Do vulnerable groups access prevention services? Cervical cancer screening and HIV testing among homeless migrant women in the Paris metropolitan area

Introduction Homeless migrant women, facing adverse living conditions and barriers to legal status, are at risk of cervical cancer, HIV infection and may encounter barriers to screening services. We investigate factors associated with each screening in a population of migrant women in France and aim to determine the mean time since last HIV testing according to duration of residence in France. Methods We use data from the DSAFHIR study (Rights and Health of Migrant Women in Emergency Housing) investigating health and migration experience of homeless migrant women housed in emergency housing hotels in the Paris Metropolitan area in 2017. We computed multivariate logistic regression models to investigate no lifetime cervical cancer screening (CCS) and no lifetime HIV test. We used linear regression models to analyze time since last HIV test. Results We included 469 women. 46% of respondents had no lifetime CCS, 31% had no lifetime HIV test. Both screenings were associated with educational attainment and French proficiency. Compared with duration of residence < 1 year, duration ≥ 7 years was associated with a lower likelihood of no lifetime CCS (adjusted Odd Ratio = 0.17; 95% CI = 0.07–0.39). Compared to women born in North Africa, women born in West (aOR = 0.15; 95% CI = 0.07–0.33) and East Africa (aOR = 0.06; 95% CI = 0.02–0.20) were less likely to have no lifetime HIV test. Time since last HIV test increased for each additional year spent in France (coef = 0.21; 95% CI = 0.09, 0.33). Conclusion While access to CCS remains poor for recent migrants, HIV testing is more likely to occur shortly after migration.


Introduction
Homeless migrant women, facing adverse living conditions and barriers to legal status, are at risk of cervical cancer, HIV infection and may encounter barriers to screening services. We investigate factors associated with each screening in a population of migrant women in France and aim to determine the mean time since last HIV testing according to duration of residence in France.

Methods
We use data from the DSAFHIR study (Rights and Health of Migrant Women in Emergency Housing) investigating health and migration experience of homeless migrant women housed in emergency housing hotels in the Paris Metropolitan area in 2017. We computed multivariate logistic regression models to investigate no lifetime cervical cancer screening (CCS) and no lifetime HIV test. We used linear regression models to analyze time since last HIV test.

Results
We included 469 women. 46% of respondents had no lifetime CCS, 31% had no lifetime HIV test. Both screenings were associated with educational attainment and French proficiency. Compared with duration of residence < 1 year, duration � 7 years was associated with a lower likelihood of no lifetime CCS (adjusted Odd Ratio = 0.17; 95% CI = 0.07-0.39). Compared to women born in North Africa, women born in West (aOR = 0.15; 95% CI = 0.07-0.33) and East Africa (aOR = 0.06; 95% CI = 0.02-0. 20

