An investigation of implicit bias about bending and lifting

Introduction

Bending and lifting are commonly reported risk factors for low back pain (LBP) described in the literature [1, 2]. Findings from early in vitro [3] and in vivo [4] studies led to an interpretation that lifting tasks should be performed with a “straight-back” posture (i.e. extended lumbar spine) to minimise the risk of potential damage and LBP. This has influenced manual handling advice across the globe and it is used in manual handling regulations as best practice to date [5, 6]. Not surprisingly, there is a common understanding amongst pain-free people and people with LBP, as well as healthcare professionals that bending and lifting with a round-back (i.e. flexed lumbar spine) is dangerous and it is potentially harmful for the back [7], [8], [9].

However, recent systematic reviews have questioned the long-held belief that bending and lifting with a round-back is dangerous and should be avoided [10, 11]. Saraceni et al. [10] reported there is no evidence that lifting with a flexed lumbar spine is a risk factor for LBP onset or persistence. In addition, the degree of lumbar spine flexion during lifting is not a differentiator between people with and without LBP. In fact, Nolan et al. [11] suggested that the lifting style presented by people with LBP may in fact reflect the lifting advice provided during manual handling training (i.e. more squat and straight back).

Beliefs are thought to influence behaviour [12], [13], [14]. Beliefs can be described as a mental process where a person adds meaning to an experience (e.g. fear, pain and movement) [15]. Thus, beliefs may be held explicitly, meaning that the person can access it deliberately, or they can be held implicitly (implicit bias), through automatic associations in memory, which the person may not be aware of [16, 17]. An example of this is found in the study of Caneiro et al. [18] that demonstrated that what people say (self-report) does not always reflect what they ‘think’ (implicit bias) about movement and perceived threat to the body. In that study, people self-reported low to moderate fear of bending, but when asked in a context of time pressure using a computer based task that minimised time for reflection, participants implicitly associated images of bending and lifting with a round-back with words that denote danger. Further studies in this field, found the same implicit bias in pain-free people [19], and also in healthcare professionals that provide treatment for people with LBP [20].

However, these studies have some limitations as they (i) used a single group, precluding direct comparisons studies; (ii) did not include people who had recovered from LBP, limiting the evaluation of whether recovering from LBP would affect beliefs or not; and (iii) only included people from a high-income country (Australia), limiting generalization of results to other countries. Replication studies that address these limitations are needed.

Therefore, the aims of the current study were to overcome these limitations. Specially:

1. To evaluate whether there is an implicit bias between back posture (straight-back vs. round-back) and safety (safe vs. danger) related to bending and lifting in pain-free people, people with LBP, and with a history of LBP in a middle-income country (Brazil) [21];

2. To evaluate whether implicit biases and explicit beliefs about the back differ between these three groups;

3. To explore correlations between implicit and explicit measures of beliefs towards vulnerability of the back (bending safety beliefs, back beliefs, and fear of movement) within these groups.

Based on previous studies [18, 19], we predicted to find an implicit bias between round-back posture and danger in pain-free people and people with LBP, but not in people with a history of LBP. We expect no difference between implicit bias and explicit beliefs between pain-free people and people with LBP. People with a history of LBP are predicted to present implicit biases and explicit beliefs different from the other two groups, as they have had LBP and recovered from it.

Methods

Outcome measures

Implicit biases

The IAT assesses the strength of association between two different target categories through measurement of reaction times, indicating implicit bias [23]. The IAT is a well stablished measure, which was adapted by Caneiro et al. [18] to assess associations between bending and lifting posture and safety in a group of people with LBP, pain-free subjects [19], and physiotherapists [20].

