December 2013

Using OSHA inspection data to analyze respirator protection program compliance


The survey does have some shortcomings regarding surveillance. The most detailed industry level at which it presents findings is the two-digit SIC; more-fine-grained industry categories would be useful. Respondents were not asked to specify the actual levels of exposures or whether the facility had been cited by OSHA for RP violations.

In addition, we don’t know whether respirator use was occasional or continuous. Finally, we do not know how reliable the answers provided by respondents were or whether some establishments that should have been using respiratory protection were not.

In this report, we use only the survey’s estimates of what percentage of establishments in each industry required respirator use during the prior 12 months. We compare this figure with the percentage of establishments in each industry that had been cited by OSHA for RP violations from 1999 to 2006. Although these measures are different, the comparison does help us see whether the OSHA data provide a similar picture of industries using respiratory protection.

Measuring rates of noncompliance. One issue was what measure to use to assess the prevalence of noncompliance. To compare our data with the RUPS data, we used a measure of the number of establishments where violations were cited. (During the period we looked at, some establishments received multiple inspections.) One measure of noncompliance prevalence could be the number of establishments with RP violations divided by all establishments in the industry (in the 47 states in our sample). That would be a valid estimate assuming the establishments actually cited for noncompliance were the only ones that were noncompliant; however, that assumption is untenable because only a minority of establishments were inspected and some of those that were not inspected had violations.

A second potential measure would be the percentage of inspected establishments that were cited for RP violations during this period. This second measure would be a valid one if inspected and uninspected manufacturing establishments in each industry were equally likely to be cited for RP violations. This assumption is reasonable, although it ignores the fact that, while most inspections are “safety inspections” rather than “health inspections,” over 90 percent of the RP violations were cited in health inspections. Apparently, in safety inspections the inspectors are generally not looking for respirator program violations.

Following through on that logic, a third measure could be RP violations only in health inspections.8 This third measure would overestimate the prevalence of noncompliance as some violations were identified in safety inspections. However, it seems likely to be a better measure than the others.

Two other factors further complicate the estimates. First, it is likely that there will be RP violations that are missed even in the health inspections. Second, because of entrances and exits, the total number of establishments in existence during the 8-year period we are examining exceeds the number that existed in 1997, the year we employ for the denominator, perhaps by as much as 50 percent.9 These two factors work in opposite directions. The first reduces the establishments-with-violations number in the numerator. The second reduces the establishments number in the denominator. We will use measure 3 in our discussions, but the reader should keep in mind the uncertainties.10 Also, to the extent that we are concerned with the relative position of the industries, the choice between measure 2 and measure 3 poses fewer problems; the correlation between them for the 18 industries we examine is +0.93.


8 Here and elsewhere in the report, the figures for the percentage of establishments with health inspections that had violations refers only to violations cited in the health inspections.

9 Commerce Department figures indicate that about 10 percent of all establishments enter in a given year and about 10 percent exit. New establishments are less likely to survive, so an estimate of a 50-percent increase over an 8-year period seems plausible, although probably high. See table 739 in Statistical Abstract of the United States, 2009.

10 The uncertainties include whether workplaces generating complaints about health hazards are more likely to have RP violations than are those without complaints. And while OSHA does try to find workplaces with health problems, the targeting system for programmed health inspections during this period allowed a great deal of area-office discretion.

prev page3next page

View full article
About the Author

John Mendeloff

John Mendeloff is a professor at the University of Pittsburgh and the director of the RAND Corporation Center for Health and Safety in the Workplace.

Maryann D’Alessandro

Maryann D'Alessandro is the director of the National Personal Protective Technology Laboratory, Centers for Disease Control and Prevention.

Hangsheng Liu

Hangsheng Liu is a policy researcher at the RAND Corporation.

Elizabeth Steiner

Elizabeth Steiner is a research assistant at the RAND Corporation.

Jessica Kopsic

Jessica Kopsic is a technical research analyst at MDRC.

Rachel Burns

Rachel M. Burns is a project associate at the RAND Corporation.