Introduction
Cervical cancer is the ninth most common cancer affecting women in the WHO European region, with 67,000 new cases and 31,000 deaths in 2020 [1]. In France, around 3000 new cases and 1100 deaths are reported each year [2]. Large disparities exist between regions in Europe, with higher incidence rates in Central and Eastern Europe compared with Northern and Southern Europe [1]. New HIV diagnoses among women totaled 50,000 cases in the WHO European Region in 2018 [3], with 1700 occurring in France in 2018 [4]. Large disparities exist within the region, with greater incidence and mortality rates in Eastern Europe [3]. In Europe, migrant women from low-income countries are at higher risk of developing cervical cancer than non-migrant women, due to higher prevalence of human papillomavirus (HPV) in their country of origin and insufficient access to screening [5,6]. They are also more likely to be diagnosed at an advanced stage of the disease, leading to worse health outcomes than the host population [6]. In France cervical cancer screening (CCS) is carried out by Pap smear. At the time of the study, it was recommended every three years for women aged 25 to 65 years, after two initial tests one year apart. HPV tests were introduced in 2019, to be carried out every five years for women above 30 years old. Self-sampling HPV testing was recommended in the French guidelines in 2019 [7] but were not available at the time of the study. Pap smears are carried out by gynecologists, general practitioners, and midwives. The test is reimbursed under the National Health Insurance (NHI), with no upfront payment for lowincome patients [2]. Migrant women with legal residency status are eligible to the NHI. Undocumented migrant women are eligible to a specific health coverage plan for undocumented people, also covering CCS at no cost for patients. CCS coverage was 59% in France in 2015-2017 [8], with large regional and social disparities. Previous studies have found that low educational level, low income and being unemployed were risk factors for being overdue for screening [9,10]. Migrant women in the Paris area were more likely to have delayed and no lifetime CCS compared to women of French origin [11].
In the EU, it was estimated that approximately 40% of new HIV cases were reported among migrants [4]. Migrant women in France constitute the group with the second highest incidence rate for new HIV infections [12,13]. In 2017, 75% of new infections through heterosexual transmission concerned immigrants, primarily from Sub Saharan Africa [12]. As in other European countries [14], the high infection rate in this group is due to high HIV prevalence in countries of origin, but also due to HIV infections in France: it was estimated that 30% of HIV-positive migrant women from Sub Saharan Africa are infected after migration [15]. Migrant men and women are also more likely to be diagnosed late, negatively impacting treatment, mortality [16] and virus transmission [12,17]. With economic deprivation increasing infection risk [18], it is crucial that migrant women be tested repeatedly in the years following migration. They are indeed specifically targeted for screening: while the French National Authority for Health recommended at least one lifetime HIV screening for the general population, it recommended yearly HIV testing among migrants from countries with high HIV prevalence, specifically Sub Saharan Africa and the Caribbean [13]. HIV testing is available in France as blood tests and rapid tests in various medical and non-medical settings, reimbursed under the NHI, and available for free in screening centers and non-governmental organizations (NGOs).
While much is known concerning both HIV testing and cervical cancer screening among migrant populations, we focus here on a group of migrant women who encounter several layers of deprivation, living in housing instability, financial stress, and uncertainty regarding administrative status. The health care system in France, with universal health care coverage, is meant to be accessible to all. In the case of these two screenings, the public health objective is to reach out to and include the most disadvantaged groups [13,19]. Analyzing both screenings together helps to gauge the capacity of universal health care systems to be inclusive of the most vulnerable groups in our societies. With different access and target populations, the two screenings illustrate and nuance access to primary prevention programs for marginalized groups.
Our objective was to identify factors associated with CCS and HIV testing in a population of vulnerable migrant women. We further aimed to determine the mean time since last HIV testing in this group, according to duration of residence in France.

Data
We used data from the DSAFHIR study (Rights and Health of Housed Migrant and Refugee Women), carried out in 2017.

Population
The study investigated the health needs and obstacles in accessing healthcare services of homeless migrant women housed in low-end hotels as an emergency housing option in the Paris Metropolitan area in France (N = 469). Women aged 25 to 40 are over-represented in this population, as housing women with children is a priority of the emergency housing system. Migrant families face obstacles accessing resources and service providers from hotels situated in distant suburban areas. Moreover, as immigration policies have restricted access to longterm residence permits, access to public healthcare coverage is hampered, further challenging access to care.

Data collection and study setting
A sampling based on a two-stage non-random selection was used to select geographical zones and hotels. All migrant women aged 18 years or more housed in the selected hotels were eligible. The questionnaire was administered in the respondents' hotel rooms, in April and May 2017. Respondents were offered to participate in sexual health interventions and a follow-up survey eight months after inclusion to evaluate the interventions. We used here only data collected at inclusion. The design of the study was detailed elsewhere [20].

Data tool
The questionnaire was administered face to face in ten different languages (French, Russian, Romanian, Arabic, Bambara, Soninke, Georgian, Armenian), by multilingual female surveyors. As much as possible, questionnaire sections of existing surveys were used. Prior to data collection, the questionnaire was pilot-tested, and it was reviewed by a committee of women with lived experience of emergency housing in hotels.

Ethics
We paid particular attention to the quality of informed consent. The multilingual surveyors made sure respondents understood the goal of the study and the possibility to refuse to participate. In particular, we insisted that participating or refusing to participate would not have any impact-positive or negative-on their housing situation.
Ethical approval for the DSAFHIR study was granted by the People Protection Committee for medical research on March 30, 2017 (CPP reference number: IDRCB number 2016-A02005-46).

CCS.
Respondents were asked whether they ever had a Pap smear, and time (more or less than 3 years prior) and country of last test (see Fig 1)

Independent variables
We tested the association between each screening and variables informing on socio-demographic characteristics, access to healthcare services and general health. Socio-demographic characteristics included age, educational level, duration of residence in France, relationship status, self-assessed knowledge of French language and paid work in the last 12 months (long or short-term, declared or not). We investigated access to care through medical insurance coverage, visit to a general practitioner (GP) and to a gynecologist in the last 12 months, having given birth in France and owning a public transportation card. General health was investigated through a measure of self-perceived physical health. Current depression was assessed using the Mini International Neuropsychiatric Interview (French version 5.0.0, 1998) [21].