This IAT has five stimuli (images) representing each target (“round-back” and “straight-back”) and attribute category (“safe” and “danger”). The words selected to represent the attribute category ‘Safe’ were: harmless, certainty, protecting, confident, secure; and ‘Danger’ were: alarming, vulnerable, risky, damaging, threatening. The “danger” words were chosen as they are frequently used to describe danger associated with movement [12, 24], and the “safe” words have the opposite meaning while matching in length, frequency, and emotionality [25]. To represent the target categories, ten side view images of males and females standing, bending and lifting an object with a round-back (target category ‘Round-back’) or with a straight back (target category ‘Straight-back’) were used.

The IAT is a computer-based test consisting of 7 phases (Figure 1) [18]. For each phase, the participant was instructed to assign a stimulus (image/word displayed in the centre of the screen) to its category (displayed in the left and right upper hand corner of the screen) by pressing the left or right “Shift” keys, as quickly as possible, while avoiding mistakes. Feedback (“correct” or “wrong”) was provided to participants on each trial. The first two phases are designed for familiarization with the test. In phase 1 (20 trials), participants sorted each of the 10 images twice, into the categories “Round-back” and “Straight-back”. In phase 2 (20 trials), participants sorted the 10 words twice into the categories “Safe” and “Danger”. In phases 3 and 4 (20 and 40 trials each) participants sorted words and images into the combined categories (e.g. Danger/Round-back and Safe/Straight-back). In phase 5 (20 trials) participants sorted images with the location of the categories switched. In phases 6 and 7 (30 and 40 trials each) the category combinations of phases three and four were reversed (e.g. Danger/Straight-back and Safe/Round-back). These sequence of category combinations was counterbalanced across participants to avoid sequence effects. At the beginning of each phase, instructions were given about how to perform the task in that phase [23]. In all phases a black background was used with words presented in bold, 20-point Arial font in white lower case, and images presented embedded in a white square image of 800 × 800 pixels.

Figure 1:

Schematic representation of the Implicit Association Test. Reprinted from “Implicit evaluations and physiological threat responses in people with persistent low back pain and fear of bending”, by J.P. Caneiro, 2017, Scandinavian Journal of Pain, Volume 17, 355–66. Copyright 2017 by De Gruyter. Reprinted with permission.

IAT data processing

Presentation of the task and reaction time recording was controlled by the software DMDX [26]. Each trial began with the display of a fixation cross for 1,000 ms followed by a stimulus (word or image) for 1,000 ms. Interval between trials was 1,000 ms. Reaction time was measured from the time the stimulus appeared on screen to the time when the “shift” key was pressed by the participant. Reaction times are shorter when strongly associated categories are placed together in the same top corner (congruent condition) than in different corners (incongruent condition) [23]. Responses were considered errors if participants were too fast (responses shorter than 100 ms) or if they chose the wrong answer (e.g. press the shift key that indicates round-back for an image of a person with straight-back posture or press the shift key that indicates “safe” for the word “vulnerable”). Responses on error trials were replaced with the individual participant’s average response time plus two times the standard deviation for the congruent and incongruent conditions (error penalty of 2 standard deviations, Lane et al. [27]).

A bias score, called IAT_D-SCORE, was calculated as suggested by Lane et al. [27]. Negative scores indicate implicit bias towards a straight-back posture as dangerous and positive scores indicate implicit bias towards round-back as dangerous. The IAT exhibits adequate reliability and internal, construct and predictive validity [23, 28].

Bending Safety Belief

The BSB was developed by Caneiro et al. (2017) aiming to assess explicit beliefs specifically about posture and safety of the back while bending and lifting [18]. Participants were given a paper-based questionnaire with two images of a person lifting a box, one with a “round-back” and the other with a “straight-back”. The same question (adapted from English to Brazilian-Portuguese) was asked about the two images: “What is the risk of this person to hurt his/her back?”. At the bottom of each image the participant was asked to score on a 11-point Likert scale (where 0=safe and 10=danger), how safe or dangerous he/she thinks that bending and lifting posture is. This is a clinically relevant question when assessing people’s beliefs about bending, how they bend and if there is a perception of danger to this activity (i.e. safety of bending). A final score for perceived danger to the back, called BSB_thermometer, is obtained by subtracting the score given to the image illustrating bending with a “round-back” from the image illustrating bending with a “straight-back”. Score range is from −10 to 10. Therefore, the closer the score is to 10 the higher is the participant’s perceived danger of bending with a round-back, and the closer it is to −10 the higher is the participant’s perceived danger of bending with a straight-back.

