Premier Pas was a multicentric cross-sectional study carried out in the Parisian and Bordeaux regions of France between February and April 2019. This study, which was based on multidisciplinary approaches, sought to better understand the experiences of UMs living in France and to assess their health status and access to rights and healthcare. The inclusion criteria consisted of being undocumented and over 18 years old. The questionnaire was translated into fourteen different languages to interact with participants: French, English, Spanish, Russian, Albanian, Portuguese, Arabic, Dari, Chinese, Tamil, Pashto, Bengali, Pulaar, and Bambara [17–19]. Individuals were recruited from 63 sites that provide support or assistance for individuals experiencing social deprivation and immigrants [19]. The structures included “Espace Solidarité Insertion,” which offers daytime shelter, social and health services for homeless people, various NGOs, and Hospital Health Access Sites (“Permanence d’Accès aux Soins de Santé”). The latter provide social assistance and healthcare to vulnerable populations, but their geographic scope is limited to certain cities. Additionally, another type of structure included was local health insurance centers (“Caisses Primaires d’Assurance Maladie”), which operate at a departmental level and deliver various social subsidies and indemnities for registered beneficiaries for claims ranging from sickness, pregnancy, disability or death. Point of Access to Rights (“Point d’accès aux droits”), which are free and permanent reception sites, provide information and resources for legal or administrative problems and were also included as recruitment sites. Finally, three more recruitment sites were included: public baths, free health centers of Doctors of the World (“CASO”) and maternal and child protection centers (“Protection Maternelle et Infantile”), which is a departmental service that is responsible for protecting the health of mothers and newborns.

In the first stage, we described the prevalence of different types of pain using percentages and then compared the categories of pain according to different mental health conditions via a chi-square test. The data were weighted on the basis of the probability of inclusion.

In the second step, we performed univariate analyses via logistic regression for musculoskeletal pain and the variable “all pain,” with pain as the dependent variable and mental health as the main independent variable. When an association was significant in the univariate analysis, we performed multivariate analyses. A difference was considered significant when the p-value of the crude odds ratio (OR) for the univariate result was <0.200.

To address collinearity and avoid overadjustment in the multivariate analyses, we used a hierarchical regression method. Blockwise selection was used, with pain as the dependent variable and mental health as the independent variable, and the different categories were added step by step. In this six-step process, we added each group of variables sequentially, monitoring how the coefficients changed to ensure that the models remained robust. To avoid weighing down the models and reducing their performance, we added only the variables with a p-value of the crude OR of less than 0.200 to the model. Alternatively, within a group of variables for which all the crude ORs were greater than or equal to 0.200, we selected the variable with the smallest p-value.

For power purposes, the age categories “18–29 years” and “30–39 years” were grouped together into the category “18–39 years” in some analyses when there were small numbers in one category.

In the final step, analyses were performed to stratify by gender, as the literature has shown that associations between pain and mental health are influenced by gender [36, 37].

The threshold of significance was 0.05 and 95% confidence intervals were used. All analyses were performed with Stata software.

Table 2 presents crude and adjusted associations between pain (all pain and musculoskeletal pain) and mental health. Individuals who were depressed, those with sleep disorder and those with anxiety were more likely to report pain. Individuals with PTSD had significantly higher likelihood of overall pain, at bivariate level and not in multivariate one. No difference was observed between musculoskeletal pain and PTSD.

Multivariate analyses were performed when the univariate association was significant. The multivariate analyses controlling for socioeconomic indicators revealed that sleep disorder was associated with higher odds of overall pain (adjusted odds ratio (aOR) = 2.83, 95% CI [1.39–5.76], p = 0.040). Sleep disorder was also associated with higher odds of musculoskeletal pain (aOR = 2.53, 95% CI [1.20–5.33], p = 0.014).

As shown in Table 3, among women, those with depression had significantly higher odds of overall pain after controlling for socioeconomic variables (aOR = 4.85, 95% CI [1.53–13.36], p = 0.007). No associations were found between mental health and musculoskeletal pain among women.

Among men, those with sleep disorder had significantly higher odds of overall pain after controlling for socioeconomic variables (aOR = 2.80 95% CI [1.05–7.48], p = 0.040). Sleep disorder (aOR = 7.14, 95% CI [1.60–31.87], p = 0.010) and anxiety (aOR = 0.12, 95% CI [0.02–0.82], p = 0.031) were associated with musculoskeletal pain.

