APPENDIX ! MATERIAL SUBMITTED FOR THE HEARING TESTIMONY BY JOHN MENDELOFF FOR THE HOUSE GOVERNMENT OPERATIONS SUBCOMMITTEE ON MANPOWER I am pleased to be asked to testify before the Subcommittee about occupational safety and health information systems. Much of my work has focused on how better information could allow OSHA to make more intelligent regulatory choices. I have conducted evaluations of OSHA's effectiveness in preventing injuries, worked on redesigning its inspection program to improve its costeffectiveness, and recently have been examining the potential uses of the data on OSHA's health inspections, which reside in the agency's computerized Management Information System (MIS). I have published Regulating Safety: An Economic and Political Analysis of Occupational Safety and Health Policy and several articles about the topics described above. I have been supported in this research by OSHA, NIOSH, the Department of Labor, the Office of Technology Assessment, and the Federal Interagency Task Force on Workplace Safety and Health. I now teach at the University of California at San Diego. I will divide my testimony into sections on safety and health. My focus will be on ideas that could provide more useful information. Safety The Bureau of Labor Statistics survey of occupational injuries and illnesses was established to identify differences in injury rates among industries and over time. For lost workday injuries, the survey succeeds reasonably well, I believe. It correlates roughly with the more reliable state workers' compensation data sets. The rates for non lost workday injuries (and thus the total recordable rate) are much less reliable because of employer uncertainty about where the line between recordable cases and "first aid" cases is to be drawn. (213) Unfortunately, the data from the survey have very limited value for policymaking purposes. They provide almost no insight into causal factors in accidents, in particular into the role of OSHA standards and OSHA enforcement. As a partial remedy, BLS has been developing its "supplementary data system" (SDS), essentially a compilation of somewhat upgraded and standardized data sets from state workers' compensation programs. These do include information about the type of accident as well as some characteristics about the accident victims and will prove useful for accident researchers. Policymakers are often confronted by "information systems" that generate massive amounts of data, but very little useful information. In part, the problem is that no one has thought through what questions were important to answer and what data might help to answer them. This is exacerbated because the information policymakers need is extremely specific and detailed. Let me give several examples. When OSHA is considering the promulgation of new safety standards, it needs to know very precisely what the likely effects of the required changes would be. Again and again, hearings on these standards reveal that the needed information is not available. The BLS survey is no help. State workers' compensation data can provide some insight into the number of times a certain type of injury has occurred--e. g., how many times forklift trucks have overturned and killed the driver. However, OSHA needs to know whether all forklift trucks overturn or only smaller ones; at what speeds they overturned; and other similar facts. strategy would be for OSHA to co-operate with one or two of the larger states which have good workers' compensation data systems and piggyback an interdisciplinary team on that system. For example, suppose that OSHA knew the four or five most likely standards that it would be addressing over the next few years and some of the key factual issues that they would raise. This team of engineers and biostatisticians could be reviewing the relevant accident reports, One and following up--either by phone or by site visits--to gather the detailed information that would be needed to resolve those issues. No state has the financial incentive to conduct the proper level of research on its own. There are two more general insights raised by this example. The first is that improving existing data sources will often be more productive than setting up entirely new ones. In its standard on punch presses, OSHA tried to require reports of all punch press accidents to federal OSHA. Employers didn't comply and the system collapsed, even while routine reporting of such accident to state WC systems continued. The second point is that federal agencies, including OSHA, often have a view that the only data that is worth analyzing is national data and they thus fail to invest in strategies that enrich the analysis of state data. The effectiveness of OSHA's safety program has been the subject of many studies, whose conclusions range from those that found small effects to those that found none. For policymakers, even the former provided few insights because they were designed to give "summative" evalutions--i.e., was there an effect?--rather than "formative" evaluations--which would provide insights about which programs worked better in different circumstances. Better designed studies may generate more insights, especially if they can tie together the particular violations that are cited in inspections and the specific types of injuries that occur before and after inspections. This will require linking together, in at least one state, the inspection data in the MIS with the accident data in the WC reporting system. However, an even more important change is needed: a willingness on OSHA's part to conduct true enforcement experiments in which the evaluation design is considered before the program is implemented. The few experimental programs that have been attempted were not carried out in ways that would facilitate evaluation: the number of sites were too few and the length of time too short to draw clear conclusions. Sadly, as a result we have learned remarkably little about which programs will be most effective in detecting and deterring violations, much less in preventing injuries. One difficulty is that OSHA has very little information about the relation between violations of standards and accidents. One of the only sources for this information comes from accident investigations, which include a review of what role violations may have played. Currently, OSHA only conducts them in the case of fatalities and accidents causing more than 4 hospitalizations. Even with these, it has conducted very little analysis. hospitalizations in its investigations. California has covered many more In a recent article (Journal of Occupational Medicine, May 1984) I have shown that analysis of that data illuminates issues such as the following: 1) which violations should be cited as "serious"; 2) in which industries and size classes of plants are violationrelated serious injuries occurring; 3) which violations can be detected by inspectors; and 4) for which types of serious accidents are current standards not relevant. This type of information can be used to inform standard-setting and enforcement practices, and can be provided to employers and workers to help them focus on the most serious violations in their industries. Health The BLS survey is generally conceded to have little value as an indicator of the magnitude of the occupational illness problem. Its central weakness comes with diseases with long latency periods--e.g. cancer and obstructive lung diseases--but even with more acute problems--e.g. lead poisoning--few believe that it captures more than a fraction of the cases. Especially for the long latency diseases, what is needed is a measure of current exposures, Ideally, we would be able to track whether exposures were decreasing, in what types of plants they were highest, and whether they declined after inspections. In later |