3D peak and cumulative low back and shoulder loads and postures during greenhouse pepper harvesting using a video-based approach

Abstract

BACKGROUND: In agricultural field work many tasks have been cited as high priority risk factors for the development of work-related musculoskeletal disorders (WRMDs). Although video-based biomechanical approaches have been effective in documenting the physical demands and risks associated with various occupational and non-occupational tasks, to date, this method has yet to be used to document jobs such as crop harvesting in a greenhouse environment.

OBJECTIVE: To document and assess the postural characteristics and 3D peak and cumulative low back and shoulder loads associated with greenhouse pepper harvesting using a video-based posture sampling approach.

METHODS: Nine male (28.2 (4.1) years) pepper harvesters from a greenhouse in Southwestern Ontario, Canada were videotaped during a normal shift. 3DMatch was used to document working trunk and shoulder postures, from which 3D peak and cumulative forces and moments were quantified.

RESULTS: On average, workers spent the majority of their time in neutral trunk postures (lateral bend: 99.1%; axial twist: 59.9%; flexion: 89.8%). Consistent results were found for the left and right shoulder, with the arms held in a neutral flexion posture 50% of the time or more. Four participants experienced peak L4/L5 compression forces (between 4116.3 N and 5937.0 N) which exceeded the NIOSH Action Limit (3400 N) during the cart pushing/pulling task, but remained below the threshold during picking. Mean cumulative L4/L5 extension and shoulder flexion moments ranged in magnitude from 18.5 Nm to 28.2 Nm, and between 19.4 Nm and 23.2 Nm, respectively, across all tasks.

CONCLUSIONS: The postural characteristics and biomechanical loads associated with greenhouse pepper harvesting were quantified with a video-based biomechanical approach. Further investigations of the physical risk factors for low back and shoulder musculoskeletal disorders is warranted in pepper harvesting, given the postures and loads documented in this study.

Keywords

Video-based posture assessment biomechanical loads vegetable harvesting agriculture work-related musculoskeletal disorders

1 Introduction

Reducing the risk of workplace injury by understanding how the external demands of a job impact the internal demands placed on a worker has been a main focus of occupational biomechanics research, especially among jobs involving manual materials handling (MMH) [1 –5]. The importance of assessing and documenting physical demands in the workplace cannot be understated when it comes to the identification of risk factors associated with work-related musculoskeletal disorders (WRMDs) (e.g., force, posture). Detailed knowledge of these potential risk factors is critical for the health and safety of workers as specific ergonomic parameters can be implemented for proactive and/or reactive job, workplace, equipment and tool design. Furthermore, an understanding of the physical demands faced by workers will also protect companies from potential financial losses resulting from decreased productivity [6, 7], as well as medical treatment and indemnity compensation claims [8] that result when a worker is injured or off work.

A variety of approaches have been used to document physical demands in the workplace; these are generally grouped into three main categories: self-reports, observational techniques, and direct measurements [9]. Self-reports are used extensively due to their applicability to a wide range of work scenarios and low cost when used on a large scale [10, 11]. However, the validity and reliability of self-reports for assessing many workplace physical demands are still relatively poor [12 –14]. In contrast, direct measurement approaches (e.g., electromyography) may be more accurate for assessing the internal demands resulting from workplace activities, but the equipment needed can be relatively expensive, requires specific training in order to operate, and interferes with workers performing their tasks normally; all factors which can limit their use in many work environments [15]. Many observation-based methods offer a trade-off between efficiency and accuracy for measuring physical demands in the workplace [9]. Specifically, video-based posture sampling approaches that are integrated with biomechanical models to estimate internal loads on workers have been reported to be valid [16 –18] and reliable [19, 20] for studying common workplace tasks. A key benefit of video-based approaches is that they can be used to evaluate work tasks in a field setting with minimal interference to the worker. Furthermore, video provides a permanent visual record of the work being investigated, which can be used for detailed analyses offline, or to verify anything about the work performed after the fact.

However, video-based posture sampling approaches can be expensive in terms of data collection and processing time, particularly for large scale studies [9, 21]. Consequently, cost-effective measurement strategies have been proposed to optimize the allocation of monetary resources [22, 23]. With appropriate training, it has been shown that video analysts can make faster posture assessment decisions and reduce the number of posture misclassification errors they make, which improves the overall efficiency of such approaches [24]. One video-based approach, 3DMatch [25], requires operators to match body segment postures observed from recorded video images with posture categories displayed on a computer interface. Compared to other video-based approaches, one of the practical advantages of 3DMatch is that the posture-matching technique used eliminates the need to digitize external markers placed on the person of interest. Moreover, with 3DMatch, three-dimensional (3D) movements can be analyzed from two-dimensional video; a key advantage that allows this video-based approach to assess asymmetrical loading scenarios (e.g., axial twisting) instead of being limited to strictly planar motion (i.e., flexion or extension). Biomechanical loads such as compression and shear forces and moments about the L4/L5 joint are estimated within 3DMatch using a low back model developed by McGill et al. [26]. The duration of time spent in several posture categories (i.e., neutral, mild, severe) is also output for joints of interest, including the low back and shoulders. The validity of 3DMatch has been established against various biomechanical methods, including a 3D electromagnetic tracking approach (FASTRAK^TM) [27] and an electromyography-assisted modeling approach [28]. Additionally, the manual digitization of video data using 3DMatch has been determined to be reliable [29, 30], regardless of camera viewing angle [31].

Video-based approaches have been employed to document low back and shoulder physical demands during numerous occupational tasks across a wide range of industries: automotive assembly [32 –34], manual patient handling activities [35, 36], food and drink services [37], and a collection of tasks from surface construction, metal- and meat-processing, and refuse collection [1]. In addition to occupational tasks, common household activities have also been documented using video-based approaches to assess low back [38] and shoulder demands [39]. However, the physical demands associated with many labor-intensive tasks in the agriculture industry have not been assessed with video-based approaches to date, which is surprising considering that the need to identify biomechanical risk factors associated with WRMDs in this sector has been highlighted in several studies [40 –42]. Only relatively recent research on wool harvesting tasks (e.g., shearing and wool handling) has used video-based documentation to quantify the physical demands faced by agricultural workers [43, 44].