Back Pain Attitudes Questionnaire

The Back-PAQ was developed to assess attitudes and beliefs about the back of the general public, people with LBP and healthcare professionals [29]. The Back-PAQ was cross-culturally adapted to Brazilian-Portuguese (Back-PAQ-Br) and shows excellent internal consistency (α=0.92, CI 0.91–0.92) and adequate construct validity [30].

The Back-PAQ-Br has a long version (34 items) and a short version (10 items). Each item is rated on a 5-point Likert scale, ranging from ‘False’ to ‘True’ (intermediate labels: ‘Possibly False’, ‘Unsure’, ‘Possibly True’). Responses are scored from −2 (True) to 2 (False) with 11 items reverse scored (items 1, 2, 3, 15, 16, 17, 27, 28, 29, 30 and 31). Scores range from −68 to 68, with negative scores reflecting unhelpful beliefs whereas positive scores reflect helpful beliefs. The questionnaire covers five different dimensions about the back: (1) vulnerability of the back; (2) relationship between back pain and injury; (3) activity participation during back pain; (4) psychological influences on back pain and; (5) prognosis of back pain [29]. In this study, the long version was used as well as a subscale created by Caneiro et al. [19]. This subscale is composed of items 1–12, 14 and 21, which are items that represent either “vulnerability” or “protection” of the back (Back-PAQ-Br_Danger) [19]. Scores range from −28 to 28, with negative scores representing a perception of vulnerability of the back or that the back needs protection. This allows direct comparison to previous studies [18, 19] and also to other explicit and implicit measures within this study.

Tampa Scale of Kinesiophobia

The TSK was developed to measure patient’s fear of movement, injury or re-injury [31]. It was later adapted to be used in the general population (TSK-G) [32] and healthcare professionals (TSK-HC) [33]. Only the TSK for patients has been cross-culturally adapted to Portuguese [34]. As the factor structure of the TSK, TSK-G [32] and TSK-HC [33] are similar and due to the fact that the Brazilian-Portuguese version of TSK has a strong correlation to Back-PAQ-Br (r=0.72) [30], the TSK was used in all three groups (pain-free, history of LBP, and current-LBP). The TSK consists of 17 items. On a 4-point Likert scale each item is rated as “strongly disagree”, “disagree”, “agree” and “strongly agree”. Items 4, 8, 12 and 16 are reverse scored and final scores range from 17 to 68 [34]. Higher scores indicate higher fear of movement. The Brazilian-Portuguese version of the TSK has shown adequate internal consistency (α=0.95) and reliability (ICC=0.80) [35].

Results

There were 188 participants included in the study. Twelve participants were excluded because they did not complete the online questionnaires, one participant was excluded due to a computer error during the experiment, and one participant was excluded because of low literacy creating difficulties in understanding the meaning of the words in the IAT test. Therefore, data of 174 participants were analysed in the study (mean age 42.8 ± 14.8 years; range: 19–76). Group allocation and sociodemographic data are provided on Table 1. There was no statistical significant difference between groups for BMI, gender and education. There was a significant difference in average age (p=0.015), occupation (p=0.003) and physical activity level (p=0.041) between groups. Tukey HSD post-hoc test indicated that the pain-free group was significantly younger than the current LBP group (38.8 ± 15.2 vs. 46.6 ± 14.0, p=0.013). Adjusted standardized residual indicated that the pain-free group comprised a significant larger number of students and the current LBP group had significant bigger number of healthcare professionals. Also, adjusted standardized residual indicated that the history of LBP group comprised a significant larger number of physically active participants.

