SAFETY SCIENCE Safety Science 31 (1999) 31±57 The classi®cation of accident data M. Lortie*, P. Rizzo Biological Sciences, Universite du QueÂbec aÁ MontreÂal, PO Box 8888, Station Centre-ville, MontreÂal, QueÂbec, Canada H3C 3P8 Received 7 November 1996; accepted 14 July 1998 Abstract The objectives of this paper were to document how accident data are usually classi®ed, whether this system makes it possible to classify all the data contained in the accident reports, and to examine the classi®cation problems encountered. The ®rst part reviews the variables retained and descriptors used by the accident studies published over the past 10 years. This synopsis showed that the types of data considered and the manner in which they were classi®ed varied greatly between the studies. Data on the accident circumstances (e.g. activity and incidents) were seldom considered, while accident and injury data were extensively analyzed. The second part analyzes the vocabulary and data reported by injured handlers in 580 accident descriptions. Possible grouping vocabulary strategies were explored and the importance of the implicit nature of data was evaluated. This revealed that the vocabulary used by the injured was both rich and variable. For example, over 80 terms were used to describe one activity. While some grouping strategies to classify data could be developed for the worksite or incidents, it was particularly dicult to identify a logic for grouping activity data. Also, the analysis showed that many important data are of an implicit nature. A literal or automatic classi®cation of terms may, therefore, lead to signi®cant biases. Furthermore, although data on incidents were frequently reported, this type of data is generally disregarded by most accident studies. Finally, the paper discusses various classi®cation problems that emerged. # 1999 Elsevier Science Ltd. All rights reserved. Keywords: Accident; First accident report; Classi®cation; Incident; Activity * Corresponding author. 0925-7535/99/$Ðsee front matter # 1999 Elsevier Science Ltd. All rights reserved. PII: S092 5-7535(98)0005 3-8 32 M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 1. Introduction Industrial accidents are very costly. For example, in Quebec, 127 883 compensatable work accidents were reported in 1995 for a direct cost of nearly $1.2 billion (CSST, 1995). An understanding of the underlying causes of accidents is essential for their prevention. As pointed out by Larsson (1990c) and the EUROSTAT task force (Clarke et al., 1992), the actual collecting of accident information is insucient to meet prevention objectives. In fact, several models have been proposed to extensively analyze accident circumstances (e.g. KjelleÂn and Larsson, 1981; Harvey, 1984; La¯amme, 1988) in order to develop pertinent preventative actions. However, these models rely on detailed post-accident investigations, while prevention strategies in industry are often based on the retrospective analysis of brief accident reports, usually completed following the injury. These brief reports are used to build national statistics or to provide an analytical focus on a speci®c industry, sector or job. Knowledge of a sector or job usually allows for more sophisticated analyses. In the ®rst case, several studies have already dealt with the dierences between the systems, from de®ning what an accident is (for more extensive reference and review see Clarke and Glendon, 1988; Clarke et al., 1992), to examining the loss of information tied to how accidents are classi®ed. For example, the task force indicated the diculty in comparing falls from heights because not every system dierentiates falls from the same height; some include falls from vehicles, others do not. However, their main focus was the harmonization of statistics among countries rather than data extraction from accident reports. Which data can be extracted, how it can be organized, and the problems encountered in analyzing them are issues which are rarely discussed. Moreover, Strandberg (1985) and Manning et al. (1988) have shown that some types of accidents, in particular the loss of balance, were long underreported in accident studies due to a classi®cation bias. Since only the event immediately preceding the injury was generally retained, earlier events in the accident chain, describing a loss of balance, were lost in the classi®cation system. However, the results of accident analysis are widely used to justify not only prevention but also research orientations. Thus, system harmonization is only one of several problems to be solved. In addition, various published studies usually present only the ®nal results of the classi®cation analysis. The scope of data actually contained in accident records and the reasons for classifying the data in a particular manner are seldom mentioned. On the whole, little attention has been paid to establishing a topology of problems associated with this analysis or to possible classi®cation strategies. The ®rst objective of this paper was to document how data were extracted from accident reports and whether the retained variables could cover the overall data present in these descriptions. The second objective was to document the classi®cation problems encountered. First, studies published on accident data covering the past 10 years were reviewed (Part 1). Subsequently, a typical accident database was analyzed in order to document the type of facts or events reported, how they could be structured, and the classi®cation problems encountered (Part 2). M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 33 2. Part 1: Review of the accident analysis studies 2.1. Methodology 2.1.1. Selection of material Initially, eight major journals (Ergonomics, Journal of Occupational Accidents/ Safety Science, International Journal of Industrial Ergonomics, Journal of Safety Research, Journal of Occupational Medicine, Occupational Medicine, Applied Ergonomics, and British Journal of Industrial Medicine) were systematically reviewed over a 10-year period (1986±1995) to identify studies analyzing accident reports (n=72). Studies centered primarily on vehicle accidents, and fatalities were a priori excluded. Secondly, only studies centered primarily on the analysis of accident circumstances and using ®rst injury records or national databases were retained. Studies focusing on non-occupational accidents (n=2), analyzing data obtained mainly through detailed investigative-type questionnaires or interviews (n=17), and dealing primarily with injury and ®nancial data (n=16) were excluded. Five studies using a brief questionnaire, similar in format to a typical accident record, were nevertheless included for a total of 37 articles. 2.1.2. Analysis performed on the selected studies The articles were pooled into two groups based on whether the study used the original accident reports or data already encoded through a national database system [e.g. SDS (Supplementary data system), ISA (Information system on accidents)]. Studies in both groups were also sub-divided into three categories according to whether the main results presented covered all accident types, predominantly musculo-skeletal type accidents (e.g. overexertion, manual handling, back injuries) or speci®c accident types other than the previous one (e.g. falls). The principal study characteristics (number of data ®les analyzed, target population, main focus of the studies) and the retained variables are presented in Table 1. These variables refer essentially to the accident circumstances (site of the accident, the worker activity, other) and to the injury/accident process (injured body part, nature of injury, type of injury, source of injury, cause of injury, cause of accident, type of accident, agency of the accident). In some cases, references have been made in the result section to data corresponding to variables not presented in the methodology section. Only those variables de®ned in the methodology section were retained. 2.2. Results As can be seen from Table 1, and as was expected, no variable was systematically used since the studies had dierent objectives; some presented a general overview of the accident situation, others focused on a speci®c type of injury/accident. 