Various types of tasks in agricultural field work have been cited as high priority risk factors for WRMDs, including the repetitive handling of heavy loads, repeated or sustained non-neutral postures of the trunk (i.e., forward flexion and axial twisting), and highly repetitive force exertions by the trunk and upper extremities [45, 46]. In crop farming, these strenuous tasks are common daily practice. Consequently, occupations in crop farming have been noted as potentially high risk for developing WRMDs, especially in greenhouse environments. Greenhouses may have elevated physical demands associated with factors such as working in confined spaces in high temperatures and humidity, maintaining high work productivity rates for extended time periods, and performing multiple, complex tasks that involve various static and dynamic loading scenarios [47 –49].

In Canada, approximately 300,000 workers were employed within the agriculture industry in 2011 [50]; over 36,000 of those workers were employed in greenhouse operations specifically, with more than one-third being located in Ontario [51]. Of that total labour force, approximately 39% were permanent workers and 61% were seasonal workers. The province of Ontario has consistently led Canada in the production of greenhouse peppers [51], in which total pepper sales increased 24.2% to $389 million from 2012 to 2013; 31.1% of the total greenhouse vegetable and fruit sales that year [52].

Considering the success of greenhouse pepper cultivation in Ontario and across Canada, detailed assessments of the physical demands associated with greenhouse pepper harvesting should be carried out to facilitate its continued productivity and potentially improve employee turnover. However, this has yet to occur. One of the main reasons for this is the confined work environment in most greenhouses, whereby rows of pepper plants are placed as close together as possible, while still allowing enough room for harvesting carts to be pushed and pulled along a track between the rows. In addition, to accommodate for the continuous growth of the plants over a season, trellising systems are used to ensure that the plants grow vertically, maximizing the space within the greenhouse. These space-saving techniques create a challenging postural environment for the workers who prune the plants and pick the peppers. As a result, workers are forced to assume increasingly awkward postures while performing acts of repetitive overhead reaching as the season progresses, both of which are known to be predisposing biomechanical risk factors for developing WRMDs of the neck, shoulder and low back [53]. With respect to data collection using a video-based approach, acquiring an adequate field of view to document the tasks performed by the workers in this physical environment is challenging, especially given the rapid pace at which they work. Additionally, the variable lighting and high humidity common to greenhouses may further impact the quality of video records.

Pepper harvesters are also responsible for pushing and pulling large carts in which the peppers are collected along a pipe and rail system that run between each row of plants and over concrete areas that separate sections of plants within the greenhouse. Pushing and pulling the carts could be potentially hazardous for workers and may be reflective of the increased risk of developing shoulder and low back pain, which has been reported by individuals who regularly push and pull heavy loads [54, 55]. Furthermore, although the mass of each pepper may be relatively small during picking, loads of a smaller magnitude should not be ignored as possible health risks. Newell and Kumar [56] demonstrated that even though the magnitude of instantaneous loads may be negligible, the cumulative stress experienced when repetitively sustaining small loads over a long duration can cause muscle fatigue, which can lead to inflammation, soreness, and possible injury [57].

Taking into consideration the physically demanding nature of greenhouse pepper harvesting and its association with WRMD risk, along with the economic value of pepper cultivation, evaluation of the tasks performed by pepper harvesters in a greenhouse environment is both important and warranted. To date, there has been no research that has documented the physical demands of pepper harvesting in greenhouses, and it is unclear if a video-based approach would be feasible for this purpose, given the physical environment characteristic of such workplaces. Therefore, the purpose of the present study was to document the postural characteristics and 3D peak and cumulative low back and shoulder loads associated with pepper harvesting in a greenhouse environment using a video-based posture sampling approach.

2 Methods

2.1 Participants

Nine male pepper pickers (mean (SD) age: 28.2 (4.1 years)) from one of the largest greenhouse operations in Southwestern Ontario, Canada volunteered for this study. The participants’ mean (SD) height, body mass, and time employed at the greenhouse were 172.1 (7.0) cm, 74.7 (7.8) kg, and 2 (16.1) years, respectively. The procedures were described verbally to potential participants prior to their involvement on site at the greenhouse. The pickers who agreed to participate signed a consent form approved by the Research Ethics Board at the university associated with the research team before data collection began. Participants received a gift certificate to a local store for their involvement.

2.2 Data collection

Prior to the start of data collection, participant age, height, body mass, and time employed at the greenhouse were recorded. Each participant was video-taped with a hand-held camera (30 frames/sec) during a typical 8.5 hour work shift while completing several cycles (∼5) of his daily duties at the greenhouse. A full cycle of pepper harvesting took between 13 and 19 minutes to complete, depending on the number of peppers on the plants, and the skill level of the harvester. A mix of morning and afternoon data collections occurred to best accommodate the workers and provide a balanced picture of the job over a normal shift. The two main tasks documented for a single cycle of pepper harvesting included: 1) picking the ripe peppers from the plants, and 2) pushing/pulling the cart in which the harvested peppers were collected. These tasks included five subtasks: i) walking to get an empty cart; ii) pushing the cart to the desired row; iii) picking the peppers, which involved holding the pepper in one hand and cutting the stem by using a small knife with the other; iv) pushing/pulling full carts of peppers; and, v) inputting the row and cart number into a computer. Throughout a cycle, the carts were manually transported under two separate conditions. When picking the peppers, participants transported the carts along a metal pipe and rail system positioned between each row of peppers (Fig. 1A). The carts were also transported along concrete areas between sections of plants, which served as a staging area for the carts (full and empty) (Fig. 1B).

