Occupational Injuries in a Commune in Rural Vietnam Transitioning From Agriculture to New Industries

Abstract

Objectives. We explored the impact on work-related injuries of workers splitting time between industry and agriculture, a common situation in developing countries.

Methods. In 2005, we administered a cross-sectional survey to 2615 households of Xuan Tien, a developing rural community of Vietnam, regarding self-reported injuries and hours worked for 1 year. We defined groups as working in industry, agriculture, or a mix of both.

Results. Overlapping employment (part time in agriculture and up to full time in industry) increased the risk of injury in both agricultural and industrial work. This pattern held across all work groups defined by the relative amount of time worked in agriculture. Those working fewer than 500 hours annually in agriculture had an agricultural injury rate (872 per 1000 full-time equivalents) that was more than 4 times higher than the average rate overall (203 per 1000) and the rate for workers employed only in industry (178 per 1000).

Conclusions. Working in agriculture for short durations while working in industry increased the risk of injury substantially in both types of work.

Throughout human history, agriculture has been central to survival and development. Even today, half of the world's workforce works in agriculture despite a reduction in the developed world to only 10% of those employed working in this sector.^1–3 During the process of industrial development, a transition from employment in agriculture to work in other industries may be expected as workers shift to higher-paying work in factories and as technology and automation replace some of the manual labor processes of plowing, planting, irrigating, and harvesting. The impact of this transition on the burden of occupational illness and injury is not well understood, but agricultural work in all countries remains one of the most dangerous occupations (27.9 deaths per 100 000 agricultural workers in the United States, more than 7 times the overall average death rate of all US workers).⁴

Despite the estimated 1.3 billion people employed in agricultural production worldwide, there are few studies investigating the burden of agricultural injuries in developing countries¹ or the impact of industrial development on worker health and the risk of injury.⁵ Most studies use a macro approach that focuses on broad regions or national statistics, a level of analysis that does not consider risks faced by a single worker engaged in different tasks during normal work life, for example, harvesting rice and running a lathe.^6,7

Notably, however, even as nations modernize, agricultural work remains central to their economies, often involves the entire family, and frequently serves to supplement work in industrial sectors.¹ Available studies to date, including those carried out in China and Korea,^6,8 have focused on either agricultural workers or industrial workers without considering an intermingling of the 2 fields of employment as workers respond to the transition from agricultural to industrial work.

We conducted an injury surveillance study using a complete community sample in Xuan Tien, Vietnam, in 2005. Previously, we reported that work in manufacturing was more hazardous than was work in agriculture in this community.^9,10 We explored the effect of the overlap of agricultural and industrial work on work-related injuries in the Xuan Tien commune. Almost half of the workforce works part time in agriculture and works as much as full time (2000 hours per year) in another industry in this commune, which is undergoing rapid industrialization. We designed our approach to the collection and analysis of surveillance data to consider the employment patterns of the area, which showed substantial work simultaneously in the traditional agricultural and emerging industrial sectors.

METHODS

Xuan Tien is a rural community in northeastern Vietnam with more than 10 000 residents. Residents include 5485 workers, many of whom are employed in 816 family-owned businesses—predominantly manufacturing. Manufactured products include electric rice-threshing machines, electric concrete mixers, clothing for export, and bamboo products such as paper fans.⁹ Despite this significant and growing industrial base, the calendar of rice cultivation continues to dominate community life, with 2 cycles of rice planting and harvesting each year.

In Xuan Tien, industrial workplaces are often located on the first floor of a home, and heavy, electrically powered equipment, furnaces, and other dangerous machines are frequently located within 3 meters of the family living space. Work and nonwork life are intermingled in time and space. During the harvest season, workers are likely to spend time in the fields in addition to their regular manufacturing jobs, sometimes even temporarily replacing those primary jobs.

