Experience of discrimination during COVID-19 pandemic: the impact of public health measures and psychological distress among refugees and other migrants in Europe

Marchietal. BMC Public Health (2022) 22:942 https://doi.org/10.1186/s12889-022-13370-y RESEARCH Experience ofdiscrimination duringCOVID-19 pandemic: theimpact ofpublic health measures andpsychological distress amongrefugees andother migrants inEurope Mattia Marchi1,2, Federica Maria Magarini1, Antonio Chiarenza1, Gian Maria Galeazzi1,2*, Virginia Paloma3, Rocío Garrido3, Elisabeth Ioannidi4, Katerina Vassilikou4, Margarida Gaspar de Matos5, Tania Gaspar5, Fabio Botelho Guedes5, Nina Langer Primdahl6, Morten Skovdal6, Rebecca Murphy7, Natalie Durbeej8, Fatumo Osman8,9, Charles Watters10, Maria van den Muijsenbergh11, Gesine Sturm12, Rachid Oulahal13, Beatriz Padilla14, Sara Willems15, Eva Spiritus‑Beerden16, An Verelst16 and Ilse Derluyn16 Abstract Background: The COVID‑19 pandemic has had a disproportionately hard impact on refugees and other migrants who are often exposed to the virus with limited means to protect themselves. We tested the hypothesis that during the COVID‑19 pandemic, refugees and other migrants have suffered a negative impact on mental health and have been unjustly discriminated for spreading the disease in Europe (data collection from April to November 2020). Methods: Participants in the ApartTogether Survey (N = 8297, after listwise deletion of missing items final N = 3940) provided data regarding to their difficulties to adhere to preventive recommendations against COVID‑19 infection (CARE), self‑perceived stigmatization (SS), and psychological distress (PD). Structural Equation Modeling was used to investigate PD as a mediator in the pathway linking CARE to SS, while adjusting for the housing and residence status. To improve confidence in the findings, single hold‑out sample cross‑validation was performed using a train/test split ratio of 0.8/0.2. Results: In the exploratory set (N = 3159) SS was associated with both CARE (B = 0.200, p < 0.001) and PD (B = 0.455, p < 0.001). Moreover, PD was also associated with CARE (B = 0.094, p = 0.001) and mediated the effect of CARE on SS (proportion mediated = 17.7%, p = 0.001). The results were successfully replicated in the confirmation set (N = 781; total effect = 0.417, p < 0.001; proportion mediated = 29.7%, p < 0.001). Follow‑up analyses also found evidence for an opposite effect (i.e., from SS to CARE, B = 0.132; p < 0.001), suggesting that there might be a vicious circle between the self‑perceived stigmatization and the access to health care and the use of preventive measures against COVID‑19 infection. © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Open Access *Correspondence: [email protected] 2 Dipartimento di Salute Mentale e Dipendenze Patologiche, Azienda USL‑ IRCCS di Reggio Emilia, Via Giovanni Amendola 2 –, 42122 Reggio Emilia, Italy Full list of author information is available at the end of the article Page 2 of 14 Marchietal. BMC Public Health (2022) 22:942 Introduction The COVID-19 pandemic has had a disproportionately hard impact on refugees and other migrants who are often exposed to the virus with limited means to protect themselves. Basic public health measures, such as social distancing, maintaining hand hygiene, and self-isolation are not possible or extremely difficult to implement insituations where many people are in close contact and gather in large groups, such as in refugee camps and the overcrowded housing of many other immigrants with low income or insecure status. Such conditions may have a considerable impact also on the mental health, with individual and public health implications [1,2]. Previous research has identified migrants’ barriers in accessing both physical and mental health care, which likely worsened during the pandemic [3,4]. These may be summarized within three groups: (1) cultural barriers, including mental health stigma and knowledge of dominant models of health; (2) structural barriers, including scarce culturally and linguistically accessible information, unstable accommodation, and financial strain; and (3) barriers specific to the migrant experience, including the consequences of postponement of decisions on their legal status, resettlements, or border closures. In spite of worldwide calls to action for inclusion of refugees and other migrants in the pandemic response [5, 6], there have been reports of increased stigmatization and xenophobia across the world since the onset of the COVID-19 pandemic [7,8]. Skovdal etal. [9] argue that this may be explained by how risk communication about COVID19 constructs new social norms about how to act and behave in public, which may inadvertently contribute to a blaming and shaming of those who are unable to comply. This disproportionately affects already stigmatized groups, such as refugees and other