Due to the space constraints during picking, participants were video-taped from behind and in front. During the pushing and pulling of carts on the concrete areas between the row sections, the workers were able to be video-taped from different angles due to the relative openness of the areas. Two investigators were present during each data collection; one captured the actions of each task on videotape and the second recorded things such as the masses of, and hand forces applied to, objects manipulated by the participants. Since the workers were paid via piece-work (i.e., the amount of money that they earn is dependent on their productivity), the data collection team did not interfere with the workers’ progress by stopping them during their work to obtain hand forces for each subtask (e.g., pushing a cart). These forces were obtained from participants using a digital force gauge with appropriate adaptors (Chatillon model DFM 100, AMETEK TCI, Florida, USA), during a mock up of each subtask, at break time or after the shift was finished. Mock ups for pushing/pulling the cart involved participants interacting with carts that were filled to four levels (empty, 1/3 full, 2/3 full and full), and that were being pulled over two floor conditions (pipe and rail system and concrete floor). The mean value of three exertions for each subtask was used to represent the hand force required to do each subtask.

Since the mass of a pepper was the primary variable contributing to the loads sustained by harvesters in this study (i.e., when individually picked from the plant or when pushed/pulled within the cart), it was important to determine the mean mass of a single pepper, given that it was not feasible to weigh every pepper handled by the participants. The mass of a plastic bin full of peppers was measured using a bathroom scale. The mass of the bin was subtracted from the full bin mass to determine the mass of the peppers within the bin. The mean mass of each pepper was determined by dividing the mass of all peppers within the bin by the number of peppers contained. This procedure was repeated three times with different peppers. The average pepper mass was later used during data analysis as an input into the biomechanical model.

2.3 Data analysis

Video data captured during the data collection process was manipulated and reduced to a sampling rate of 3 frames/s [58] using Adobe Premiere Elements Software (v.1.0.). The video data were then converted to AVI (Audio Video Interleave) format and imported into 3DMatch software to estimate the peak and cumulative 3D biomechanical loads on the low back and shoulders [25].

For each frame of video analyzed, corresponding trunk, neck, arm, and forearm postures that most closely matched the participant’s posture observed in the video image were selected from a series of available posture categories (Fig. 2). Participant height, body mass, and hand forces (magnitudes and directions) determined during data collection were input into 3DMatch as necessary for each participant. Recent recommendations provided by NIOSH, with regard to assessing postural stress of the trunk and upper limbs using observation-based posture assessment approaches, were followed [59].

All physical demands variables determined for the low back and shoulders were estimated from the 3D biomechanical model incorporated within the 3DMatch software. The model output low back loads, including peak and cumulative L4/L5 compression and shear forces (i.e., anterior/posterior and medial/lateral joint and reaction shear forces). Peak and cumulative flexion/extension, lateral bend, and axial twist moments were also documented about the L4/L5 joint. Left and right shoulder loads included peak and cumulative anterior/posterior, cranial/caudal shear forces, and peak and cumulative adduction/abduction, internal/external rotation, flexion/extension moments. For the purposes of this study, only compression forces and moments at the L4/L5 and shoulder joints were reported. The mean cumulative loads and moments for the low back and shoulders were determined for all tasks performed within one full cycle (13 to 19 minutes). The loads for one representative cycle were then extrapolated to estimate the loads experienced over an entire work day by multiplying them by the number of cycles that occurred in an 8.5 hour day (standard work day). The assumption that the work performed in each cycle was representative of the work performed across the entire shift was seen to be reasonable given the repetitive nature of the work.

In addition, the percent time that each participant’s trunk and shoulders spent in a neutral, mild or severe posture category was documented. Trunk positions were determined in three posture ranges for flexion (neutral: <20°, mild: 20°–45°, severe: >45°), lateral bend (neutral: <15°, mild: 15°–30°, severe: >30°), and axial twist (neutral: <15°, mild: 15°–30°, severe: >30°). Shoulder postures were determined in varying degrees of flexion (neutral: <20°, mild: 20°–90°, severe: >90°) and axial twist (neutral: <15°, mild: 15°–30°, severe: >30°).

3 Results

The mean mass of a single pepper was determined to be 0.193 kg and the mass of a fully loaded cart of peppers ranged between 85 kg and 97 kg. The total number of peppers picked by each participant during a regular 8.5 hour shift ranged from 13,832 to 24,240 peppers, which corresponded to a total mass of peppers harvested per participant of between 2670 kg and 4678 kg per shift.

3.1 Peak loads

Peak L4/L5 compression forces experienced by the harvesters during cart pushing/pulling ranged from between 1787.5 N and 5937.0 N (mean of 3584.3 N), which were approximately 1.7 times greater on average than during picking (1757.4 N to 2353.6 N; mean of 2070.5) (Fig. 3). Four participants exceeded the NIOSH [60] Action Limit for lumbar spine compression (3400 N) from the loads sustained during cart pushing/pulling.

The mean peak L4/L5 moments experienced during the pushing/pulling task were greater in magnitude than the corresponding moments during picking, with the exception of the mean right lateral bend moment (Fig. 4). On average, the mean (SD) peak L4/L5 extension moments (pushing/pulling: 109.0 (22.1) Nm; picking: 88.9 (29.5) Nm) were greater in magnitude by between 2.1 and 7.4, and 1.8 and 3.5 times for the pushing/pulling and picking tasks, respectively, compared to all other moments calculated about the trunk. Specifically, when pushing/pulling the cart, the mean (SD) peak moments for trunk axial twist were moderately high in both directions (left: 56.5 (28.1) Nm; right: 61.3 (25.6) Nm).