Survey Design and Administration

We began in 2005 with a population survey of more than 99% of the households in the commune. Twenty-one trained community residents (mainly volunteer health care paraprofessionals) administered a household questionnaire to the heads of 2647 households between April and December 2005. During survey administration, most family members were present and contributed to the survey. Volunteers received a 2-day training session on administering the questionnaire, which included human participant protection issues.

The household questionnaire consisted of 3 parts. The first requested general information about each household, for example, addresses and characteristics of family-owned businesses. The second requested details of work history and work duration (average hours worked per day, average days worked per week, and months worked in the past year) for every job of each household member.

In the third section, we asked the head of household about “hurts” to all family members in the last year: “Was [name] hurt in the past year enough to need care or to disrupt his or her normal activities for at least 1 day or longer?” In our questionnaire, we used the Vietnamese equivalent of the English term “hurt” instead of “injury” because it included a broader, more comprehensive set of conditions. Through discussions with our Vietnamese colleagues and pilot testing, we determined that asking about đau (hurts) was also more likely to capture overexertion events leading to injury, which we wanted to include in our estimates. For those reporting a đau, we obtained information on the type of đau and details of the incident from a series of open-ended narrative questions. To assist understanding, we have used the English term “injury” in this article when describing the study findings.

We gathered data for the entire previous year to improve capture of severe but rare events (e.g., amputations) as well as to represent all months in the year. This inclusion was particularly important because work activities and hazards can vary considerably from month to month as a result of the agricultural cycle. We used a decentering translation technique that gives equal weight to both languages to translate the English questionnaire into Vietnamese and then back into English to confirm both accuracy and cultural appropriateness.¹¹

Inclusion and Exclusion Criteria

A certified coder classified all reported injuries according to the International Classification of Diseases, 10th Revision (ICD-10)¹² by nature of injury and body part. We included injuries for which the ICD-10 classification began with S or T (injury, poisoning, and certain other consequences of external causes). Before including cases assigned an ICD-10 diagnosis code of M or R (mainly musculoskeletal conditions), 2 researchers (H. M. W. and J. L. W.) independently reviewed each to determine whether they qualified as a predefined injury using the description of activities at the time of injury, events following injury, and age. Generally, the ICD-10 M or R conditions we excluded were those with no external attribution in the narrative, pain persisting for a long period, or age precluding employment. For the 27% of M and R injuries for which reviewers did not agree, the reviewers conferred to reach consensus. An independent reviewer also assigned work-relatedness for each reported injury after reviewing all supporting information.

Work Groups

We used a modified International Labour Organization industry classification scheme to classify the industry of each job reported and the job at the time of each reported injury.¹³ We created 4 work groups using work status in agriculture and primary job industry (PJI) for the year before the survey. We determined status for work in agriculture as “yes” for any time worked in agriculture over the year or “no.” We determined the PJI assigned to each worker by the industry sector in which the individual worked the most hours in the year. When participants did not provide work hour information (0.03%; n = 155), we imputed work hours according to mean work hours in comparable industry groups.

The 4 work groups were group I: performed solely agricultural work; group II: PJI was in agriculture but worked in other industries; group III: PJI was in a nonagricultural industry but worked some time in agriculture; and group IV: performed no agricultural work. We also broke down work groups III and IV for detailed analyses of workers with a PJI of manufacturing, construction, wholesale or retail trade, and other (all other industry sectors with fewer than 100 workers, e.g., services, transportation, and public administration). Further details of the study site, survey design, and administration have been described previously.⁹

Data Analysis

All comparisons (proportions and rates) use hours worked in the denominator converted to full-time equivalents (FTEs) calculated as cumulative hours worked divided by 2000 hours per year. We have described the demographics of work in Xuan Tien in terms of the number of workers by age and gender, hours worked, reasons for not working, and work status in agriculture. We computed stratified gender-specific rates for each work group for injuries in agriculture and injuries in nonagricultural work. We also subdivided rates by hours spent in agricultural work during the prior year.