2.2.1. Accident site variables In 24 of the 37 studies reviewed, at least one variable referred to the accident site. In 17 of these, the work site was used as an independent variable to stratify worker 208 736 154 597 178 23 901 662 442 712 416 Larsson (1990b) Norrish and Cryer (1990) Harker et al. (1991) La¯amme et al. (1991) Al-Arrayed and Hamza (1995) Lortie (1987) Niskanen and Lauttalammi (1989) Brun (1990) Yassi et al. (1995) 417 Sullivan and Shimizu (1988) Waller et al. (1989) 451 No. of ®les Doyle and Conroy (1988) Based on accident descriptions (21) Study nursing, tertiary care hospital electric utility construction industry orderlies; geriatric hospital various industries engine workshops various industries commercial ®shermen metal workers carpenters law enforcement person farming Target population Table 1 Main characteristics of the studies reviewed (headings de®ned in footnote) 2: overexertion 2: back injury handling x x x x 2: patient handling 2: material handling x x x x x AS x x x x x x x WA x x x I Variable on accident circumstances 1 1 1 1 1 1 1 1 Focus of study x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x NI T M SI I/AC TA AA IB O Variables on the injury/accident process 34 M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 232 657 395 83 1695 37 314 108 Inancsi and Guidotti (1987) Sinks et al. (1987) Reesal et al. (1989) Larsson (1990a) Ma et al. (1991) Miller (1992) Larsson and Rechnitzer (1994) ToÈrner et al. (1995) NA NA 146 Pines et al. (1987) Conroy (1989) Oleske et al. (1989) Based on national databases (16) 109 Olson and Gerberich (1986) automotive parts industry meat packing industry construction industry ®shing industry various industries farmers wood-bamboo furniture industry agriculture welders various industries various industries various industries 1 1 1 3: serious accident 3: forklift truck accident 3: machinery accident 3: mechanical injury 3: hand injury 3: eye injury 3: cold injury 3: burns 3: traumatic amputations x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x (continued overleaf) x x x M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 35 No. of ®les 14 156 16 700 229 NA 26 923 7427 5647 26 450 60 000 38 000 990 8981 Study Brewer et al. (1990) Leigh et al. (1990) O'Connor et al. (1993) Kisner and Fosbroke (1994) Andersson et al. (1990) Leigh et al. (1991) Engkist et al. (1992) Bobick and Myers (1994) Aghazadeh and Mital (1987) Stout-Wiegand (1987) Oleske et al. (1992) Cross and Walters (1994) Table 1Ðcontd coal mining industry various industries various industries various industries agriculture nurses' aides underground coal mining various industries construction industry agriculture coal mining various industries Target population 3: vibration back 3: foot injury 3: forklift trucks accident 3: handtools injury 2: sprain/strain injury 2: overexertion back accident 2: sprain/strain back injury 2: overexertion 1 1 1 1 Focus of study x x x x x x x x AS x x x x x x x x x WA I Variable on accident circumstances x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x NI T M SI I/AC TA AA IB O Variables on the injury/accident process 36 M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 2397 mining industry 3: contractor workers 24 x 25 x 3 17 10 2 12 x 13 14 x x 13 29 8 x Table headings: AS, accident site; WA, worker activity; I, incidents; NI, nature of injury; T, type; M, mechanism; SI, source of injury; I/AC, injury/accident cause; TA, type of accident; AA, agency of accident; IB, injured body part; O, other. 1, All accidents; 2, musculo-skeletal injuries; 3, other speci®c focus. NA, not available. Total Blank et al. (1995) M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 37 38 M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 populations when comparing accident pro®les. In 12 of them, the site referred only to the general workplace, such as the type of industrial sector (e.g. agriculture, mining), while ®ve identi®ed departments/sections within a plant (e.g. diesel vs petrol workshop in automotive plant). Although it could be argued that this type of variable also refers to the general work activity, it was considered here as a site variable. In eight studies the work site was used as a dependant variable, making it possible to situate the worker within his workplace (e.g. bedroom or washroom on a hospital ward) at the time of the accident. In three cases, however, no results were presented on the accident site. 2.2.2. Activity variables At least one such variable was present in 25 studies. As seen for the work site, an activity variable was also used to stratify groups, essentially according to worker occupation, with descriptors proposed by the US Census Bureau of Occupation Codes or some equivalent (n=19). Fourteen studies considered data referring speci®cally to the activity performed at the time of the accident, eight of them using both approaches. In most of these cases (11/14), the studies using original accident reports were the ones which de®ned an activity variable. It is dicult to establish general trends, since the data were classi®ed according to dierent study objectives. As can be seen from Table 2, the importance given to the work activity varied, the number of de®ned variables varying between one and eight, with seven studies de®ning at least two. The choice of variable titles was also quite diversi®ed, with over 17 dierent titles identi®ed (e.g. activity, main task, handling operation, construction phase, type of work, handler's actions). As expected, the descriptors associated with these dierent variables were also dierent: data on handler's actions cannot be grouped in the same way as data on type of work. However, an attempt was made to classify the descriptors based on whether they described the task (the general work objectives), a sub-task (a component of the task), or a speci®c action. Certain descriptors clearly referred to either the task (e.g. assembling, maintenance, loading/unloading, formwork), the sub-task (e.g. grinding, cutting, feeding into machine) or the action (e.g. holding post, push/pull, hold back, pulling arm). In some cases, it was found dicult to classify the descriptors without a knowledge of their contextual use (e.g. handling can refer either to a task, a sub-task or an action). In many cases, descriptors of dierent natures were grouped under the same variable, rendering the contents heterogeneous (e.g. equipment used, movement executed). Overall, while everyone appeared to have developed his own logic for grouping data, this logic was rarely developed. 2.2.3. Incident variables Incidental data are the events or facts reported in the accident process that were either unexpected or unusual to the task being performed; they often constitute the factor initiating the accident. Only three studies de®ned an incident variable and only few results were presented. M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 39 Table 2 Variables and descriptors used to classify activity data (n=14) Authors Variables de®ned and their descriptors Olson and Gerberich (1986) 1 1: activities performed at time of injury: routine operations, unjamming, cleaning, loading/unloading, servicing/ repairing Lortie (1987) 4 1: handling operation: on-site handling, transfer, lift, installing in bed 2: action: lift, pull, push, turn, hold back, etc. 3: handling site: bed, chair, other 4: equipment used: quilt, basin Niskanen and Lauttalammi (1989) 3 1: construction phase: excavation, foundations, frame and roof structures, etc. 2: work operation: formwork, concrete reinforcing, heating, plumbing, etc. 3: activity: manual/carrying, handling in storage, pushing/pulling Reesal et al. (1989) 1 1: work activity: chipping, grinding metal particles, welding, cutting, gouging, preparing equipment, air blast, etc. Brun (1990) 2 1: task: material handling, equipment utilization, ascend/ descend, climb 2: equipment used at time of accident: handtool, ladder, electric cable, etc. La¯amme et al. (1991) 8 1: main task: handling, assembling, machining 2: sub-task: moving, carrying, handling, adjusting, ®xing, tightening 3: movement performed: lift material, lose grip, lose balance, manipulation of tool 4, 5, 6: object related to main task, sub-task and movement 7: type of task achieved: no examples presented 8: type of duties: no examples presented Larsson (1990b) 1 1: activity: feeds in/takes out, adjust/cleans machine Larsson (1990a) 2 1: type of work: woodworking, food processing, farming 2: external agency/activity: forest work, ®rewood production, timber processing, moving of animals Leigh