Consistent with the low back moments, the mean peak shoulder moments experienced during the pushing/pulling task were greater in magnitude than the corresponding moments during picking in almost all cases. With respect to individual loads, the mean peak flexion moments for both shoulders exceeded all other moment values by factors of between 1.1 and 7.3 times and 5.1 and 31.9 times for the pushing/pulling and picking tasks, respectively (Fig. 5A, B). The mean (SD) peak flexion moment for the picking task was greater for the left shoulder (90.9 (28.0) Nm) than the right shoulder (59.5 (22.4) Nm). Conversely, the mean peak flexion moment for the right shoulder (83.2 (27.8) Nm) was greater in magnitude than the left shoulder (74.8 (26.3) Nm) for the pushing/pulling task. In addition, the mean (SD) peak extension moments for both shoulders were also found to be relatively high, especially for the right shoulder (78.3 (41.8) Nm).

3.2 Cumulative loads

All nine participants experienced greater cumulative L4/L5 compression forces when performing the picking task (range from 16.9 MN·s to 28.9 MN·s) compared to the pushing/pulling task (range from 8.6 MN·s to 12.1 MN·s) for an entire shift (Fig. 6).

Mean (SD) cumulative L4/L5 extension moments far exceeded all other cumulative moments for both the picking (28.2 kNm·s (5.2)) and pushing/pulling (18.5 kNm·s (8.7)) tasks (Fig. 7). With respect to the shoulders, the mean cumulative moments for the left and right shoulders were comparable in magnitude (Fig. 8A, B). The mean (SD) cumulative flexion moments were the greatest in magnitude for both the left shoulder (picking peppers: 21.8 (10.3) kNm·s; pushing/pulling carts: 21.1 (12.8) kNm·s) and right shoulder (picking peppers: 19.4 (7.8) kNm·s; pushing/pulling carts: 23.2 (13.2) kNm·s).

3.3 Percent time in postures

Participants spent a considerable amount of their time working in neutral trunk postures for each of the three positions analyzed (lateral bend: 99.1%; axial twist: 59.9%; flexion: 89.8%), and only a limited amount of time in mild (lateral bend: 0.9%; axial twist: 24.2%; flexion: 8.8%) or severe (lateral bend: 0%; axial twist: 15.9%; flexion: 1.4%) trunk postures (Fig. 9).

Consistent results were found for the left and right shoulder, with the arms held in a neutral flexion posture 50% of the time or more for both tasks. It was found that mild to severe flexion postures were more frequently adopted than adduction and abduction postures overall, in which significantly more time was spent in mild flexion (left shoulder: 41.5%; right shoulder: 42.9%) than severe flexion (left shoulder: 1.6%; right shoulder: 5.8%).

4 Discussion

The aim of the present study was to document the postural characteristics and 3D peak and cumulative low back and shoulder loads sustained by pepper harvesters in a greenhouse environment with the use of a video-based posture sampling approach, such as 3DMatch. Two main harvesting tasks were analyzed: 1) picking the ripe peppers from the plant, and 2) pushing/pulling the cart in which the harvested peppers were collected. Four of the nine participants sustained peak L4/L5 compression forces that exceeded the NIOSH Action Limit (3400 N) during the pushing/pulling task, whereas the picking task resulted in significantly greater mean cumulative L4/L5 compression forces for all participants. Across both tasks, mean peak and cumulative L4/L5 extension moments were found to be much greater in magnitude than all other peak moments calculated. With regard to the left and right shoulders, mean peak and cumulative flexion moments had the highest magnitudes for each of the tasks. The mean peak extension moment for the right shoulder was also found to be relatively high for the pushing/pulling task in particular. Overall, participants spent a majority of their time working in neutral trunk and shoulder postures.

4.1 Low back

Mean peak L4/L5 compression forces for the cart pushing/pulling task varied considerably among the nine participants (Fig. 3); four sustained values that significantly exceeded the NIOSH Action Limit of 3400 N, putting them at an elevated risk of low back injury, while the remaining five had values well below the limit. This clear division between the workers evaluated could be the result of several contributing factors. It has been reported that greater low back compressive loads occur during pulling compared to pushing activities [61 –63]. Workers were free to maneuver the carts in the way that best suited themselves. Therefore, exposure to higher compressive forces may be dependent on how workers chose to manipulate the carts (i.e., push or pull) when loaded. In addition, several studies have shown that the effect of mechanical loading on the low back for pushing/pulling tasks is significantly influenced by the handle height on a cart [61 –65]. Considering that the cart handle used by the workers was at a fixed height, increased loading on the lumbar spine could have occurred for certain participants depending on where the handle was situated relative to their height (e.g., waist versus shoulder height). In order to substantiate these claims, further investigation into the manipulation techniques and cart specifications involved with transporting harvested loads of peppers in a greenhouse is needed.

Although the peak compression forces on the low backs of workers during the pepper picking task were not overly hazardous, high mean L4/L5 cumulative loads were observed (Fig. 6). Given that cumulative loading of the trunk from workplace exposures is a well-documented risk factor for low back pain [32 , 66–68], the physical demands of greenhouse pepper harvesting may be associated with long-term repercussions for workers. In comparison, the mean cumulative L4/L5 compression forces observed for pepper picking (ranging from 16.9 MN·s to 28.9 MN·s) were generally higher in magnitude than cumulative compressive forces in jobs from other industries that also involve object manipulation [32 , 66]. Moreover, when the mean cumulative L4/L5 compression forces from the pushing/pulling task are combined with those from the picking task, the daily exposure in terms of cumulative compression for the pepper pickers in this study (range from 27.0 MN·s to 39.2 MN·s for 8.5 hours) is considerably higher than workers in the automotive industry who reported low back pain (21.0 MN.s over an 8 hour shift) [32].

Both harvesting tasks were also found to have high mean peak and cumulative L4/L5 extension moments (Figs. 4, 7). Consequently, these results demonstrate that throughout the course of a shift, greenhouse pepper harvesters have to counteract high external demands associated with trunk flexion, which lends support to prior research identifying trunk flexion as one of the main risk factors for low back pain among agricultural field workers [44–46 , 69].