migrants who are typically discriminated against for spreading disease to host populations, despite little evidence to support this [10]. Actually, the transmission is more likely within refugee and migrant populations, e.g., due to crowded living conditions and scarce access to basic sanitation. Previous studies have documented the role of racism in health inequalities, defining race as a “pathogenic factor” for both physical and mental illness [11,12]. Compelling models on racial health disparities suggest that racism negatively impacts both individual wellbeing and health care utilization in many ways [13]. The manifestations of the phenomenon are varied, from institutional racism (e.g., in judicial or educational systems) to interpersonal racism (e.g., stigma, social exclusion), also within health care services when immigrant users “express disappointment with services that dismiss their concerns and fail to attend to their priorities” [14]. The perception of being discriminated against and excluded from social interactions has been linked with both increased psychological distress and worse physical health [15–18]. In addition, studies on mediating factors suggested that the negative effect of stigmatization on physical health might be in part exerted through increased psychological distress [19–21]. In other infectious disease epidemics, social stigma and self-perceived discrimination have been found to amplify the dangers related to infection and engagement with preventative measures, treatment, and care. For example, people who perceived more HIVrelated stigma had a two-fold probability of delayed treatment seeking and higher viral load at first healthcare contact [22–24]. Qualitative studies in the field of HIV continue to find different forms of stigma to negatively affect engagement with medical care [25]. The situation is even more alarming among undocumented and migrant populations, where added to disease-related stigma, people might fear losing social support or not being resettled – compounded by experiences of prejudice and discrimination linked to race and their migrant or refugee status [26]. Such conditions may create a vicious circle, negatively impacting mental health, psychological selfregulation, and treatment adherence – all of which carry important public health implications. According to the theory of Social Stigma [27, 28], in a prejudice-based context, the victims of racism are not passive actors of that circumstance, instead, they actively face the perception of being the object of discrimination, employing different coping strategies affecting their social interactions and the overall social inclusion process. In the context of COVID-19, an effect of social stigma and self-perceived discrimination, could be that refugees and other migrants are afraid of seeking treatment or disclosing symptoms leading to risks to wider public health outcomes, including for host populations [29]. Conclusions: Refugees and other migrants who had more difficulties in accessing health care and preventive measures against COVID‑19 infection experienced worse mental health and increased discrimination. These negative effects appeared to be stronger for those with more insecure housing and residence status, highlighting from one side the specific risk of insecure housing in the impact of COVID‑19 upon mental health and infection protection, and for another side the need to proper housing as a strategy to prevent both COVID‑19 and mental distress. Keywords: COVID‑19, Migrants, Social stigma, Mental health, Public health Page 3 of 14 Marchietal. BMC Public Health (2022) 22:942 Using the European data of the ApartTogether Survey [30], we aim to better understand the connection between the difficulties to adhere to preventive recommendations against COVID-19 infection (CARE) and self-perceived stigmatization (SS), focusing on the role of psychological distress (PD). Specifically, we hypothesize that PD might be a mediator in the relationship between CARE and SS, and that there might be a bidirectional relationship between the self-perceived stigmatization and the access to health care and the use of preventive measures against COVID-19 infection. Methods Study design andpopulation The plight of refugees and other migrants during the (ongoing) COVID-19 crisis has been highlighted in the media, and by online stories, blogs, reports, and posts, often with regional or national focus and not reported by refugees and migrants themselves. The ApartTogether is a perception-based survey, prompted by the WHO Global Programme for Health and Migration in collaboration with a large European consortium of academics, and was the first inquiry into the social impact of the COVID-19 pandemic on refugees and other migrants globally [30]. The survey gave refugees and other migrants an online platform to self-report the perceived impact of