Global Estimates of Fatal Occupational Accidents

Global Estimates of Fatal Occupational Accidents Jukka Takala Data on occupational accidents are not available from all countries in the world. Furthermore, underreporting, limited coverage by reporting and compensation schemes, and non harmonized accident recording and notification systems undermine efforts to obtain worldwide information on occupational accidents. This paper presents a method and new estimated global figures of fatal accidents at work by region. The fatal occupational accident rates reported to the International Labour Office are extended to the total employed workforce in countries and regions.

For areas not covered by the reported information, rates from other countries that have similar or comparable conditions are applied. In 1994, an average estimated fatal occupational accident rate in the whole world was 14. 0 per 100,000 workers, and the total estimated number of fatal occupational accidents was 335,000. The rates are different for individual countries and regions and for separate branches of economic activity. In conclusion, fatal occupational accident figures are higher than previously estimated.

The new estimates can be gradually improved by obtaining and adding data from countries where information is not yet available. Sectoral estimates for at least key economic branches in individual countries would further increase the accuracy. (Epidemiology 1999;10:640-646) Keywords: occupational accidents, fatality rates, statistics, recording and notification systems. The International Labour Office (ILO) collects and publishes global accident figures and rates that are based on national recording and notification systems. The ILO also supports member states in the enhancement of their recording and notification systems for occupational accidents and diseases. 2 However, reasonably reliable data may only be obtained ftorn a rather limited number of countries, ie, from about one-third of the 174 ILO member states. The information is not based on harmonized recording and notification systems, underreporting is common, and in many countries the reporting and compensa- tion systerris cover only selected economic activities, leaving out major sectors, such as griculture, which are known to have higher-than-average accident-frequency rates. Furthermore, some countries cover commuting accidents, traffic accidents at work, and occupational diseases, whereas others do not. Fatal occupational accidents are better reported than nonfatal ones in developing countries, but the same 1 limitations apply. Underreported figures and low estimates are currently used as a baseline for priority setting at the national level, which leads to preventive action that is less dim appropriate.

In addition to a better picture on fatal accidents, rough estimates of nonfatal accidents could be derived from the data on fatal accidents. Studies in the United States,3 Australia,4 Zimbabwe,5 Finland,6 and the European Union7,8 show that, if reporting is reliable enough, a rather constant ratio exists between fatal and nonfatal From the Occupational Safety and Health Branch, Working Conditions and Environment Department, International Labour Office, 4 Route des Morillons, CH-1211 Geneva 22, Switzerland. ©1999 by Epidemiology Resources Inc. ccidents leading to absence from work. This was shown first by the classical work of Heinrich9 as early as 1931. Earlier estimates have been based on a crude global fatality rate (8 per 100,000 workers) obtained from existing sources of industrialized countries. No attempts have been made to estimate such rates regionally. These earlier figures have been shown, by new information from developing countries, to be underestimated. New sources are based on real counts of fatalities in selected ILO member states and range from 1. 5 to 5 times higher1 than the old estimates.

Objective The objective of this paper is to establish a better method for estimating the number of fatal occupational accidents, in particular at global, but also at regional and national levels. This estimate can be based on available data, and it would provide a basis for setting policies and priorities for prevention of such accidents, as well as for funding and national decision making. More realistic estimates would demonstrate a visible difference corn, pared with the actual, and often unrealistic, reported data, which in turn would motivate countries to improve their recording systems.

The need for international harmonization of reporting methods and criteria would be more easily understood. What Is Reported.. Source Data Usually the ILO member states report both absolute numbers and frequency rates of fatal occupational accidents calculated for 1,000 workers. The frequency rates are more useful than absolute numbers for preventive purposes and for comparisons within and outside the 640 1 Epidemiology September 1999, Vol. 10 No. 5 FATAL OCCUPATIONAL ACCIDENTS 641 reporting field.

The ILO receives reports that usually contain information on fatal accidents at work and include those accidents that occur in traffic while the worker is carrying out work duties. Fatal commuting accidents and fatal occupational diseases are often but not always separately, reported although the latter figures are very poorly given. The reliability, comparability, and transparency of data are reduced, if these figures are not given separately. The data have been obtained from the annual compilation of the ILO Yearbook of Labour Statistics. Detailed accident statistics are also submitted to the ILO, and these are included in the CISDOC database. 10 This database was used for the study to retrieve countryspecific published data. Employment figures have been obtained from the ILO Yearbook of Labour Statistics,1 the United Nations De, velopment Programme’s Human Development Report 1994,11 and ftom the Global Burden of Disease and Injury Series of Murray and Lopez. 12 When either employment or fatality rate was not available for the year in question, data were used from the closest year in which they were available.

