Potentially avoidable hospitalization for asthma in children and adolescents by migrant status: results from the Italian Network for Longitudinal Metropolitan Studies

This is a retrospective longitudinal study based on dynamic cohorts of residents in six cities participating in the IN-LiMeS [22‐24]: Torino, Venezia, Reggio Emilia, Modena, Bologna, and Roma. Each cohort represents one city. Data from the participating cohorts can be analyzed together to create a pattern of outcomes in relation to sociodemographic characteristics in Italy. For this analysis, participants were recruited and followed-up between 01/01/2001 (21/10/2001 for Torino) and 31/12/2013 (31/12/2014 for Venezia).

The participants in the study were all 2–17-year-old individuals registered as residents in the Population Registries of at least one of the six cities included in the study during the follow-up period (N = 1,256,826). The age selection followed the criteria of the Agency for Healthcare Research and Quality (AHRQ), and the case definition for pediatric asthma [25, 26]. Non-resident immigrants were not included, while refugees and asylum seekers were included after having obtained a permit for residence. In 2015, about 86% of immigrants in Italy were registered as resident [27]. To calculate person-time-at-risk, we built datasets containing all time periods of residence for everyone who resided in one of the aforementioned cities for at least 1 day during the study period (residential segments). Net person-years (PYs) at risk were calculated excluding time periods that occurred between emigration and re-immigration in each city. The entry date for the cohort was the registration date in the Population Register from birth or immigration after 01/01/2001 (21/10/2001 for Torino), while the exit date was for the 18th birthday, emigration, death or end of the study (31/12/2013 or 31/12/2014 for Venezia) whichever was the less recent. Record-linkage procedures between demographic and hospital data were performed deterministically using anonymous keys to guarantee the privacy of individuals. Detailed description of the method can be found elsewhere [22‐24]. Figure 1 displays the total number of residential segments; the number of segments with valid anonymous keys for record-linkage to hospital data; the number of eligible residential segments and subjects, and the final study population with valid information on entry and exit dates and exposure for each cohort.

We focused the statistical analysis only on the comparison between Italians and immigrants coming from HMPCs because the latter represent the majority of vulnerable immigrants in relation to health assistance and because the percentage of immigrants from HDCs was very low (it ranged between 0.13% in Reggio Emilia and 0.90% in Roma). We calculated crude (CHRs) and direct age-standardized hospitalization rates (SHRs) using the 2011 Italian census population. CHRs and SHRs were calculated from the total population for each cohort as well as for the total study period (2001–2014). Then, we estimated age and calendar-year-adjusted hospitalization rate ratios (HRRs), and the related 95% confidence intervals (95%CIs) for immigrants from HMPCs compared to Italians, using the negative binomial regression, which allowed us to model the number of occurrences of an event when it has overdispersion, with PYs as the exposure. We stratified the analysis by gender because males were found to have higher asthma prevalence than females during childhood, but lower prevalence and better outcomes at later ages [29]. In addition, we stratified the analysis by cohort to acknowledge any unmeasured contextual factors of the cohort that could have influenced the associations. To provide a global measure of effect we performed a random effect Individual Participant Data (IPD) meta-analysis [30]. Thus, we estimated meta-analytic hospitalization rate ratios (MHRRs) and related 95%CIs. We performed the Cochran heterogeneity test to assess whether the differences observed in the effect measures of each cohort were due to chance and, to quantify such heterogeneity, we calculated the I² statistic, which describes the percentage of variability across studies [31]. We also estimated MHRRs by geographical area of origin; but in this case we could not stratify by gender because in some combinations of cases by gender and area of origin there were no observations. We considered two-sided p-values less than 0.05 as statistically significant. We managed the data using the software SAS Enterprise guide 7.1 and analyzed the data using the Stata software, version 13, commands nbreg, and metan (with options randomi).

The study included person-level-data-linkage across Population Register and Hospital Discharge Records.

The data were linked using anonymized keys and then analyzed in aggregate form. We assessed the linkage quality by calculating the record linkage percentage that was greater than 99.1% (Fig. 1).

Figure 1 displays the selection of the population included in the study. We selected 1,256,826 2–17-year-old subjects during the follow-up time period (18.6% in Torino, 5.4% in Venezia, 4.1% in Reggio Emilia, 4.1% in Modena, 7.0%, Bologna, and 60.9% in Roma) (Table A1, additional material) who contributed with 8,070,415 PYs. The percentage of individuals coming from HMPCs was 13.7%, with percentages ranging between 11.0% in Roma and 22.2% in Reggio Emilia, while those coming from HDCs were only 0.7%. Among the immigrants, the highest percentage of individuals was observed among Central-Eastern Europeans (41.3%) and the lowest among Sub-Saharan Africans (7.3%). Among the participants, there was the same proportion of males and females. The mean age calculated over the whole study period was 9.3 years. The number of subjects enrolled was 608,240 in 2002 and 673,964 in 2013 (2002 and 2013 were the first and last calendar years with complete follow-up for all the cohorts).