COVID19 on their lives including the preventive measures recommended. It is not a systematic survey of the incidence of COVID-19 among refugees and migrants nor an assessment of the prevalence of symptoms or immunity, rather it aims to give refugees and other migrants a voice to better understand their situation and specific difficulties they face during the COVID-19 pandemic, to provide information for policy makers and inform the health systems on preparedness and inclusiveness [31]. Data were collected from the ApartTogether online global survey, which ran from April 2020 until November 2020. The current study used the European data from 46 different countries and mainly focused on three different parts of the ApartTogether Survey, consisting of: difficulties to adhere to preventive recommendations against COVID-19 infection (CARE), experience of stigma and self-perceived discrimination (SS), and mental health (i.e., the level of self-reported psychological distress [PD]). In total, N =8297 refugees and migrants hosted in Europe entered the survey, however given the participants could stop the survey at any time, not everyone completed all the items, thus the number of full entries used in this study was N =3940. Study questionnaire Participants completed questions about sociodemographic characteristics, including age, gender (i.e., male, female, non-binary), education, housing situation (i.e., living in a house or apartment, asylum center for asylum seekers - meant as someone who is seeking international protection but whose claim for refugee status has not yet been determined - refugee camp for refugees waiting for resettlement, or on the street), residence status (i.e., citizen, permanent documents, temporary documents, no documents), and working situation (i.e., student, employed, unemployed since the corona-crisis, or already unemployed before corona-crisis). To investigate the difficulties in adhering to preventive recommendations against COVID-19 infection (i.e., the CARE latent construct), participants were asked how frequently they engage in handwashing, keeping physical distance from others, covering nose and mouth in public, avoiding using public transport and going out of the house. The possible answers were structured on a 4-levels Likert-type scale consisting of: “No, because I am not able”, “No, because I don’t want”, “Yes, sometimes”, and “Yes, all the time”. In order to focus on the overall utilization of precautions, the answers were coded as “Never” (i.e., including both “I don’t want” and “I am not able” replies – yet description of all possible answers is available in the Supplement), “Sometimes”, “Always”. In addition, participants were asked if they would contact a doctor if they or a person close to them would develop COVID-19 symptoms. Further, participants reported changes in their mental health from COVID-19 pandemic (i.e., the PD latent construct) through indicating how often they experienced feelings of depression, anxiety, loneliness, anger, irritation, hopelessness, unpleasant reminders of past traumatic experiences, physical reactions to stress, sleeping problems, and substance use, such as alcohol and drugs. Responses were collected over a Likert-type scale on 4-levels: “Never”, “Seldom”, “Sometimes”, “Always”. Alongside, they reported also on a binomial “worse-scale” (as “Yes” or “No”) if these mental health problems worsened in respect to before COVID19. To maximize the sample size, we primarily focused on the “frequency” questions which had many more entries (e.g., if a person replied “never” in the frequency question, in all probability left the “worse-scale” empty). Finally, to measure the experiences of discrimination of respondents (i.e., the SS latent construct), six items were investigated: being differently treated because of your origin, being called names because of your origin or religion, being avoided, other people being anxious of you, being unfairly treated by the police, being treated with kindness. Respondents were asked to indicate on a binomial scale, as “Yes” or “No”, whether they felt their treatment by others had worsen since the outbreak of the COVID-19. Page 4 of 14 Marchietal. BMC Public Health (2022) 22:942 Statistical analysis The analyses were conducted with structural equation modeling (SEM) using the lavaan [32], tidySEM [33], and SEMsens [34] packages (versions: 0.6-10; 0.1.3.1; 1.0.1, respectively) in R (2021.09.2 + 382 “Ghost Orchid” Release for macOS) [35]. Missing data rates ranged from 14 to 47% and were listwise deleted, because multiple imputation was not possible for ordinal factors in the lavaan package, therefore the final N used in the analyses was 3940. Previous to the SEM, confirmatory factor analysis was performed on the full set of complete data (N =3940) to determine if the theorized measurement models have an acceptable fit for each latent