We computed confidence intervals (CIs) of incidence rates (IRs) assuming a Poisson distribution of reported counts. We performed Poisson regression analyses to determine whether uncontrolled relative rates among work groups were maintained after controlling for gender, age, and cumulative annual work hours. Because we found no confounding by age and work hours, we have reported on gender-stratified IRs and 95% CIs. We computed P values for the rate ratios of the gender-stratified IRs using the χ² test. We analyzed the information using SAS version 9.0 (SAS Institute, Cary, NC).

RESULTS

The survey was completed by 2615 households (out of 2647 attempted) with 10 671 family members, for a response rate of 99%. We excluded 255 family members who no longer lived in Xuan Tien, which left 10 416 reported residents of Xuan Tien.

Characteristics of Work

Half the residents of Xuan Tien (5485 of 10 416) reported working at some job in the past year for a total of 11 516 000 work hours, which is equal to 5758 FTEs. Participants reported the main reason for not working as being too old or too young to work.⁹ The distribution of workers and FTEs was essentially equal by gender and by age within gender; the large majority (82% of workers, 85% of FTEs) were aged 18 to 55 years (Table 1). Forty-five percent of workers worked more than 1 job, usually a combination of work in 1 of the 816 family-owned businesses and in agriculture.

TABLE 1.

Number of Work-Related Injuries and Incidence Rates by Age and Gender: Self-Report of Injury in the Previous Year in Xuan Tien Commune, Vietnam, 2005

		Annual FTEsa			Self-Reported Injury in the Previous Year
Demographic	Workers, No.	All	Nonagriculture	Agriculture	Annual No.	IR/1000 FTEsa (95% CI)
Overall	5485	5758	4760	1003	1169	203 (177, 229)
Men, age	2629	2857	2561	296	825	289 (269, 309)
≤ 17 y	90	84	82	2	23	274 (161, 385)
18–24 y	407	431	421	11	105	243 (197, 290)
25–34 y	668	785	738	46	209	266 (230, 302)
35–44 y	568	671	605	66	203	303 (261, 344)
45–54 y	528	584	489	95	183	313 (268, 359)
55–64 y	224	205	159	46	80	390 (304, 475)
≥ 65 y	140	91	62	29	22	242 (141, 343)
Missing age	4	5	4	1
Women, age	2845	2889	2182	707	344	119 (106, 132)
≤ 17 y	83	86	81	5	7	81 (21, 141)
18–24 y	455	450	396	54	34	76 (50, 101)
25–34 y	727	770	621	149	74	96 (74, 118)
35–44 y	633	732	537	195	91	124 (99, 150)
45–54 y	521	509	332	178	86	169 (133, 205)
55–64 y	236	203	128	75	34	168 (111, 224)
≥ 65 y	185	133	82	51	17	127 (67, 188)
Missing age	5	6	5	1	1
Missing gender	11	12	18	1

Note. CI = confidence interval; FTEs = full-time equivalents; IR = incidence rate.

^aFTEs were calculated as cumulative hours worked divided by 2000 hours per year.

Work in Agriculture

In Xuan Tien, 21% of workers were exclusively engaged in agriculture (group I, n = 1143), whereas 37% mixed agricultural work with other work (group II, n = 106 and group III, n = 1924). The remainder did not work in agriculture (group IV, n = 2312).

Patterns of agricultural work differed by gender. Among women, 73% worked some time in agriculture compared with 42% of men, and women worked 2.4 times more cumulative hours in agriculture over the year than did men. Women working in agriculture worked every day of the week throughout the year, whereas men often worked for shorter periods, for less than half the week, and for less than 6 months of the year.

Mixing Agriculture With Other Work

Gender differences were also reflected in the type of nonagricultural work for those who mixed agricultural with nonagricultural work. Although more than half of men and women worked in manufacturing (56% of men, 52% of women), the type of manufacturing varied by gender. Men dominated in the manufacture of machinery and equipment (44% men, 11% women), whereas women worked mainly in the manufacture of food products (53% women, 18% men). The second most common nonagricultural employment for those with mixed employment was construction for men (20%) and wholesale or retail trade for women (31%).