Statistical analyses
We ran bivariate analyses to identify factors associated with each of the two screenings using Chi2 tests. Variables with p-value < 0.2 were selected for multivariate analysis in logistic regression models. Variables were then backward-selected manually at the 0.2 threshold. Knowledge of French language was removed from multivariate analyses because of correlation with region of origin and educational level.
We ran the same analyses fitting maximum-likelihood two-equations probit models (bivariate probit) with simultaneous estimation of equations associated with the two outcomes of interest [22]. Bivariate probit model is a generalization of the probit model used to estimate two correlated binary outcomes jointly. The method adjusts for the potential residual correlation of our two outcomes of interest, and yields in this case unbiased estimates. Because the pvalue of the residual correlation coefficient rho was greater than 0.05 (p = 0.19), therefore not confirming the correlation of error terms after controlling for the independent variables, and because our results showed no difference from the results estimated with simple logistic regression models, we chose to present the multivariate logistic regression models for each outcome (see S1 Table for bivariate probit model).
We computed mean number of years since last HIV test according to duration of residence. P-value was computed using a simple linear regression predicting time since last HIV test based on duration of residence as a categorical variable.
A simple linear regression was calculated to predict time since last HIV test based on duration of residence as a continuous variable. Covariates (age, educational level, relationship status, access to health professionals) were chosen according to hypotheses suggested by the literature of variables associated with both HIV screening and duration of residence and likely to confound their association and were then added into a multivariate linear regression model.
Because of different no lifetime HIV test rates between women from Sub Saharan Africa and women from other regions, and because HIV testing recommendations differ for the two groups [13], we analyzed the timing of HIV testing according to duration of residence separately for women born in Sub Saharan Africa and women from other regions.

Analytical sub-samples
Concerning CCS, we included respondents aged 25 to 65 years, corresponding to the age range for CCS in the French screening guidelines. We excluded 10 respondents with non-valid value for CCS realization (missing = 1; 'I don't know' = 9). Analytical sample for CCS included 387 respondents.
HIV testing analyses included respondents of all ages. We excluded respondents with missing values (n = 2) or answering 'I don't know' (n = 37) for number of lifetime HIV tests. Analytical sample for HIV test included 430 respondents.

PLOS ONE
For analyses of time since last HIV test, analyses were carried out on 268 respondents with a valid date of last HIV test.
All analyses were performed using Stata 15.

Population description
One third of respondents were aged below 30 years and 46% were aged between 30 and 40 years (Table 1). One third of the respondents had no schooling or only primary education, and 18% pursued higher education. 56% of the respondents were born in Sub Saharan Africa. 41% of the sample were undocumented, and 21% were uninsured. Two thirds (70%) of respondents had not worked in the past year. French language proficiency was unevenly distributed, with 41% of respondents with a very good command of French and 20% with very poor proficiency. Two thirds of the sample had given birth in France. 33% of respondents had no GP visit in the past year, and 57% had no gynecologist visit. 56% of the respondents reported good self-perceived health status, while 45% had current depression.

CCS
Screening participation rates. 46% of respondents had no lifetime CCS ( Table 2). The great majority of Pap smears took place in France, with only 4% of respondents having a last Pap smear abroad (not shown in tables).
Factors associated with no lifetime CCS. In bivariate analyses, no lifetime CCS was significantly associated with socio-demographic characteristics such as educational level, French language proficiency, relationship status and region of origin. Two thirds (69%) of respondents with no schooling had no lifetime CCS, versus only one third (30%) of respondents with higher education (p = 0.001). 60% of respondents with a very poor proficiency in French had no lifetime CCS, versus only 36% of those with very good command of French (p = 0.003). Women in a relationship were more often never screeners (52% vs 36%, p = 0.002). Respondents with longer length of residence had significantly lower rates of no lifetime screening (p < 0.0001). Respondents with no health coverage were more often never screeners (54%) than respondent with national health coverage (38%, p = 0.05). Access to healthcare professionals was associated with no lifetime screening: 41% of women who had visited a GP in the last year and 35% of women who had visited a gynecologist were never screened, while 54% of women with no GP (p = 0.01) or gynecologist (p < 0.0001) were never screened. 41% of women who had given birth in France had not been screened, while this was the case for 53% of women who had not given birth in France (p = 0.02).
In multivariate analyses, educational level, relationship status, duration of residence and no visit to a gynecologist remained associated with no lifetime CCS (Table 2). Compared to women with secondary education, women with higher education were less likely to have no lifetime CCS (adjusted Odd Ratio = 0.46; 95% CI = 0.23-0.93). Single or separated women were less likely not to have been screened (aOR = 0.46; 95% CI = 0.26-0,81). Women with duration of residence of 3 years or more were less likely to have never been screened, compared with recently arrived women. Women who had visited a gynecologist in the past 12 months were also less likely not to have been screened (aOR = 0.49; 95% CI = 0.30-0.80).