Discussion

The aims of this study were to evaluate (i) implicit biases between back posture and safety related to bending and lifting in pain-free people, people with a history of LBP and people with current LBP in a middle-income country, (ii) whether implicit biases and explicit beliefs about the back were different between the groups, and (iii) to explore the correlations between implicit and explicit measures within the groups.

Our results indicate that participants in all three groups displayed a strong implicit bias between images of bending and lifting with a “round-back” and words representing “danger”. Caneiro et al. had similar results when assessing people with persistent LBP [18], pain-free people [19], and physiotherapists in a high-income country [20]. We also investigated beliefs of people with a history of LBP, and no difference was found among the three groups regarding implicit biases. It would be plausible to expect that a person who experienced LBP but recovered from it could have adjusted their beliefs (i.e. the back is not as vulnerable; or, round-back lifting is not dangerous), once the pain experience had subsided. However, our findings suggest this was not the case; people with a history of LBP also associated lifting with a round-back with danger. Thus, independent of the presence of pain, and history of pain with and without recovery, all groups implicitly associated lifting with a round-back and danger. This reinforces the notion of a shared societal belief that round-back lifting is dangerous irrespective of LBP status. Considering that information provided by the media, healthcare professionals, ergonomic advisors, family, and friends can influence a person’s belief [12, 38, 39], it is not surprising that even pain-free people could hold an implicit bias that round-back bending and lifting is dangerous that pre-existed an experience of LBP [18], [19], [20].

Participants in this study also reported unhelpful explicit beliefs about the back as shown by the self-reported measures. This is consistent with studies internationally, where people believe the back is vulnerable and easily harmed [9, 24, 39, 40]. Particularly, people with current LBP and pain-free people perceive bending and lifting with a round-back as dangerous and that it may cause LBP [10, 18, 19, 40, 41]. These explicit beliefs may negatively impact on a person’s behaviour and potentially lead to avoidance of physical and social activities [42, 43], altered movement patterns (e.g. co-contraction of paraspinal and abdominal muscles) [41, 44, 45], and increase pain and disability [46], [47], [48].

Exploration of correlations between implicit (IAT) and explicit measures (BSB_thermometer, TSK, Back-PAQ-Br and Back-PAQ-Br_Danger) indicated very weak and non-significant correlations. These findings are in line with a large meta-analysis which showed that correlations between implicit and explicit measures are often weak or negligible [49]. Caneiro et al. found significant, but weak correlations between IAT and BSB in a sample of pain-free people [19] and physiotherapists [20]. These small differences may be due to sample size as well as cultural characteristics of the samples used in both studies (Brazilians vs. Australians). Explicit beliefs have been shown to influence motor behaviour (i.e. limiting movement due to enhanced protective strategies) [41, 44]. While speculative, if explicit beliefs can drive movement behaviour, it is plausible to infer that implicit biases may also influence movement behaviour. For instance, in a given context such as when under time pressure, a person who may say (explicit belief) that round-back lifting is safe, may lift an object with a straight back posture. In this example, context may favour implicit biases to guide behaviour, influencing how a person performs the task [50]. This highlights the importance of context when evaluating beliefs. For instance, in the clinic, enquiring about beliefs during the interview when the patient is asked to reflect on their beliefs, vs. during the physical examination, when the patient is actually performing a threatening and/or provocative task that may expose their implicit biases.

Studies suggest that regardless if people live in Australia, New Zealand, Brazil, Nigeria or Malawi, people believe that the back is vulnerable [9, 19, 51]. More specifically, our findings suggest people in Australia and Brazil have the same implicit bias in relation to posture (i.e. how one bends and lifts) and safety (i.e. risk of harm). However, there is a lack of evidence to support this “truth” [52], [53], [54]. World leaders in the field of LBP have recently posed a global call for action to change how LBP is conceptualised and managed [55, 56]. As they pointed out, there is a need to change the narrative about LBP using evidence-based information in the hope of influencing care delivery. A global effort from policy makers, public and private healthcare providers, and social services is necessary [56].

