Manual handling has been associated with an increased risk of back disorders [1
], mainly lower back pain [2
]. In the industrial sector, most tasks have been automated, although manual handling is still being carried out when it cannot be avoided. Furthermore, people are not only exposed to manual handling at work, but there also are many situations in daily life where it is required and can increase the risk of back disorders. Hence, European Union directives, such as Council Directive 90/269/EEC and national legislations of member states have been outlining appropriate MMH training [3
]. Considering the requirement for MMH training in workplaces, numerous studies have been focused on developing, implementing, and evaluating the efficacy of manual handling training on reduction of lower back pain or back injury prevention [4
]. It has been observed that fewer studies have been conducted on the effectiveness of manual handling training in industries outside of the healthcare sector [5
Different systematic reviews have concluded that the majority of manual handling training procedures are not effective to reduce lower back pain or back injury [5
]. This lack of effectiveness is associated with one-dimensional interventions, lack of transferability of training, and a lack of training based on the theory of changing health behaviour. Moreover, studies tend to evaluate the effectiveness of training on long-term results, such as reduction in musculoskeletal disorders [5
], and have omitted intermediate variables. Hogan et al. [7
] pointed out that there is the need to evaluate the effect of manual handling training on intermediate variables, such as knowledge, behaviour change, and training transferability. To date, few studies have evaluated the effect of MMH training on knowledge and behaviour change [5
In the last decades, studies focused on MMH training have evaluated different training methods. They tended to combine lifting training with back school, lumbar support, verbal feedback, practice, biofeedback, warm-up exercise, and ergonomic redesign of equipment. In addition, physical exercise has been highlighted as a component of MMH training [5
] because it helps to improve leg strength and consequently avoid back loading [8
However, one of the current challenges is how to actively involve employees in the training, taking into account that the higher the involvement in the training, the more effective it will be [9
]. In this sense, a training method that could emphasize the active role of the employee in the training is that of self-observation. Self-observation is a relevant method to promote behaviour change, because it influences self-confidence, self-awareness, and self-efficacy, and has an emotional impact [10
]. Self-observation has been evaluated in interventions focused on behaviour change, such as in the treatment of attention deficit hyperactivity disorder, autism, emotional disturbances, speaking ability, and so on [12
], and for skill acquisition in sport [13
]. Moreover, self-observation has been applied in different training programs to improve the interaction skills of professionals [15
], and to increase one’s own teaching [16
]. Nonetheless, there is little evidence of self-observation as a formative method in training for risk prevention and health promotion at work, and specifically in MMH training. As far as we know, self-observation (specifically self-modelling) has been applied as a training method for reducing musculoskeletal risk among office workers using computers, showing positive effects [17
]. It has also been implemented to improve movement awareness among nursing students, who highlighted its usefulness and its cognitive impact with regard to changing body postures during patient transfer [18
]. In all these cases, self-observation was done through video.
In the learning process of motor activities (e.g., MMH technique), video self-observation can provide two types of external feedback: knowledge of results and knowledge of performance. Usually, video self-observation tends to be used to give knowledge of performance [20
]. However, video self-observation can provide a lot of information that could have a negative effect on employee attention. This is why video self-observation should be complemented with external feedback from the technician, who should give transitional information that reports what is well done [21
], what should be improved, and how it should be done—also called feedforward [10
]. We call this type of external feedback hetero-observational feedback and feedforward. Even so, it is important to consider that the employee will not always receive these external feedbacks because the training takes a particular period of time. Therefore, the training should also promote the use of intrinsic feedback (information provided from intrinsic sources, such as vision and proprioception) with the proposal of making the employees aware of their movements and positions adopted. Schmidt and Lee [20
] suggested that it is this intrinsic feedback that one must learn to interpret, because it will always be available to the learner. Thus, by promoting intrinsic feedback, employees will be capable of regulating themselves once the training has finished.
Additionally, another way to conduct self-observation in a natural environment is through ambulatory assessment (e.g., using a self-report questionnaire) [22
]. Studies have shown that ambulatory assessment as an intervention component promotes self-awareness and behavioural change [19
]. However, it is important to take into account that these behaviour changes can also be caused by reactivity [22
The positive effects of self-observation, hetero-observational and intrinsic feedbacks have been widely studied in sport skill acquisition [20
], even though there is a lack of studies on risk prevention and health promotion.