HIV
Screening participation rates. 31% of respondents had no lifetime HIV test (Table 3). Factors associated with no lifetime HIV test. In bivariate analyses, women with no schooling were never tested for HIV twice as often as women with higher education (41% vs 20%, p = 0.02). Only 18% of women with very good command of French language had never been tested for HIV, compared to 49% of women with poor language proficiency and 43% of women with very poor proficiency (p<0.0001). Women in a relationship had never been tested more often than single women (43% vs 17%, p<0.0001). Women from Sub Saharan Africa had the lowest rates of no lifetime HIV test. Owning a public transportation pass was associated with a lower rate of no HIV testing (23% vs 36%, p = 0.008). 27% of women who had given birth in France had not been tested, while this was the case for 40% of women who had had not given birth in France.
In multivariate analyses, educational level, relationship status, region of origin and owning a public transportation card were associated with no lifetime HIV testing (Table 3). Compared to women with secondary education, women with higher education were less likely not to have been tested (aOR = 0.43; 95% CI = 0.20-0.95), but there was no difference for women with no schooling. Compared to women from North Africa and the Middle East, women from West Africa and East or Central Africa were less likely not to have been tested (aOR = 0.15; 95% CI = 0.07-0.33 and aOR = 0.06; 95% CI = 0.02-0.20). Women with no public transportation card were more likely not to have been tested (aOR = 1.72; 95% CI = 1.00-2.98).

Time since last HIV test
The mean number of years since last HIV test was 1.85 years in the total sample and increased from 1.35 years in newly-arrived migrant women to 2.56 years in women who had been in France for 7 years or more (p = 0.04) (