In spite of the significant peak and cumulative low back loads during pepper harvesting, participants spent the vast majority of their shift working in neutral trunk postures (Fig. 9). Although these results seem contradictory, they do coincide with the findings from Azar et al. [38], who used equivalent data collection and analysis methods to those used in this study, for evaluating various non-occupational tasks. The relatively high low back loads in these studies, compared to the previous work by Norman et al. [32] for workers in the automotive industry, is likely due to the differences in models and integration methods utilized to calculate the cumulative loads between the studies [1 , 70–72].

4.2 Shoulders

Biomechanical analysis of the shoulders showed similar patterns of loading for each of the two harvesting tasks. The magnitude of the mean peak and cumulative flexion moments about the left and right shoulders were greater than in the other directions assessed (Figs. 5A/B, 8A/B). These results are in agreement with observations from automotive assembly [33] as well as non-occupational tasks [39]. When picking peppers above shoulder height, workers often elected to stand on the rails between the rows of plants and adopt awkward postures of excessive overhead reach with both hands (i.e., one to grasp the pepper and the other to remove it from the plant with a knife). Comparable techniques involving a high volume of overhead work have been documented during tomato, pear, and apple harvesting, reinforcing that static and repetitive flexion above shoulder height are prominent risk factors for shoulder disorders [48 , 74]. When pushing/pulling the cart, comparably high peak and cumulative flexion moments were observed about the left and right shoulders. This is consistent with symmetrical postures utilized by most workers when pushing the heavy loads with two hands. It should be noted that, although shoulder postures involving flexion greater than 45° are associated with musculoskeletal disorders of the shoulder [53 , 76], risk of developing a WRMD could not be determined for this study since the broad range of the mild flexion posture category in 3DMatch (i.e., 20°–90°) does not provide enough detail regarding the shoulder flexion angles adopted by the participants.

The right shoulder was also exposed to very high mean peak extension moments during the pushing/pulling task. Unlike the symmetrical postures observed for pushing the cart, when participants pulled the peppers carts, they typically assumed an asymmetrical posture characterized by the use of only one hand, trunk axial twist, and shoulder extension (refer to Fig. 2). One-handed pulling techniques are commonly used in many occupations involving MMH. However, research on the shoulder mechanics and loads associated with one-handed pulling tasks is scarce. It has been established that volitional postures that are freely chosen generate greater pulling strength than standardized postures [77]. Therefore, it can be speculated that participants opted to use this one-handed pulling posture because it may have catered to their strengths (i.e., right hand dominance), and allowed them to execute the task within the confines of the greenhouse environment, even if it put them at increased risk of injury.

The percent time analyses for the shoulder were comparable to the low back in that high peak and cumulative loads were reported about the shoulder even though, on average, participants spent the majority of their shift in neutral postures. These findings were consistent with the observations from automotive manufacturing [33] and non-occupational tasks [39], emphasizing the parallels between greenhouse pepper harvesters and other populations noted to be at risk of WRMDs.

4.3 Limitations

Although the current study presents promising results for the feasibility of using video-based posture sampling to document the physical demands of greenhouse pepper harvesting, some important limitations must be acknowledged. With this approach, it should be highlighted that the cumulative loads on low back and shoulders over an 8.5 hour shift were obtained by extrapolating the loads documented from single work cycles (13 to 19 minutes). Actual shift values may be smaller or larger than extrapolated [70]. Furthermore, posture classification errors are always a concern when using video-based posture assessment methods such as 3DMatch, especially when analysts must classify body postures near the boundaries between two posture categories [78]. Guidelines for posture matching [59] were used to optimize analyst performance in terms of posture classification error and the speed of posture classification. Viewing angle constraints in the greenhouse were sometimes problematic as participants’ hands were not always visible when video-taping the picking task within the confined rows of pepper plants. In these cases, the investigators had to estimate the position of the hands based on the anatomical alignment of the participants’ shoulders, arms, and forearms. This has been shown to result in significant differences in ratings of posture metrics compared to clearly visible postures [79]. However, in field settings, such problems may prove to be inevitable to some degree.

Regarding the experimental design, only a small sample of male harvesters from one greenhouse operation in Southwestern Ontario, Canada was observed in the study. Data collection was also limited to being within a short period of time during the pepper harvesting season, in order to reduce the intrusion on the workers. Pepper picking was only video recorded at a single height. As the pepper plants grow during the season, pickers are required to pick while standing on a scissor lift platform. Picking height was not evaluated, but it was indicated by several workers that the position they are in during picking did not differ significantly as the height of the lift increases above the ground. Therefore, it was assumed that the postures captured in the current study were representative of the normal postures assumed by the workers. Although the above limitations may reduce the generalizability and interpretation of the results, it was the purpose of the current study to determine if a video-based approach was feasible in a greenhouse environment. Biomechanical loads experienced on the low back and shoulders, and the amount of time spent in specific posture ranges about these joints were effectively determined for this group of workers during the allotted observation period, clearly demonstrating the viability of this type of approach for similar agricultural work. However, it should be noted that additional work needs to be conducted on a larger sample and across more facilities throughout the entirety of a harvesting season, in order to be more confident in establishing the risk associated with manual greenhouse harvesting operations.

5 Conclusions

The results of the present study illustrate that common tasks associated with greenhouse pepper harvesting may expose workers to an increased risk of developing WRMDs. With the use of a video-based posture sampling approach (3DMatch), postural characteristics, peak and cumulative low back and shoulder loads, were well documented. It was found that even though greenhouse pepper harvesters spent the majority of their shift working in neutral trunk and shoulder postures, their exposure to peak and cumulative loads and moments at the trunk and shoulder were still considerable, in relation to other occupations. This may suggest a greater risk of injury with respect to the forces experienced during pepper harvesting in comparison to the postures adopted. However, this would need to be determined in future work using an appropriately designed study. Moreover, potentially injurious levels of peak L4/L5 compression forces were experienced during pushing/pulling for some participants. It was also found that workers adopted different posture strategies for each of the harvesting tasks in order to maintain a high level of task productivity, even if it compromised joint safety. Taken together, these findings demonstrate that future research is necessary to confirm the key issues raised in the current study, and determine appropriate interventions that will reduce the biomechanical loads and risk experienced by greenhouse peppers harvesters.

