We included published original scientific papers that measured causal beliefs about non-specific LBP and reported results from this domain that could be isolated from other beliefs domains. Population were adults from non-clinical populations with or without non-specific LBP, health-care providers and clinical (i.e., care seeking) populations of patients with non-specific LBP. We excluded studies testing psychometric properties of questionnaires or transcultural adaptations.

Our interpretation of causal beliefs was conceptualized by the common-sense model, which implies that the perception of what caused LBP should be distinguishable from beliefs that according to the common-sense-model relate to other domains. We defined causal beliefs as: a) a perceived cause of LBP, b) a perceived trigger of a new onset of LBP, or c) a perceived risk factor for LBP. Any quantitative measure or data from quantified text responses capturing a causal belief was included. Thus, studies measuring causal beliefs by text responses were only included if the researchers categorized and quantified the text responses in the studies. Studies that in the abstract mentioned causal beliefs specifically or unspecified beliefs were included for full text assessment. Studies measuring beliefs that were specified as other types of beliefs than causal beliefs (e.g., fear avoidance beliefs or kinesiophobia beliefs) were not included.

Furthermore, only peer-reviewed articles written in English were included. We did not use any restriction on time period.

We searched the following electronic databases: Embase, Medline, PsychInfo, and CINAHL. The search strategy was developed in collaboration with a librarian from University Library of Southern Denmark and was initially developed for Embase and then adapted to the other databases. Keywords and search terms were identified from preliminary searches and reading of articles related to the subject. The search combined words of LBP with words for causal beliefs, using both keywords and subject headings (Supplementary material A). The search was conducted on January 10, 2022.

Duplicates were removed in Endnote before uploading citations to Covidence review software (Veritas Health Innovation, Melbourne, Australia) for screening and data extraction. Prior to screening of titles and abstracts the inclusion criteria were tested by the entire review team in a pilot screening of a small test sample of 30 titles and abstracts. Two rounds of pilot screening were completed to achieve the desired 75% agreement threshold.

The screening of titles and abstracts as well as full-text assessments were done double-blinded with SG screening the entire sample and TJ, KB, and JD splitting the sample between them. Disagreements between the reviewers were settled through discussion to reach consensus.

The extracted data included study characteristics (population, setting, country, and aim), causal beliefs measurement tool, causal belief items, and outcomes investigated for cross-sectional or longitudinal associations with causal beliefs. The outcomes investigated were only extracted in cases where the association could be linked to the isolated causal belief item. SG and AK piloted the extraction tool and made modification before moving on to the final extraction. Extraction was done independently by SG extracting the entire sample of studies and AK, TJ, JD, and KB splitting the sample between them.

Prior to data extraction consensus between the review team was made to determine which items from the identified questionnaires were considered causal beliefs (Supplementary material B). Each member independently voted for each item whether they deemed it to measure a causal belief based on face validity. The votes were compared, and disagreements settled by discussions in the entire review team to reach consensus.

Data were exported from Covidence and handled in Stata/MP V.17. (StataCorp Texas, USA). Extracted data were organized in tables and visualized in bar charts made in Microsoft Excel (Microsoft Corporation, Redmond, WA). The causal belief items extracted from the studies were categorized into mutually exclusive categories based upon face validity of the beliefs. The first half of the items were categorized in a consensus forum between SG, JD, KB, and AK. The remaining were categorized by SG whereupon all authors commented and agreed upon the final categorization (for resulting categories see supplementary material C).

We identified the following questionnaires from which causal beliefs were obtained: Pain Attitudes and Belief Scale for Physiotherapists (PABS-PT) (7 studies) in which 7 items were deemed to be causal beliefs, Back pain attitudes belief scale (Back-PAQ) (5 studies, 2 items), Illness Perception Questionnaire (IPQ) (8 studies, 1 section), Attitudes to Back Pain Scale in Musculoskeletal Practitioners (ABS-MP) (2 studies, 1 item), Neurophysiology of pain questionnaire (NPQ) (1 study, 5 items), and the Worker Attribution Scale (WAS) (1 study, 1 section). Additionally, questions based on two of “Deyo's myths” regarding low back pain were used in 12 studies. For the remainder of the studies, eight measured causal beliefs using modification or adaptations of other questionnaires and 32 measured causal beliefs by other non-validated questionnaires or items specifically developed for the purpose of the study. Fig. 2 shows the use of the measurements within the investigated populations.