et al. (1990) 2 1: type of mining: underground versus open-cut 2: activity: handling, roofbolting, equipment repair, conveyor work, walking/running Leigh et al. (1991) 3 1: activity: handling, roofbolting, conveyor work, equipment repair, metal trades 2: speci®c activity: unassisted lifting, walking, pulling arm 3: type of mining: underground versus open-cut Ma et al. (1991) 1 1: job task: woodworking operations, material handling, maintenance Miller (1992) 1 1: handlers' actions at the time the accident: positioning post-driver, holding the post upright, checking the post (continued overleaf) 40 M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 Table 2Ðcontd Authors Variables de®ned and their descriptors Blank et al. (1995) 1 1: activity: normal operations, maintenance and repairs, movements on foot, carrying/lifting ToÈrner et al. (1995) 1 1: activity: hauling of crawl, shooting of crawl, hawling other type of gear, climbing/descending ladder or stairs, work in galley, loading, unloading, repair work/work by the wharf, cleaning of ship/equipment, etc. 2.2.4. Injury and accident variables All reported variable titles containing the word `injury' and `accident' were recorded; as seen in Table 3, 26 dierent titles were identi®ed. In many cases, the same type of data was referred to by dierent titles. For example, the descriptors relating to the variables `nature of injury' and `diagnosis of injury' were identical. The titles referring to the same data or descriptors were, therefore, grouped together to form eight sub-groups. They are identi®ed in Table 3 by the most commonly reported title. For the injury they are the nature, type, mechanism, source and cause; for the accident they are the cause, type and external agency. Since most of the studies used American National Standards Institute (ANSI, 1969) codes (voluntarily or involuntarily), at least in some form, the classi®cation system proposed by the ANSI Z.16.2 standard was used to compare the dierent variables de®ned. ANSI is a classi®cation system de®ning categories (equivalent to the term variable used in this paper), codes and subcodes (equivalent to the term descriptor used here). In fact, most of the descriptors corresponded to ANSI codes found under three ANSI categories: the nature of injury, de®ned as the identi®cation of the injury in terms of its principal physical or more simply medical diagnosis; the accident type, which identi®es the event directly resulting in the injury; and the source of injury, which identi®es the object, substance, exposure or bodily motion which directly produced or in¯icted the injury. Since this was insucient to cover the descriptors reviewed, three other types of data were identi®ed: the external agency of the accident, the activity, and the incident data. The latter two dier from those previously identi®ed, which were speci®cally de®ned under an activity or incident variable. The present case concerns activity or incident descriptors found under accident or injury variables. Two important points emerge from Table 3. First, one given variable may regroup descriptors clearly referring to dierent types of data. For example, the descriptors under the `cause of injury' and `cause of accidents' can refer to the medical diagnosis, the event resulting in the injury, the object in¯icting the injury and/or the activity at the time of the accident. Second, identical descriptors can also be found under dierent variables. For example, descriptors referring to the event resulting in the injury may be found under 11 dierent variable titles, belonging to ®ve of the eight classes of variables identi®ed. Sometimes, authors even used dierent titles interchangeably within the same paper. For example, `type of injury' was found to M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 41 Table 3 Variables and data considered for the injury and accident process (n=37) De®ned variables Data related to: Event Object, Object, Activity Particular Medical diagnosisa resulting equipment, equipment, at the circumstance/ in the substance substance time of unexpected events injurya in¯icting associated injury the injurya with the injury Related to injury Nature of injuryb (14)c Diagnosis of injury (2) 16 Type of injury (11) 6 Mechanism of injury (2) 4 1 2 1 1 1 2 Source of injury (7) Agency of injury (2) Material causing injury (1) Object causing injury (2) Cause of injury (6) Event causing injury (1) Event preceding injury (1) Action causing injury (1) Related to accident Cause of accident (3) Main event causing accident (1) Type of accident (8) Type of incident (1) Accident event (1) 12 1 8 4 1 4 3 1 10 Principal external agency (2) Agency of the accident (2) Vehicle of energy transfer (1) Harmful agency (1) Handtool type (1) Equipment/vehicle involved (1) Reported machinery/ equipment/material (1) Equipment or material used (3) External agency/ activity (1) Total a b c 25 28 19 1 1 1 1 13 1 1 15 6 4 Corresponds to the ANSI category: nature of injury, accident type, source of injury. The most commonly reported title. Number of studies. 42 M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 be used interchangeably with `nature of injury, type of accident or mechanism of injury'. 2.2.4.1. Injury variables. In the ®rst group, `nature of injury', all the descriptors referred to the physical characteristics of the injury. Generally, between ®ve or six ANSI codes or similar expressions were usedÐe.g. fractures, sprains/strains, cuts or lacerations. The treatment of these data appeared to be very homogeneous. In the case of `type of injury', no formal de®nition was found in the 11 studies using this variable. However, the descriptors listed often referred to the ANSI ones corresponding to either the nature of the injury or to the type of accident. Between three and six descriptors were generally employed with `contusion, sprain/ strain, fracture, laceration' (nature of the injury), `struck by' and `overexertion' (accident type) being the most common. In rare cases, the authors added other terms to those proposed under ANSI codes, such as `manual handling', an activity descriptor. The `mechanism of injury' was reported only twice, without being formally de®ned. The descriptors were also similar to those listed under the ANSI `accident type'. Additional descriptors also referred either to the activity (e.g. transferring patient, pushing, pulling) or to incidental events (e.g. technical shortcomings in equipment, loss of control caused by another person, collapse of equipment). For the `source of injury', most of the descriptors used were derived from the ANSI source of injury category. Between ®ve and ten descriptors were generally used. As previously indicated, the `cause of injury' was remarkable in the sense that it grouped together descriptors referring to data of dierent nature, such as the event resulting in the injury, and/or the object/equipment involved in or causing the injury, and/or the medical diagnosis of the injury, and/or the activity performed at the time of the accident, and/or particular circumstances or unexpected events arising during the execution of the task. No formal de®nition was presented in the studies. However, most studies used standard ANSI codes. Some also combined the cause and the source in the same descriptor (e.g. struck by ropes). The descriptors found under the three titles used, `event causing injury, event preceding the injury, action causing injury', referred primarily to the `main event causing accident', which corresponds to the ANSI de®nition of `type of accident'. 2.2.4.2. Accident variables. The `cause of accidents' (n=4) essentially grouped together the same descriptors found under the `cause of injury'. The `type of accident', de®ned by ANSI as the event which directly resulted in the injury or how the object or source of injury contacted the person, was considered in ten studies, under three dierent titles. Most of the reported descriptors are those outlined in ANSI codes for accident types, with the terms `falls, overexertion, struck by/against, caught in/between' being the most commonly used. Some added other descriptors, such as `cut/laceration, sprains/strains' and `working surface', descriptors which are also used under other variables such as `the nature and the source of injury'. M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 43 Finally, the last group, `external agency of the accident' (n=13), proposes similar descriptors to those used for the source of injury, excepting that here the agent is a factor in the accident process, which does not necessarily in¯ict the injury. Six of the 13 studies speci®cally recorded the equipment used at the time of the accident, one included descriptors referring to the characteristics of the working environment (e.g. unstable, wet ground; ANSI hazardous condition category). 