Therefore, the aim of this study was to assess the effect of the previous components on worker’s knowledge and behaviour of the MMH technique by employing a randomised controlled trial. We use the name SsObserWork to identify the intervention. Specifically, we tested and evaluated the effect of the main components of SsObserWork, which are systematic self-observation (SSO), hetero-observational feedback and feedforward (HFF), and intrinsic feedback. It was hypothesized that these three main components improved employee knowledge and behaviour of the MMH technique.
2. Materials and Methods
2.1. Study Design
A parallel randomised trial of two groups was conducted among blue-collar workers recruited from a food processing company in Catalonia, Spain. Participants were randomly allocated to the SsObserWork or control groups. To evaluate the intervention effect, we adopted a methodological complementarity perspective [27
], combining experimental design with observational methodology. Observational methodology is essential in this study in order to evaluate behaviour, to implement the self-observation component, and to plan the data collection [28
]. In this sense, according to observational methodology, the study adopted a follow-up, nomothetic, and multidimensional design [28
The Ethics Committee of the Autonomous University of Barcelona approved the study protocol (PRO1742). Prior to commencement of the intervention, the participants were made aware of their right to withdraw from the study at any time and steps to safeguard information. In accordance with the principles of the Declaration of Helsinki, the participants were informed that they were being recorded. They were shown the location of the video cameras, which were positioned directly to minimize reactivity bias. Informed consent was also obtained.
2.2. Participants and Procedure
We used a nonprobability sampling. A convenience sample was initially composed of 103 blue-collar workers at a leading company in the Spanish meat sector, certified by the food safety regulations of the International Food Standards (IFS) and the British Retail Consortium (BRC). Completing the questionnaire and participating in the study was voluntary. The questionnaires were collected during September and October 2015. From the first sample, purposive sampling was applied, using the information provided in the questionnaire. The inclusion criteria were (i) to be older than 18 years old, (ii) to be employed as a blue-collar worker, (iii) not suffer any chronic bone, muscle, or joint disease in the trunk, and/or chronic or acute pain diagnosed by a specialist, and (iv) not suffer any chronic or acute knee joint disease diagnosed by a specialist. Of the 103 employees who completed the questionnaire, 65 were eligible and willing to participate. The 65 participants were randomly assigned to the SsObserWork or control group (Figure 1
The SsObserWork intervention was designed taking into consideration the Health Belief Model [32
] and trans-theoretical model of behaviour change [33
]. The aim of the intervention was to raise awareness and promote the adoption of proper postural back habits during MMH tasks, considered as complex tasks and a risk factor to develop lower back pain [2
]. We considered that by adopting a proper bask posture during MMH, it could be transferred to other working and daily life tasks.
The SsObserWork intervention was made up of components, formative activities, and didactic materials, which are described in Appendix A
. The intervention consisted of two sessions and a three-week follow-up between sessions. The intervention was implemented by a trained physical activity specialist. Figure 2
shows the intervention structure and how the components, formative activities, didactic materials, and data collection were distributed orderly. As shown, SSO (including self-report questionnaire), HFF, and intrinsic feedback are the components that evaluated their effect. Derivatively, MMH practice (which promotes intrinsic feedback) and messages to remind workers to answer the self-report questionnaire were only implemented in the SsObserWork group. Instead of the SsObserWork components, the control group received a standard MMH training based on theoretical information. The theoretical information consisted of explaining how to perform a MMH task and without any type of practice. Each SsObserWork and control session was carried out individually with each participant during November and December 2015.
