Article Text

Download PDFPDF

The utility of injury narratives for prioritising future prevention activities in small and medium enterprises in rural Vietnam
  1. Helen Marucci-Wellman1,
  2. Joanna L Willetts1,
  3. Nguyen Bich Diep2,
  4. Ta Thi Tuyet Binh2
  1. 1Center for Injury Epidemiology, Liberty Mutual Research Institute for Safety, Hopkinton, Massachusetts, USA
  2. 2The National Institute for Occupational and Environmental Health, Hanoi, Vietnam
  1. Correspondence to Dr Helen Marucci-Wellman, Liberty Mutual Research Institute for Safety, 71 Frankland Road, Hopkinton, MA 01748, USA; helen.wellman{at}


Objectives Injuries are a leading cause of work-related disability and death in rapidly developing countries such as Vietnam. The authors' objective was to demonstrate the utility of detailed injury narratives, derived from a household survey, in providing information on the determinants of work-related injuries to inform potential intervention targets.

Methods In a cross-sectional survey administered to 2615 households of a rapidly developing community of Vietnam where many workers engage in both agriculture and industrial work, the authors collected information about self-reported work-related injuries, annual hours worked in each industry and narrative text describing the circumstances of each injury. The authors used a customised coding taxonomy to describe injury scenarios.

Results Several intervention themes emerged, including the implementation of machine guarding, the use of cut resistant gloves and safety glasses which would benefit the small- and medium-sized enterprises. Calculation of incidence rates using full-time equivalents, stratified by work group, provided some unexpected observations of the risks of working in agriculture; workers who work in agriculture in addition to another industry are at an increased risk of fatigue or overexertion and other consequences of working too hard in their agricultural activities.

Conclusions A lack of aggregate injury statistics makes it difficult for the owners of small- and medium-sized enterprises to recognise a priori the most effective safety interventions. This analysis of detailed injury narratives with an appropriate taxonomic basis offers the ability to focus on the level of cause, activity and source and may inform the choice of various potential interventions at the workplace or enterprise level.

  • Incidence
  • injury
  • intervention
  • narrative text
  • surveillance
  • work-related
  • public health
  • international
  • methods
  • occupational

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Southeast Asia has become a prominent force in the global market producing much of the world's rice and material goods. Vietnam is one of the fastest growing of the emerging economies, with a potential real dollar growth rate in Gross Domestic Product of almost 10% per year.1 Since the opening of the economy in 1986, many farming villages have organised themselves into working communities involving large numbers of family-owned small-to-medium sized enterprises (SMEs),2 which produce, sell and export traditional crafts and manufactured products. Many are concentrated in a single mono-industry, such as the Ceramic Village, the Silk Village or the Metallurgical Village.3

Historically, as economic development occurs, workers migrate from working in the fields to the factories. Despite the rapidly developing Vietnamese economy, 80% of the 12 million rural households have remained totally or partially dependent on agriculture with as little as 7% abandoning their agriculture work.2 Many work primarily (full time) in SMEs, and also part time in the fields4 ,5 particularly at harvest times. As development occurs, female subjects are assuming more of the daily tasks in agriculture as male persons work for more time in the new industrial jobs.4

With economic development, injuries have grown to be the leading cause of mortality reported by hospitals.6 With the present lack of an effective national injury registration system, any estimate of the scale of occupational injuries is based solely on employer reporting and is believed to be incomplete. The lack of aggregate injury statistics makes it difficult for the owners of local SMEs to recognise the most pressing need for effective safety interventions in their enterprise. Given their scarce resources, enterprise owners need such information to prioritise interventions and address hazards in the most effective way.

Work-related injury incidence is often reported by stratifying and ranking by age, gender, industry and occupation.7 This approach, while useful in identifying the most hazardous industries, provides little help in selecting the best prevention strategies to improve safety conditions. Furthermore, it does not allow for a comprehensive understanding of workers' risks, given that many perform work in multiple industries throughout the year. We previously reported an effect of such a pattern: working in agriculture, even for short durations, while primarily working in another industry, substantially increases the risk of injury in both types of work.5

The objective of this study was to demonstrate the utility of detailed injury narratives, analysed by a novel, customised coding taxonomy, in providing information on the most important factors leading to work-related injuries in Xuan Tien. By using the workers' occupation and this additional narrative information, our intention was to: (1) learn specifically what people were doing and how they got hurt in their primary work and (2) identify additional factors from working primarily (full time) in SMEs and also part time in agriculture. Such analyses can facilitate better matching of resources to injury prevention interventions in this rapidly developing community.