Self-Reported Injuries in the Past Year

Participants reported 1562 injuries for the year before the interviews; 50 injuries did not satisfy the case definition and were excluded. The 1512 qualified injuries resulted in an IR of 145 injuries per 1000 residents. The large majority (n = 1169, 77%) occurred during work activities, yielding an annual IR for work-related injuries of 213 per 1000 workers and 203 per 1000 FTEs. Of the 1023 injured individuals, 886 reported only 1 injury, 128 reported 2 injuries, and 9 reported 3 injuries. Of workers who reported only 1 injury, 45% and 43% worked in groups III and IV, respectively, whereas 55% and 39% of workers who reported more than 1 injury worked in these groups.

Injury Incidence Patterns

Overall, men had more than twice the number of work injuries per 1000 FTEs than did women (289 vs 119; Table 1). Work injury rates gradually increased with age and were highest for those aged 55 to 64 years among men and those aged 45 to 64 years among women; rates dropped for both men and women aged 64 years and older. The highest incidence of work injury was in manufacturing, followed by construction, with agriculture a close third: 245 per 1000 FTEs, 199 per 1000 FTEs, and 192 per 1000 FTEs, respectively.

Injury Impact Related to Agricultural Work

Stratified gender-specific rates from injuries in agricultural and nonagricultural work by work group are provided in Table 2. The overall injury incidence for those working entirely in agriculture (group I) or not at all (group IV) was similar and significantly lower than for those who worked predominately in agriculture (group II) or predominantly in industry (group III; P < .01).

TABLE 2.

Number of Work-Related Injuries and Incidence Rates by Work Groups: Self-Report of Injury in the Previous Year, Xuan Tien Commune, Vietnam, 2005

			Injuries in Nonagriculture Work		Injuries in Agricultural Work
Work Group	Description	Gender	Annual No.	IR/1000 FTEsb (95% CI)	Annual No.	IR/1000 FTEsb (95% CI)
I–IV	All	All	976	205 (191, 215)	193	192 (165, 219)
		Men	713	278 (269, 309)	112	378 (308, 448)
		Women	263	121 (106, 132)	81	115 (90, 140)
I	Only work was in agriculture	All			93	171 (136, 206)
		Men			51	286 (208, 365)
		Women			42	115 (80, 150)
II	Primary worka was in agriculture but worked in nonagriculture	All	14	400 (190, 610)	8	108 (33, 183)
		Men	3	434 (0, 926)	4	344 (7, 682)
		Women	11	392 (160, 623)	4	64 (1, 128)
III	Primary worka was in nonagriculture but also some work in agriculture	All	463	234 (213, 256)	92	238 (189, 286)
		Men	289	382 (338, 426)	57	538 (398, 677)
		Women	174	144 (122, 165)	35	125 (84, 166)
IV	Did not work in agriculture	All	488	178 (163, 194)
		Men	416	232 (210, 255)
		Women	72	76 (59, 95)

Note. CI = confidence interval; FTEs = full-time equivalents; IR = incidence rate.

^aPrimary work assigned to workers on the basis of the industry sector in which they worked the most hours in the year.

^bFTEs were calculated as cumulative hours worked divided by 2000 hours per year.

Gender differences.

Table 2 shows that men had a much higher rate of injury than did women in nonagricultural work (men, 278 per 1000 FTEs; women, 121 per 1000 FTEs) and agriculture work (men, 378 per 1000 FTEs; women, 115 per 1000 FTEs). Although the IRs varied substantially by gender for all work groups, gender did not significantly modify the risk of injury by work group in the Poisson regression model (P > .1).