Overall, clear associations between several mental health conditions and pain were revealed after controlling for SDHs. These results are consistent with the international literature [62, 72]. Similar to our study, a Swedish study revealed that a higher risk of chronic and severe pain among immigrants was significantly influenced by depression [73]. In another cohort study of Syrian refugees in Norway 12 months after their arrival in Northern Europe, poor mental health was a predictor of chronic pain at follow-up [72]. Finally, studies also support that trauma and pain, especially joint pain, are connected [74].

The rationale for including the “all pain” variable was that our study addressed the experience of pain and its relationship with mental health. The experience of pain, rather than its mechanism or cause, was the focus of this work, and this grouping made sense to us. Moreover, the inclusion of this variable increased the power of our analyses. This grouping and the results are also coherent from the perspective of screening UMs who visit clinics for pain. When UMs visit a clinic because they are experiencing pain, they should be screened for mental health problems, as they are more at risk.

We stratified the data by gender to verify whether the relationship between mental health and pain varied according to gender on the basis of the literature [36, 37]. Our results confirmed this hypothesis, as some associations remained only for women or men. Depression was associated with higher odds of overall pain among women, and sleep disorder was associated with higher odds of overall pain and musculoskeletal pain among men. The associations found in our study were in line with those described in the literature in the general population [36, 37, 75].

Surprisingly, after controlling for socioeconomic indicators, a negative association appeared for men. Anxiety was associated with lower odds of musculoskeletal pain among men. The association usually reported in the literature is that anxiety is a risk factor for musculoskeletal pain [76]. Nevertheless, among veterans, it appears that when anxiety is associated with PTSD, pain perceptions can be reduced [74]. In our study, no association between anxiety and PTSD was found, possibly because of a lack of power. Some associations, such as that between sleep disorder and musculoskeletal pain, among women have been described in the literature but were not found in our population [36, 77].

In addition to musculoskeletal pain and headaches, men had lower rates of reported pain than women did. One reason for this result could be the variation in pain verbalization depending on gender. Pain may be less easily expressed by men not because of biological differences but because of gender performativity [78]. Moreover, the verbalization of anxiety also appears to be gender biased. Male participants appear to be less inclined to express their anxiety but would be more comfortable doing so in certain contexts, such as in online forums or to women in their close circle (to their spouse, mother, etc.) [79].

In our study, the multivariate models were adjusted for SDHs. The role of SDHs in pain has been described, and migration seems to play a role [39, 58, 80, 81]. However, how migration affects the association between pain and mental health is unclear. The allostatic load theory explains how migration-related exposures play a role in the accumulation of somatic damage and increase the risk of several diseases [82, 83]. Migration-related stressors such as poor safety, a poor physical environment or poor access to healthcare are associated with chronic pain among immigrants [72].

Our results highlight the importance of identifying when pain is experienced by UMs not only because of its prevalence but also because it can be associated with mental health conditions. Our study was observational, so a causal pathway cannot be established between pain and mental health. However, it seems that they can both maintain each other mutually. The “mutual maintenance theory” suggests that chronic pain may exacerbate and maintain PTSD, whereas PTSD may exacerbate and maintain chronic pain [84]. This mutual maintenance occurs through specific mechanisms, such as reminders of trauma, depression and reduced activity levels. This relationship seems to extend further and has an impact on treatment for both mental health conditions and pain. Physical pain may interfere with a patient’s ability to respond to treatment, and the presence of a mental health condition could also interfere with effective pain management [74].

The pain expressed by the participants was self-assessed, and the causes of pain were not explored. Several underlying hypotheses may explain the cause of pain. The pain experienced could be linked to an unperceived or undiagnosed health problem, and reduced access to healthcare may increase the risk of delayed diagnosis or even non-diagnosis of the cause of pain [26]. In cases where pain can be explained by a physical or organic cause, whether a mental health problem exists must still be investigated, as it may accompany or even maintain the pain [84]. There are situations where pain cannot be medically explained and may either be part of a psychiatric syndrome or remain medically unexplained [85]. Regardless of the cause of the pain, its treatment can be challenging. Few studies have evaluated the impact of pain treatment on reported pain and mental health among immigrants [86, 87]. A study conducted in Norway among refugees from Syria aimed to assess two different interventions to reduce pain and posttraumatic symptoms: the “Physiotherapy Activity a pain Awareness Intervention” (PAAI), which consists of a combination of psychomotor and general physiotherapy, and the “Teaching Recovery Technique” (TRT), which involves cognitive behavioral therapy [86]. The PAAI had no effect on either chronic pain or mental health symptoms after 8 and 12 weeks. One reason that the authors proposed was that the intervention was not sufficiently adapted for the participants and their situations. A large proportion of participants in the intervention group could not attend the sessions. The qualitative analyses revealed that it was difficult to combine the sessions with other activities of daily life, thereby potentially contributing to additional stress [88]. The TRT significantly improved general mental health but did not help reduce chronic pain [89].