2.2.5. Injured bodypart An injured bodypart variable was de®ned in most of the studies (29/37), using half a dozen or so ANSI codes. The majority of the bodypart descriptors were treated individually. However, in 17 studies some descriptors were grouped. The groupings of descriptors covering adjacent large body areas (e.g. lower limb vs trunk vs upper limb), in particular the neck, shoulder and hip joints, were documented and compared to the grouping proposed by ANSI, which groups them under the trunk, shoulders, chest, back, abdomen and hips. Most of the studies ®rst separated the back and shoulders from the trunk. When the shoulder was grouped with other bodyparts it was with the upper limb and/or with the cervical spine (e.g. neck± shoulder±arm) and/or with the upper part of the trunk (e.g. shoulder±chest±arm). The hip was mainly grouped with the lower limbs (n=6) and less frequently with the abdomen area (n=2). The neck was included with the head, back, trunk or shoulder girdle. Thus, the trunk could include dierent body parts, the boundaries often being dictated by the bodyparts which were not treated as separate data. Generally, the neck, back and shoulders, and occasionally the chest, were treated separately. Therefore, the trunk per se was mainly considered to be the bodypart inferior to the back or the front part (e.g. chest, abdomen), not the whole of the body core as de®ned in ANSI. 2.3. Discussion The studies reviewed had dierent objectives and analyzed both original reports and preclassi®ed data from national databases. Since a variability with respect to the treatment of the data was expected, comparing these studies to a given standard would thus have been unjusti®ed. For example, studies catering to sprain/strain injuries did not de®ne a `nature of injury' variable but instead emphasized data on activities such as manual handling. Generally, much emphasis was placed on the data relating to the injury and accident process, the site or activity data being used to stratify the injured workers according to the type of industry and occupation. Analysis of the data centered on the accident circumstancesÐmore speci®cally the activity at the time of injury, the worksite and the incidents were less frequently covered. This focus is far from that favored in the accident models referred to in the introduction. Essentially, these investigative approaches are based on the identi®cation of the incidental factors and/or unusual events that either triggered the accidental process or were associated with it, as well as a comprehensive understanding of the activity performed at the time of the accident. As pointed out by Larsson (1990c) and the EUROSTAT task 44 M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 force (Clarke et al., 1992), the actual collecting of accident data is insucient to meet prevention objectives. This is true on both a national and a sectorial basis. This review shows that little attention is paid to the characterization of incidents or activities. When using original accident reports few variables were designed to cover activity data. In addition, the variability in the treatment of these data was such that comparisons between studies were impossible. As shown, ANSI codes were used in some form or other by most of the studies, either voluntarily or involuntarily. This standard was sponsored in part by the National Safety Council in the 1960s to provide a method of recording key facts about work accidents in a form that could be analyzed to show general patterns of the accident process. It was not the objective of the summaries either to indicate speci®c accident prevention measures that should be taken or to record faults or responsibilities. The standard de®nes eight categories: nature of injury, part of body aected, source of injury, accident type, hazardous condition, agency of the accident, agency of accident part, and unsafe acts. The ®rst four categories were mentioned in the results section above. The hazardous condition classi®cation identi®es the hazardous physical condition or circumstances which permitted or occasioned the occurrence of the accident type (e.g. defects of agencies, environmental hazards). The agency of the accident identi®es the object, substance or premises in or about which the hazardous condition existed (e.g. ladders, conveyors). The agency part is a speci®c of the agency. The unsafe act classi®cation identi®es violations of a commonly accepted safe procedure which directly permitted or occasioned the occurrence of this accident type (e.g. operating or working at unsafe speeds). No paper retained similar variables to these last four. Also, as a system intending to cover all accidents in all sectors, the list of codes (three digits) is of course far more extensive than the one used in the quoted papers. However, most papers used the same codes. Even though similar codes or descriptors were used, in particular for the nature of injury, injured bodypart, source of injury and accident type, their groupings were variable. Also, many studies added other descriptors, sometimes of a similar nature, sometimes conceptually dierent. Few provided an explanation for the chosen classi®cation strategy. Grouping various types of data together, as in the cause of injury may in fact have been an attempt to more eectively identify the accident circumstances by combining the ANSI `accident type' descriptors with those referring to the activity and to the incidents. The European task force, referred to in the introduction, compared how occupational accidents were reported in 12 European countries. Only three variables dealing the accident description material were gathered from nine or more countries: the body parts (12), the type of injury (11), and the source of injury (9). Therefore, anyone interested in accidents has no choice but to develop other classi®cation systems. Unfortunately, between the investigative models and national statistics proposals, there are few systematic proposals of classi®cation system that would met the intermediate needs. As indicated previously, this leads to fewer proposals and subsequent problems in comparing studies results. M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 45 3. Part 2: Analysis of the data contained in a series of accident reports 3.1. Methodology 3.1.1. Study material The material consisted in 580 accident reports involving manual material handlers. The main task of these handlers was to load and unload goods from vans. Subtasks involved driving lift trucks and handling various equipment or structural elements (e.g. changing the fork on the lift truck or installing the plate joining the dock to the van). The accident reports, which were completed either by the foreman or health and safety department employees, were literal transcriptions of the accident process as described by the injured worker. Only those reports completed in French (the large majority) were retained. However, they contained many English terms, which is common in industry. These reports varied from being very brief to very detailed. This database has already been analyzed to identify the main accident pro®les (Lortie et al., 1996). This present study focused on the vocabulary and the manner in which the accidents were reported by the workers. 3.1.2. Data analyzed As reported in Part 1, most of the studies use similar descriptors for the injury, accident and bodypart variables (e.g. accident type, nature of injury). Attention was, therefore, focused on data describing the accident circumstances. All the terms or expressions characterizing the worksite, the activity/object handled and the incidents were systematically recorded. The worksite corresponds to any data making it possible to identify the worker geographically with respect to his activity. This could include terms referring to various areas of the workplace (e.g. dock, van) or to a piece of equipment (e.g. to be standing on a forklift truck). The handling activity refers to what the handler was doing at the time of the accident, which implies both an action verb (e.g. lift, push) and its direct complement or the object of the action, generally the handled object. The incidental data includes any reported unexpected or unusual event, in function of the work process. In this case, the events were generally reported using short sentences most often consisting of three or four words. 