During the study period, 3787 hospital discharges for asthma were registered among 2–17-year-old participants, 3304 among Italians (87%) and 471 (12%) among immigrants from HMPCs. Among immigrants, the highest percentage of discharges for asthma was observed for subjects from Asia (38%) and the lowest for those from Sub-Saharan Africa (11%). Males showed higher percentages of hospitalization than females (60% among Italian males and 55% among immigrant males). The mean age at admission was 8.2 years (8.3 among Italians and 5.0 among immigrants). The number of hospitalizations ranged from 426 in 2002 to 182 in 2013, with a clear decrease among Italians and a steady pattern among immigrants.

Age-standardized rates (per 1000 PYs) ranged from 0.18 (95% CIs:0.13–0.24) in Modena to 0.54 (0.46–0.61) in Venezia among Italians, and from 0.14 (0.35–0.54) in Modena to 0.68 (0.60–0.75) in Roma among immigrants. Among Italians, gender specific rates were always higher among males, from 0.19 (0.12–0.27) in Modena to 0.63 (0.57–0.70) in Torino, compared to females whose rates were from 0.17 (0.10–0.25) in Modena to 0.48 (0.38–0.59) in Venezia. Among immigrants, rates ranged from 0.05 (0.00–0.14) in Modena to 0.80 (0.46–1.13) in Reggio Emilia among males, and from 0.24 in Modena (0.03–0.45) and Bologna (0.06–0.42) to 0.66 (0.58–0.79) in Roma among females (Table 1).

Compared to Italian counterparts, excess-age and calendar-year-adjusted asthma PAH rates were found only among immigrant males in Bologna (HRR:2.42; 95%CI:1.53–3.81) and Roma (1.22; 1.02, 1.45), and among immigrant females in Torino (1.56; 1.10–2.20) and Roma (1.82;1.50–2.20). Heterogeneity was observed among areas of origin in the HMPCs group, though estimates were imprecise, due to low or null numbers of cases.

Meta-analytic results showed a higher asthma PAH rate among immigrants compared to Italians among females only (MHRR:1.48; 95%CI:1.18–1.87). Results among males also showed a higher asthma PAH rate among immigrants compared to Italians (MHRR:1.26; 95% CI:0.96–1.71), though not statistically significant. Also, the statistic I² showed a substantial level of heterogeneity across studies for males (65.3%; p = 0.013) (Fig. 2).

Compared to Italians, 63 to 113% meta-analytic excess overall (males and females) PAH was found for all the areas of origin, except for immigrants coming from Central-Eastern Europe (MHRR:0.80; 95% CI:0.65–0.98), with some heterogeneity across cohorts (Table 2).

As among females, immigrant children and adolescents have higher PAH than Italians, while among males they show higher PAH than Italians in the cohorts of Bologna and Roma. A possible interpretation of these findings is that they might reflect a combination of increasing asthma prevalence after immigration, more severe asthma symptoms, a lower ability to manage asthma symptoms and lower care; in addition to more barriers in accessing primary care and increased utilization of emergency units. Barriers may occur at patient level, provider level and system level [40], and may depend on problems with language (such as difficulties to describe an illness or to understand medical advice), bureaucratic aspects (e.g. lack of proper documentation to access the health service), cultural (e.g. different recognition of symptoms or perception of an illness), or organizational (e.g. difficulties to combine own-work-time with opening-time of health provider) [1]. In addition, immigrants may experience specific barriers depending on the local context where they live, which may differently affect the use of health services. In a systematic review on the development of asthma and allergic diseases in relation to international immigration, which involves moving from low income to Western countries, the Authors found that although the prevalence of asthma among immigrants was lower than among non-immigrants, it increased steadily with length of residence [41]. This may suggest that environmental factors are involved in the development of asthma among the usually healthy immigrants, and that barriers accessing primary care [20] may be related to the observed increased rate of hospitalization. Also, the severity of the disease, which was found to be related to the ethnic background [42], may be involved to explain the higher hospitalization for asthma. In the United States, asthma attacks were found higher among African American children [43], and the risk of ED visits were found higher among non-whites and Hispanics, compared to non-Hispanic whites [44]. Another explanation may be that PAH is the consequence of a low ability to manage asthma among immigrants. Racial and ethnic minorities disparities for pediatric asthma care were observed in the United States as consequences of the interactions between the patient family and the healthcare provider [45].