construct (i.e., CARE, PD, and SS). Next, the total sample was randomly split into exploratory and confirmation sets using a split ratio of 0.8/0.2 (i.e., N =3159 in the exploratory set, and N =781 in the confirmation set). We performed a mediation analysis using diagonally weighted least squares (DWLS) estimator to test whether CARE exerts an effect on SS both directly and indirectly through PD. The bias-corrected bootstrap 95% confidence interval (95% CI) for the indirect effect was tested to be entirely above zero based on 5000 bootstrap samples applying the adjusted bootstrap percentile method (BCa). Moreover, we evaluated the model results’ sensitivity to an omitted confounder using the ant colony optimization algorithm (based on k = 100, and maximum iterations = 1000) to automatically search for sensitivity parameters that would lead to a change in the study conclusions. The following fit indices were considered as indicating good fit: Root Means Square Error of Approximation (RMSEA) < 0.05 (< 0.08 indicates adequate fit); Standardized Root Means Square Residual (SRMR) < 0.08; Tucker–Lewis Index (TLI) and Comparative Fit Index (CFI) > 0.95 [36–39]. Cronbach’s alpha was used to test the reliability of the latent variables, deeming a value of 0.70 or higher as a satisfactory level of reliability [40]. The variance covariance matrix of the full set of data (N =3940) is available as Supplementary Table1. Results Description ofthesample The main characteristics of the sample analyzed in the SEM are displayed in Table1 and detailed description of the study questionnaire answers is displayed in Supplementary Table2. Since there may be important country-specific differences in the access to health care and in the legal requirements in the COVID-19 response impacting participants’ adherence, although not fully investigable with inferential statistics due to the heterogeneous distribution of participants across the countries, we provided in the Supplementary Table 3 description of participants’ adherence to recommendations against COVID-19 by country. Latent variable estimation The CARE dimension was built from the following questions: “What precautions are you taking to avoid transmitting the coronavirus?” and “In case I or one of your family members (or a person close to you) develops symptoms, I would contact a doctor or health care provider?”. The answers were homogeneously ordered so that more difficulties in adhering to preventive measures against the virus and in accessing care give rise to the CARE score. The Cronbach’s alpha for this latent variable was 0.73. PD was estimated using the frequency answers (i.e., “No”, “Seldom”, “Sometimes”, “Always”) to the question: “Since the coronavirus and the corona measures, have you been bothered by the following problems?”, except for the item 2 (i.e., “worried”) that has been removed because collinear with item 3 (i.e., “anxious”). The answers were homogenously ordered, so that a more frequent engagement in that mental health experiences gives raise to the PD score. The Cronbach’s alpha for the latent variable PD was 0.91. Finally, the following question investigated experience of SS: “Since the corona-crisis, do you feel differently treated by others than before?”. Reporting a worsening in the perceived discrimination than before the coronacrisis increases the SS score. The Cronbach’s alpha for SS was 0.95. Supplementary Table4 summarizes the contribution of the observed items to each latent variable. Mediation analysis intheexploratory set From the mediation analysis conducted in the exploratory set (N =3159), CARE influenced the level of SS both directly and indirectly through its effect on PD. Table2 and Fig.1 (Panel A) show that subjects who less adhere to preventive recommendations against COVID-19 infection experienced more stigmatization (unstandardized regression beta [B] = 0.200; p < 0.001) and psychological distress (B = 0.094; p = 0.001). In addition, the people who reported higher PD also reported to perceive higher stigmatization (B = 0.455; p <0.001). The bootstrap 95% CIs for the indirect effect through PD (B = 0.043) were entirely above zero (0.018 to 0.067), consistent with a significant contribution of CARE to SS, through PD. The goodness of fit measures for this model were: CFI = 0.985; TLI = 0.983; RMSEA = 0.045; SRMR = 0.065, all indicating a good fit. Page 5 of 14 Marchietal. BMC Public Health (2022) 22:942 Mediation analysis intheconfirmation set Table2 and Fig.1 (Panel B) show the mediation model in the confirmation set (N = 781). Both the direct (B = 0.239; p = 0.002) and the indirect effect - through PD - (B = 0.124; bootstrap 95% CI: 0.056 to 0.192) of CARE on SS were confirmed. The goodness of fit indices were indicating good fit: CFI = 0.988; TLI = 0.986; RMSEA = 0.042; SRMR = 0.078. The relationship between PD and SS was fully retained also if