There is strong evidence that negative back pain beliefs exist in people with and without LBP [9, 12, 24]. However, for people with LBP, those with more negative beliefs report greater levels of disability and care seeking compared to people with LBP with more positive beliefs [57, 58]. There is also evidence that educational interventions at a population level may reduce levels of disability [59]. However, in a recent clinical trial pain education resulted in a change of pain beliefs but not levels of pain and disability [60], suggesting that other factors mediate this relationship [61]. Together these findings support that once a person develops LBP, negative beliefs may (but not always) act as a mediator for greater levels of disability.

Changing patient’s beliefs about LBP is important to set people on a recovery journey [39, 48]. However, people’s beliefs are also influenced by clinician’s advice and information most readily available to the public (e.g. internet search and friends’ experiences) [12, 38, 39]. Hence, all stake holders, including physiotherapists, may need a tool that educates people in an interactive manner (as opposed to passive education) to reinforce the patient’s new knowledge and that could be widely disseminated (e.g. online interactive education tool).

This study has some limitations. First, there were statistically significant differences between pain-free and current LBP groups’ mean age, the pain-free group had more students, and the current LBP group had more healthcare professionals than the other groups. These, in combination with the number of comparisons (risk of false-positive results), might have influenced the Back-PAQ-Br_Danger results. Second, recall bias can be an issue when using self-reported questionnaires that require participants to recollect information about past experiences. This may have influenced participants’ responses to the questionnaires, and the consequent categorisation of participants into the studied groups; this may have been an issue particularly in the groups ‘history of LBP’ and ‘pain-free’. Third, due to a potential time difference in completion of the IAT and BSB questionnaires and the online survey (0–2 weeks), it is possible that participants might have changed their LBP status. Fourth, participants of the current LBP group were not sub-classified as having acute or chronic LBP. It is possible that the duration of LBP could influence one’s beliefs (implicit and explicit) about the back. Fifth, we used the same version of the TSK to assess fear of movement in people with and without back pain, and healthcare professionals [32]. This was done as Brazilian translations of the TSK for lay population and the TSK for healthcare professionals do not exist. However, this is unlikely to impact on our findings as we have used the TSK general to assess beliefs of the general public, people with LBP and healthcare professionals successfully in a previous study [30]. Sixth, the majority of our sample had post-graduate education and were office workers. Lastly, the use of computer-based test and online questionnaires might have limited our sample to a higher literacy level of participants, potentially influencing the generalisability of our findings to the general Brazilian population.

Our study has strengths. Having three groups allowed for direct comparison of implicit biases and explicit beliefs. The study was powered to assess implicit bias differences between groups. Also, it assessed beliefs of a Brazilian sample, allowing comparison of beliefs to an Australian sample.

Conclusions

People with current LBP, people with a history of LBP and pain-free people have an implicit bias that round-back bending and lifting is dangerous, irrespective of group allocation. These findings support previous studies of implicit evaluation of beliefs performed in a high-income country. The belief that the back is at risk when rounded during bending and lifting is likely a shared societal belief.

Implications

Our results show that what people say (deliberate thinking) may not always correlate with what they think (implicit biases). This suggests that, in order to achieve a long term and in-depth change of beliefs and attitudes, patient education should not only be cognitive but also behavioural. Such an approach may include, for instance: (i) gradually exposing patients to activities that are feared and avoided in order to elicit their implicit biases, creating an opportunity to address misconceptions about LBP; and (ii) interactive learning tools such as a communication quiz that requires patients to make a choice and reflect on the answer (these tools could be online and in different languages to maximize reach). Given the similarities with previous studies in a high-income country, it may suggest that a global effort to change beliefs about the back and pain may be warranted.