Duration of residence and disease occurrence among refugees and family reunited immigrants: test of the ‘healthy migrant effect’ hypothesis
Danish Research Centre for Migration, Ethnicity and Health, Section of Health Services Research, Department of Public Health, University of Copenhagen, Copenhagen, Denmark
Section of Immigrant Medicine, Department of Infectious Disease, University Hospital Copenhagen, Hvidovre, Denmark
Corresponding Author Marie Norredam, Danish Research Centre for Migration, Ethnicity, and Health, Department of Public Health, University of Copenhagen, Øster Farimagsgade 5, Building 10, 1014 Copenhagen K, Denmark. Tel.: +45 3532 7632; Fax: +45 3532 7629; E-mail: email@example.com
The ‘healthy migrant effect’ (HME) hypothesis postulates that health selection has a positive effect on migrants' health outcomes, especially in the first years after migration. We examined the potential role of the HME by assessing the association between residence duration and disease occurrence.
We performed a historical prospective cohort study. We included migrants who obtained residence permits in Denmark between 1 January 1993 and 31 December 2010 (n = 114 331). Occurrence of severe conditions was identified through linkage to the Danish National Patient Register. Hazard Ratios (HRs) were modelled for disease incidence by residence duration since arrival (0–5 years; 0–10 years; 0–18 years) adjusting for age and sex.
Compared with Danish-born individuals, refugees and family reunited immigrants had lower HRs of stroke and breast cancer within 5 years after arrival; however, HRs increased at longer follow-up. For example, HRs of stroke among refugees increased from 0.77 (95% CI: 0.66; 0.91) to 0.96 (95% CI: 0.88; 1.05). For ischaemic heart disease (IHD) and diabetes, refugees and family reunited migrants had higher HRs within 5 years after arrival, and most HRs had increased by end of follow-up. For example, HRs of IHD among family reunited migrants increased from 1.29 (95% CI: 1.17; 1.42) to 1.43 (95% CI: 1.39; 1.52). In contrast, HRs for TB and HIV/AIDS showed a consistent decrease over time.
Our analyses of the effect of duration of residence on disease occurrence among migrants imply that, when explaining migrants' advantageous health outcomes, the ruling theory of the HME should be used with caution, and other explanatory models should be included.
L'hypothèse de “l'effet du migrant en bonne santé” (EMS) postule que la sélection de la santé a un effet positif sur les résultats de santé des immigrants, en particulier dans les premières années après la migration. Nous avons examiné le rôle potentiel de l’EMS en évaluant l'association entre la durée de séjour et la survenue de la maladie.
Nous avons réalisé une étude prospective de cohorte historique. Nous avons inclus les immigrants qui ont obtenu un permis de séjour au Danemark entre le 01 janvier 1993 et le 31 décembre 2010 (n = 114 331). La survenue de conditions sévères a été identifiée grâce aux liens avec le registre national danois des patients. Les rapports de risque (HR) ont été modélisés pour l'incidence de la maladie selon la durée de résidence depuis l'arrivée (0 à 5 ans, 0 à 10 ans et 0 à 18 ans) et ajustés pour l’âge et le sexe.
Comparativement aux personnes nées au Danemark, les réfugiés et les immigrants de regroupement familial avaient un HR inférieur d’AVC et de cancer du sein dans les 5 ans après l'arrivée. Cependant, le HR augmentait sur un suivi plus long. Par exemple, le HR pour l’AVC chez les réfugiés passait de 0,77 (IC95%: 0,66–0,91) à 0,96 (IC95%: 0,88–1,05). Pour la maladie cardiaque ischémique (MCI) et le diabète, les réfugiés et les immigrants de regroupement familial avaient un HR plus élevé dans les 5 ans après l'arrivée et la plupart des HR augmentaient à la fin du suivi. Par exemple, le HR pour la MCI chez les immigrants de regroupement familial a augmenté de 1,29 (IC95%: 1,17–1,42) à 1,43 (IC95%: 1,39–1,52). En revanche, les HR pour la tuberculose et le VIH/SIDA ont montré une diminution constante au cours du temps.
Nos analyses de l'effet de la durée de résidence sur la survenue de la maladie chez les immigrants impliquent que, pour expliquer les résultats de santé avantageux pour les immigrants, la théorie courante de l’EMS doit être utilisée avec prudence et d'autres modèles explicatifs devraient être inclus.
La hipótesis del “efecto del inmigrante sano” (EIS) postula que la selección de salud tiene un efecto positivo sobre los resultados de salud de los inmigrantes, especialmente en los primeros años tras la migración. Hemos examinado el papel potencial de la EIS evaluando la asociación entre el tiempo de duración de la residencia y la aparición de enfermedad.
