, 50 weeks per year)
State rates
Unlike at the national level, at work and average hours data are not available at the state level. State rates by industry can be imputed by using national-level average hours and at work information to calculate the total annual number of hours for each worker industry group, regardless of ownership. State rates are not available by demographic, occupational groups, or for public and private sectors.
The rate represents the number of fatal occupational injuries per 100,000 full-time equivalent workers and was calculated as:
where:
• NS = number of fatal work injuries in the state
• EHS = total hours worked by all employees in the state during the calendar year
• 200,000,000 = base for 100,000 equivalent full-time workers (working 40 hours per week, 50 weeks per year)
State rates by industry were imputed by using national-level average hours and at work information from CPS to calculate the average annual number of hours for each employee, because these data are not available at the state level. EHS (total hours worked by all employees in the state during the calendar year) was calculated as:
where:
• ES = State employment
• HWN = average annual number of hours for each employee at the national level
• The N subscript denotes a national number; an S subscript denotes a state number.
Table 1 presents data to be used in an example:
Table 1
| Industry |
A
Employment
(in thousands)
EN
|
B
At work
(Employment in thousands)
AWN
|
C
Average
annual hours
(Weekly hours x 50)
HN
|
D
Total hours
(Annual, in thousands)
EHN=AWN*HN
|
E
Adjustment
(Annual total hours,
per employee)
HWN = EHN / EN
|
| All industry |
145,362 |
139,824 |
1,945 |
271,957,680 |
1,871 |
| Construction |
10,974 |
10,558 |
1,980 |
20,904,840 |
1,904 |
For the national rates, compute:
To get EH, the total hours for the year, multiply at work and average hours. This first step involves columns B, C, and D in table 1. So to get the national denominator:
Then take the total number of fatal injuries in 2008, 5,084 (5,214 minus the number of workers under the age of 16, volunteers, and members of the resident military), and divide by EHN and multiply by 200,000,000:
This would be read as 3.7 fatal occupational injuries per 100,000 full-time equivalent workers.
For the state rates, the information for columns B, C, and D in table 1 is not available, so to impute the state average hours, divide the national total hours (column D) by the national employment (column A) to get the average annual number of hours per employee (
, column E, is the imputed number).
The last step, to get the denominator for the state rate, is to multiply the average annual number of hours per employee HWN by the state employment, using data from California, presented in table 2, as an example:
Table 2
| State |
Employment
(in thousands)
ES
|
Fatalities
NS
|
| California total |
17,045 |
448 |
| California construction |
1,294 |
67 |
To get the state rate, the calculation is as follows:
Using our values from table 1 and table 2, the calculation is as follows:
This would be read as 2.8 fatal occupational injuries per 100,000 full-time equivalent workers.
CFOI rates prior to 2008
CFOI fatal injury rates published by BLS for the years 1992 through 2007 were employment-based rates and measured the risk of fatal injury for those employed during a given period, regardless of hours worked. The following is the formula for calculating a fatality rate from 1992 through 2007:
where:
• N = the number of fatally injured workers, 16 years and older
• W = the number of employed workers, 16 years and older.
For example, in computing the 2005 national fatality rate from 2005 CFOI:
And, from CPS, 2005 annual averages, plus resident military figures derived from the U.S. Department of Defense:
So the fatality rate is calculated as:
Comparisons of national and state rates
CFOI uses the Local Area Unemployment Statistics (LAUS) data to calculate state rates, because the CPS does not produce employment estimates by state.
State industry rates are not directly comparable to national industry rates. Due to the difference in the way LAUS and CPS calculate their employment data, state rates include government workers in their respective industry sector and are not broken out separately, both the numerator and denominator include a different group of workers than that of the national rates.
If users decide to add up all the states in one industry and average out their rates to compare it to the national average, they will not get the national average due to this data difference, with the all-ownership/state rates most likely being slightly higher because of the added fatal injuries and different employment data. For more information on fatal injury rates and how they can be used, see our Frequently Asked Question on Rates.