Materials and methods

Study site

This cross-sectional study was conducted in Xuan Tien Commune of Vietnam. It was selected because it includes many small family-owned businesses and is undergoing rapid economic development where work life continues to revolve around the growing and harvesting of rice.

Survey design and administration

Twenty-one community residents, mainly healthcare volunteer (HV) para-professionals, administered a cross-sectional survey between April and December 2005 to all households in the Xuan Tien Commune (2647 households); 32 households did not respond (99% participation rate).8 ,9 We provided the HVs with a 2-day training course covering human study subject protection and administration of the questionnaire. The survey was administered to the head of household of each residence who reported on all members of the household.

The survey collected demographics for each resident, including age, gender and details of their work for the previous year: industry, occupation, and work duration (average hours/day, average days/week and months worked in the last year) for every job. Five thousand four hundred eighty-five residents were reported to have at least one job in the last year.

The head of household was asked to report whether each working household member had been injured in the last year using this question: ‘Was <name> hurt in the past year enough to need care or to disrupt their normal activities for at least one day or longer?' A decentering translation technique that gives equal weight to both languages was used to translate the English questionnaire into Vietnamese and then back to English to confirm both accuracy and cultural appropriateness.10 (Note: We used the Vietnamese equivalent of the English term ‘hurt’ instead of ‘injury’ because ‘injury’ in Vietnamese related to a very specific type of injury.) Data were gathered for the entire previous year to improve capture of severe, but rare events (eg, amputations) as well as to accommodate known cyclical variations as work activities and hazards vary considerably due to the annual agricultural cycles. Further details of the study site, survey design and administration have been described previously.5 ,8 ,9

Classification of industry and occupation at the time of injury

A modified International Labour Organisation industry classification scheme was used to classify the industry and occupation of each job reported, and the job at the time of each reported injury.11 The industry where each worker worked the most hours the previous year was designated as their primary job industry. We created three groups of workers based on whether their primary job was in agriculture or another industry and whether they had any exposure to agricultural work: Group I: only or primarily agriculture work; Group II: primary job in non-agriculture industry but some work in agriculture; and Group III: primary job in non-agricultural industry (no work in agriculture).5

Occupational injuries

Inclusion/exclusion criteria

Injury inclusion criteria and work-relatedness have been described previously.5 Briefly, all reported injuries were classified according to International Classification of Disease (ICD-10)12 nature of injury by a certified ICD-10 coder. Injury cases that were identified as injury or poisoning (S or T) were included; cases that were musculoskeletal (ICD-10 code of M or R) were independently reviewed for exclusion (n=160 reviewed, n=40 excluded). The main criteria for exclusion were if no external attribution could be determined or if the case indicated pain had persisted for a long period of time. Work-relatedness was identified by an independent reviewer based on survey information and responses to direct questions about the injury occurring while working or commuting.

Narrative coding: activity, cause and source leading to injury

For each reported injury, the head of household was asked to describe the type and circumstances of the injury; the responses were recorded as narrative text.

We developed a customised, narrative text coding scheme and protocol to provide a clear description of the scenarios leading to occupational injuries in this community. Our coding approach, while in line with previous attempts to investigate occupational injury scenarios,13 ,14 was tailored to this unique environment. We coded each injury narrative into three categories: (1) Activity: the specific activity that the worker was performing at the time of the injury, for example, carrying, welding and using hammer; (2) Cause: the manner in which the injury occurred, for example, slipped, cut or struck by; and (3) Source: the object, substance or exposure that directly caused the injury, for example, a machine, heavy object and splinter.