building PD using the dichotomous “worse scale”, yet the association with CARE became not significant, likely due to loss of power (the missing rate for PD on the “worse scale” was 80%; data available on request). However, the PD “worse scale” was significantly associated with PD “frequency scale” (covariance = 0.487; r = 0.99; p <0.001). Sensitivity analysis To evaluate the sensitivity of our SEM to potential missing confounders, we used the Ant Colony Optimization algorithm [41–43]. Specifically, through that algorithm, we introduced for each path in the model a “phantom variable” which serves as a potential missing confounder. We relied on the changed p-values that are obtained with the introduction of the phantom variables. Note that NAs occur when there is no change in the p-value for any of the tested phantom variable path coefficients. The results of this analysis are displayed in Supplementary Table5. Following Kolenikov’s [44] recommendation that parameters with greater than 10% change can be considered sensitive to misspecification, we found that all the paths might be sensitive to potential confounders. However, when evaluating the change in p-values, we found that only the path from CARE to SS was no longer significant when the confounder is added. One implication of this result is that, at worse, the relation between CARE and SS is fully mediated by PD. Table 1 Characteristics of the sample that was analyzed in the SEM Abbreviations: SD Standard deviation, COVID-19 Coronavirus disease 2019 Variable (N =3940) Frequency (%)/Mean (SD) Missing rates Gender: 0 Male 1924 (48.8%) Female 2004 (50.9%) Non‑binary 12 (0.3%) Age (years) 34.7 (10.3) 0 Education: 178 (4.5%) No schooling 57 (1.4%) Primary school 171 (4.3%) Secondary school 852 (21.6%) Higher education 2682 (68.1%) Housing situation: 165 (4.2%) House/Apartment 3500 (88.8%) Asylum center 155 (3.9%) Refugee camp 91 (2.3%) On the street 26 (0.7%) Residence status: 159 (4.0%) Citizen in this country 860 (21.8%) Permanent documents 1140 (28.9%) Temporary documents 1524 (38.7%) No documents 255 (6.5%) Work situation: 617 (15.7%) Student 691 (17.5%) Employed 1813 (46.0%) Unemployed since the corona‑crisis 104 (2.6%) Already unemployed before corona‑crisis 715 (18.1%) Have you tested positive for the COVID‑19? (current November 2020) (No/ Yes) 3528 (89.5%)/347 (8.8%) 65 (1.6%) Page 6 of 14 Marchietal. BMC Public Health (2022) 22:942 Exploring other models As a follow-up to the sensitivity test, two variations on the main model were assessed. To better distinguish the adherence to preventive measures against COVID19 and the access to care (namely, question 19: “In case I or one of your family members develops symptoms, I would contact a doctor or health care provider?” [answer coded as flipped, “Yes/No”, thus higher score indicates that people would not go to a doctor in case of corona symptoms]), we fitted a first model where this question was removed from the latent variable CARE (new variable called CAREi), and modelled as the outcome of SS. So, the theorized model would add investigation of the impact of SS on the health care utilization. As can be seen in Table3 and Fig.2, our findings suggest that SS had a negative impact on the health care access, shown as an increased likelihood to not consult a doctor in case of COVID-19 symptoms (B = 0.353; p <0.001). The Cronbach’s alpha for that newly conceptualized CAREi latent variable increased up to 0.88, and the fit indices were still indicating good fit: CFI = 0.985, TLI = 0.983, RMSEA = 0.045, SRMR = 0.062. In order to alleviate any concerns about the direction of the theorized main model, which is going from CARE to SS, the second sensitivity model consisted in reversing the order of the predictor and the outcome, thus replying to the question: does psychological distress mediates also the relationship between stigmatization and the use of preventive measures against COVID-19 spread? As can be seen in the Supplementary Table6, and Supplementary Fig.1, we found significant direct effect of SS on CARE (B = 0.132; p <0.001), supporting the likely existence of a vicious circle between the self-perceived stigmatization and the access to health care and the use of preventive measures against COVID-19 infection, though not significant indirect effect through PD (B = 0.007; bootstrap 95% CI: − 0.007 to 0.021) suggesting a unidirectional path from CARE to SS through PD, which is likely attributable to the specificity of the latent class CARE as conceptualized in the questionnaire. The fit indices were indicating good fit: CFI = 0.985, TLI = 0.983, RMSEA = 0.045, SRMR = 0.065. Finally, to investigate the impact of potential confounders, some complementary models were assessed. Given previous work on the same sample [31], we included as covariates the housing situation and the residence status, gathered beneath the latent variable