Asthma-Like Symptoms in Homeless Children in the Greater Paris Area in 2013: Prevalence, Associated Factors and Utilization of Healthcare Services in the ENFAMS Survey

Introduction Asthma remains poorly studied in homeless children. We sought to estimate the prevalence of asthma-like symptoms (ALS) and to identify the factors associated with ALS and healthcare service utilisation. Materials and Methods A cross-sectional survey of a random sample of sheltered homeless families was conducted by interviewing 801 parents of children (0–12 years) in 17 languages. ALS were defined as wheezing or night cough without fever during the previous year. Poisson regression models with robust error variance were used to compute prevalence ratios (PR) for factors associated with ALS and healthcare service utilisation for ALS. Results The prevalence of ALS among the children was 19.9%. Poor housing sanitation was significantly associated with ALS, as being born in the European Union. Most of the children with ALS had used healthcare services (85.4%). The main barriers to accessing such services were having lived in France for less than 49 months, having difficulties in French and living in poor housing conditions. Conclusion ALS prevalence seemed lower than in the general child population, possibly because of the children's origins. Environmental factors associated with ALS point to the need to improve the indoor environment of family shelters. The relatively high rate of healthcare service utilisation should not overshadow existing barriers.


Background
Asthma prevalence among children in the general population is approximately 10% in Western countries, such as the United States and France, but appears to be much lower in Africa and Northern and Eastern Europe [1,2]. In the latest French national health surveys in schoolchildren, 9.8% of those aged 5 years and 14.4% of those aged 10 years had asthma. The prevalences of asthma-like symptoms (ALS) were, respectively, 10.7% and 10.1% for wheezing and 12.6% and 16.5% for night cough without fever [3,4].
Asthma prevalence can reach as high as 40% among sheltered homeless children, according to several studies conducted in the United States [5][6][7][8]. Because homeless families are often housed in dilapidated buildings, they may be exposed to a poor indoor environment (moulds, dust mites, pets, mice, etc.), which is detrimental to their health, especially their respiratory health. Since the indoor environment is known to have a significant impact on childhood asthma [9][10][11][12], these factors deserve to be studied in this population. Also, since socioeconomic conditions and racial and ethnic disparities lead to inequalities in access to asthma treatment and care [13][14][15][16][17], homeless families may be at risk for under-or inappropriate healthcare service utilisation.
In 2013, given the dramatic increase in the number of homeless families in the Paris region over the previous ten years (exceeding that of lone homeless persons in 2010 [18]) and the complete lack of epidemiological data on homeless children, the ENFAMS survey (Enfants et familles sans logement [Homeless Children and Families]) specifically focused, for the first time in France, on sheltered homeless children and families in the Greater Paris Area.
The aims of our study were: 1) to estimate the prevalence of ALS among sheltered homeless children in this area; 2) to identify environmental and socioeconomic factors associated with ALS; and 3) to identify factors associated with healthcare service utilisation among children with ALS.

The ENFAMS survey
The methodology used in the ENFAMS survey has been described elsewhere [19]. Briefly, the survey was conducted between January and May 2013 in a random sample of 801 homeless families with 566 children aged 0-5 years and 235 children aged 6-12 years in the Greater Paris Area (a region of 762 km 2 with 6.7 million inhabitants). The survey employed a threestage random sampling procedure. First, the sampling frame of the 796 existing accommodations for the homeless in the region were randomly selected using a stratification based on the type of facility, its distance from the centre of Paris and its distance from the nearest subway or railway station. Four types of facility have been considered: emergency housing centres which house homeless people or families for periods ranging from one night to a few months; social reinsertion centres which the objective is to help families to access or recover personal and social autonomy proposing individualised, globalised accompaniment for all families facing difficulties; centres for asylum-seekers which cater for people whilst their applications for asylum are being processed by the French refugee office; and social hotels which are part of the emergency housing centres initiative, but offer real private areas for each family (rooms with one or two beds instead of dormitories), a decent level of comfort, washing facilities and private kitchens or kitchens shared just by a few families. Second, 801 families were randomly selected from among those with at least one child under 13 years of age living with them. Third, one child under 13 was randomly selected. Family participation rate was 79%; non-participant adults were younger (mean age: 33 years vs. 38), more often men (15.3% of all those who refused vs. 4.6% of those who accepted) and had more than two children under care more frequently than participants (31.7% vs. 23.1%).
A face-to-face questionnaire was administered, in one of 17 languages, to the mother (or the father, if the mother was not there) by a trained interviewer and a psychologist. All questionnaires and information tools have been entirely translated from French into 8 languages (English, Arabic, Armenian, Bulgarian, Mongolian, Romanian, Russian, Tamil) by professional translators and validated by reverse translation. Then they were also translated orally in 8 other languages (Spanish, Italian, Portuguese, Serbian, Soninke, Wolof, Bambara and Lingala) by bilingual psychologists and survey interviewers. The questionnaire collected data on the parents' demographics, socioeconomic status, migration history, residential mobility and health, and the selected child's demographics, health, diet, sleep patterns and physical activities.
The participants provided their written informed consent to participate for themselves and their child in accordance with French law. This study was approved by two ethics boards, the Comité de protection des personnes (CPP) in Île-de-France and the Comité consultatitf sur le traitement de l'information en matière de recherche dans le domaine de la santé (CCTIRS), and by the Commission nationale de l'informatique et des libertés (CNIL), which regulates individual data protection.