Baseline characteristics of participants. An ad hoc questionnaire was made to select the sample. It consisted of demographic, social, and health questions, and it included the Standardised Nordic Questionnaire [34
], SF-12v2 Health Survey [35
], and the Utrecht Work Engagement Scale [36
] and stages-of-change Items [37
MMH behaviour: MMH behaviour was measured with an ad hoc instrument called the MMH-SsObserWork instrument. This observational instrument allows to identify and describe the body positions adopted continually during MMH tasks by using a set of criteria and categories. The criteria are: feet, knee joint, back, elbow joint, load position, and interaction between back tilt and move around. At the beginning of each session, workers had to lift, carry, and lower five boxes (8 kg each) while they were recorded from the sagittal plane. The video camera was positioned at the height of the worker’s hip. The MMH-SsObserWork instrument had to be used using a software application for the record and computation of observational data (e.g., Lince) in which the video recording can be displayed. The MMH-SsObserWork instrument has a very good inter-observer reliability [38
]. According to Bakeman et al. [39
] for time-event data, the global kappa index based on time units ranged from 0.90 to 0.97 for all criteria, and the global kappa index based on events ranged from 0.72 to 0.87 for all criteria.
MMH Knowledge: MMH knowledge was measured with a instrument developed and based on the MMH-SsObserWork instrument [38
]. This instrument consists of five criteria related to the main parts of the body involved in the MMH task (feet, knee joint, back, elbow joint), and the load position. Each criterion has its categories that describe the different positions that can be adopted. There are 16 categories in total. Categories are represented with pictures and a brief description. Workers had to indicate, for each criterion, which position was the most recommended to adopt during the MMH task by choosing a category. They had to indicate the most recommended for each MMH phase (lifting, carrying, and lowering). Hence, there were three identifications for each participant.
2.5. Data Management
To analyse MMH knowledge, we calculated the number of recommended positions identified for each criterion, and then the score change was computed in each criterion comparing the different moments (pre and post the first session; pre-first session and post-second session). In this sense, it was implied that a positive value meant an increase in the frequency of recommended positions identified.
To analyse MMH behaviour, the Lince software [40
] was used to generate the observational record for each worker and session. The observational record provided information of time duration (frames and milliseconds) of each category for each criterion. This observational record was exported to MS-Excel. We did a recording of categories for each criterion, determining if they were placed in the recommended or non-recommended position, taking into account the literature review of MMH technique. As has been justified in previous works [38
], the recommended positions established were: feet placed asymmetrically, one beside the load and the other behind it; knees in a semi-squat position (moderate flexion); a neutral back position at any front inclination of back; load placed closed to the body; arms extended or slightly flexed, and do not start or finish carrying phase while lifting and lowering phases are being carried out. The relative duration in which each criterion was found in the recommended position was calculated (time unit of recommended category as regards total duration).
The change scores for the different dependent variables (MMH knowledge and behaviour) were computed in order to control the hypothetical baseline effect.
2.6. Statistical Analysis
A basic statistical description had been made for each dependent variable, by group and time moment. The aim was to characterize the empirical distribution obtained. We assessed the assumption of normal distribution of variables using the Shapiro-Wilk test.
During the three-week follow-up, only 25 out of 31 participants from the SsObserWork group that initiated the follow-up answered the self-report questionnaire. Thirteen participants answered it between 11–15 days, 7 participants answered it between 6–10 days, and 5 participants answered it between 1–5 days. Due to the fact that there was irregular participation in answering the self-report questionnaire, we tested whether there were statistical differences in the dependent variables among those participants from the SsObserWork group who answered the self-report depending the number of days. For non-normal distributed variables, the differences were assessed using the Kruskal-Wallis test. For normally distributed variables, the differences were assessed using one-way analysis of variance.
The components effect on outcomes (MMH knowledge and behaviour) was assessed using independent samples t-test or Mann-Whitney U test on change scores, depending on the assumption of normality. The Mann-Whitney U test has been used when normality cannot be assumed. We specify the statistical test applied (t-test or Mann-Whitney) in Tables 2 and 3. Regarding the change score, a positive value indicates a change in the direction defined by the hypothesis.