Methods Accident-reporting schemes of industrialized countries and of economies in transition cover a higher proportion of the employed population than those of developing countries, but even in industrialized countries, some industries and sectors, such as agriculture and the selfemployed, are often excluded. Information from developing countries is often limited to a small proportion of the economically active workforce. When reported rates, in fatal accidents per 100,000 workers, are applied to the whole economically active workforce, much higher-and more reliable-numbers of fatal occupational accidents are obtained.

To estimate the rates for countries where no reliable data are available, rates from similar or comparable countries have been used. EXAMPLE The given occupational fatality rate for Finland was 3. 2 per 100,000 in 1993. 6 The equivalent covered population was 1. 729 million employees (reporting base). The number of occupational fatalities within this group was 56. The size of the employed labor force, actively employed, was, however, 2. 064 million, which includes the self-employed-in particular, farmers, who are not included in the usual reporting scheme. When the rate of 3. is applied to all employed, a calculated but realistic figure, 66, is obtained. Ten fatalities were discovered that were not reported to the 11-0. This figure (10) was then compared with the separately reported number of occupational fatal accidents for the 131,000 insured farmers, which was 6. The remaining number of ex, pected but not reported fatal accidents (4) for the 104,000 actively employed (=self-employed) workers who were not covered by the reporting scheme looks very reasonable (3. 8 per 100,000 workers, whereas the overall rate was 3. 2 per 100,000).

This method thus produced almost precisely accurate results for the fatal accidents of self-employed workers, which are normally not reported to the 11-0. In addition, the detailed statistics also reported 14 commuting fatalities and 92 fatal occupational diseases, of which almost all were caused by asbestos. The method is not expected to be equally accurate for countries for which the missing portion of information is proportionally much higher. Results The results were classified according to the World Bank regions, because regional employment figures were available and comparable countries are reasonably well grouped together.

Table 1 covers the Established Market Economies (EMEs) and provides well reported data. The total number of fatal occupational accidents was estimated (calculated) to be 19,662, whereas the reported number was 16,320. For some countries, specifically repotted data were used to cover missing information. For example, the fatality rates for Norway are average rates from 1985 to 1989. 13 The European Union rate (5. 89 per 100,000 workers) was calculated for 19947 on the basis of 13 countries according to the model of Eurostat. Data from the Netherlands and the United Kingdom were not used. The resulting rate (5. 9 per 100,000 workers) was slightly lower than the latest available given rate from the EU for the previous year, 1993 (6. 10 per 100,000 workers). This change may reflect either random fluctuation of the number of accidents from year to year or a real decline and improvement in work environment in the European Union. However, even the fairly well organized reporting systems in Europe miss about 3,000 occupational fatalities (9,027 expected and 5,977 reported). For some countries, such as Japan, the fatality rate was not reported from the original source but was calculated by the author.

This provides another source of error, as the given number of fatalities may well be obtained from a smaller population than the full active employment. Such errors are producing more conservative results and can be corrected once information becomes available. Major disasters can also be detected, such as the Baltic ferry disaster, which doubles the Swedish figures for the year 1994 from the usual. A similar expected increase is not visible in the Estonian figures, which may be caused by different compensation practices. Reported figures and corresponding frequency rates are never overestimated, as they are based on facts.

However, the mixing of commuting accidents (and rarely diseases) may cause a source of erroneous increase. This is likely to be offset by the more frequent sources of erroneous decrease. Table 2, on Former Socialist Economies of Europe (FSE), is already somewhat less reliable, given that more data are missing from those countries. Therefore, no meaningful resulting reported data to the ILO for the whole region can be given. The expected number of fatal occupational accidents (15,563) is, however, realistic. 642 Takala Epidemiology September 1999, Vol. 10 No. 5 TABLE 1.