Comparisons of state rates to other state rates
State industry rates are not comparable to other states because of the large differences in the industry composition of employment by state. For example, comparing rates for a state with a large agricultural economy with that of a state with a large industrial economy would be ill-advised because agriculture has one of the highest fatal injury rates while manufacturing has one of the lowest. In addition, the number of fatalities and their circumstances can vary markedly within a state from one year to the next, in part reflecting single incidents involving multiple deaths, such as airplane crashes and natural disasters.
CPS data limitations
There are several limitations of using CPS data in CFOI rate calculations:
• Employment sampling errors: The CPS data used to calculate rates are estimates based on a sample rather than a complete count. Therefore, the CPS estimates and fatality rates have sampling errors. The measurement error in the fatality rates is a result of the sampling error of the CPS data used as the denominators. The rates calculated using the CPS may differ from those that would have been obtained from a census of employed persons. See Explanatory Notes and Estimates of Error in the February 2004 Employment and Earnings for an explanation of CPS sampling and estimation methodology, and standard error computations. The relative standard errors of the CPS estimates can be used to approximate confidence ranges for the fatality rates.
• Primary job versus job at the time of incident: The CPS annual average employment data used in the rate calculations count workers according to their primary job, whereas CFOI uses the job held when fatally injured, which may differ.
• State of residence versus state of incident: CPS counts workers by their state of residence, whereas CFOI counts workers by state of incident.
• The annual average of hours worked represent total hours at work for CPS respondents, including those that work more than one job. Total hours worked for respondents with multiple jobs will be recorded in the occupation and industry of the primary job.
• Rates are calculated at the level of detail available from the CPS data. In addition, because the methodology change in 2008, rates are only calculated for those occupations and industries that met minimum thresholds of having at least 15 fatal injuries and 40 million hours worked for national data. The threshold for calculating state fatal injury rates is 5 fatal injuries. See the following section on using fatality rates to evaluate risk for more discussion on the importance of thresholds.
Using fatality rates to evaluate risk and dangerous jobs
Fatal injury rates depict the risk of incurring a fatal work injury for workers in a given worker group expressed as the proportion of fatal injuries per total hours worked annually per 100,000 full-time equivalent workers. This allows risks to be compared among different worker groups. Occupations with few fatal injuries and low employment in the reference year are removed from annual fatal injury rate analyses. Therefore rates produced in prior years may not appear in future years if they do not meet this threshold.
The article “Dangerous Jobs” illustrates the difficulty and impracticality of measuring fatal injury rates for occupations with few workers. The article explains why numbers of deaths, fatal injury rates, and other factors should be considered together when analyzing the danger of particular jobs and uses the occupation elephant trainer in a hypothetical case to illustrate the point.
Example: Is elephant trainer a dangerous job?
• On the one hand, because few workers are employed as elephant trainers, a small number of fatal injuries to elephant trainers would make the fatal injury rate extremely high for a single year, despite their low number of deaths. On the other hand, in most years, this occupation incurs no deaths, rendering their fatality rate 0 and ranking them among the least at risk for incurring a fatal injury.
• “Elephant trainer” is a hypothetical occupational classification. The classification BLS uses groups these workers with either “artists and performers” or “animal caretakers,” both of which include many more people than just elephant handlers.
There are many other elements that factor into any definition of a “dangerous job” such as the likelihood of incurring a nonfatal injury, the potential severity of that nonfatal injury, the safety precautions necessary to perform the job, and the physical and mental rigors the job entails.
Because there is no universal definition of “dangerous” or “hazardous,” the Injuries, Illnesses, and Fatalities (IIF) program does not frame occupations as the “most dangerous” in a particular year. The IIF program also has certain minimum thresholds that must be met for a fatal injury rate to be published. As such, fatal injury rates are not calculated for many occupations that have a relatively small number of fatal work injuries and employment. Please see “Using fatality rates to evaluate risk and dangerous jobs” for more information.
United States Department of Labor