Fig. 2.

The frequency of used questions / questionnaires distributed by population. ABS-MP, Attitudes to Back Pain Scale in Musculoskeletal Practitioners; Back-PAQ, Back pain attitudes belief scale; IPQ, Illness Perception Questionnaire; NPQ, Neurophysiology of pain questionnaire; PABS-PT, Pain Attitudes and Belief Scale for Physiotherapists.

(0.19MB).

A total of 308 unique causal belief items were identified and categorized into 15 mutually distinct categories. All categories are explained in Table 2, and an in-depth description of items included in each category can be found in Supplemental Material C: Full list of items. The most prevalent investigated category was causal beliefs related to “structural injury or impairment”, which was investigated in 45 (56%) of the studies. The second and third most prevalent categories were related to “lifting and bending“ (26 studies [32%]) and “mental or psychological” (24 studies [30%]) (Fig. 3).

Fig. 3.

The frequency of studies investigating each category of causal beliefs distributed by population.

(0.26MB).

Among the frequently used questionnaires, PABS-PT contained items from the categories “structural injury or impairment”, “mental or psychological”, and “unknown”. Back-PAQ contained only items from “structural injury or impairment”. The questions based on Deyo's myth contained items from “structural injury or impairment”, “lifting and bending”, and “unknown”. IPQ had, due to its free text option, the capability to contain all the categories of causal beliefs created for this review.

Twenty-eight studies investigated an association between causal beliefs and other factors. Twelve studies (43%) were conducted in the general population, 6 (21%) in clinical populations, 6 (21%) among health-care providers, 2 (7%) in mixed populations, and 2 (7%) among health-care students. Cross-sectional associations were reported in 22 studies (Table 1). The most common cross-sectional associations investigated were with pain,17-21 sex,22-25 disability,17,18,26 and care seeking.20,27,28 Longitudinal associations were investigated in 8 studies. The longitudinal association most commonly investigated was reporting LBP29,30 (Table 1).

The review followed a stringent method and was reported in accordance with current guidelines to ensure high transparency with the choices made in the process. A main concern was that it is not clear cut what constitutes a causal belief, and we consider it a strength that our definition of causal beliefs was based on the common-sense model and the question “what caused my LBP” or “what causes LBP”. However, in the review process we realized that beliefs related to triggers of back pain and contributing factors relate to this domain and thus were eligible for inclusion. For instance, the question “what do you believe may have triggered your LBP?”37 and also questions where participants rated how important they believed different items were in causing back pain, were both deemed to be a measure of causal beliefs.38 As these types of beliefs were discovered in the review process, specific search terms for these were not included in our search strategy. We acknowledge that relevant studies may have been missed on this account, but do not consider this a major flaw because we used a broad search strategy and screened a large number of studies.

We were strict on not including aggravating factors as causal beliefs. However, aggravating factors overlap with causal beliefs. For instance, the item from Back-PAQ “Stress in your life (financial, work, relationship) can make back pain worse” was deemed as measuring aggravating factors and not as a causal belief. This distinction may have favored biomedical beliefs and specific structural causes of LBP while items reflecting psychosocial causes may more often be presented as aggravating factors than as an initial cause. It can be argued that a focus on “contributing factors” to LBP would have been more inclusive but would also make the differentiation from other belief domains less clear. The overlap between domains made the isolation of causal beliefs challenging in some studies. Additionally, many studies had a vague description of methodology and how they measured beliefs. Thus, some subjective interpretation was inevitable.

We did not look for gray literature as we decided to limit the scoping review to peer reviewed literature. Thus, additional knowledge regarding measuring of causal beliefs may exist. However, we have no reasons to believe this would change the general findings of this review.