Fatal Occupational Accidents, Including Traffic Accidents at Work: Established Market Economies (EMEs) Reported Fatality Rate, (F,) per 105 Employment (E), Millions Fatalities (F, X E) 554 936 3 2,414 83 79 276 6,600 228 264 74 66 1,636 1,712 166 46 900 19 213 258 1,228 243 354 8,716t 9,027t 19,662 Fatal Accidents Reported to ILO 324 724 3 2,414 71 ou 238 6,588 160 115 75 56 1,082 1,712 56 iv 649 19 213 258 1,008 234 291 5,907 5,977 16,320 Country Notes 1994* 1994* 1995* 1994 1995I 1995* 1994§ 1995** 1994 1994 1994 1993 1991 1994 1994¶ 1994 1994¶ 1992 1994** 1994 1994II 1994** 1994 1993 1994

Australia 7. 00 7. 921 Canada 6. 93 13. 506 Iceland 2. 20Y 0. 137 Japan 3. 74Y 64. 570 New Zealand 5. 30 (1993) 1. 560 Norway 3. 80 (1985-1989) 2. 079 Switzerland 7. 30 3. 776 United States 124. 900 5. 30Y 3. 742 Austria 6. 10 3. 772 (1992) Belgium 7. 00(1991) 2. 555 Denmark 2. 90 Finland 3. 20 2. 064 France 7. 40 22. 110 Germany 4. 70H 36. 076 Greece 4. 40 3. 790 Ireland 3. 60 1. 268 Italy 20. 002 4. 50 0. 190 (1990) Luxembourg 10. 00H 6. 692 Netherlands 3. 10H 4. 458 Portugal 5. 80H 12. 042 (1995) Spain 10. 20 3. 928 Sweden 6. 20 25. 317 (1993) United Kingdom 1. 40 147. 88 European Union 5. 89H§§ (author calculation) 6. 10 (1993 EU) 147. 988 European Union (given rate)**** 366. 455 5. 37 Total EMEs (author calculation) Italics = data not available. * Including diseases. H Calculated, not reported. ICommuting accidents and diseases included. §Suva, Luzern. ||National Safety Council. ¶ Commuting reduced. # Based on Belgian figures. ** Including commuting but without Public Administration sector. HH Major shipping disaster. IIWork traffic accidents included. §§Based on 13 countries (United Kingdom and the Netherlands omitted). ****Eurostat 1997-2.

Figures on both India and China (Table 3) are much less reliable and are based on figures from other countries and regions: Malaysia (11. 0 per 100,000) in the case of India and FSE (11. 1 per 100,000) in the case of China. It is likely that the real figures are (considerably) higher than estimated and are expected to increase closer to the levels of countries in the group Other Asia and Islands (OAI) as the Indian and Chinese economies continue to grow. The highest rates (23. 1 per 100,000) and absolute fatality figures (80,586) are found in OAI countries as shown in Table 4.

The reason is that the rapidly industrializing countries such as the Republic of Korea, Thailand, and Indonesia have reported high fatal accident frequency rates. These rates are, however, obtained from a relatively small part of the employed workforce, which is covered by the compensation schemes. The biggest component of the usually noncovered employed work, force is agriculture, which is clearly more dangerous than the average of all industries in the whole world. The given fatality rate in the Republic of Korea in agriculture (90 per 100,000) is close to three times higher than the average fatality rate (34 per 100,000).

Methods in agri- culture may, however, differ, and the average rate in agriculture may be not so high in all other countries of the region. The component of the Table 5 that covers SubSaharan Africa is almost solely based on the frequency rate in Zimbabwe. Togo figures are close to those in Zimbabwe. The resulting absolute number of fatalities is not much more than an informed guess before further information becomes available. Farmers’ fatality rates are, however, providing some indicators as most of the actively employed in the region are occupied in agriculture.

Burkina Faso had an agricultural fatality rate of 99. 7 per 100,000. Guinea-Bissau had a 6-year average agricultural fatality rate of 90 per 100,000 but this was obviously based on a very small population covered. Latin-American results are based on figures from the biggest country, ie, Brazil. Other available data (from Bolivia, Colombia, Mexico, Panama, and Guatemala) confirm that the rate for the region (13. 5 per 100,000) is reasonable. The Middle Eastern Crescent has been based on a weighted average of Turkey and Egypt, which are reasonably large samples of the region.