3.1.3. Analysis performed 3.1.3.1. Frequency and variation in vocabulary or phrases used. All dierent terms/ phrases referring to one of these four types of data, i.e. the work site, the activity, the object of activity and the incident, were recorded. The number of terms referring to the same type of data for each accident description (intra-description) and for the whole of the database (inter-description) were recorded. For example, the following description ``while loading the van, I lifted the plate and I forced'' contains two activity terms (loading, lifted), one handled object (plate), one site term (van) and one incident term (I forced). 3.1.3.2. Possible grouping strategies for the terms recorded. As the recorded vocabulary was highly variable, dierent strategies of grouping for the four data categories 46 M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 were explored. First, synonyms were identi®ed. Second, terms that either were used indierently (e.g. pop, van) or were essentially a speci®c of another term (e.g. van, trailer, semi-trailer) were grouped together. Since this covered only a limited number of situations additional grouping strategies were explored, which are explained in the results section. 3.1.3.3. Contextual signification of the terms used. The goal was to determine whether the meaning of a given term varied in the function of the context in which it was used. The underlying question being: could a term be classi®ed independently of its context? This analysis was performed for the handling activity terms only. The context was de®ned as being the verb's direct object, essentially the handled object. First, the identi®ed complements were grouped into ®ve classes based on whether they referred to: a group of diversi®ed merchandise (e.g. to load the van), a group or a sequence of a same object (e.g. to pile boxes), one given object (e.g. to tilt a drum), handling equipment, or other objects (e.g. to lift the door). Second, these same complements were regrouped in function of some of their descriptive characteristics (e.g. heavy or large objects, objects cylindrical in form). In each step it was noted whether the activity verb was used exclusively for a given class. Finally, the complements reported with the verbs `lift' and `slip' were speci®cally analyzed, since they are frequently considered in the literature. 3.1.3.4. Analysis of the explicit and implicit nature of the data. Certain data are implicit in the sense that they are not actually written down but they can be deduced from the accident description. Two cases of implicit data were considered. In the ®rst, the data was deducible from another category of data. For example, in `handler is loading merchandise', the activity data makes it possible to deduce that the worker is inside the van, i.e. it provides work site data. In the second case, data are explicitly reported, but their interpretation or meaning is implicit. For example, in `handler used excessive force to lift the plate (joining the van to the shipping dock)', which describes a incident, it can be deduced that the plate was either jammed or defective. Of course, these analyses imply a minimum, yet not extensive, knowledge of the workplace. Therefore, the portion of the accident descriptions which allowed for implicitly deducing other data were extracted. 3.2. Results 3.2.1. Variability of the vocabulary used The workers used a very extensive vocabulary when describing their accidents. 41 and 84 dierent terms were used to describe the accident site and the activity, respectively. From the 84 activity terms recorded, 61 referred speci®cally to manual handling activity. Other terms mainly described worker displacements or other marginal activities. One hundred dierent phrases/terms were used to describe an incident and 134 dierent handled or manipulated objects were recorded (Appendix A). As can be seen from Table 4 a term referring either to the accident site or to the activity was reported in nearly 90% of the accident reports, while a term or phrase M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 47 Table 4 Frequency analysis of the vocabulary reported for the dierent types of data Type of data Accident site Activity Handling objects Incidenta a No. of dierent terms reported Percentage of reports with at least one term Percentage of reports with more than one term 41 84 135 100 86 90 76 63 34 23 20 28 Can be short locutions. describing the handled object or an incident was used in 76 and 63% of the cases, respectively. Moreover, for all four data categories, 20±34% of the reports contained at least two terms or phrases referring to the same category. In the case of the `site' and `activity', the second term generally served to specify the ®rst (e.g. load truck and lift box), whereas for the `incident' variable, the second phrase was often the result of the ®rst (e.g. the ¯oor was icy and I slipped; the plate was jammed and I forced). In the case of the handled objects, one object was usually reported except for the following two situations. The ®rst describes situations in which the object and equipment were handled together (e.g. drum handled with drum buggy). The second, and most frequent, describes situations in which a dual term was used to describe the object, usually one term referring to the object, the other to a characteristic (e.g. bundle of steel, crate of glass). 3.2.2. Grouping of the reported terms Obvious grouping strategies were ®rst identi®ed. Certain terms, English and French synonyms, represented between 5 and 10% of the terms reported for each data category (e.g. dock and quai, warehouse and entrepoÃt, bundle and paquet, crate and caisse, drum and baril). Some terms were also clearly used interchangeably (e.g. dock and warehouse, reel and rim, drum and 45 gallons). Terms which were merely a speci®c of another term, such as dierent models of motorized handling equipment or trailers (e.g. van, trailer, semi-trailer), were also considered as equivalents. Therefore, even though 41 `work site' terms were reported, they could be easily be regrouped into a few dierent sites as shown in the Appendix AÐfor instance, the dock, dock-van junction, van, or motorized handling equipment. Moreover, since few could be considered as obviously equivalent in the case of an activity or incident, other grouping strategies were explored. In the case of the activity it was dicult to identify one logical basis for sorting, the major reason being that the terms changed meaning depending on the object handled. Despite this polysemantism (which will be developed later), some grouping was attempted. A ®rst grouping was based on whether the term referred to the goal/ objective to be performed (e.g. pile up, load, unload, move, replace, reload) or to a speci®c action or object displacement (e.g. push, pull, turn, tilt, roll, hold). A second attempt was based on whether or not the handled objects maintained continuous 48 M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 contact with the surface during the handling process, (e.g. drag, slide, pivot, roll, pull vs throw, carry, deposit, pile). Other possibilities explored were based on the phase of the handling process (e.g. pick-up, transport/transfer, deposit) or the spatial context of handling (e.g. `to gather' refers to the ground while `hang' refers to a certain height). None of these groupings allowed for a classi®cation of the majority of the activity terms. In the case of the handled objects an initial simple grouping distinguished between the nature of the objects, i.e. whether they were goods, structural elements, manual equipment, or parts of motorized equipment. It was also easy to dierentiate the 105 terms describing goods as boxes and objects other than boxes. As the latter represented most of the objects reported, possible grouping strategies were tested to identify sub-classes of goods. Strategies were based on grouping the goods based on their