that we called socio-demographic hardship (SDH). For constructing the SDH latent variable, the housing status was dichotomized (house yes/no), the residence status was set on a three-level scale (permanent/temporary/no documents), therefore a higher SDH score was indicating socio-demographic instability. The SDH dimension resulted strongly associated with both PD and SS (B = 0.350; p <0.001; and B = 0.274; p <0.001, respectively). The effect of PD on SS is confirmed (B = 0.338; p <0.001), as well as the direct effect of CARE on SS (B = 0.097; p = 0.015), yet the indirect effect through PD became no significant (B = -0.009; bootstrap 95% CI: − 0.030 to 0.011). From a theoretical perspective, it could be argued that SDH is a condition predisposing to increased difficulties in adhering to preventive measures against COVID-19. At the same time, being in disadvantaged living conditions may result in disengagement attitudes, based on not trusting the system or health care providers, expressed also with the refusal of governmental instructions of COVID-19 prevention. That, added to the fact that SDH is also strongly associated with both PD and SS, is suggesting a possible multiple mediator model where CARE and PD mediate the relationship between SDH and SS. As can be seen in Table4 and Fig.3, when testing that model, we could confirm that SDH exerts an effect on SS both directly (B = 0.265; p < 0.001) and indirectly through PD (B = 0.113; bootstrap 95% CI: 0.087 to 0.139) and CARE (B = 0.019; bootstrap 95% CI: 0.005 to 0.033). Table 2 Results of the mediation model in the exploratory set (i.e., 80%; N = 3159 ‑ above), and in the confirmation set (i.e., 20%; N = 781 ‑ below) Abbreviations: 95% CI: 95% bias-corrected bootstrap confidence interval, SS Self-perceived stigmatization, CARE Difficulties to adhere to preventive recommendations against COVID-19 infection, PD Psychological distress Regressions Unstandardized estimate (95% CI) Standardized coefficient p-value Exploratory set (80%) Direct effects Outcome model: SS (R2 = 0.182) CARE 0.200 (0.126; 0.274) 0.151 < 0.001 PD 0.455 (0.397; 0.514) 0.387 < 0.001 Mediator model: PD (R2 = 0.007) CARE 0.094 (0.039; 0.148) 0.083 0.001 Indirect effect (proportion mediated) PD (17.7%) 0.043 (0.018; 0.067) 0.032 0.001 Total effect 0.243 (0.165; 0.321) 0.183 < 0.001 Confirmation set (20%) Direct effects Outcome model: SS (R2 = 0.221) CARE 0.293 (0.104; 0.482) 0.172 0.002 PD 0.457 (0.357; 0.558) 0.408 < 0.001 Mediator model: PD (R2 = 0.032) CARE 0.272 (0.120; 0.423) 0.179 < 0.001 Indirect effect (proportion mediated) PD (29.7%) 0.124 (0.056; 0.192) 0.073 < 0.001 Total effect 0.417 (0.215; 0.620) 0.245 < 0.001 Page 7 of 14 Marchietal. BMC Public Health (2022) 22:942 Fig. 1 Mediation model in the exploratory set (Panel A:80%;N = 3159) and in the confirmation set (Panel B:20%;N = 781). The estimates reported are the unstandardized regression coefficients. *p < 0.05; **p < 0.01; ***p < 0.001. Abbreviations: CARE: difficulties to adhere to preventive recommendations against COVID‑19 infection; PD: psychological distress; SS: self‑perceived stigmatization Page 8 of 14 Marchietal. BMC Public Health (2022) 22:942 The fit indices were indicating a good fit: CFI = 0.983, TLI = 0.981, RMSEA = 0.044, SRMR = 0.066. Cronbach’s alpha for the latent variable SDH was 0.75. Discussion This study set out to better describe the effect of the COVID-19 pandemic on refugees and other migrants in Europe. Specifically, we aimed to model the relationship between the difficulties to adhere to care and preventive measures against further spread of the COVID-19 infection, experience of self-perceived stigmatization, and psychological distress. We found that increased difficulties in following the preventive measures against COVID-19 are associated with greater psychological suffering and experience of discrimination. Moreover, psychological distress mediated the effect of preventive measures use on self-perceived stigmatization. The use of a single hold-out sample cross-validation strengthens our confidence in these findings. The relationship between the self-perceived stigmatization with both the adherence to preventive measures against COVID-19 infection and the psychological distress was robust after the adjustment for known conditions of social vulnerability, such as insecure housing situation and residence status. Although the relationship between stigma and health is well known in vulnerable populations [21,45,46], this is the first study to confirm that access to care and to preventive measures influence the reporting of experience of discrimination, both directly and indirectly through its effect on psychological wellbeing. In our models, psychological distress was defined as a condition of hyperarousal and anxious-depressive symptoms. Therefore, within the specific context of the COVID-19 pandemic, the relationship