The difficulties of diagnosis have made it clear that serious illnesses, such as those related to carcinogens, are not adequately identified and, therefore, are underreported in our survey and in other data. The present state of medical science and the difficult problems of identifying causality suggest great difficulties in moving ahead in this area.

Now, Mr. Chairman, I have included in my statement a series of answers to specific questions that you raised in your letter to me and I would be glad to comment on those later, should you wish it.

I think it is important to recognize that over the past dozen years, we have accomplished a great deal. I might also point out, Mr. Chairman, that the Bureau of Labor Statistics is celebrating its 100th birthday next Wednesday, and I am pleased to report to you that in the very early days of the Bureau's existence, it was interested in occupational injuries and health and did do some survey work, even that early.

But we have accomplished a great deal, especially over the past dozen years. For the first time, a coherent, scientific, nationwide statistical system for occupational injuries and illnesses has been put into place. It provides essential information for the public and for policymakers, for employers and workers.

The data have proved useful for compliance programs and employers have been provided with guides that enable them to compare their performance with that of other establishments. The employer guides were developed and printed in response to industry requests for more feedback from survey results.

I would also like to point out that the Bureau has made two program innovations that provide additional perspective on the results of the annual survey: One, the Supplementary Data System, draws on workers' compensation information to develop a data base on the kinds of injuries that are suffered and where they occur. A research effort is now under way in the Bureau to combine those data and the annual survey data to yield national estimates on the numbers of injury cases by characteristics such as types of accidents and nature of injuries.

Another innovation is a series of work injury reports analyzing specific types of injuries and accidents. These analyses come from information secured from employees who have been identified from workers' compensation reports. Over the years, in brief, we at the Bureau of Labor Statistics have conducted a vigorous and, I believe, innovative program and we welcome any suggestions for further improvement.

[The prepared statement of Ms. Norwood follows:]

Statement of Janet L. Norwood
Commissioner of Bureau of Labor Statistics

Before the

Manpower and Housing Subcommittee
of the Committee on Government Operations
House of Representatives
Congress of the United States

June 20, 1984

Mr. Chairman and Members of the Subcommittee:

I am pleased to have the opportunity to testify on the annual survey of occupational injuries and illnesses conducted by the Bureau of Labor Statistics for the past 12 years. Your special concern this morning is with the quality

of the data obtained in this survey, and this is a concern that is always uppermost in our minds as well. The reputation of the BLS has been established on the quality of the work that it produces in all areas of its activities. Before turning to some of the specific questions raised in your letter, let me provide some background about the nature of the survey and of the quality issue.

THE MANAGEMENT OF QUALITY

The Bureau has long been concerned with the maintenance of quality in all of its operations, including issues involved in survey design, data collection, processing and error structure. Over the last few years, we have developed a new, compre

hensive approach to quality management in the Producer Price Index program which can serve as a prototype for other programs in the BLS. In fact, we have already begun this year to develop a quality management program for the Consumer Price program as a part of the current CPI Revision activities. These programs are intended to follow all parts of the index programs-from definition of concept, through the survey process itself, to the publication of results. We believe that quality control of the entire management of a survey constitutes an important step not only toward the goal of developing statistical estimates of sampling and non-sampling error, but also toward the goal of more efficient allocation of program resources.

In the case of Occupational Safety and Health (OSH) data, quality management consists essentially of two major tasks. The first task is the full responsibility of the BLS. It begins with the design of the most appropriate conceptual and statistical methodology for the development of data and the steps required to see that the survey design is fully and accurately carried out. It includes a system of regular review processes, as well as special survey measurement and analysis, to ensure that the data are accurately and properly collected. This work also frequently involves research on concepts and clarificaion of definitions to ensure that the survey properly represents the phenomenon to be measured.

The second task is the assessment of such non-statistical matters as the maintenance of the recordkeeping system. The quality of that system is important because the safety and health records kept by establishments are the source of the survey data gathered by BLS. However, the role of the BLS in this second task is more limited than it is in the first. As the statistical arm of the Department of Labor, BLS assists the Department's Occupational Safety and Health Administration (OSHA) by clarifying definitions, responding to inquiries about what is to be recorded, and helping as best we can to bring about consistency in the application of the recordkeeping rules. BLS, however, is a statistical agency; the Bureau does not engage in regulatory or enforcement activities.

and enforcement are the responsibility of OSHA.

Regulation

The first task--quality management of the survey design and implementation process--is the one that BLS customarily carries out. It draws on the special expertise and experience of the Bureau. We believe that we have developed a very good statistical design and survey implementation program for the OSH survey, including a probability sample of 280,000 units, with a 95 percent response rate, a detailed program for screening and editing all reported data (with callbacks for schedules failing an edit), an ongoing review and monitoring of State Grant Agencies' procedures, and the publication of sampling

errors (variances) for survey estimates (less than 0.5 percent in 1982). We use tne current state of the art in these activities and are extremely proud of these aspects of the OSH survey. Of course, this does not mean that our procedures are perfect or that they should not be periodically reviewed. Even though the overall survey meets today's high statistical standards, we believe that from time to time special investigations should be undertaken to check on the effectiveness of ongoing operations.

The second task is a much more difficult one.

The group

:

of government and outside experts who advised the Department of Labor more than a decade ago on establishing a recordkeeping system agreed that the most feasible source of occupational safety and health data is the employer. BLS must rely on each company to record injuries and illnesses and to classify them properly rather than on its own trained field representatives, such as those who gather consumer price data. As a result, the problem of maintaining accuracy and consistency is greater. This dependence is not unique to the OSH survey.

1

Payroll and earnings data have been gathered successfully from employers for decades. But the recording of illnesses and injuries is not an ordinary part of business accounting; the recording of occupational illnesses and injuries frequently relies on judgments by persons who often have no special training for