Cliff’s δ statistics [41
] were applied to assess component effect size for data that did not pass the normality test. The amount of effect sizes was interpreted as trivial (<0.147), small (between 0.147 and 0.33), medium (between 0.33 and 0.474), or strong (>0.474) [42
]. For data that passed the normality test, effect sizes were examined using Cohen’s d
statistics. Effect sizes were assumed as trivial (<0.20), small (between 0.20 and 0.49), medium (between 0.50 and 0.79), or large (>0.80) [43
]. Analyses were performed using SPSS version 23 (SPSS Institute, Cary, NC, USA). However, R programme version 3.3.3 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria) was used to analyse effect sizes. Statistical significance was set at p
We conducted a supplementary analysis focused on delving into the temporal structure of employee behaviour, specifically with participants of the SsObserWork group who experienced an improvement in back position. The type of observational instrument and observational design that we used and adopted allowed to conduct a T-pattern analysis. This type of analysis allows to detect hidden or non-obvious temporal patterns in behaviour that are not always visible. Our aim was to detect T-patterns that involved a recommended back position, and we compared them between sessions of the SsObserWork group. From the observational record of each SsObserWork participant and session, we could detect the criteria co-occurrences during the MMH period (we called it a forward event). To detect temporal regularities in the order of event occurrences, we used the detection algorithm developed by Magnusson [44
], and implemented in THEMETM
software (Patternvision Ltd., Reykjavík, Iceland). This detection algorithm first identifies significant (non-random) recurrences of any two events within a similar temporal configuration (critical interval) in real-time behavioural data and then proceeds to identify a hierarchical relationship with any other antecedent or subsequent events. This statistical method of detecting temporal patterns (T-patterns) of related behavioural events provides behavioural structures that may not be identifiable by traditional sequential methods [46
]. Data was analysed using Theme 6.0. Default temporal patterns search parameters were used, an acceptable level of significance was set at 0.005, minimum occurrences at 3, and minimum percentage of samples (workers on this data) in which a pattern must occur, was set at 51%. The results have been validated by simulation, through randomization of data on five occasions, with acceptance only of patterns for which the probability of the randomized data coinciding with the real data is zero. The T-pattern differences between sessions were assessed using Pearson’s chi-squared test.
We hypothesised that the implementation of SSO, HFF, and intrinsic feedback would lead to improving employee knowledge and behaviour of the MMH technique. This hypothesis was confirmed. The SsObserWork group improved their knowledge and behaviour compared to the control group that received standard training.
As regards employee knowledge of the MMH technique, there was an increase in the recommended positions identified in both groups and comparing between times (before and after the first session, and comparing identifications at the beginning of the first session with ones conducted in the second session). The results showed that both interventions (SsObserWork and standard training) had a positive effect on employee knowledge, but knowledge of the recommended back position was significantly higher in the SsObserWork group, and this improvement was maintained between sessions. The SSO with the HFF allowed workers to focus their attention on detecting and discerning recommended and non-recommended back positions, compared to a general explication of how to perform an MMH task. The fact that workers could be observers of their own behaviour was more relevant because it had an emotional impact and gave them the opportunity to focus on qualitative aspects of their behaviour, identifying recommended and non-recommended positions.
As regards the control group, the results showed a decrease in the number of recommended identifications of back position between the identifications made at the beginning of the first session and with those made in the second session. In this case, the technician explained the instrument again to the control group in the second session, as was done with the SsObserWork group. This is why it is difficult to find an explanation for this result. However, we hypothesise that the standard training gave them general information on back position that could be retained for a short period (in the same session), but this information could be confused after a longer period. In fact, few studies have evaluated the intervention effect on knowledge of the MMH technique that could help reach a clearer conclusion. Additionally, some of these studies did not describe the questionnaire and some others used general questions about MMH [7
Some studies observed that their intervention had an effect on the knowledge of MMH technique, but it did not have an effect on their behaviour [7