Conflict of interest

The authors have no conflict of interest to report.

Footnotes

Acknowledgments

Thanks to the Occupational Health Clinics for Ontario Workers (OHCOW) in Windsor Ontario Canada for their support and to the greenhouse workers for agreeing to participate in this study.

References

Jäger

, Jordan

, Luttmann

, Laurig

. DOLLY GrouEvaluation and assessment of lumbar load during total shifts for occupational manual materials handling jobs within the Dortmund Lumbar Load Study –DOLLY. Int J Ind Ergo 2000;25(6):553.

Allread

, Wilkins

3rd , Waters

, Marras

. Physical demands and low- back injury risk among children and adolescents working on farms. J Agric Saf Health 2004;10(4):257.

Burdorf

, Windhorst

, van der Beek

, van der Molen

, Swuste

PHJJ

. The effects of mechanised equipment on physical load among road workers and floor layers in the construction industry. Int J Ind Ergon 2007;37(2):133.

Jang

, Karwowski

, Quesada

, Rodrick

, Sherehiy

, Cronin

, Layer

. Biomechanical evaluation of nursing tasks in a hospital setting. Ergonomics 2007;50(11):1835.

Okunribido

, Magnusson

, Pope

. The role of whole body vibration, posture and manual materials handling as risk factors for low back pain in occupational drivers. Ergonomics 2008;51(3):308.

Burton

, Conti

, Chen

, Schultz

, Edington

. The role of health risk factors and disease on worker productivity. J Occup Environ Med 1999;41(10):863.

Alavinia

, Molenaar

, Burdorf

. Productivity loss in the workforce: Associations with health, work demands, and individual characteristics. Am J Ind Med 2009;52(1):49.

Mroz

, Carlini

, Archer

, Wegener

, Hoolachan

, Stiers

, Shore

, Castillo

. Frequency and cost of claims by injury type from a state workers’ compensation fund from through Arch Phys Med Rehabil (2014;95(6):1048.

David

. Ergonomic methods for assessing exposure to risk factors for work-related musculoskeletal disorders. Occup Med 2005;55(3):190.

10.

Wiktorin

, Karlqvist

, Winkel

. Stockholm MUSIC I Study GrouValidity of self-reported exposures to work postures and manual materials handling. Scand J Work Environ Health 1993;19(3):208.

11.

Wiktorin

, Hjelm

, Winkel

, Köster

. Stockholm MUSIC I Study GrouReproducibility of a questionnaire for assessment of physical load during work and leisure time. J Occup Environ Med 1996;38(2):190.

12.

Barrero

, Katz

, Dennerlein

. Validity of self-reported mechanical demands for occupational epidemiologic research of musculoskeletal disorders. Scand J Work Environ Health 2009;35(4):245.

13.

Hansson

, Balogh

, Byström

, Ohlsson

, Nordander

, Asterland

, Sjölander

, Rylander

, Winkel

, Skerfving

. Malmö Shoulder-Neck Study GrouQuestionnaire versus direct technical measurements in assessing postures and movements of the head, upper back, arms and hands. Scand J Work Environ Health 2001;27(1):30.

14.

Balogh

, Ørbæk

, Ohlsson

, Nordander

, Unge

, Winkel

, Hansson

; Malmö Shoulder/Neck Study Group. Self-assessed and directly measured occupational physical activities—influence of musculoskeletal complaints age and gender. Appl Ergon 2004;35(1):49.

15.

van der Beek

, Frings-Dresen

. Assessment of mechanical exposure in ergonomic epidemiology. Occup Environ Med 1998;55(5):291.

16.

Sutherland

, Albert

, Wrigley

, Callaghan

. A validation of a posture matching approach for the determination of 3D cumulative back loads. Appl Ergon 2008;39(2):199.

17.

Chang

, McGorry

, Lin

, Xu

, Hsiang

. Prediction accuracy in estimating joint angle trajectories using a video posture coding method for sagittal lifting tasks. Ergonomics 2010;53(8):1039.

18.

Coenen

, Kingma

, Boot

CRL

, Faber

, Xu

, Bongers

, van Dieën

. Estimation of low back moments from video analysis: A validation study. J Biomech 2011;44(13):2369.

19.

Neumann

, Wells

, Norman

, Kerr

, Frank

, Shannon

. Trunk posture. Int J Ind Ergon 2001;28(6):355.

20.

Coenen

, Kingma

, Boot

CRL

, Bongers

, van Dieën

. Inter-rater reliability of a video-analysis method measuring low-back load in a field situation. Appl Ergon 2013;44(5):828.

21.

, Buckle

. Current techniques for assessing physical exposure to work-related musculoskeletal risks, with emphasis on posture-based methods. Ergonomics 1999;42(5):674.

22.

Mathiassen

, Liv

, Wahlström

. Cost-efficient measurement strategies for posture observations based on video recordings. Appl Ergon 2013;44(4):609.

23.

Rezagholi

, Mathiassen

, Liv

. Cost efficiency comparison of four video-based techniques for assessing upper arm postures. Ergonomics 2012;55(3):350.

24.

Weir

, Andrews

, van Wyk

, Callaghan

. The influence of training on decision times and errors associated with classifying trunk postures using video-based posture assessment methods. Ergonomics 2011;54(2):197.

25.

Callaghan

, Jackson

, Albert

, Andrews

, Potvin

. The design and preliminary validation of ‘3DMatch’ – A posture matching tool for estimating three dimensional cumulative loading on the low back. Proceedings of the 34th Annual Conference of the Association of Canadian Ergonomists. London, Ontario, Canada, 2003 [CD ROM].

26.

McGill

, Norman

, Cholewicki

. A simple polynomial that predicts low-back compression during complex 3-D tasks. Ergonomics 1996;39(9):1107.

27.