Outcomes and explanatory variables
Based on the ISAAC questionnaire criteria [20], ALS were defined as wheezing and/or night cough without fever during the previous year.
Healthcare service utilisation (yes/no) was defined as any medical consultation, emergency department visit or hospitalisation for respiratory symptoms during the previous year. Children hospitalised for respiratory problems were considered severe cases.
Explanatory variables were the parents' demographics (age, length of residence in France, and number of children), socioeconomic status (administrative status, education level, employment status, household income per consumption unit (CU), and difficulties in French) and health-related characteristics (self-reported physical health, smoking status, healthcare service utilisation [defined as at least one medical consultation during the previous 12 months], and health insurance status); housing characteristics (the type of homeless housing facility, poor sanitation including the presence of mice or cockroaches, cold, dampness, poor-quality bedding, a lack of space or crowding as reported by the parent, and the length of stay in the present housing facility); and the child's demographics (age, sex and region of birth) and medical characteristics (atopic dermatitis, defined as an itchy rash having appeared and disappeared intermittently since birth, and the frequency of respiratory symptoms).

Statistical analyses
Three populations were considered in our analysis. First, ALS prevalence was computed for all the children included in the ENFAMS survey. Second, the factors associated with ALS were analysed among the children with no missing data for any of these factors. Third, the factors associated with healthcare service utilisation were analysed among the children with ALS (again, those with no missing data) (Fig 1).
To account for the unequal probability of inclusion and to provide valid estimates, the complex sample design was redressed by weighting all our estimates and models by the inverse of the inclusion probability of each participant. Factors associated with previous-year ALS and healthcare service utilisation among children with ALS were explored using Poisson regression models, from which prevalence ratios (PRs) were obtained [21]. A stepwise variable selection procedure was performed to determine the final models (adding and removing thresholds of 0.10). Age was forced into both models, while sex was forced only into the analysis of factors associated with ALS. By way of an exploratory analysis of factors associated with ALS severity, we used the same procedure to study those associated with hospitalisation. All the analyses were performed using R software, version 3.1.2, with the survey package.

Results
Eight hundred and one children were included in the ENFAMS survey, with an extrapolated estimate of 17,661 (95% CI = [16,969]) children under 13 housed with their parents in homeless facilities.

Population's characteristics
The children's median age was 4 years, 53.2% were girls, and 73% were born in the European Union (61.7% in France). For 95.3% of them, the interviewed parent was their mother ( Table 1). Half of the children had a parent who was under 32 years of age, who had been living  in France for less than 3.5 years and/or who reported a monthly household income under 227 €/CU. Two-thirds had parents who were legal residents (64.8%) but who had difficulties understanding, speaking, reading or writing French (64.9%). For the most part, the children were living with their parents in social hotels (76.5%), and this in poor conditions: 34.2% experienced cold, 36.5% experienced dampness and 45.5% experienced poor sanitation. They had been in their present housing facility for a median of 8 months.
Weighted prevalence of wheezing, night cough without fever and either symptom (ALS) The data for wheezing, night cough without fever and either symptom concern, respectively, 792, 789 and 790 of the 810 children included in this survey. Among the respondents, the prevalence and 95% confidence intervals were 14 (Table 2). Overall, these proportions seemed to be higher among girls than boys and varied with age, with the prevalence of wheezing being higher among the children aged 0-1 year and 6-7 years than in the other age groups. With regard to the children who had had either symptom, boys were more frequently affected than girls at age 8-9 compared to other age groups (Fig 2).