The rate in agricul- 1 Epidemiology September 1999, Vol. 10 No. 5 FATAL OCCUPATIONAL ACCIDENTS 643 TABLE 2. Fatal Occupational Accidents, Including Traffic Accidents at Work: Former Socialist Economics of Europe (FSE) Country Reported Fatality Rate (Fr) Employment (E), Millions Fatalities (Fr X E) 97 417 232 127 257 75 427 89 101 45 192 840 763 9,450 129 28 2,294 14,524 15,563 Fatal Accidents Reported to ILO 400 27 2 73 257 61 151 89 128 66 645 482 6,770 121 25 2,300 Notes 1994 1991 1994 1993* 1994 1994H 1994 1995 1995 1995 1994* 1994 1994 1994 1994 1994 1994 Albania 0. 51 (1991) Belarus 8. 90 4. 696 Bosnia and Herzegovina Bulgaria 11. 43 2. 032 Croatia 11. 43 1. 108 (1993) Czech Republic 4. 32 5. 945 Estonia 11. 60 0. 649 Hungary 11. 40 3. 751 Latvia 7. 40 1. 205 Lithuania 6. 00 1. 675 Macedonia 6. 00 0. 396 Moldova 11. 43 1. 681 Poland 5. 70 14. 747 Romania 7. 00 10. 914 Russian Federation 13. 80 (1995) 68. 484 Slovakia 6. 00 2. 147 Slovenia 3. 20 0. 882 Ukraine 12. 00 19. 119 Yugoslavia (Serbia) Unknown FSE (10 countries) 11. 43 127. 060 FSE 11. 10 140. 282t Italics = data not available; calculated based on average of other countries in the region. Including commuting accidents. H Including commuting accidents and diseases. ISource: World Bank. tural work is again higher than average. The last part of Table 5 shows an overall absolute figure of fatalities in the region to be 41,850. World figures are summed in Table 6. The world rate is 14. 0 per 100,000 workers, and the estimated absolute number of fatal occupational accidents comes close to 335,000. The earlier ILO estimate has been 220,000 fatalities, which is clearly shown to be underestimated. 14

How TO GET FIGURES OF NONFATAL ACCIDENTS? The ratio between fatal occupational accidents and accidents causing 3 or more days’ absence from work varies with a wide range: 1 per 10, Africa1; 1 per 600, ILO earlier estimate15; 1 per 933, U. S. ratio of fatal accidents to all disabling accidents in 1992 (6,529 per 6. 09 million)3; 1 per 1,019, Finland [55 per 56,072 accidents causing 3 days or more absence from work in 1994; 1 per 975 in 1993 (56 per 54,623)]; 1 per 1,818, and 55 per 99,980, for all reported accidents in Finland; and 1 per 2,029, U.

S. ratio of fatal to all nonfatal injuries in 1992 (6,529 per 13. 247 million). A low ratio indicates underreporting of minor accidents, which is very common. No cases of overreporting are expected. A reasonably conservative ratio, 1 per 750, between the fatal accidents and nonfatal accidents could be a basis for a new ILO estimate of nonfatal occupational accidents. This would elevate the ILO estimate of the number of accidents to 250 million per year in the Discussion HOW TO IMPROVE THE ACCURACY OF THE FIGURES OBTAINED?

The estimated or obtained industry (economic branch)-specific frequency rates and corresponding employment figures from regions or countries that have similar or comparable conditions would provide a more accurate basis for calculations. if, for example, rates for agriculture, construction, transport, and the informal sector could be obtained from representative countries or regions, and if these sectoral rates would be weighted with the actual numbers of sectorally employed workers, the country estimates would be much better. TABLE 3.