Sutherland

, Albert

, Wrigley

, Callaghan

. A validation of a posture matching approach for the determination of 3D cumulative back loads. Appl Ergon 2008;39(2):199.

28.

Parkinson

, Bezaire

, Callaghan

. A comparison of low back kinetic estimates obtained through posture matching, rigid link modeling and an EMG-assisted model. Appl Ergon 2011;42(5):644.

29.

Sullivan

, Bryden

, Callaghan

. Inter- and intra-observer reliability of calculating cumulative lumbar spine loads. Ergonomics 2002;45(11):788.

30.

Cann

, Connolly

, Ruuska

, MacNeil

, Birmingham

, Vandervoort

, Callaghan

. Inter rater reliability of a posture matching assessment approach: A pilot study with food service worker. Ergonomics 2008;51(4):556.

31.

Sutherland

, Albert

, Wrigley

, Callaghan

. The effect of camera viewing angle on posture assessment repeatability and cumulative spinal loading. Ergonomics 2007;50(6):877.

32.

Norman

, Wells

, Neumann

, Frank

, Shannon

, Kerr

. A comparison of peak vs cumulative physical work exposure risk factors for the reporting of low back pain in the automotive industry. Clin Biomech 1998;13(8):561.

33.

Seaman

, Albert

, Weldon

NRE

, Croll

, Callaghan

. Biomechanical shoulder loads and postures in light automotive assembly workers: Comparison between shoulder pain/no pain groups. Work 2010;35:39.

34.

McClellan

, Albert

, Fisher

, Seaman

, Callaghan

. Shoulder loading while performing automotive parts assembly tasks: A field study. Occup Ergon 2008/2009;8:81.

35.

Ulin

, Chaffin

, Patellos

, Blitz

, Emerick

, Lundy

, Misher

. A biomechanical analysis of methods used for transferring totally dependent patients. SCI Nurs 1997;14(1):19.

36.

Jäger

, Jordan

, Theilmeier

, Wortmann

, Kuhn

, Nienhaus

, Luttmann

. Lumbar-load analysis of manual patient-handling activities for biomechanical overload prevention among healthcare workers. Ann Occup Hyg 2013;57(4):528.

37.

Dainty

, Alcorn

, Ferguson

, Gregory

. Prevalence of occupation related pain among baristas and an examination of low back and shoulder demand during the preparation of espresso-based beverages. Ergonomics 2014;57(8):1192.

38.

Azar

, Godin

, Andrews

, Callaghan

. Three-dimensional peak and cumulative L4/L5 spine loads and trunk postures during non-occupational tasks. Occup Ergon 2010;9(3):127.

39.

Azar

, Iley

, Godin

, Callaghan

, Andrews

. Three-dimensional peak and cumulative shoulder loads and postures during non-occupational tasks: A preliminary investigation. WORK 2014;47(1):73.

40.

Litchfield

. Agricultural work related injury and ill-health and the economic cost. Environ Sci and Pollut Res 1999;6(3):175.

41.

Fathallah

. Musculoskeletal disorders in labor-intensive agriculture. Appl Ergon 2010;41(6):738.

42.

Kirkhorn

, Earle-Richardson

, Banks

. Ergonomic risks and musculoskeletal disorders in production agriculture: Recommendations for effective research to practice. J Agromed 2010;15(3):281.

43.

Gregory

, Pal

, Carman

, Milosavljevic

, Callaghan

. An examination of shoulder postures and moments of force among different skill levels in the wool harvesting industry. Int J Occup Saf Ergon 2009;15(4):409.

44.

Gregory

, Laughton

, Carman

, Milosavljevic

, Callaghan

. Trunk postures and peak and cumulative low back kinetics during upright posture sheep shearing. Ergonomics 2009;52(12):1576.

45.

Meyers

, Faucett

, Tejeda

, Kabashima

, Miles

, Janowitz

, Duraj

, Smith

, Weber

. High risk tasks for musculoskeletal disorders in agricultural field work. Proceedings of the 44^th Annual Meeting of the Human Factors and Ergonomics Society. San Diego, California, USA, 2000.

46.

Meyers

, Miles

, Faucett

, Janowitz

, Tejeda

, Weber

, Smith

, Garcia

. Priority risk factors for back injury in agricultural field work: Vineyard ergonomics. J Agromed 2004;9(2):433.

47.

Maeda

, Okazaki

, Suenaga

, Sakurai

, Takamatsu

. Low back pain related to bowing posture of greenhouse farmers. J Hum Ergol 1980;9(2):117.

48.

Palmer

. Musculoskeletal problems in the tomato growing industry: “Tomato trainer’s shoulder?” Occ Med 1996;46:428.

49.

Callejón-Ferre

, Pérez-Alonso

, Sánchez-Hermosilla

, Carreño-Ortega

. Ergonomics and psycho-sociological quality indexes in greenhouses, Almería (Spain). Span J Agric Res 2009;7(1):50.

50.

Statistics Canada. Table 282-0008 – Labour force survey estimates (LFS), by North American Industry Classification System (NAICS), sex and age group (CANSIM database) [Internet]. Ottawa: Statistics Canada; 2016 [updated: 2016 Jan 8; cited: 2016 Feb 7]. Available from: http://www5.statcan.gc.ca/cansim/a26?lang=eng&retrLang=eng&id=2820008&&pattern=&stByVal=1&p1=1&p2=37&tabMode=dataTable&csid=

51.

Statistics Canada. Greenhouse, sod and nursery industries, 2011 (cat. no 22-202-X) [Internet]. Ottawa: Statistics Canada; 2012 [cited 2015 Mar 28]. Available from: http://www.statcan.gc.ca/pub/22-202-x/22-202-x2011000-eng.pdf.

52.

Statistics Canada. Greenhouse, sod and nursery industries, 2013 [Internet]. Ottawa: Statistics Canada; 2014 [cited 2015 Mar 28]. Available from: http://www.statcan.gc.ca/daily-quotidien/140424/d140424b-eng.htm.