physical shapes/characteristics, (e.g. long, cylindrical, voluminous, bundle, stackable objects). However, it was dicult to group the terms under these sub-classes since many objects possessed more than one of the de®ned qualities. For example, a bundle of pipes may be grouped according to its characteristic `bundle' or its format `long'. In the case of incidents a ®rst trial consisted in classing the terms into two main categories depending on whether they described external/environmental factors or activity related factors. As shown in Table 5, it was possible to de®ne a typology that allowed for some logical classi®cation. The terms describing the incidents related to the working environment could be grouped according to some typology for dysfunctioning, i.e. objects/equipment that were either defective or not as they were planned to be, unusual or atypical ¯oor conditions, dicult access conditions and instabilities of either the handling equipment or the merchandise (e.g. a buggy that moved while unloading goods, a box that fell). The incidents related to the activity, as numerous as the previous ones, could also be regrouped into dierent classes according to the nature of the incident, such as a problem controlling an item of merchandise, loss of balance, an unusual eort, an error or omission by another person. Most data could be ®tted into one of these classes. The question as to whether to group the terms under activity or environment categories is debatable and will be dealt with in the discussion. As previously mentioned, two incidents may be reported, one often being the consequence of the other (e.g. the van advanced, the worker lost balance). This grouping strategy does not dierentiate between cause and consequence. However, the description usually makes it possible to identify the relationship between the reported incidents. This data was implicit. 3.2.3. Contextual signi®cation of the terms recorded As stated in the methodology, the handled objects were ®rst grouped into ®ve large classes to verify the importance of the exclusivity of the terms. As can be seen from Table 6, 23 of the 61 handling terms were used exclusively in association with one class. For example, 15 terms were used only when referring to a single object (e.g. to roll, to pivot, to pull). However, the 23 terms were found in less than 5% of all the accident descriptions, non-speci®c terms such as `loading, unloading' being used the most often. Loading or unloading was used to refer to the overall task M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 49 Table 5 Groupings of the incident related terms and phrases Related to working environment Related to the activity 1. Defectiveness (a) Breakages of merchandise (e.g. wrapping broke, bag pierced) 1. Diculty in controlling merchandise (a) Explicitly reported (e.g. I dropped a pole) (b) Defective or poor functioning of equipment (e.g. plate was jammed) (b) Implicitly reported (e.g. furniture piece tilted during manouvre) (c) Protruding object (e.g. piece of wood with nail on it) 2. Floor conditions presenting diculties 2. Errors or ommisions performed by/with another person (a) Related to coordination (e.g. partner lifted too fast) (a) Slippery ¯oor (e.g. oil on the ¯oor) (b) Objects on ¯oor (e.g. stepped on piece of wood) (b) With contact (e.g. X backed up with lift truck and hit me) (c) Holes in the ¯oor (e.g. hole on the ¯oor of the van) (c) Without contact (e.g. X forgot to place junction plate) (d) Instable surface (e.g. the van advanced) 3. Instabilites (a) Horizontal (e.g. buggy moved) 3. Unusual eort (a) Explicitely reported (e.g. forcing to lift merchandise) (b) Vertical (e.g. a box fell) 4. Working space presenting particular diculties (a) Lack of space (e.g. I was squeezed between two desks while unloading) 4. Loss of balance (b) Dicult to access goods (e.g. box stuck between two rolls) (b) As a consequence of (e.g. placed foot in hole and fell) (a) Without apparent reason or explanation Table 6 Exclusivity of the associations in function of the nature of the object handled Classe of complements No. of terms associated to complement No. of terms associated exclusively Group of various goods Group/series of the same object One given object Equipment handled Other 5 22 47 11 11 2 2 15 2 2 Total 96 23 50 M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 (e.g. I was loading a van), to a series of the same object (e.g. loading tires) or to a speci®c object (e.g. loading a drum). In a second attempt, the handled objects were grouped according to some of their physical characteristics (e.g. stackable, heavy or voluminous, cylindrical in form). Only 12 terms were found referring to a speci®c characteristic (e.g. pile up, unpile, manoeuvre, roll), representing approximately 10% of the total accidents. Therefore, few terms appeared to have a speci®c meaning. As the terms `lift' and `slip' were frequently reported, their utilization was more speci®cally analyzed. This analysis showed that both terms may have had several meanings, as indicated in Table 7. Slip could refer to a loss of balance, to a displacement of an object or to a handling action. (In French, `slip and slide' are both translated by the same word, `glisser'). The term lift, even though it always referred to a handling process, could describe very dierent types of handling actions (e.g. to lift a drum, which mainly means to tilt or roll the drum; to lift a box). Therefore, in both cases, an understanding of the context was needed to correctly interpret the meaning of the word. 3.2.4. Implicit nature of the data Data on the accident site, the activity and the incident could be deduced from the accident descriptions in 30, 7 and 25% of the cases, respectively. With respect to the work site, it was generally the reported activity that made it possible to deduce Table 7 Possible meanings of the terms `lift' and `slip' Context of usage Meaning Slip I slipped on the deck a loss of balance The box slipped from above the box moved and eventually fell The buggy slipped describes that the buggy moved horizontally when in it was expected to remain ®xed In laying down the box, it slipped implicitly implies a problem in the deposit phase Slip the boxa means to slide the box by pushing or pulling it Lift Lift a box implies a complete displacement of an object usually performed in three phases: pick-up, transfer and deposit Lift a bundle of pipes may refer to an object that is partially lifted and then pulled Lift a barrel with a buggy describes the nature of the load transfer, which consists of inserting the buggy underneath the barrel and then to tilt it Lift a junction plate to pivot the plate; this is accomplished by pulling the plate, which is fastened to a hinge, towards oneself a In French `glisser la boite' means to slip or slide the box. M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 51 where the worker was located. For example, `lifting the plate' implies that the worker was close to the exterior door of the dock. As for the `activity', sometimes an inference could be made from the object handled. For example, `lift a barrel' actually means to tilt it. In the case of incidents it was generally the data from both the activity and the event causing the injury that permitted an inference. The loss of control while handling objects is a typical example of such an inference. For instance, ``while handling a gas cylinder with a trolley, my hand was caught in between the trolley and the cylinder'', refers implicitly to a problem controlling the handled object. As a fact, a direct reference to ``I had a control problem'' was rare. Less frequently, it was the meaning of the reported incident that was implicit. For example, ``I fell between the dock and the van'' refers implicitly to the fact that the plate joining the dock to the van was not in place. 