Fatal Occupational Accidents, Including Traffic Accidents at Work: India and China Country India* ChinaH Reported Fatality Rate (F) 11. 0* 11. 1H Employment (E), Millions 334. 000 614. 690 Fatalities (F~ X E) 36,740* 68,231H Fatal Accidents Reported to ILO 310 7,235 Notes Human Development Report 1994, labor force, employment; F, based on Malaysia total (11. 0) 1994 State,owned enterprises reported * Italics = data not available; rate based on estimate and Malaysian average (11. 0), Indonesia total (43. 7), Bangladesh manufacturing (10. 0), India mining (42. ), Republic of Korea agriculture (90. 0), and Kazakhstan agriculture (15. 6). H Italics = rate based on Former Socialist Economics of Europe average calculated value. 644 Takala Epidemiology September 1999, Vol. 10 No. 5 TABLE 4. Fatal Occupational Accidents, Including Traffic Accidents at Work: Other Asia and Islands Country Bangladesh Cambodia Comores Reported Fatality Rate (F) 11. 00* 11. 00 11. 00 40. 00 10. 90 43. 70 11. 00 34. 00 11. 00 11. 00 11. 00 11. 00 11. 00 36. 26 19. 20 11. 00 10. 50 11. 00 19. 20 11. 00 23. 12 56. 160 3. 780 0. 220 0. 240 2. 915 78. 04 10. 170 19. 837 2. 470 0. 173 7. 645 16. 817 8. 240 33. 047 25. 166 1. 920 1. 649 5. 148 32. 095 34. 050 339. 840 Employment(E), Millions 47% 43% 38% 35% 1992 45% 55% 119. 5 8. 8 0. 6 0. 7 22. 6 4. 5 Fatalities (F, x E) 6,178 415 24 96 318 34,131 1,118 6,745 272 19 840 1,850 906 11,984 4,832 211 173 566 6,162 3,746 80,586 Fatal Accidents Reported to ILO Notes 1994H 1994H 1994H 1992H 1994I 1992′,tt 1994H 1994 1992H 1994 1995 1994 1992H 1992 || 1947¶ 1992H 1994 1994 1994 1992H 1994 Fiji Hong Kong Indonesia Korea, Democratic Republic of 255 2681 2662 16 534

Korea, Republic of Laos Macau Malaysia Myanmar Nepal Pakistan Philippines PNG Singapore Sri Lanka 40% 47% 20. 6 4. 1 220 65 820 Thailand Vietnam 47% 69. 5 Total Italics = data not available; rate based on Malaysian average rate. * Dockers Fr = 41. 0. H Human Development Report, Labour Force, 199411 I Including commuting accidents and diseases. § Based on approximately 6 million employees; includes commuting accidents. || Manufacturing F, = 17. 5. ¶Based on Thailand. IIIncluding commuting accidents. whole world, if the estimates are derived from the figures presented above.

Using a higher ratio experienced in Finland and the United States (1 per 1000) would result in a world figure of 335 million. In fact, the recently presented ILO, WHO, and United Nations figures (250 million occupational accidents and 160 million occupational diseases) are still based on relatively conservative estimates. 16 How To GET ALL WORK-RELATED FATALITIES? Commuting accidents are generally not included in the results given above. The ratio of occupational accidents (at the workplace and in traffic at work) to commuting 17 accidents is 68%/32% (or, more precisely, fatal acciTABLE 5.

Fatal Occupational Accidents, Including Traffic Caribbean, and Addle Eastern Crescent Reported Fatality Rate (F) per 105 Sub-Saharan Africa 21. 0 Latin American and Caribbeant 13. 5 Middle Eastern Crescentg 22. 5 Employment (E), Millions 218. 400 195. 000 186. 000 dents at work/fatal traffic accidents at work/fatal corn, muting accidents: 49. 3%/18. 6%/32%). If the estimate for the world, 335,000 fatal occupational injuries caused by accidents, is taken as a baseline, the estimate for fatal commuting accidents will be 158,000 (Figure 1).

The ratio of work-related traffic fatalities from all traffic deaths is 7. 67%17 (traffic at work/commuting: 2. 77%/4. 89% of all traffic fatalities). In the United States, the ratio of fatal highway work accidents is 3. 04 from all fatal road accidents ,18 ie, slightly higher than that in Finland. The ratio of fatal commuting accidents (of all fatal traffic accidents) in the United States is most likely also higher and is probably clearly higher for developing countries, where proportionally less time is Accidents at Work: Sub-Saharan Africa, Latin America and