53.

Bernard

. Musculoskeletal disorders and workplace factors: A critical review of epidemiologic evidence for work-related musculoskeletal disorders of the neck, upper extremity, and low back. Cincinnati, OH, Department of Health and Human Services, National Institute for Occupational Safety and Health (NIOSH), 1997.

54.

van der Beek

, Frings-Dresen

, van Dijk

, Kemper

, Meijman

. Loading and unloading by lorry drivers and musculoskeletal complaints. Int J Ind Ergon 1993;12(1):13.

55.

Hoozemans

, van der Beek

, Frings-Dresen

, van der Woude

, van Dijk

. Pushing and pulling in association with low back and shoulder complaints. Occup Environ Med 2002;59(10):696.

56.

Newell

, Kumar

. Comparison of instantaneous and cumulative loads on the low back and neck in orthodontist. Clin Biomech (Bristol, Avon) 2005;20(2):130.

57.

Visser

, van Dieën

. Pathophysiology of upper extremity muscle disorders. J Electromyogr Kinesiol 2006;16(1):1.

58.

Andrews

, Callaghan

. Determining the minimum sampling rate needed to accurately quantify cumulative spine loading from digitized video. Appl Ergon 2003;34:589.

59.

Lowe

, Weir

, Andrews

. Observation-based posture assessment: Review of current practice and recommendations for improvement. US Department of Health and Human Services, Centers for Disease Control and Prevention, National Institute for Occupational Safety and Health, Cincinnati, OH: U.S. DHHS (NIOSH) Publication 2014-131 (2014). http://www.cdc.gov/niosh/docs/2014-131/default.html

60.

NIOSH, Work Practices Guide for Manual Lifting, NIOSH Technical Report No. 81-122, US Department of Health and Human Services, National Institute for Occupational Safety and Health, Cincinnati, OH, 1981.

61.

Lee

, Chaffin

, Herrin

, Waikar

. Effect of handle height on lower-back loading in cart pushing and pulling. Appl Ergon 1991;22(2):117.

62.

Hoozemans

, Kuijer

, Kingma

, van Dieën

, de Vries

, van der Woude

, Veeger

, van der Beek

, Frings-Dresen

. Mechanical loading of the low back and shoulders during pushing and pulling activities. Ergonomics 2004;47(1):1.

63.

Knapik

, Marras

. Spine loading at different lumbar levels during pushing and pulling. Ergonomics 2009;52(1):60.

64.

Al-Eisawi

, Kerk

, Congleton

, Amendola

, Jenkins

, Gaines

Jr . The effect of handle height and cart load on the initial hand forces in cart pushing and pulling. Ergonomics 1999;42(8):1099.

65.

Lett

, McGill

. Pushing and pulling: Personal mechanics influence spine loads. Ergonomics 2006;49(9):895.

66.

Kumar

. Cumulative load as a risk factor for back pain. Spine 1990;15(12):1311.

67.

Seidler

, Bolm-Audorff

, Heiskel

, Henkel

, RothKuver

, Kaiser

, Bickeboller

, Willingstorfer

, Beck

, Elsner

. The role of cumulative physical work load in lumbar spine disease: Risk factors for lumbar osteochondrosis and spondylosis associated with chronic complaints. Occup Environ Med 2001;58:735.

68.

Coenen

, Kingma

, Boot

, Twisk

, Bongers

, van Dieën

. Cumulative low back load at work as a risk factor of low back pain: A prospective cohort study. J Occup Rehabil 2013;23(1):11.

69.

May

, Scribani

, Wyckoff

, Bauer

, May

, Wyckoff

, Jenkins

. An ergonomic assessment of the long handle blueberry harvesting rake. Am J Ind Med 2012;55(11):1051.

70.

Callaghan

, Salewytsch

, Andrews

. An evaluation of predictive methods for estimating cumulative spine loading. Ergonomics 2001;44(9):825.

71.

Callaghan

, Keown

, Andrews

. Influence of dynamic factors on calculating cumulative low back loads. Occup Ergon 2005;5(2):89.

72.

Waters

, Yeung

, Genaidy

, Callaghan

, Barriera-Viruet

, Abdallah

, Kumar

. Cumulative spinal loading exposure methods for manual material handling tasks. Part 2: Methodological issues and applicability for use in epidemiological studies. Theor Issues Ergon 2006;7(2):131.

73.

Sakakibara

, Miyao

, Kondo

, Yamada

. Overhead work and shoulder-neck pain in orchard farmers harvesting pears and apples. Ergonomics 1995;38(4):700.

74.

Fulmer

, Punnett

, Slingerland

, Earle-Richardson

. Ergonomic exposures in apple harvesting: Preliminary observations. Am J Ind Med 2002;(Suppl 2):3.

75.

Silverstein

, Bao

, Fan

, Howard

, Smith

, Spielholz

, Bonauto

, Viikari-Juntura

. Rotator cuff syndrome: Personal, work-related psychosocial and physical load factors. J Occup Environ Med 2008;50(9):1062.

76.

Roquelaure

, Ha

, Rouillon

, Fouquet

, Leclerc

, Descatha

, Touranchet

, Goldberg

, Imbernon

; Members of Occupational Health Services of the Pays de la Loire Region. Risk factors for upper extremity musculoskeletal disorders in the working population. Arthritis Rheum 2009;61(10):1425.

77.

Chaffin

, Andres

, Garg

. Volitional postures during maximal push/ pull exertions in the sagittal plane. Human Factors 1983;25(5):541.

78.

Andrews

, Arnold

, Weir

, van Wyk

, Callaghan

. Errors associated with bin boundaries in observation-based posture assessment methods. Occup Ergon 2008;8:11.

79.

Trask

, Mathiassen

, Rostami

. Partly visible periods in posture observation fromvideo: Prevalence and effect on summary estimates of postures in the job. Appl Ergon 2015;49:63.