3.3. Discussion Substantially more data on the accident circumstances were identi®ed in this analysis compared to that set out in Part 1. Assuming that these accident reports were not qualitatively atypical, it can be acknowledged that a certain loss of data occurs during the analysis of accident descriptions. This is particularly the case for data pertaining to an activity or an incident. The activity at the time of the accident was considered by less than half the studies, which generally retained less than ®ve descriptors. In this database, 84 dierent terms, 61 of which referred to manual handling alone were recorded. These terms appeared to be highly informative. However, no simple grouping strategy was identi®ed, which may partly explain why these data are generally treated super®cially by the accident studies. On the other hand, the analysis of manual handling cannot be con®ned to only three or four generic terms without excessively oversimplifying the information reported. If both the handling object and the activity are considered, more data could be extracted from these accident reports. In respect of incidents, data were practically left untreated by most of the studies, even though the ANSI standard guidelines propose a category for its treatment. Incidents were frequently reported in these descriptions. We may, therefore, assume that the underreporting in literature of data on incidents may be owing to the fact that they are simply left untreated. However, contrary to the other data, the `accident site' data was less informative here: they were highly correlated to the activity and most of the handling of goods was performed in the van. For this database, a worksite variable seemed to be useless. This vocabulary analysis revealed several classi®cation problems which are surprisingly rarely reported or discussed in the literature. The ®rst problem was tied to the considerable variability of vocabulary, which leads to the need to develop sorting strategies in order to make the data more treatable. In some cases, the grouping was easy: some terms were either synonyms, used indierently or simply speci®cs of another term. In other cases, such as concerning incidents, a reasonably coherent typology could be developed which would make it possible to class most of the data. The structure set out in this paper was one of many possibilities. The criteria used to 52 M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 dierentiate activity and environmental incidents is debatable, since both are often inter-related. For example, a four-wheeled trolley that moves while a worker unloads merchandise, was classi®ed as environmental. This may, however, be due to several factors: the driveway slope causing the van to be unlevel, the manner in which the handler positioned his trolley, as well as the handler's direct actions on the trolley while handling the goods. Also, the determination of what is considered an incident is not clear, since the accident itself is dysfunctional. In a sense, a loss of balance without speci®c circumstances may be considered no more of an incident than an overexertion injury. In this paper, the inclusion criteria for incidents included any unusual event prior to the mechanism of injury. A second problem was one of choice when several terms are reported on the same type data. In some cases the choice was simple because the second term was only a speci®c of the ®rst. This was less obvious when the two terms belonged to dierent classes, such as `a bundle of pipes'. Selecting one term instead of another may result in an unfortunate loss of data. For example, with reference to the incidents, choosing to retain only the initial event in a previous study (Lortie et al., 1996) resulted in a loss of data regarding loss of balance, as also reported by Manning et al. (1988) and Strandberg (1985). A third classi®cation problem was related to the fact that the signi®cation of data may be implicit or contextual. A ®rst implication is that the literal or automatic classi®cation of the terms may lead to a signi®cant bias, as seen for the term lift. The European task force (Clarke et al., 1992) reports examples where an automatic classi®cation of terms leads to misinformation on the accident process. Also, such loss results in the systematic exclusion of certain data which, for whatever reason, are rarely literally reported. The loss of control while handling is an example of such systematic exclusion. In these cases, workers report circumstances describing a problem with the control of the merchandise without actually mentioning a control diculty. Accident studies rarely report this issue, while on-site incident observations (Lortie and Pelletier, 1996) and interviews with dock handlers (Authier and Lortie, 1993) con®rm its importance. However, it is clear that analyzing implicit information is more dicult than classifying `readily ascertainable facts'. The former is probably more subjective and requires a minimum amount of knowledge of the work performed. On the other hand, the consequences of not considering these data outnumber those associated with their treatment. 4. Conclusion The data present in accident reports are variable and complex. Most of the accident data studies reviewed analyzed only a portion of the data available, most often barely treating data pertaining to incidents and activities. The analysis of an accident database revealed several important classi®cation problemsÐan issue seldom addressed in the literature. Despite the limits and diculties associated with the analysis of ®rst accident reports, this material remains a mine of information which deserves closer attention. M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 53 Acknowledgements A scholarship was awarded by the Institut de Recherche en Sante et SeÂcurite du Travail du QueÂbec to conduct this study. References Aghazadeh, F., Mital, A., 1987. Injuries due to handtools. Applied Ergonomics 18(4), 273±278. Al-Arrayed, A., Hamza, A., 1995. Occupational injuries in Bahrain. Occupational Medicine 45(5), 231± 233. ANSI, 1969. American National Standards Method of Recording Basic Facts Relating to the Nature and Occurrence of Work Injuries. ANSI: Z16.2-1962/R1969, New York. Andersson, R., Kemmlert, K., Kilbom, AÊ., 1990. Etiological dierences between accidental and nonaccidental occupational overexertion injuries. Journal of Occupational Accidents 12, 177±186. Authier, M., Lortie, M., 1993. Assessment of factors considered to be important in handling tasks by expert handlers. International Journal of Industrial Ergonomics 11, 331±340. Blank, V.L.G., Andersson, R., LindeÂn, A., Nilsson, B.-C., 1995. Hidden accident rates and patterns in the Swedish mining industry due to involvement of contractor workers. Safety Science 21, 23. Bobick, T.G., Myers, J.R., 1994. Agriculture-related sprain and strain injuries, 1985±1987. International Journal of Industrial Ergonomics 14, 223±232. Brewer, R.D., Oleske, D.M., Hahn, J., Leibold, M., 1990. A model for occupational injury surveillance by occupational health centers. Journal of Occupational Medicine 32(8), 698±702. Brun, J.-P., 1990. De l'analyse des donneÂes aÁ une deÂmarche d'expression des travailleurs: une approche integreÂe de la preÂvention. Journal of Occupational Accidents 12, 307±319. Clarke, R.K., Glendon, I. 1988. How Occupational Accidents and Diseases are Reported in the European Community. Oce for Ocial Publications of the European Communities, Luxembourg. Clarke, R.C, Darby, K.V., King, K.L., King, R.P., Otto, E.H., 1992. MeÂthodologie pour l'harmonisation des statistiques europeÂennes sur les accidents de travail, seÂrie E: MeÂthodes, Project manager: Khalili-Araghi, H.. Oce des publications ocielles des CommunauteÂs europeÂennes, Luxembourg. CSST, 1995. Rapport annuel d'activiteÂ. Commission de la Sante et de la SeÂcurite du Travail, MontreÂal. Conroy, C., 1989. Work-related injuries in the meatpacking industry. Journal of Safety Research 20(2), 47±53. Cross, J., Walters, M., 1994. Vibration and jarring as a cause of back injury in the NSW coal mining industry. Safety Science 17, 269±274. Doyle, Y., Conroy, R., 1988. A one year survey of accidents on Irish farms and their medical outcome. Journal of Occupational Accidents 10, 199±208. Engkist, I.-L., Hagberg, M., LindeÂn, A., Malker, B., 1992. Over-exertion back accidents among nurses' aides in Sweden. Safety Science 15, 97±108. Harker, C., Matheson, A.B., Ross, J.A.S., Seaton, A., 1991. Accidents in the workplace. Occupational Medicine 41, 73±76. Harvey, M.D., 1984. Theories of Accident Causation. Alberta Workers' Health, Safety and Compensation, Alberta, Canada. Inancsi, W., Guidotti, T.L., 1987. Occupation-related burns: ®ve-year experience of an urban burn center. Journal of Occupational Medicine 29(9), 730±733. Kisner, S.M., Fosbroke, D.E., 1994. Injury hazards in the construction industry. Journal of Occupational Medicine 36(2), 137±143. KjelleÁn, U., Larsson, T.J., 1981. Investigating accidents and reducing risksÐa dynamic approach. Journal of Occupational Accidents 3, 129±140. La¯amme, L., 1988. ModeÁles et meÂthodes d'analyses de l'accident du travail: de l'organisation du travail aux strateÂgies de preÂvention. SyGeSa LteÂe, MontreÂal. 