Hemos realizado un estudio de cohortes histórico y prospectivo. Hemos incluido inmigrantes que obtuvieron permisos de residencia en Dinamarca entre 01.01.1993 y 31.12.2010 (n = 114 331). Se identificó la aparición de condiciones severas a través del Registro Nacional de Pacientes de Dinamarca. Se modelaron los cocientes de riesgos instantáneos (HRs) para la incidencia de enfermedad según la duración de la residencia, desde la llegada 0–5 años; 0–10 años; 0–18 años) ajustando para edad y sexo.
Comparados con individuos nacidos en Dinamarca, los refugiados y los inmigrantes de familias reunificadas tenían un menor HR de accidente cerebrovascular isquémico (ACV) y cáncer de mama dentro de los 5 años posteriores a su llegada; sin embargo, los HRs aumentaban a mayor seguimiento. Por ejemplo, los HRs para ACV entre refugiados aumentaba del 0.77 (IC 95%:0.66; 0.91) al 0.96 (IC 95%:0.88; 1.05). Para la enfermedad cardiaca isquémica (ECI) y la diabetes, los refugiados y los inmigrantes de familia reunificadas tenían unos HRs mayores dentro de los 5 años posteriores a la llegada, y la mayoría de los HRs habían aumentado al final del seguimiento. Por ejemplo, HRs de ECI los inmigrantes de familias reunificadas aumentó del 1.29 (IC 95%:1.17; 1.42) al 1.43 (IC95%:1.39; 1.52). En con traste, los HRs para TB e VIH/SIDA mostraban una disminución consistente a lo largo del tiempo.
Nuestro análisis sobre el efecto de la duración de la residencia sobre la aparición de enfermedad entre inmigrantes sugiere que, cuando se explican los resultados ventajosos para los inmigrantes, la teoría de descarte del EIS debería utilizarse con precaución, y deberían incluirse otros modelos explicativos.
Previous studies showed that incidence and mortality rates of migrants as compared to local-born populations greatly vary according to the disease studied. (Razum et al. 1998; Lanting et al. 2008; Agyemang et al. 2009; Rechel 2011; Rechel et al. 2013), but that levels of all-cause mortality are relatively low in many migrant populations (Bos et al. 2005; Gadd et al. 2006; Ronellenfitsch et al. 2006; Norredam et al. 2012b; Tarnutzer & Bopp 2012). The lower overall mortality rates have been especially attributed to two types of health-selective migration.
First, the ‘salmon bias’ or ‘remigration bias’ hypothesis implies that severely ill migrants or those who expect to die in the near future tend to return to their countries of origin in order to be cared for by their relatives in a familiar environment (Razum & Twardella 2002; Spallek et al. 2011). Consequently, the mortality of migrants is underestimated to the extent that outmigration is selective to ill health. Although ‘salmon bias’ is often mentioned as a potential factor underlying the lower overall mortality among migrants compared with local-born individuals, the hypothesis is facing increasing challenge (Agyemang et al. 2012a).
Second, the ‘healthy migrant effect’ hypothesis (HME) implies that the health advantages of migrant populations are explained by the selection of the healthiest individuals into migration (Spallek et al. 2011; Agyemang et al. 2012a). The hypothesis originally compared migrants to those remaining in the emigration country, but it is now often applied as an explanation for favourable health outcomes among migrants relative to local-born individuals in host countries. The hypothesis has been difficult to investigate due to lack of data on the health of populations in the countries of origin. As an alternative, a few studies in immigration countries have tried to examine the HME. The few studies originated from the USA and only provided it with weak support (Abraido-Lanza et al. 1999; Rubalcava et al. 2008; Bostean 2013; Lau et al. 2013). Recently, the HME has been challenged by Agyemang et al. (2012a) because it seeks to explain health outcomes for a number of very different diseases with a number of different risk factors and because it tends to be applied to all migrants without acknowledging that people migrate for very different reasons.
To gain insight into the role of HME, we assessed the effect of duration of residence on disease outcomes among recently arrived migrant groups compared with local-born Danes. If the theory holds, one would expect non-communicable diseases to be lower among migrants than among locally born individuals. This advantage would be apparent in the years shortly following arrival but would decline over time. Opposite patterns may hold for communicable diseases because the HME is generally thought to apply to non-communicable diseases in particular.
