disease. Neurotoxic disorders are particularly difficult to monitor (NIOSH, 1986; Friedlander and Hearne, 1980), and present surveillance systems are not very effective. Neurotoxic disorders are easy to overlook or misdiagnose, because their signs and symptoms often develop slowly and subtly. And some can be reversible or self-limiting. The patterns of effects produced by many neurotoxic chemicals are similar, consisting of common, nonspecific complaints, often mimicking diseases with other etiologies. The association between an observed health effect and the culpable environmental agent is rarely obvious. Clinical tests used in diagnosis are usually too imprecise or insensitive to identify the full spectrum of adverse health effects. A latent period between exposure and overt response will further complicate determination of causation.

Several existing data systems have been suggested as potentially useful for surveillance for neurotoxic disease that might be attributable to environmental exposures (Gable, 1990). They include vital records, health surveys, specifically designed surveillance systems, and the U.S. Census.

Since 1971, the Bureau of Labor Statistics (BLS) has conducted the Annual Survey of Illness and Injury with a well-constructed (and well-defined) probability sample; the most serious problem with this data base for the detection of neurotoxic illness is that the occurrence of illnesses, especially those involving latent periods, is vastly under-reported. BLS also maintains a Supplementary Data System in which data from the states' workers' compensation programs are compiled; the data gathered differ from state to state, and the lack of total reference populations precludes the calculation of rates.

The National Center for Health Statistics (NCHS) compiles data (from death certificates) on every death in the country through the National Death Index. The basic data on death certificates are fairly uniform across states, and demographic cause-of-death information and some occupational data are almost always available. Occupational data are of uneven quality, which usually depends on whether anyone uses them for research purposes. The categories ''housewife'' and "retired" are often reported in the place of out-of-home employment or one-time major employment. Data from the Bureau of the Census are appropriate as denominators in calculating rates. NCHS makes the encoded mortality data available to researchers on computer tapes. The National Death Index is valuable for determining the vital status of subjects in epidemiologic followup studies; for subjects who have died, researchers are directed to the appropriate state for detailed information. NCHS also collects data from the states on every birth and fetal death, including birthweight, Apgar scores, congenital anomalies, and complications of pregnancy or birth. The birth records are not as uniform across states as are death certificates, and only a few states record information on parental occupation. NCHS has conducted the National Health Interview Survey, a large (120,000 persons), stratified probability sample of the civilian, noninstitutionalized population, every year since 1957. In addition to standard demographic information, both occupational status and data on illnesses, injuries, disabilities, and use of medical services are gathered. The fourth NHANES, a probability sample of the civilian, noninstitutionalized population also conducted by NCHS, is currently being carried out on a full-scale basis (the first was carried out in 1970). Physical examinations, laboratory tests, and responses to a questionnaire yield a detailed medical picture, but occupational information was not gathered in the previous surveys. In the current survey, job histories are being requested, and neurotoxicity is among five work-related conditions about which information is sought.

Other data bases are of potential value for the detection of neurotoxic disease. Regional poison-control centers, insurance