Discussion
We found that 46% of the respondents had never been screened for cervical cancer, and 31% had never received an HIV test. Women with higher educational attainment, with better  French proficiency and women who were not in a relationship were more likely to have received the screenings. Women with a long duration of residence and women who had visited a gynecologist in the last year were more likely to have received a Pap smear, while women from Sub Saharan Africa were more likely to have been tested for HIV. Because data is scarce concerning this specific population, it is difficult to make international comparisons. However, studies from across Europe have consistently reported low CCS participation among migrant women, and their results show that the most vulnerable migrant women are the most likely not to be screened: migrant women with low socio-economic status [23][24][25][26][27], low social support [26,28], and with short duration of residence [23,25,28]. Concerning HIV testing, a study carried out in 2015 in several European countries [14] has found that access to HIV testing for migrant women was better for women born in Africa, for women with permanent residency and for women who had received prenatal care in the host country.
Analyses from the SIRS (French acronym for health, inequalities and social ruptures) cohort, a representative sample of adult French-speaking population of the Paris Metropolitan area, found that 9% of the total eligible population had no lifetime CCS [10], and that 26% of immigrant women had no lifetime CCS [11] in 2010. Because the cohort data were collected only in French, women encountering obstacles to CCS due to a poor command of French and generally poor knowledge of the healthcare system were excluded, accounting for the higher rate of no lifetime screening in our results. In a sample of homeless women housed in emergency shelters in the Paris metropolitan area in 2013, researchers found no lifetime CCS rates similar to our own [29].
In 2018, after this survey was conducted, an organized screening program for cervical cancer was initiated in France, where women with delayed screening are invited by post. While it is undoubtedly beneficial for increasing screening participation rates, it may not reach migrant women in unstable housing, whose affiliation to the national health insurance system is periodically jeopardized by precarious administrative status. Self-sampling has been shown to improve screening participation for under-screened women [30] and will soon be available in France [7]. Although self-sampling might prove a helpful option, it would not reach women with no health coverage and might be difficult to carry out for women with little command of written French. Outreach services, such as mobile screening clinics, might improve access to CCS. Additionally, as a large share of this population has given birth in France, increasing CCS in prenatal care might be an efficient way to reach homeless migrant women.
Concerning HIV testing, a study conducted in the general population in the Paris metropolitan area in 2010 found that 22% of women aged 18 to 54 had no lifetime HIV test [31]. Analyses from the SIRS cohort [32] found that 33% of women were never tested, and 41% among immigrant women. The better outcome in our population can be explained by the fact that women from Sub Saharan Africa represented 56% of our sample (versus 26% of immigrant women included in the SIRS cohort [33]) with a low rate of no-lifetime HIV test contributing to drive down the global rate of no lifetime HIV test in our sample and likely accounting for the disparity. The lower likelihood of no lifetime HIV test among women from Sub Saharan Africa has been noted before [33] and reflects the French HIV testing guidelines of repeated HIV testing in migrants from Sub Saharan Africa and the Caribbean. It can also reflect testing opportunities before migration, in countries with high HIV prevalence.
Women with higher education were more likely to have received both screenings, consistent with a large body of literature on the role of educational attainment on health [34,35] and health promotion behaviors [36,37]. This speaks to the increased health literacy, knowledge about infections and testing opportunities among people with higher education, and increased ability to request it or to be receptive when testing is offered. It could also result from differentiated practices of health professionals, possibly prompter to interact with more educated patients. A previous study carried out in the Paris region has found that even among women with a recent medical visit [10,38], higher educational attainment was associated with higher CCS rates. It should be noted that compared to women with secondary education, women with no schooling were not more likely to have no lifetime CCS or no lifetime HIV test. Previous studies investigating hepatitis B [39,40] or HIV [32,41] screening have found that people in precarious living conditions could be screened more often, resulting from public health and service providers' policies to target vulnerable groups.
We found that owning a public transportation card, allowing people to access service providers, is associated with more HIV testing. This resonates with a public debate in the Paris metropolitan area where special public transportation discount for undocumented migrants was cancelled by conservative regional leadership in 2016 [42]. Our results illustrate the direct impact of such misguided policies on public health and the fight against HIV/AIDS in particular.
Women who had been in France for 3 years or more were more likely to have been screened for cervical cancer than women who migrated within the year. This is in line with generally better access to care gained with longer time in the host country, and more opportunities for repeated screening. However, duration of residence was not associated with lifetime HIV testing: greater familiarity with the health care system with longer duration of residence did not provide benefits for HIV testing. This suggests that HIV testing occurs mostly shortly after migration and before migration. Our results confirm that time since the last test increased with each additional year spent in France. This is consistent with previous research that has shown that the median first-time HIV test occurred during the second year in France among Sub Saharan migrants [41] and that time since last HIV test increased with number of years spent in France [17]. This is problematic because it was estimated that one third of HIV positive migrant women from Sub Saharan Africa were infected after migration [15], and therefore need to be tested regularly in the years following migration. Women in our sample are especially exposed to the risk of HIV infection in France because they have no personal housing and live in poverty [18]: they are more likely to have risky sexual encounters, to trade sex, and to have poor negotiating power for safe sex.
Our results suggest that CCS realization demands a long-term integration in the health care system while HIV testing is restricted to recently arrived migrants, as if the risk of infection existed only upon arrival. A long duration of residence in the host country might act as a signal to health professionals that HIV test is superfluous. Long-term integration in the health care system seems beneficial for CCS, but detrimental to HIV testing.

Limitations and strengths
As respondents lived in the Paris metropolitan area, our results can not be extended to migrant women living in other regions of France, where screening possibilities may differ. Because other CCS method exist in some respondents' countries of origin, such as visual inspection with acetic acid, asking only about Pap smear realization may restrict our measure of lifetime cervical cancer screening occurring before migration. As data was self-reported, recall and reporting biases may exist. To help with comprehension, surveyors systematically explained what the Pap smear is and how it is carried out.
The survey was conducted in several different languages, in a population who usually remains unheard and out of reach of large public surveys. There was no medical setting bias, and we could reach women with very poor or no access to health services.

Conclusion
We found poor access to cervical cancer screening for recently arrived migrant women and social inequalities in accessing both screenings. Our results argue for early offer of CCS and for repeated and sustained testing for HIV. Better awareness of migrant women's poor living conditions may help health professionals understand the heightened risk of infection and offer relevant screening.
Supporting information S1