For Cause we used the ICD-10 E-codes.12 The taxonomy for the Activity code was designed specifically to encompass the work being done in the commune of Xuan Tien. For Source we used the Bureau of Labour and Statistics Occupational Injury and Illness Classification System, adding custom source codes for objects not found in the Occupational Injury and Illness Classification System, such as bamboo.15

Two typical narratives and resultant coding taxonomy follow (translated from Vietnamese to English):

  • Example 1: While carrying iron sheets, his hand got cut for 6 cm and bloodied.

  • Activity: Carrying heavy materials and/or objects; Cause: Striking against or struck by other objects; and Source: Metal sheets, ingots and bars.

  • Example 2: Foreign body flew into eyes when lathing, the eyes were burned by lathe shavings.

  • Activity: Operating lathe; Cause: Foreign body entering into/through eye or natural orifice; and Source: Metal chips, particles and splinters.

We trained two experienced coders on the rules of each coding scheme. During the coding process, the coders met regularly with researchers to reach consensus on problematic narratives. Reliability testing on a sample of narratives showed the coders agreed on more than 80% of the codes at the highest level of detail for each of the coding schemes.


All survey information was analysed using Statistical Analysis Software V.9.2 (SAS Institute Inc.). We first presented descriptive statistics for Occupation, Activity, Cause and Source leading to injury for all workers who reported an injury in the last year in order to prioritise factors associated with injuries in Xuan Tien and provide guidance on intervention strategies for the bulk of work being done in the many high risk SMEs. Next, we stratified our results by gender because men and women clearly had different work roles.5 Finally, we stratified by the three industry exposure work groups similarly to determine if there were any further differences in the factors leading to injury for those who were exposed only to agricultural work, those who were exposed to both agricultural work and work in another industry, and those who were not exposed to agricultural work in the previous year.

Incidence rates (IRs) were computed based on the number of full time equivalents (FTEs), or 2000 work hours over a 1-year period. CIs of IRs were computed assuming a Poisson distribution of reported counts; rate ratios (RR) assumed normality when log transformed. Differences in proportions for each activity between male and female subjects were tested with a two-sided z test at the 0.1 and 0.05 levels of significance.

Human study subject protection

The household survey and protocol were approved by the appropriate institutional review boards in the USA and Vietnam.


Work type and factors associated with injuries

Five thousand four hundred and eighty-five residents of Xuan Tien reported working in 8479 jobs in the previous year, accumulating 5758 FTEs. Agriculture work was usually part time while craft and related trade work were often greater than full time (>1 FTE per job). Female subjects worked more in agricultural occupations than male subjects (722 vs 333 FTEs) and male subjects worked more in craft and related trade work than female subjects (1917 vs 1256 FTEs).

We identified 1169 injuries as work-related, resulting in an annual, work-related injury IR of 21.3 per 100 workers or 20.3 per 100 FTEs. Overall, male subjects were injured more often than female subjects (RR=2.4, 95% CI 2.1 to 2.8, table 1). In agriculture, the RR was even higher (RR=3.1, 95% CI 2.3 to 4.1) with male subjects working in animal farming having an IR six times the female subjects' rate (29.2 per 100 FTE male subjects, 4.9 per 100 FTE female subjects (95% CI 2.0 to 7.0)). (Notable also was that male subjects working in rice farming had a very high IR (45 per 100 FTE, second highest IR for males by specific occupation.))

Table 1

Incidence of self-reported work-related injury in the last year by gender and occupation (sub-categories included if the number of FTEs >10 and injuries >5)

The highest IR of injury for male subjects by specific occupation was in the brewing of alcohol and wine (seven injuries, 63 per 100 FTE), although few male subjects worked in this occupation (n=13). The third highest IR for male subjects at 39 per 100 FTE was in metal machinery crafts and related trade work, perhaps the most significant occupation because it also had the highest count of injuries (n=404).

The highest IR of injury in a specific occupation for female subjects was the manufacture of paper/bamboo fans (40 per 100 FTE) with brewing of alcohol and wine second (26 per 100 FTE) and brass bell or foundry workers third (23 per 100 FTE). Although many female subjects worked in agriculture, there were only 83 injuries (IR 12 per 100 FTE). This rate was lower than the rate for female subjects in all of the craft and related trade occupations except making rice vermicelli and sheets (seven per 100 FTE).