The key strength of this study lies in its recruitment of a diverse population and its focus on mental health issues within this vulnerable group. Moreover, to our knowledge, our study is the first to focus on UMs not exclusively recruited in healthcare settings in France. Very little is known in terms of the public health of this population. Thus, this study is highly important, as it provides insights into this hard-to-reach and hard-to-study population. One of the factors that enabled recruitment was the use of diverse languages and the inclusion of numerous sites outside care settings. Even though the sample size was not large, the results of this study are still substantial if we consider that this population is difficult to approach, especially outside healthcare settings. The results presented in this article are the first to document the relationship between pain and mental health in this population.

Unfortunately, the lack of power did not allow us to further explore the implications of PTSD and depression on the different types of reported pain. Additionally, the small sample size of women did not allow us to explore associations between mental health and musculoskeletal pain. For some results, such as the association between sleep disorder and musculoskeletal pain, the CIs were large. Thus, even if the value was significant, the uncertainty was high, and the estimate was less precise because of the size of the 95% CI. A larger sample size might have reduced this uncertainty.

Furthermore, our study does not represent all UMs across France because recruitment took place in two regions. However, our results are representative, as UMs are geographically concentrated in urban regions, particularly in the Parisian region [90]. Future research with larger samples should thoroughly investigate immigration-related factors, including a wide range of sociodemographic and health-related factors that may contribute to health status among immigrants.

Pain could have been studied in more detail, for example, by examining the intensity of the pain or by asking how the pain affected the participants’ lives. Self-assessment of subjective health, such as pain, is influenced by culture, as expectations and norms for a given health dimension can be very heterogeneous [91]. As a result, the subjective self-assessment of pain in this study did not allow us to compare subjective health states between individuals, especially to compare samples from different countries or regions. This can lead to serious misunderstandings about the health status of populations, particularly those that are the most economically disadvantaged. This bias, known as the “DIF” (Differential Item Functioning) effect, can be corrected via more advanced methods, such as “Anchoring Vignettes,” which can be used to detect and correct possible comparability problems. This method was not used in our study, as the study design was not initially intended to study pain in depth. Future studies could be carried out using this type of method to better study pain.

The PHQ-9 and PC-PTD-5 were used in different languages. In this study, interviews were conducted in French (75%), Arabic (8%), English (7%), Spanish (4%), Russian (2%), and Portuguese (2%), accounting for 99% of the questionnaires collected [19]. Although the PHQ-9 has been validated in these languages, this was not the case for the PC-PTSD-5 [92–96]. The PHQ-9 has been validated in French, Arabic, English, Spanish, Russian and Portuguese. The PC-PTSD-5 has only been validated in English but not in French or other languages [70]. For those languages, no suitable and validated scale to measure PTSD symptoms was found in the literature. Thus, PTSD symptoms may have been over- or underestimated in our study, leading to measurement bias for this indicator.

Since our study did not aim to establish causality, the lack of specific time frames does not undermine the relevance of our findings. Moreover, recall bias would have been a challenge. The PHQ-9 and PC-PTSD-5 were administered to all individuals, and anxiety and sleep disorders can rarely be ignored, as these symptoms often have a major impact on quality of life. Thus, the risk of over- or underestimating the prevalence of mental health conditions in our study was reduced. Moreover, since our study was essentially conducted outside care settings, the risk of overestimating mental health conditions if the participants were seeking health services also remained low.

Our study confirms the link between mental health and perceived pain in this large sample of UMs in France. These results could have a direct impact on the care provided to UMs. This study encourages practitioners to systematically ask questions pertaining to mental health when a patient is experiencing pain to implement targeted interventions, especially when the pain remains medically unexplained. The results call for more specific studies on the associations between certain types of pain and certain mental disorders and the development of treatment algorithms that consider the social and psychological context of UMs.