54 M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 La¯amme, L., DoÈoÈs, M., BackstroÈm, T., 1991. Identifying accident patterns using the FAC and HAC: their application to accidents at the engine workshops of an automobile and truck factory. Safety Science 14, 13±33. Larsson, T.J., 1990a. Severe hand injuries among Swedish farmers. Journal of Occupational Accidents 12, 295±306. Larsson, T.J., 1990b. Permanent occupational disabilities among Swedish metal workers in 1986. Journal of Occupational Accidents 12, 283±293. Larsson, T.J. 1990c. Accident Information and Priorities for Injury Prevention (Ipso Factum 21) Institute for Human Safety and Accident Research (Ipso), Stockholm. Larsson, T.J., Rechnitzer, G., 1994. Forklift trucksÐanalysis of severe and fatal occupational injuries, critical incidents and priorities for prevention. Safety Science 17, 275±289. Leigh, J., Mulder, H.B., Want, G.V., Farnsworth, N.P., Morgan, G.G., 1990. Personal and environmental factors in coal mining accidents. Journal of Occupational Accidents 13, 233±250. Leigh, J., Mulder, H.B., Want, G.V., Farnsworth, N.P., Morgan, G.G., 1991. Sprain/strain back injuries in New South Wales underground coal mining. Safety Science 14, 35±42. Lortie, M., 1987. Analyse comparative des accidents deÂclareÂs par des preÂposeÂs hommes et femmes d'un hoÃpital geÂriatrique. Journal of Occupational Accidents 9, 59±81. Lortie, M., Pelletier, R., 1996. Incidents in manual handling. Safety Science 21, 223±327. Lortie, M., Lamonde, F., Collinge, C., Tellier, C., 1996. Analyse des accidents associeÂs au travail de manutention sur les quais dans le secteur transport. Travail Humain 59(2), 187±205. Ma, W.S., Wang, M.J., Chou, F.S., 1991. Evaluating the mechanical injury problem in the wood-bamboo furniture manufacturing industry. International Journal of Industrial Ergonomics 7, 347±355. Manning, D.P., Ayers, I., Jones, C., Bruce, M., Cohen, K., 1988. The incidence of underfoot accidents during 1985 in a working population of 10 000 Meyerside people. Journal of Occupational Accidents 10, 121±130. Miller, K., 1992. Causes of accidents with post-drivers and their remedies. Applied Ergonomics 23(2), 101±104. Niskanen, T., Lauttalammi, J., 1989. Accidents in materials handling at building construction sites. Journal of Occupational Accidents 11, 1±17. Norrish, A.E., Cryer, C.P., 1990. Work related injury in New Zealand commercial ®shermen. British Journal of Industrial Medicine 47, 726±732. O'Connor, T.A., Gordon, J.E., Barnett, M., 1993. Agricultural injury surveillance using a state injury registry. Journal of Safety Research 24(3), 155±166. Oleske, D.M., Hahn, J.J., Leibold, M., 1992. Work-related injuries to the foot. Journal of Occupational Medicine 34(6), 650±655. Oleske, D.M., Brewer, R.D., Doan, P., Hahn, J.J., 1989. An epidemiologic evaluation of the injury experience of a cohort of automotive parts workers: a model for surveillance in small industries. Journal of Occupational Accidents 10, 239±253. Olson, D.K., Gerberich, S.G., 1986. Traumatic amputations in the workplace. Journal of Occupational Medicine 28(7), 480±485. Pines, A., Halfon, S.-T., Prior, R., 1987. Occupational accidents in the construction industry of Israel. Journal of Occupational Accidents 9, 225±243. Reesal, M.R., Dufresne, R.M., Suggett, D., Alleyne, B.C., 1989. Welder eye injuries. Journal of Occupational Medicine 31(12), 1003±1006. Sinks, T., Mathias, T., Halperin, W., Timbrook, C., Newman, S., 1987. Surveillance of work-related cold injuries using workers' compensation claims. Journal of Occupational Medicine 29(6), 504±509. Stout-Wiegand, N., 1987. Characteristics of work-related injuries involving forklift trucks. Journal of Safety Research 18(4), 179±190. Strandberg, L., 1985. The eect of conditions underfoot on falling and overexertion accidents. Ergonomics 28(1), 131±147. Sullivan, C.B., Shimizu, K.T., 1988. Epidemiological studies of work-related injuries among law enforcement personnel. Journal of the Society for Occupational Medicine 38(1±2), 33±40. ToÈrner, M., Karlsson, R., Sñthre, H., Kadefors, R., 1995. Analysis of serious occupational accidents in Swedish ®shery. Safety Science 21, 93±111. M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 55 Waller, J.A., Payne, S.R., Skelly, J.M., 1989. Injuries to carpenters. Journal of Occupational Medicine 31(8), 687±692. Yassi, A., Khokhar, J., Tate, R., Copper, J., Snow, C., Vallentyne, S., 1995. The epidemiology of back injuries in nurses at a large Canadian tertiary care hospital, implications for prevention. Occupational Medicine 45(4), 215±220. 5. Appendix A 5.1. Accident site terms (n=41) Van Dock Dock-van junction Motorized handling Other equipment camion unite coÃte de la vanne mur de l'unite vanne pop remorque semi-remorque trailer entreÂe camion porte (numeÂro) nom du quai entrepoÃt dock quai quai de chargement dock oce oce de la plateforme espace camion-plate jct dock-van jct quai-camion plaque plateforme plate poteau sortie remorque sortie quai chariot chariot eÂleÂctrique chariot eÂleÂvateur lift lift truck monte-charge tow motor transporteur fourches du chariot escalier cour debout sur: marchandise palettes skid 5.2. Activity terms (n=84) Handling terms (n=61) accrocher arreÃter attrapper basculer bloquer buer changer charger chercher coucher (buggy) deÂbarquer deÂcharger deÂfaire manipuler recharger deÂpiler manoeuvrer reculer deÂposer mettre redescendre deÂplacer oÃter relever descendre passer remplir donner piler rentrer embarquer placer renverser empiler poser replacer enlever pousser retenir faire tomber march. prendre retirer installer (e.g. adapteur) ramasser retourner lever recevoir transport rouler sortir soulever tasser tenir tirer tourner transfeÂrer transporter trier utiliser (2 roues) veÂri®er (march.) vider virer 56 M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 Displacement terms (n=14) Movement terms (n=3) deÂbarquer dea (se) deÂplacera descendrea embarquera entrera redescendrea rentrera sauter sortir dea a grimpera marcher montera passer dansa reculer (se)pencher (se) relevera (se) retournera These terms were or could have been used as handling terms. Other activities (n=16) Conduire (chariot) OpeÂrer (chariot) stripper (vanne) chaõÃner (voyage) fermer ouvrir Casser (strapping) DeÂcrocher (chaõÃne) attacher (binders) couper travailler sur reÂparer deÂplier eÂtendre (brin de scie) laver attendre 5.2.1. Objects handled (n=134) Goods handled (n=105) carton canots boite bicyclette boõÃte de poisson de textile gallon de tissus baril de cigarettes drum plate d'acier tapis meuble plate de meÂtal teÂleÂvision plateforme basculante tuyau fauteuil morceau d'aluminum de plastique bureau de fer tubes de plastic classeur pesant longueurs de meÂtal ®lieÁre caisse de kleenex contenant du metal grosse boõÃte gallon de 45 lbs moteur boiÃte contenant acide chaudieÁre soudeuse de bicyclette poubelle castings vaisselles pieÁce de machine bundle transformer boõÃte de bois de bois bloc de fonte crate de fer essieux crate d'arborite de meÂtal de meÂtal d'aluminum bonbonne de gaz de vitres de chaõÃnes cylindre de parebrises planches de ceÁdre de bois detuyaux reel de machine aÁ boule paquet de chaõÃnes reel de crop de pelles rim caisse de bois pieÁces de fer deÂvidoire de coton pieux de fer rouleau de fer de peinture barres de fer de papier de vitres de preÂlart passerelles d'acier boite aÁ lettres petit core fort sacs de lait chauerette de peat moss cage de poudre tombe de sulfate poches de toxique skid prod. chimiques palette pneus viande poeÃle frigidaire laveuse cadre de porte panneau planche/plywood feuilles/formica bob a porte cric de bois marchadise M. Lortie, P. Rizzo/Safety Science 31 (1999) 31±57 Manual handling equipment (n=12) Part of an equipment (n=6) Structural elements (n=6) 4 roues dock cart ¯at pin pin de chariot pin aÁ tapis poÃle aÁ tapis l'adapteur fourche plate plaque plate de quai rampe rampe de chargement 2 roues buggy chariot diable hand truck chariot aÁ plateau truck aÁ drum drum buggy trans-palette porte de vanne Other (n=5) strapping toile bearings binders chaine 57