Gender and specific activities, causes and sources of injury

Using or operating tools or machinery

Operating powered machinery resulted in 28% of work-related injuries for male subjects (n=231, table 2) while only 18 (5%) of female subjects were injured by this activity. Male injuries included using or operating welders (n=57, 7%), lathes (n=44, 5%) and saw/cutting machines (n=29, 4%). For injuries occurring while male subjects were using or operating powered machinery (n=231), 30% were caused by the machines (n=70); 17% were caused by foreign bodies due to chips, particles and splinters (n=40); 7% were caused from being struck by/against or caught between metal sheets (n=18); and 7% were caused by fire/flames due to welding or soldering fumes (n=17, table 3).

Table 2

Most common male/female activities at the time of the injury self-reported injuries in the last year (shown in descending count order of major coding taxonomy categories and within their sub-categories)

Table 3

Common activity–cause–source scenarios for male and female subjects (reported >5 times) for self-reported injuries in the last year

Operating non-powered machinery or hand tools injuries were slightly more common for female than male subjects (17% vs 12%, table 2). Injuries while using a knife/saw/scissors/axe were more frequent for female compared with male subjects (16% vs 6%), while male subjects were generally injured using other hand tools (eg, hammer (n=28), chisel (n=10)).

Materials or object handling

Thirty percent of all injuries to female subjects (n=102) were from materials or object handling, most often involving carrying materials (n=68, 20%); 24% of male injuries were from materials or object handling of which 15% (n=121) occurred while carrying materials (table 2). Overexertion injuries while performing materials or object handling tasks were numerous for both genders (39% and 40%, respectively, table 3).

The most frequent source and cause injury scenario for male subjects who were performing materials or object handling was being struck by/against the metal sheets while carrying an object or material (n=36, 30% of injuries, table 3). This same scenario accounted for only 13% (n=8) of female injuries during this activity. A similar analysis found, instead, that overexertion from carrying a container caused many (n=20, 29%) of the injuries to female subjects.

Physical activities NEC and overexertion

Approximately the same proportion of male and female injuries occurred during physical activities not elsewhere classified (NEC) (see table 3). Many of these injuries happened while walking, wading, shuffling or stepping (n=65 male, n=23 female) and were often caused by falling (female subjects, 9 (40%), male subjects, 22 (34%)) or for male subjects by being cut/pierced by animals or minerals eight times (eg, snail shells) or cut/pierced by chips particles or splinters an additional eight times.

Only female subjects reported overexertion injuries from sitting ‘too long’. ‘Working too hard’ was the reported activity at the time of injury for 8% of all work-related injuries to male subjects and 13% of all work-related injuries to female subjects.

Factors associated with injury by work group

The largest work group consisted of those who spent no time in agricultural work (Work Group III, 42%). There were 35% who worked in agriculture and primarily in a non-agriculture industry (Group II) and 23% of workers who worked only or mostly in agriculture (Group I). We were particularly interested in a comparison of workers who only worked in a non-agriculture industry (Group III) compared with those who worked both in a non-agriculture industry and in agriculture (Group II).

For the largest group (Group III, no agriculture work), the dominant activity at the time of injury for male subjects was using or operating tools or machinery (12.9 per 100 FTE, figure 1A,B). For female subjects it was using or operating tools or machinery and materials or object handling (2.7 per 100 FTE and 2.5 per 100 FTE, respectively, figure 1A,B).

Figure 1

A. Incidence of agriculture versus non-agriculture injuries by industry exposure groups and activity at time of injury: male subjects*. B. Incidence of agriculture versus non-agriculture injuries by industry exposure groups and activity at time of injury: female subjects*. FTE, full time equivalent.

The dominant activity at the time of injury for male subjects in Work Group I (only or mostly agriculture work) was Physical activities NEC (12.6 per 100 FTE, figure 1A,B), and for female subjects was materials or object handling (3.7 per 100 FTE).