]. The present study has not only observed effects on knowledge, but also on behaviour. The results showed that the SsObserWork group significantly increased the relative duration in which the back, knee joints, elbow joints and interaction between back tilt and move around were in recommended positions during the MMH task between the first and second session and comparing with the control group. On looking deeper into the results of the back position of the SsObserWork group, the T-pattern analysis showed a statistical increase in the events type that included a recommended back position, and this analysis allowed studying the change in the temporal structure of these events. After the intervention, it was observed that the temporal structure of behaviour was reduced, and there was a statistical increase in the percentage of patterns with recommended positions. These results suggest that the intervention not only had an effect on increasing recommended positions, but on also modifying the movement pattern in the MMH performance. Hence, the SSO, HFF, and intrinsic feedback contributed to adopting a neutral back, as well as recommended positions of knee joints and elbow joints during MMH tasks. These improvements are relevant because MMH training should not only focus on back but also on other body parts that are involved actively, and could effect the forces and posture of the back [48
The SSO allowed workers to realize the need for a change in their MMH performance by active error detection, which was augmented with the HFF that directed their attention, something that is necessary because of so much information offered in the video. Watching oneself has an emotional impact that helps to attract the attention of the workers, which does not happen when theoretical information is given without any connection to them. In fact, when workers were observing themselves, they were astonished because they figured that they adopted recommended positions. It does not mean that they reacted negatively when they watched themselves, something that happened in the study of Linnerooth et al. [49
], who associated the negative reactions of workers with non-self-observation effects. This is an issue that has to be considered because not everyone likes watching their own behaviour. Other studies have highlighted the positive effects of self-observation in other fields, such as improving teaching [16
] and sports skill acquisition [13
]. In the occupational health field, Taieb-Maimon et al. [17
] observed that combining ergonomic training with self-modelling had more positive effects on workers’ postures than traditional ergonomic training, and the improvement lasted over time. Furthermore, Backåberg, Rask et al. [18
], observed that self-observation promoted body self-awareness. In their qualitative study, they identified that participants pointed out the positive effects of describing verbally the body positions adopted at the beginning of the session [19
]. As regards the results of Backåberg et al. [18
], the present study complemented the SSO and HFF with promoting the intrinsic feedback. The aim of promoting the intrinsic feedback was to raise body self-awareness in order to make the employees capable of paying attention to their movements and self-correction, because after the training, they rely on their intrinsic feedback [26
]. In this case, the HFF provided knowledge as regards ensuring that intrinsic feedback was being interpreted correctly. This suggests that technician feedback has to be provided after interpreting intrinsic feedback to require cognitive effort [20
Strengths and Limitations
The main strength of this study is the methodological complementarity by combining elements of the experimental design with those of observational methodology. This multi-method perspective allowed integrating data from an experimental and observational design, providing richer data to explain employee behaviour change after the intervention. The observational data obtained by using the MMH-SsObserWork instrument allowed characterizing the behaviour change from two perspectives: duration in recommended positions and T-patterns detection. As far as we know, this is the first study in occupational health field that assess behaviour change by implementing T-pattern analysis, which provides relevant information of how the training effects the pattern of the workers’ behaviour. Additionally, to our knowledge, this is the first study that assessed the effect of SSO, HFF, and intrinsic feedback as components of MMH training in the industrial sector, implemented during working hours.
As regards limitations, few participants of the SsObserWork group answered the self-report questionnaire during the follow-up period. Thus, this component was not implemented as expected due to the lack of compliance by the workers, even though we sent them reminders by text messages every day. Probably, the reason for the lack of follow-up of the self-report questionnaire was the effort required by the designed reporting system. The aim of the self-report questionnaire was to promote reactivity and collect data to identify whether the SsObserWork training was transferable. However, we did not identify significant differences between workers who completed the self-report questionnaire most days with those who did not complete it. The self-report questionnaire could have helped obtain information related to training transferability into their daily working life. This is essential to know if there is to be optimal feasibility and acceptability of the intervention and to identify barriers and facilitators to transfer it [4
]. Therefore, we have to work on finding a better system to ensure worker compliance.
The next step is to assess transferability in their working and daily life, and to observe if the improvement in knowledge and behaviour will last over time. We are aware of the complexity and industriousness of the intervention’s approach, which is why our future aim is to tailor the training by focusing on key workers, such as older workers or supervisors. The objective will be focused on training them to be trainers in their professional and personal environments. Efficacy will be under well-controlled circumstances and will need to improve effectiveness in a real working-life situation. However, it is methodologically complex to implement and assess this type of training in a company if the requirement to provide behavioural evidence is maintained, as has been done in the present study.