between facing barriers to access health care and preventive measures against COVID-19 infection (as indicated by higher CARE scores) and higher psychological distress (i.e., higher PD score) could reflect a fear of catching the virus [47]. In addition, lower levels of CARE might imply that people travel despite possible lockdown-measures or interact with others without wearing a face cover; those kinds of behaviors could be viewed as unsafe or irresponsible which might elicit discriminatory behaviors (such as refusal to provide service, harassment, and bullying), avoidant reactions in the host population [48] or the perception thereof [49]. That may be a plausible explanation for the relationship found between the lack of preventive measure utilization and the self-perceived stigmatization. With respect to the indirect effect of CARE on SS through PD, it could be hypothesized that the awareness to have difficulties in following public health measures against COVID-19 may create the expectation of being discriminated by the others, nurtured by negative psychological and emotional reactions [50]. According to the Theory of Social Stigma, for dealing with the experience of discrimination, people employ either active or passive coping strategies. The former reflects motivation for struggle and change, whereas the latter consists in elusive and disengagement attitudes [27,28,51]. Our findings and previous studies suggest that difficulties at following the recommendations against COVID-19 infection are higher among those with more unstable housing and residence status, which is likely due to no available income when the guidelines implied to buy products (mask, certain soap, disinfectant, etc.) [2,52]. Also, access to information about COVID-19 preventive measures is a challenge in such conditions [9,52]. That could be experienced by migrants as the failure of their migration project and social protection programs, leading them to lose trust in the host country. That may result also in passive coping strategies, involving elusive attitudes such as the rejection of state guidance for COVID-19 prevention, or reluctance to get vaccinated [15,19,53,54]. From that perspective, the survey answers “No, because I’m not able” and “No, because I don’t want” to the question about the precautions use may be seen as two manifestations of the same psycho-social condition. A vicious circle of discrimination and poor utilization of protection measures against COVID-19 can be created, and the finding of an association between the selfperceived stigmatization and the negative healthcare seeking attitude (i.e., the answer “No, I would not contact a doctor if I or someone close to me develop COVID-19 Table 3 Multiple outcome mediation model in the full set Abbreviations: 95% CI 95% bias-corrected bootstrap confidence interval, SS Self-perceived stigmatization, CAREi Difficulties to adhere to preventive recommendations against COVID-19 infection without question 19, PD Psychological distress a Answers flipped: higher score indicates that people would not go to a doctor in case of COVID-19 symptoms Regressions Unstandardized estimate (95% CI) Standardized coefficient p-value Direct effects Outcome model: SS (R2 = 0.186) CAREi 0.205 (0.135; 0.275) 0.147 < 0.001 PD 0.456 (0.406; 0.507) 0.393 < 0.001 Outcome model: Go to Doctor a (R2 = 0.104) SS 0.353 (0.267; 0.438) 0.323 < 0.001 Mediator model: PD (R2 = 0.08) CAREi 0.109 (0.058; 0.161) 0.091 < 0.001 Indirect effects (proportion mediated) PD (8.2%) 0.050 (0.026; 0.074) 0.036 < 0.001 Total effect 0.607 (0.492; 0.722) 0.505 < 0.001 Page 9 of 14 Marchietal. BMC Public Health (2022) 22:942 Fig. 2 Multiple outcome mediation model in the full set. Legend: The estimates reported are the unstandardized regression coefficients. *p < 0.05; **p < 0.01; ***p < 0.001. Abbreviations: CAREi: difficulties to adhere to preventive recommendations against COVID‑19 infection without question 19; PD: psychological distress; SS: self‑perceived stigmatization Table 4 Results of parallel multiple mediator model in the full set Abbreviations: 95% CI 95% bias-corrected bootstrap confidence interval, SS Self-perceived stigmatization, SDH Socio-demographic hardship, CARE Difficulties to adhere to preventive recommendations against COVID-19 infection, PD Psychological distress Regressions Unstandardized estimate (95% CI) Standardized coefficient p-value Direct effects Outcome model: SS (R2 = 0.246) SDH 0.265 (0.189; 0.341) 0.274 < 0.001 PD 0.342 (0.275; 0.409) 0.293 < 0.001 CARE 0.105 (0.028; 0.181) 0.075 0.007 Mediator model: PD (R2 = 0.160) SDH 0.330 (0.275; 0.385) 0.399 < 0.001 Mediator model: CARE (R2 = 0.069) SDH 0.183 (0.130; 0.235) 0.263 < 0.001 Indirect effects (proportion mediated) PD (28.5%) 0.113 (0.087; 0.139) 0.117 < 0.001 CARE (4.8%) 0.019 (0.005; 0.033) 0.020 0.007 Total effect 0.397 (0.327; 0.468) 0.411 < 0.001