The higher meta-analytic PAH for asthma among female children and adolescent immigrants compared to non-immigrants, but not males, may be explained by factors such as gender differences of severity and prevalence at different ages [29], or of compliance to treatment and awareness. However, we also found higher meta-analytic PAH among immigrant males compared to non-immigrants, although not statistically significant, as well as higher cohort-specific PAH in some cohorts. Alternative explanations, such as some heterogeneity of the results among the cohorts and the low number of events in some cohorts, can be invoked. Therefore, further analysis is necessary to better examine and give interpretation for the gender-specific patterns observed.

We also found that Sub-Saharan Africans had the highest rate of PAH for asthma. Studies conducted in Italy on different health outcomes, such as mortality [23] or perinatal outcomes [46], showed vulnerability of immigrants from Sub-Saharan Africans. A study conducted in the United States found a consistent association of African ancestry with asthma risk [47]. Our interpretation is that socioeconomic, environmental, and genetic factors may all play a role. On the other hand, we found lower risks for immigrants from Central-Eastern countries which may depend on a higher integration of this population in Italy. Further studies are required to explain the variability observed among immigrant groups.

Finally, the results suggest some differences among the cohorts. We think that the differences may be related to unmeasured individual and contextual factors, such as the severity and prevalence of the disease, the composition of the immigrant population and their health literacy, or the accessibility to healthcare services in the local context. However, caution is needed in the interpretation as they may also be related to the very low number of events in some cohorts.

It is posited that some advantages of the present study, particularly in the creation and use of longitudinal studies, are such that they give value and validation to the Authors’ interpretation. The study encompasses 13 years and provides sufficiently enough cases to measure the associations. In addition, the study is conducted using data from six cohorts of all residents in cities located in the North and Central Italy. This data may mirror the effect of local policies on healthcare assistance to different migratory flows, and consequentially, on the amounts of prevailing PAH for asthma. The use of standardized archives of data and shared procedures for data analysis guarantee the internal validity of the results. The inclusion of cities located in the North and Central Italy, where the number of immigrants is high, support the generalizability of results to other Italian cities and to cities of other European countries with universal health systems and immigration from HMPCs.

A limit of our study is that we could not include relevant risk factors which might confound the associations or act as effect modifiers. For example, the time period of residence may have a bidirectional effect on asthma because on the one hand it may increase health literacy, and thus the ability to access primary care, and on the other hand it may increase the risk of the onset of asthma and episodes as a consequence of exposure to environmental factors [12]. Another relevant factor is the socioeconomic status which is negatively associated to an increased risk of asthma [48]. The socioeconomic status might also be an effect modifier of the association between the migrant status and asthma. Indeed, there is evidence that disadvantaged migrants have better health than advantaged non-migrants, an effect known as the epidemiological paradox [49, 50]. Finally, despite the fact that we expected the prevalence of asthma to be different among the groups compared [12, 51], which may affect the risk of hospitalization, the prevalence estimate was not available and therefore it was not possible to consider its effect on the results.

We must mention two relevant aspects concerning misclassification and its possible related information bias. Firstly, we used the birthplace instead of citizenship as a proxy measure of the migrant status for people residing in Rome before the year 2007, as citizenship at that time was not available. Thus, the offspring of immigrants in Italy would be classified as Italians, even though they had a migrant background. In 2007, 11.3% of those born in Italy had foreign parents [52]. Conversely, very few Italians were born in another country (less than 0.3%) [53]; and while it was difficult to estimate the effect of this potential misclassification, we expect that it would not have changed the direction of the association observed, because the analysis was based on a very large population. It may be useful to consider that in another study, where we calculated cause-specific and overall hospitalization rates using data from the same cohorts, we found the results for Rome to be more similar to those of all the cohorts included, after restricting the analysis to the years 2008–2013, compared to 2001–2013. However, the results were not much different from the results over the whole study period 2001–2013 [24].

Secondly, the migrant status may be misclassified for those individuals who acquired Italian citizenship during the study period. However, the percentage of individuals obtaining Italian citizenship was low due to the legislation in force, the ius sanguinis, which requires many years of residence in Italy to be eligible for Italian citizenship. According to the Italian National Institute of Statistics, about 135,814 non-EU immigrants acquired the Italian citizenship in 2017, representing only 3,7% of the non-EU immigrants living in the country [54].