Factors associated with ALS
Of the 679 children included in the analysis (Fig 1), 19.6% (n = 126) had had ALS during the previous year. After adjustment for the children's age, sex and region of birth (children born   Healthcare service utilisation for respiratory symptoms One hundred and twenty-seven children with ALS were included in the analyses of the healthcare service utilisation for respiratory problems (Fig 1). Of them, 85.4% (n = 110) had used healthcare services during the previous year: 57.8% (n = 83) had been seen by a medical practitioner, 32.9% (n = 34) had visited an emergency department, and 32.1% (n = 35) had been hospitalised. Parental socioeconomic status and the child's region of birth were both significantly associated with healthcare service utilisation in bivariate analysis (Table 4). In the multivariate analysis adjusted for the children's age, those whose parents had been in France for less than 49 months or had difficulties in French and those who were living in poor housing conditions had   (Table 5).

Discussion
The ENFAMS survey, which was conducted in 17 languages, was the first one in France among homeless families, a hard-to-reach population. Its design made it possible to estimate ALS prevalence among homeless children housed in the Greater Paris Area and to identify associated factors, even if its cross-sectional design precludes any assertion of causality. The survey did not explicitly ask about asthma diagnosis because there is no consensual definition of asthma in preschool-aged children [22,23]. We estimated that 19.9% of homeless children under 12 years of age in the Greater Paris Area had had ALS during the previous year. Those born in the European Union and those living in housing conditions with poor sanitation were more affected with ALS. Dampness was at the borderline of statistical significance. Even if most of the children with ALS had used healthcare, some barriers exist, such as a shorter parental length of residence in France, parental difficulties in French and poor housing conditions.
To the best of our knowledge, our study is the first one in France, as well as in Europe, to focus on the respiratory health of homeless children and their use of healthcare services.
Surprisingly, perhaps, the prevalence of ALS found in the ENFAMS survey tended to be lower than that observed in the general child population in previous French studies (Fig 3) [1,3,4,[24][25][26]. Indeed, only wheezing was more common in the ENFAMS survey, and then only among 6-and 7-year-olds. The fact that the survey was conducted in the winter may have resulted in a lower frequency of reported night cough without fever. Indeed, even though the parents were asked whether or not their child had usually had fever with his/her respiratory symptoms during the previous 12 months, we cannot rule out the possibility that they tended Poor housing condition was defined by at least one of the following problem: poor sanitation, coldness, dampness or poor-quality bedding. doi:10.1371/journal.pone.0153872.t004 to answer in the affirmative because of a recent winter viral syndrome, failing to mention the previous episodes of respiratory symptoms (recall bias). Furthermore, it is known that the prevalence of asthma and ALS in immigrants from low-and middle income countries is lower than that in the host population of high-income countries [27][28][29]. In our study, 27% of the children were born outside the European Union in countries were asthma prevalence is, overall, lower than in France [1]. In the United States, where asthma prevalence in homeless children can reach 40%, the situation is completely different because most of them are African-American [5,6], a subgroup of the population known for its high asthma prevalence compared to that among whites [2]). Environmental triggers, such as second-hand smoke, and indoor allergens, such as mice, cockroaches and moulds, may increase asthma symptoms [9][10][11][12]. In the ENFAMS survey, dampness and poor sanitation were found to be associated with ALS. Of course, the data on ALS and indoor environments may have been biased in the ENFAMS survey, since they were reported by the parents. However, it was previously found that parental reports regarding exposure to dampness were more reliable than onsite inspections and that self-reported allergic symptoms were consistent with medical examinations [30]. Unfortunately, these asthma triggers are frequently encountered in homeless shelter environments, but the ability of families in such facilities to control their living environment is limited [31]. On the subject of tobacco smoke, the smoking status only of the interviewed parent (the mother in most cases) was asked about, so we did not have an accurate measurement of the children's exposure. Poor parental self-reported health may have had an effect on the answers regarding the children's health, but ALS does not seem to have been overestimated in the ENFAMS survey. Indeed, the children with ALS had had atopic dermatitis more frequently, which is consistent with the literature on asthma and wheezing illness. It will be noted that atopic dermatitis in early childhood seems to increase the risk of asthma later in childhood [32]. In many schoolaged asthmatic children, wheezing occurs before or with the onset of atopic dermatitis [33]. As well, atopic dermatitis may have been overreported in the survey (it was defined as "itchy rashes appearing and disappearing intermittently since birth"), but it seems unlikely that it was reported differently between the children with and without ALS.