For both men and women, the risk in nonagricultural work was much higher if they also worked in agriculture than if they did not work in agriculture (group III vs group IV; men, 382 per 1000 FTEs vs 232 per 1000 FTEs; P < .01; women, 144 per 1000 FTEs vs 76 per 1000 FTEs; P < .01). In addition, men in group III experienced a 40% higher work injury incidence (538 per 1000 FTEs) when engaged in agricultural work than when engaged in nonagricultural work (382 per 1000 FTEs; P < .01) and a 90% higher work injury incidence than those engaged exclusively in agriculture (group I; 286 per 1000 FTEs; P < .01). Women in group III experienced a nonsignificant higher risk in nonagricultural work compared with agricultural work (144 per 1000 FTEs vs 125 per 1000 FTEs; P = .45). Also, their risk in agricultural work did not differ much from the risk for women exclusively doing agricultural work (group I; 115 per 1000 FTEs; P = .72).

Work hours in agriculture.

Among both men and women, those who worked less than 500 hours per year in agriculture had the highest rates of agriculture-related injury (P < .01). This finding was true regardless of whether the work was exclusively in agriculture or whether the primary work was in another sector. Figure 1 shows that men with limited work time in agriculture had an IR of more than 800 per 1000 FTEs during this time regardless of whether their work was mixed or only in agriculture. Similar results were found for women (not shown).

FIGURE 1.

Incidence rates for time spent by men in agricultural and nonagricultural work, by annual hours worked in agriculture: self-report of work-related injury in the previous year, Xuan Tien Commune, Vietnam, 2005.

Note. FTE = full-time equivalent. Group I only worked in agriculture; Group II was excluded because of too few injuries in agriculture hour groups to report; the primary work of Group III was in nonagriculture work, but some work was in agriculture; Group IV did not work in agriculture.

Using Poisson regression, we found that the gender-controlled rates were approximately 1.7 times higher for workers whose primary work was in nonagriculture but who worked some time in agriculture (group III) compared with those who worked no time in agriculture (group IV; P < .01). This finding was unchanged after adjusting for age and cumulative work hours.

Additionally, a mix of agricultural and nonagricultural work influenced rates in the primary job. As illustrated in Figure 1, men with a PJI other than agriculture who worked fewer than 500 additional hours in agriculture had twice the rate of injury in their PJI (396 per 1000 FTEs) compared with those who did not work in agriculture (232 per 1000 FTEs; P < .01).

Industry mix.

Figure 2 presents mixed work among men by specific work sector. We compared those who worked exclusively in each sector with those who worked in the same nonagricultural sector plus agriculture. For those employed primarily in manufacturing, the annual IR of injury arising from manufacturing work was lower for those who did not work in agriculture (293 per 1000 FTEs) compared with those who also worked in agriculture (448 per 1000 FTEs; P < .01). We found similar results for the other industry comparisons, with construction having the highest disparity in rates (413 per 1000 FTEs compared with 135 per 1000 FTEs for no agricultural work; P < .01). Women experienced similar but less striking patterns (not shown).

FIGURE 2.

Incidence rates for men stratified by industry and work status in agriculture: self-report of work-related injury in the previous year, Xuan Tien Commune, Vietnam, 2005.

Note. FTE = full-time equivalent.

DISCUSSION

We assessed the risk of part-time agricultural work injury in a rural population in Vietnam with different primary work affiliations. Dividing workers into 4 broad categories of work on the basis of whether they worked exclusively or any time in agriculture made it possible to evaluate injury rates by time spent in agriculture versus time spent in the nonagricultural work. We found that work in agriculture for short durations while also working in another industry increased the risk of injury substantially in both agricultural work and work in the PJI. These patterns remained unchanged after controlling for age, gender, and cumulative annual work hours. Moreover, this pattern held across all 4 work groups defined by the relative amounts of agricultural and nonagricultural work. Those working in agriculture for less than 500 hours annually had a very high rate of injury in their agricultural work (Figure 1). We saw this difference in both genders.