We could test these hypotheses thanks to a historical prospective cohort of migrants in Denmark. The cohort includes all refugees and family reunited immigrants who obtained residence permits in Denmark between 1 January 1993 and 31 December 2010. We could classify migrants by the reason of migration (i.e. ‘migrant status'), and we could link them to nationwide data on disease outcomes. The prospective design allowed us to assess disease incidence in relation to the migrants' duration of residence.
The study cohort was established in collaboration with the Danish Immigration Service. This authority provided data on migrants by age, sex and nationality upon arrival and migrant status (refugee vs. family reunification). The study included all migrants aged at least 18 years who obtained right of residency as refugees or through family reunification in Denmark between 1 January 1993 to 31 December 2010. These groups were chosen as they obtained long term or permanent residency as opposed to students or labour migrants. A total of 152 749 individuals were identified. First, we excluded individuals with a missing personal identification number. These were mainly family reunited individuals who obtained permission while in the emigration country but did not end up not emigrating. Consequently, such individuals never received the personal identification number needed for residence in Denmark. Second, we excluded individuals whose personal identification numbers appeared twice. These duplicated individuals were excluded, as well as individuals with invalid age, negative follow-up time or unknown nationality. This resulted in a final cohort of 114 331 migrants, including 43 992 refugees and 70 339 family reunited immigrants. Previous publications are based on an earlier version of the cohort using the same selection process but a smaller population (Norredam et al. 2010). The population was divided into five groups based on migrant status: (i) asylum seekers, (ii) quota refugees, (iii) family reunifications to refugees, (iv) family reunifications to family reunited immigrants and (v) family reunifications to Danish/Nordic citizens. The latter imply reunification with someone who has obtained Danish citizenship or citizenship in one of the other Nordic countries, either by birth or by naturalisation. Characteristics of the study cohort are shown in Table 1.
Table 1. Characteristics of the study cohort (n = 114 331), distributed by region of origin and migrant status
Total at study start
Median age at study end
Emigrations to any country
Population at study end
East Europe, Baltics, and Central Asia
84.4 (17 677)
Western Asia and North Africa, including Turkey and Iraq
86.9 (25 708)
75.9 (10 325)
South Eastern and Eastern Asia
90.5 (12 761)
87.1 (12 835)
Western Europe, Americas, Australia, and New Zealand
85.2 (37 448)
84.5 (32 936)
Family reunified immigrants
82.1 (57 727)
Family reunified to refugee
83.3 (10 488)
Family reunified to immigrant
Family reunified to Danish/Nordic citizen
80.7 (39 379)
Personal identification numbers of the study cohort were cross-linked to the Danish National Patient Register, which contains national data on discharge diagnoses and dates on all inpatient visits since 1977 and all outpatient and emergency visits since 1995. For inpatient visits, we used discharge diagnoses and for outpatient visits, we used the ongoing diagnoses, which are updated regularly. All diagnoses were classified by the International Classification of Diseases, tenth revision (ICD-10). For analyses, we distinguished between the following diagnostic categories: Diabetes (E10-E14), Stroke (I60-I69), Ischaemic heart disease (I20-I25), Tuberculosis (A15-19, B90), HIV/AIDS (B20-24), Breast cancer (C50) and Lung cancer (C34). These diseases were selected in order to include both communicable and non-communicable diseases as well as to include diseases that are known to occur either more or less frequently among migrants relative to local-born individuals.
In our study, we define ‘duration of residence’ somewhat alternatively as the duration from the date of commencement of right of residency until the first of the following events: (i) time of diagnosis for one of the selected diseases, (ii) date of first emigration, (iii) end of study (31 December 2010 or d) date of death. Consequently, we use ‘duration of residence’ to quantify the number of years a person was at risk of getting a disease. This duration varies according the disease studied. Data on dates of immigrations and emigrations were obtained from the Danish Civil Registration System. In total, during follow-up, we identified 23 875 contacts with one of the selected diagnoses. These contacts occurred among 12 082 individuals.
We estimated the hazard ratios (HRs) and 95% confidence intervals of disease outcomes among all subgroups of refugees and family reunited immigrants. The HRs were estimated by duration of residence using three time horizons: (1) 0–5 years, (2) 0–10 years and (3) 0–18 years since arrival in Denmark. This categorisation enabled an interpretation of each of the (three) results as the results that would be obtained if one looked at 5-year morbidity, 10-year morbidity and morbidity during the entire follow-up. HRs were estimated using the Cox proportional hazards models stratifying for migrant status and adjusting for age and sex. Danish-born individuals formed the comparison group. We calculated time-dependent relative cumulative hazard functions that compared nonparametric cumulative hazard function for migrants to that for Danish-born individuals. This was performed for men and women separately. We prepared plots to illustrate the observed difference between migrant and local-born individuals in disease development over time. We allowed individuals in the cohort to have one or more first-time contacts. Consequently, if an individual was first diagnosed with TB and subsequently with diabetes, both diagnoses were considered first-time diagnoses; hence, there were more contacts than individuals. Analyses were performed using SAS 9.2 statistical package (SAS Institute, Cary, NC, USA).