No socioeconomic factors were associated with ALS in the ENFAMS survey, in which the families had a very low household income and were often unemployed but had a relatively good education level. Actually, associations between these factors and asthma or wheezing illness are found inconsistently in the literature [4,[34][35][36][37]. Furthermore, the prevalence of wheezing and asthma is consistently reported to be higher in boys than in girls [4,38,39]. In our study, this difference was observed in 8-and 9-year-olds only. This could reflect the predominant role of housing and living conditions in this underprivileged population, as no difference was found in a U.S. study either, in homeless children aged 4 to 7 years [7].
Eventually, we cannot highlight in the present study the role of the stress as a contributor to ALS or to exacerbation of symptoms among children already at risk for ALS [40] when homeless children living in housing conditions are known to be exposed to many of stressors, such as community and family violence [41]. Also, another limit of our study is the absence of distinction between atopic and non atopic asthma phenotypes. Risk factors and measures of severity vary between children with different asthma phenotypes [42] but the data not allowed us to distinguish these phenotypes (the dermatitis atopic was probably not well declared and there was no other sign of atopy requested). Missing data were taking into account using a listwise deletion method. In the analysis of factors associated with ALS, 679 children were included and 122 excluded (15.2%). There was no notable selection bias (in particular no difference in ALS prevalence), except that included children had parents who were more often in regular situation (68.0% versus 47.8%, p<0.01) and were more frequently smoker (13.3 versus 3.6%, p<0.01). In the analysis of the healthcare service utilisation for respiratory problems, 127 children with ALS were included and 20 excluded (13.6%). Included children had more frequent respiratory symptoms (25.9% had symptoms more than once a week, versus 3.8% of the excluded children, p = 0.03), used less healthcare service (85.4% versus 96.8%, p = 0.05), and had parents with a higher monthly household income per consumption unit (332.3 +/-43.3€ versus 185.1 +/-43.0€, p = 0.02) but with a more frequent lack of any health insurance (p<0.01).
The overall high proportion of children with ALS who had used healthcare services at least once during the previous year may seem encouraging, but only 57.8% of them had actually been seen by a medical practitioner, which could indicate a lack of regular follow-up. Inequalities in asthma care have been reported among minority and poor children, who are at greater risk for undertreatment, suboptimal disease control, emergency healthcare utilisation and hospitalisation [5][6][7][13][14][15][16]. In France, according to a national survey among 5-year-olds, 4.6% of those with wheezing or being treated for wheezing or asthma (n = 2237) had been hospitalised during the year preceding the survey [4]. In the ENFAMS survey, the hospitalisation rate for ALS was 32.1%.
Our data suggest a socioeconomic gradient in healthcare utilisation by children with ALS in this population. In univariate analysis, a low level of education, being unemployed or inactive, an irregular administrative status, and a low or very low income are all parental factors associated with lower healthcare service utilisation by their children. A lack of statistical power may explain why this was not observed in multivariate analysis. However, in the multivariate model, a shorter length of residence in France, difficulties in French and parental non-use of healthcare services remained as barriers to accessing healthcare. This may reflect a lower level of acculturation and of knowledge concerning healthcare norms and the organisation of healthcare services in France [43].

Conclusion
ALS prevalence in homeless children seems to be lower than that observed in the general child population. This may be due to the origin of some of these children. Environmental factors associated with ALS, such as dampness and poor housing sanitation, point to the need to improve the indoor environment of family shelters in order to limit asthma triggers. Despite a relatively high rate of healthcare service utilisation for ALS, there are still barriers to accessing them for the most underprivileged and/or the least acculturated migrants that need be addressed to improve care for this vulnerable child population.