The most surprising results were for workers engaged primarily in a non-agriculture industry who also worked in agriculture. In their non-agriculture work, both male and female subjects in Group II experienced high IRs while using or operating tools or machinery or during materials or object handling (similar to the results discussed above for work group III). But the major factor contributing to injury in their agriculture work was unexpected (figure 1A,B). These workers frequently reported working too hard (male subjects 22.6 per 100 FTE and female subjects, 7.5 per 100 FTE). In comparison, this activity was infrequently reported for those only working in agriculture (Group I), those only working in a non-agricultural industry (Group III) or for Work Group II in their non-agriculture work (male subjects, 2.5 per 100 FTE and female subjects, 1.0 per 100 FTE for this activity).

In our previous paper, surveillance data showed that working in agriculture, even for short durations, while primarily working in another industry substantially increased the risk of injury in both types of work.5 The current analysis allowed us to understand that the risk of injury different, and the factors leading to injury in agriculture for those working in both industry and in agriculture were also different than would be expected, particularly the higher rate of injury from overexertion in only their agriculture work.


Using a taxonomy developed to reflect the work culture in Vietnam, we were able to capture, extract and quantify the unique injury scenarios occurring across workplaces of Xuan Tien. By doing so, we could identify potential interventions at the level of the enterprise or workplace. Specifically, by aggregating the data, we provide the owners of the SMEs with an awareness of the actual risk of specific operations. Workplace hazards are highly prevalent in low income countries where resources are limited and equipment may not be maintained. However, owners of small enterprises or family-owned businesses may not be able to realise from their experience the high risk of work-related injury because a single enterprise (with only a few workers) would perhaps only have one significant injury every few years. It is clear that without the data aggregated over many workplaces, such as the data collected in our household survey, enterprise owners would be unaware a priori how to use scarce resources for prevention.

We were able to identify several, physical interventions that would likely reduce the number of injuries in this community. Many male subjects, working in non-agriculture work, were injured while operating machinery where the machinery itself was the source of the injury. This suggests that point of operation guards would reduce injury for certain machines. We further identified welders, lathes, saw cutting machines, planing machines, polisher/grinders, rice threshing machines and powered drills as candidates for this intervention.

Through our approach, we identified the potential benefits of personal protective equipment in specific types of tasks. For example, male and female subjects were both injured frequently by handling corrugated iron sheets in metal working factories and when using knives in the manufacture of paper/bamboo fans where cut resistant gloves would be particularly effective.16 ,17 Many male subjects operating powered machinery were injured from foreign bodies, such as chips, particles and splinters, where eye protection would be valuable. Unfortunately, during our visits, we observed very few workers wearing personal protective equipment such that the integration could have substantial impact. We believe the best approach forward would be to couple findings from surveillance, such as these, with active participatory training programmes, such as the WIND, WISE and WISH programmes in use in Southeast Asia, to arrive at cost effective intervention solutions.18 These types of approaches where the importance of active worker engagement for increased safety in small-scale enterprises is emphasised have been found to be very successful in developing countries as well as in a recent similar adoption in the USA.19

The approach of focusing on the hours worked in each job and the calculation of IRs using FTE, stratified by work group rather than number of jobs, appear more appropriate for the types of employment existing in a developing, but also still very agrarian, community such as Xuan Tien. In particular, this approach provided some unexpected observations of the risks of working secondarily in agriculture. Workers who primarily worked in a non-agricultural industry but continued to work in agriculture were at an increased risk of fatigue or overexertion and other consequences of working too hard in agricultural activities. It may be that these part time agricultural workers are performing the most strenuous tasks and may not be adequately conditioned because this work is atypical of their normal work life. Given the very low level of mechanisation in farming rice, it is unlikely that the burden of working too hard can be reduced. The significance of this category was confirmed by our direct observation: large bundles of harvested rice may be carried for significant distances before being loaded onto a bicycle for delivery to the contract thresher. In our prior research we found that the IR of injury for workers working only a fourth of FTE in agriculture was much higher than for workers who worked in agriculture full time.5 The finding that these workers report that they are working too hard supports this observation.