Fatalities (F, X E) 45,864 26,374 41,850 Notes Employment 39% of 560 million,* fatality rate based on Zimbabwe, 1995 F, = based on Brazilian rate (13. 5); E = 44% of* 444 millionI F, based on weighed average rate in Turkey, 28. 3, (E = 25,537,000); in Egypt, 12. 0 (E = 15,241,400) E = 37%* of 503 million|| Italics = data not available. 11 * Human Development Report, UNDP, 1994. HOther fatality rates: Bolivia, 12. 5; Colombia, 14. 1; Mexico, 12. 0; Panama, 22. 0; and Guatemala, 25. 0. 12 I Global Burden of Disease, 1996, p. 493. § Other fatality rates: Bahrain, 16. (1995); Jordan, 8. 2; Tunis, 9. 1; Cyprus, fishing, 298. 1 (1994); Egypt, agriculture, 13. 0 (1994); Turkey, agriculture, 39. 9 (1994). 12 ||Global Burden of Disease, 1996, p. 497. Epidemiology September 1999, Vol. 10 No. 5 FATAL OCCUPATIONAL ACCIDENTS 645 TABLE 6. Fatal Occupational Accidents in the World: Estimates (All Regions) Region EME FSE IND CHN OAI SSA LAC MEC World Fatality Rate (F,) per 105 5. 3 11. 1 11. 0 11. 1 23. 1 21. 0 13. 5 22. 5 14. 0 Employment (E), Millions 366. 437 140. 282 334. 000 614. 690 339. 840 218. 400 195. 000 186. 00 2,394. 667 Fatalities F, X E 19,662 15,563 36,740 68,231 80,586 45,864 26,374 41,850 334,870 Notes World Labour Force 2. 7 billion EME = Established Market Economics; FSE = Former Socialist Economies of Europe; IND = India; CHN = China; OAI Other Asia and Islands; SSA = Sub-Saharan Africa; LAC = Latin America and the Caribbean; and MEC = Middle Eastern Crescent. consumed in leisure-time traffic (and there are fewer motor vehicles). Murray and Lopez19 estimate that 222,000 road traffic accidents occurred in 1990 in the developed regions. Of those, 7. 7% or 17,027 fatal road accidents are work related (at work in traffic or commuting accidents). Roughly the same number, or 17,000 (49. 3% of total work-related fatal accidents), are expected to take place at a stationary workplace, and 6,324 while in traffic at work, with a total at work of about 23,000. The calculated figure for such fatalities in EME countries was 19,662 in 1994. The ratio of fatal occupational accidents/fatal occupational diseases was 51%/49% calculated from all fatal occupational injuries in Finland in 1990-1994, and 40%/60% in 1994. 7 This would mean 325,000 fatal occupational diseases in the whole world. Counting all of these together gives the following figures: 335,000 fatalities at the workplace; 158,000 fatalities for comEXAMPLE muting between work and home; 325,000 fatal occupational diseases; and 818,000 total fatalities. This does not, however, cover diseases that are only partially work related, such as diseases of the heart and circulatory system. S. Hansen’s (Denmark) estimate in Working Conditions and Environment in Figures 20 is that 10% of such diseases are work related. There are about 6. 6 million deaths caused by ischemic heart disease in the world, which would mean, if Hansen’s estimate is used, 626,000 additional work-related deaths. Furthermore, the reference values in Finland cover no silicosis, as it is rarely found and practically no fatalities occur any more from this condition. Thus, the number of fatal occupational diseases is expected to be underestimated. The Australian National Occupational Health and Safety Commission estimates that there are 2,900 workrelated fatalities in Australia annually. 4 The total employment in Australia was 8. 235 million in 1995.

The frequency rate for fatal work-related injuries (work-related accidents and diseases) is 35. 2 per 100,000 workers. The extrapolated world number of work-related fatalities is 843,299. Murray and Lopez19 estimate that there were 1,129,000 to work-related fatalities in 1990. Taking into account the missing work-related diseases in the calculations above, the magnitudes of the three methods seem to match well. This number, 1. 1 million, could be considered the best available estimate of annual work-related deaths in the world. This means 3,000 deaths caused by work every day.

Figure 1 presents the categories of work-related injuries and the estimates of fatal injuries in each category. Information that will be annually accumulated and added to the ILO databases will certainly increase the accuracy of these estimates; however, radical changes are not expected. Probably a small gradual growth can be expected because of better reporting and increasing world population and active labor force. Critical Comments Although the numbers given above are based on practically all available information, precautions in interpreting them are necessary for the following reasons. . Conditions are never completely comparable from one country to other; industrial structures influence strongly the rates of accidents. This results in inaccuracies if rates of one country or region are used, in the absence of proper data, to estimate numbers in another area. For example, if mining and con- 646 Takala Epidemiology September 1999, Vol. 10 No. 5 struction industries in one country employ relatively more workers than in another country or region, figures and rates are poorly transferable. 2. Employment figures used were often inaccurate. 3.