Our findings highlight the potential impact on injury rates during the transition from traditional work patterns (working only in agriculture) to combining agricultural work with new types of work. We found that jobs in the newer industries were more dangerous than were those in traditional agriculture (count and rate of injury were higher in manufacturing than in agriculture), but this overall difference appears to mask a more complex association of injury with agricultural work.

We previously reported that manufacturing ranked as the industry with the highest injury risk in the commune.^9,10 The current analysis refines our earlier observation and indicates that manufacturing workers who also worked in agriculture might be an even more important target for safety interventions than are those solely employed in manufacturing. Although such workers spent fewer hours in agriculture than in their manufacturing work, their rate of injury in their agricultural work was significantly higher than it was in their other work. Furthermore, we unexpectedly found that work in agriculture modified (increased) the risk of injury in their manufacturing work. Thus, using only the principal job as the indicator of work injury risk would miss the additional burden in both agriculture and manufacturing that was associated with having a secondary job in agriculture.

One possible explanation is that the workers who temporarily leave their industrial work to work in the fields are those who hold lower-paying positions in industrial work (data not shown). They may, for example, tend to be laborers rather than skilled craftsmen and therefore have different exposures in their industrial work.

Alternative explanations for the high injury rates among those doing both agricultural and industrial work is fatigue from their extra work,^14,15 inexperience, unfamiliarity,¹⁶ or performance degradation as the result of heat stress¹⁷ or pesticide toxicity, but the data do not permit direct evaluation of these possibilities. From our analyses, we were able to observe that the high rates of injury were not related simply to the number of work hours, but it is still possible that fatigue was a factor. Although we could control for total annual work hours, we could not control for season-specific differences (e.g., during the rice harvest). Careful study, perhaps with work diaries, would be valuable in identifying ways to reduce these high risks, for despite the increase in manufacturing, agriculture remains an important economic sector and its impact will likely be substantial for many years.

Men and women in this developing community work in very different jobs in both agriculture and industry. Although women work more hours in agriculture throughout the year, men are likely working in higher-risk jobs both in agricultural (during high-risk seasons) and nonagricultural work, with more than double the incidence of injury compared with women regardless of work group. However, we found the higher risk associated with mixed agricultural and industrial work for both women and men.

One potential study weakness is recall bias that may have resulted in an undercount of the total number of injuries.^18,19 Asking for injury recall in a shorter time window (e.g., 1 month) is a potential solution but has its own limitations because this approach could miss important seasonal patterns for injuries in agriculture (and in other industries when work there is disrupted by agricultural work).⁹ We administered most surveys from June to September 2005, and recall was likely poorer the further away participants went in their memories from this window.^18,19 However, the consistent trends throughout these analyses strengthen the reliability of the findings and suggest that the rates we report are likely to be a representative sample of the most severe injuries in each of the work groups. Many of the injuries reported here likely were not formally treated; hence, we were unable to validate by matching self-reported injuries to medical treatment.

Although the World Bank has identified the need to promote development rooted in local culture, reflecting the knowledge and needs of the local population,²⁰ few reports have attempted to understand how community structure and work culture (e.g., multiple jobs and part-time work) in a largely agrarian society may modify work-related health risks during industrial development.

We found that the pattern of intermingled agricultural and industrial work in Xuan Tien was particularly risky and contributed a heavy burden to the health of the commune. Approximately 80% of the world's workforce lives in developing countries,²¹ with agricultural workers representing up to 60% of the total workforce. Few statistics exist on the rate of injury, and even fewer exist on the rate of work-related injury in developing countries. Studies on workers in developing countries usually reflect rates exclusively for agricultural or industrial workers.^6,8 Our findings suggest that methods of surveillance should evolve from assessing variations of injury rates separately in each distinct industrial group to assessing variations of injury rates relevant to the complex nature of work in developing countries. Surveillance methods should be designed to investigate work characterized by multiple jobs, the home as work environment, self-employment, and working for long hours.