The study was approved by the Danish Data Protection Agency. Further ethical approval regarding registry-based research is not required in Denmark. The data set was made available and analysed in an anonymous form by remote online access to the data set stored at Statistics Denmark.
Table 1 shows characteristics of the study cohort distributed by nationality upon arrival and migrant status. Tables 2 and 3 show adjusted HRs of specific disease outcomes among refugees and family reunited immigrants at different moments during follow-up.
Table 2. Hazard ratios (HRs) of disease outcomes among all refugees, quota refugees and asylum seekers during 0–5, 0–10 and 0–18 years' follow-up, respectively. HRs are adjusted for age and sex. Danish-born individuals form the reference group
Table 3. Sex-specific hazard ratio of disease outcomes among all family reunified immigrants and immigrant subgroups during 0–5, 0–10 and 0–18 years' follow-up, respectively. HRs are adjusted for adjusted for age and sex. Ethnic Danes form the reference group
Refugees and family reunited immigrants had lower HRs of stroke 5 years or less after arrival than Danish-born individuals; however, HRs had increased by end of follow-up. For example, HRs of stroke among refugees increased from 0.77 (95% CI: 0.66; 0.91) to 0.96 (95% CI: 0.88; 1.05). For ischaemic heart disease (IHD), refugees and family reunited immigrants had higher HRs within 5 years of arrival than Danish-born individuals. For 0–18 years, HRs had increased for family reunited immigrants, whereas they had decreased for refugees. For example, HRs of IHD among family reunited immigrants increased from 1.29 (95% CI: 1.17; 1.42) to 1.43 (95% CI: 1.39; 1.52). Refugees and family reunited immigrants had higher HRs of diabetes within 5 years of arrival than Danish-born individuals, yet HRs had increased even more at 0–18 years' follow-up for both migrant groups. Figure 1 supplements these results by illustrating the relative cumulative hazard of diabetes over time as residency obtained for migrant men and women stratified by sex and migrant status. All subgroups showed similar tendencies with rising incidence for diabetes with increasing duration of residence.
Refugees and immigrants had lower HRs of breast cancer 5 years or less after arrival compared with Danish-born individuals; however, HRs had increased by end of follow-up. For example, HRs of breast cancer among family reunited immigrants increased from 0.44 (95% CI: 0.34; 0.56) to 0.69 (95% CI: 0.60; 0.79). Figure 2 shows the relative cumulative hazard of breast cancer over time since residency was obtained stratified by migrant status. All subgroups showed similar tendencies towards rising incidence for breast cancer with increasing duration of residence. Results for lung cancer showed a different picture than did breast cancer. Results were stratified by sex due to significantly different tendencies. Both refugee and immigrant men had higher HRs of lung cancer within 5 years of arrival than Danish-born individuals, and HRs decreased during follow-up. For example, HRs of lung cancer among refugee men decreased from 1.69 (95% CI: 1.27; 2.25) to 1.51 (95% CI: 1.27; 1.80). In contrast, HRs of lung cancer among refugee women decreased from 0.89 (95% CI: 0.59; 1.33) to 0.74 (95% CI: 0.57; 0.95).
In total, refugees had a significantly higher HR of tuberculosis (TB) at all points in time, but differences decreased over time though they remained significantly higher than among Danish-born individuals. Refugees and immigrants had higher HRs of TB within 5 years of arrival than Danish-born individuals; however, HRs had decreased by end of follow-up. For example, HRs of TB among family reunited immigrants decreased from 39.50 (95% CI: 31.60; 43.37) to 25.76 (95% CI: 21.93; 30.25). Figure 3 shows the relative cumulative hazard of TB over time as residency for migrant men and women stratified by sex and migrant status. Regarding HIV/AIDS, all subgroups of refugees and immigrants had higher HRs of HIV/AIDS 5 years or less after arrival compared with Danish-born individuals; however, HRs had decreased by end of follow-up. For example, HRs of HIV/AIDS among refugees decreased from 6.13 (95% CI: 4.89; 7.67) to 4.22 (95% CI: 3.55; 5.02).