Underreporting is likely even in the most advanced countries. 4. Only limited information is available for the two most important sectors in developing countries, the agricultural and informal sectors. These cover more than 50% of the world active employment. Additional specifically planned studies should be undertaken to reveal more accurate information. 5. The number of unemployed was estimated, but the number of underemployed has not been taken into ac, count. It is likely that a major proportion of the underemployed are engaged in the informal sector, but details are not available.

Acknowledgments The staff of the ILO Occupational Safety and Health Branch has contributed greatly to this paper. Mary Anderson has efficiently processed the tables and the text, and K. Mikkola, of the Ministry Health and Social Affairs in Finland, has provided statistical advice. References 1. Yearbook of Labour Statistics (in English, French, and Spanish). 55th Geneva: International Labour Office, 1996. 2. Recording and Notification of Occupational Accidents and Diseases: Code of Practice. Geneva: International Labour Office, 1996. 3. Leigh JP, Markowitz SB, Fahs M, Shin Q Landrigan P.

Occupational Injury and Illness in the United States. Arch Intern Med 1997;157:1557-1568. 4. Australian Occupational Health and Safety. Statistics Bulletin No. 1. ed. Trends Over Recent Years. Sydney, National Occupational Safety and Health Commission, 1997. 5. Leewenson R. Health Impact of Occupational Risks in the Informal Sector in Zimbabwe. Geneva: International Labour Office. Available at http:ll ~. ilo. org/public/english/90travai/sechyglpapers/inhimb/index. htm 6. Accidents at Work 1993. Official Statistics of Finland (in Finnish, Swedish and partially English).

Helsinki: Statistics Finland, Ministry of Labour, 1996. 7. Accident at Work in the European Union in 1993: Initial Results. Eurostat. Statistics in Focus 199712. Luxembourg: European Communities, 1997. 8. Accidents at Work in the European Union in 1994. Eurostat. Statistics in Focus, 199812. Luxembourg: European Commission, 1998. 9. Heinrich HW. Industrial Accident Prevention. Ist ed. New York: McOrawHill, 1931. 10. International Occupational Safety and Health Information Centre. CIS: CISDOC Bibliographic Database (printed, on-line, and CD-ROM versions, in English and French).

Geneva: International Labour Office, 1997. 11. United Nations Development Programme. Human Development Report 1994, New York: United Nations, 1994. 12. Murray CJ, Lopez AD. The Global Burden of Disease, Global Health Statistics, World Health Organization, World Bank. Cambridge, MA: Harvard School of Public Health, 1996. 13. Fatal Accidents at Work: Fatal Accidents in the Nordic Countries over a Ten-Year Period. Copenhagen: Danish Work Environment Service, 1993. 14. International Labour Office GLO)/World Health Organization (WHO).

Methods for Health Impact Assessment in Environmental and Occupational Health: Report on WHO/11,0 Consultation. Geneva: ILO/WHO, 1998020. 15. Programme and Budget for 1998-99. Geneva: International Labour Office, 1997. 16. Annan KA. Occupational health and safety: a high priority on the global, international and national agenda (Editorial). Afr Newslett Occup Health Safety 1997;7:51 (also published in Asian-Pacific Newslett Occup Health Safety 1997;4:59). Geneva: International Labour Office; Helsinki: Institute of Occupational Health. 17.

Accidents at Work 1994: Official Statistics of Finland (in Finnish, Swedish and partially English). Helsinki: Statistics Finland, Ministry of Labour, 1997. 18. Accident Facts, 1996 Edition. Itasca IL: National Safety Council, 1996. 19. Murray Q1-, Lopez AD. Summary: The Global Burden of Disease, World Health Organization, World Bank. Cambridge, MA: Harvard School of Public Health, 1996;1S. 20. Sosiaali-ja terveysministerid. Tybolot numeroina. (Ministry of Social Affairs and Health. Working Conditions and Environment in Figures. ) Tampere, Finland: Ministry of Social Affairs and Health, 1997.