Our study was limited in that we did not collect the amount of exposure time of a particular activity nor do we have any observational data on how workers divided their time between activities of their different jobs. However, we were able to calculate IRs using the overall time exposed to agriculture and non-agriculture work for the denominator value, assuming that workers from a particular industry or industry exposure group would have similar activity exposures. Despite this limitation, it appears that male and female subjects may be doing very different tasks within the same occupation (eg, farming) which accounts for some of the observed differences in injury IRs by gender.

Additionally, the narratives were all translated from Vietnamese and during that translation we expect some of the meaning or context may have been lost. Activity and Cause could only be coded as closely to the defined categories as the translated descriptions allowed. For example, initially we realised working too hard was not a valid activity, but as this was mentioned so often we decided to capture it as a specific activity, recognising that workers who spent time in other defined activities may also have been working too hard but did not include it in their narratives. Similarly, during the collection of the census data, some occupation descriptions included more detail than others. Welding was often mentioned as an occupation within the metal working trade, but other occupations such as being an assembly worker or machine operator were not. Our results, therefore, can only represent what was described in the narrative and we have no way of evaluating non-response.


Our analysis of detailed injury narratives and an appropriate taxonomic basis appears to offer some advantages over more traditional analyses. In particular, the ability to focus on cause, activity and source may inform interventions at the level of the workplace or enterprise. With detailed and aggregated data, SME owners and workers in developing countries can better identify and intervene on common workplace risks to enhance safety, productivity and quality of life.

We identified the potential need for machine guarding and personal protection for specific machines and processes associated with manufacturing. Our anecdotal experience with enterprise owners in Xuan Tien indicates that when risks are identified, the owners of the many SMEs are prepared to introduce improvements which are within their limited resources, and they are proficient at fabricating workplace modifications.

The cooperation and communication among all levels of government, healthcare and enterprises in Vietnam, as well as the development of a multicultural team of researchers from the USA and Vietnam, enabled rich lessons to be transferred back and forth in data collection, aggregation, translation and analyses. It also added a unique richness to the interpretation of findings on occupational injury surveillance and the prioritisation of prevention activities in this rapidly developing country.

What is already known on the subject

  • Vietnam is one of the fastest growing of the emerging economies where agriculture still plays a central role.

  • Since the opening of the market economy in 1986, many farming villages have organised themselves into working communities involving large numbers of family-owned small-to-medium sized industrial enterprises.

  • With economic development, injuries have grown to be the leading cause of mortality reported by hospitals in Vietnam, yet no local or national statistics exist with regard to work-related injuries.

What this study adds

  • In rural but rapidly developing areas of Vietnam, owners of small and medium enterprises, often comprising family members and their neighbours, are unaware a priori which workplace hazards are the most important to address, because there are so many hazards and so few injuries per enterprise.

  • The use of detailed injury narratives, analysed by a customised coding taxonomy, provided information on the most important factors leading to work-related injuries and enabled the identification of several, physical interventions that would likely reduce the number of injuries in this type of community.

  • The approach of stratifying by work group defined from work hours spent in each job over the year led to the discovery that workers who primarily worked in a non-agricultural industry but continued to work in agriculture (a common occurrence in rural developing communities) were at an increased risk of fatigue or overexertion and other consequences of working too hard in their agricultural activities.


The authors would like to give special thanks to all research staff, health staff and business owners who supported this project at the National Institute and in Xuan Tien and, especially, Dr Ngo Quynh Sang, who was the main physician and study coordinator instrumental in the success of the study. We would like to thank Tom Leamon who carefully orchestrated the study and built the collaboration between the research institute and all levels of Vietnam from the local level in Xuan Tien to the ministry level and contributed greatly to the study design. We would like to thank David Kriebel and David Wegman who were instrumental in the design of the surveillance study and surveys. Finally, we would like to thank David Lombardi, Larry Hettinger and Theodore Courtney for their careful review and Peg Rothwell for her helpful editorial input on the final version of the paper.



  • Competing interests None.

  • Ethics approval The household survey was submitted to and approved by the institutional review boards at the Liberty Mutual Research Institute for Safety and the National Institute for Occupational and Environmental Health in Vietnam.

  • Provenance and peer review Not commissioned; externally peer reviewed.