Insight into the healthy migrant effect is important because it is often invoked to explain the lower mortality of migrants. Using unique data set on a cohort of refugees and family reunited migrants and Danish-born individuals, this study shows that migrants had significantly or non-significantly higher HRs for most disease outcomes within the first 5 years of arrival relative to Danish-born individuals. Moreover, the occurrence of stroke, IHD, diabetes and breast cancer increased with increasing duration of residence, while the occurrence of HIV/AIDS, TB and lung cancer decreased.
The effect of the HME
The HME is invoked as a selection mechanism that is hypothesised to contribute to migrants' health profile in immigration countries and especially to the lower incidence, prevalence and mortality from several non-communicable diseases (Spallek et al. 2011). Our results generally do not support the HME theory for the following reasons.
First, compared with Danish-born individuals, both refugees and family reunited immigrants had a higher disease burden already within 5 years of arrival for most diseases except stroke and cancer. As other studies have shown, this is unsurprising for infectious diseases such as TB and HIV/AIDS because many migrants come from countries with a high incidence of these conditions (Lillebaek et al. 2002; Williamson et al. 2009). However, the high occurrence of IHD and diabetes appears surprising in view of the healthy migrant hypothesis. It implies the need to move beyond the governing theory of the HME when explaining migrants' advantageous health outcomes. Indeed, other explanatory models are increasingly being recognised, including the role of past exposures (Spallek et al. 2011; Agyemang et al. 2012a). Immediate health disadvantages include the effect of trauma and the migration process as such on mental health (Norredam et al. 2009), the effect of exposures to infectious agents on later infectious disease occurrence (Norredam et al. 2012a) in immigration countries, and the interplay between mental health disorders such as PTSD (Post-Traumatic Stress Disorder) with non-communicable diseases (Kinzie et al. 2008; Agyemang et al. 2012b).
Second, the HME is hypothesised to wear off over time. Our results showed that the risk indeed increased over time for all disease outcomes apart from TB, HIV/AIDS and lung cancer. However, even for TB and HIV/AIDS, disease incidence remained high and significantly increased compared with Danish-born individuals. These results imply that the wearing off component of the HME holds better than the component predicting low mortality rates at onset.
Third, the HME has been challenged because it is often applied to all migrants without acknowledging that people migrate for a number of very different reasons. A uniform result for all migrant groups is thus assumed without acknowledging that not only healthy individuals migrate, but that people also migrate because of political persecution, famine, poverty, family reunification and maybe even poor health. We were able to analyse by seven different subgroups of migrants. We found some general tendencies across subgroups of refugees and family reunited immigrants. For example, the risk of HIV/AIDS was increased in all migrant groups. Simultaneously, however, we observed great variation in the risk of the different subgroups compared with that among Danish-born individuals. For example, refugees had a five times higher HR of HIV/AIDS after the entire follow-up period than Danish-born individuals, whereas that among family reunited immigrants was 12 times higher.
Methodological strengths and limitations
The main strength of the study is its unique ability to identify and follow a large cohort of migrants based on information from the Danish Immigration Service, which allowed us to divide by specific migrant status and region of origin. We furthermore linked to the nationwide National Danish Patient Registry, obtaining data on inpatient contacts for a number of diagnoses (Lynge et al. 2011).
Several factors may have influenced our results. First, although the study was based on a relatively large cohort of migrants, absolute numbers became somewhat small when specific diseases were investigated and when we stratified by migrant status. Consequently, we were unable to go beyond the major diagnostic categories that we analysed and thus cannot generalise further from our findings. This is a common limitation to register-based studies on ethnicity and health (Norredam et al. 2011).
Second, we did not control for socio-economic position (SEP). In Denmark, education is most often used as a proxy for SEP income, but data on education for migrants are considered inconsistent and of low validity as education obtained outside of Denmark is not registered. Instead, personal income may be used in settled migrant groups. However, we chose not to adjust for income as the personal income among newly arrived migrants initially while resettling in the immigration country is not representative of their SEP and might therefore skew the analyses.
Finally, we are aware that we were limited in including only refugees and family reunited immigrants in this explorative study as our historic retrospective cohort only included individuals who received permanent residence permission or permission for up to 7 years (which refugees receive before they receive permanent residence permission). Indeed, these groups make up the largest non-Western populations in Denmark and many other European countries of individuals receiving permanent residence permission. Consequently, the cohort includes neither ‘old’ labour immigrants from the 1960s and 1970s nor ‘new’ labour immigrants from the 1990s and onwards; albeit these are both groups where the HME might have been even more influential. It is possible to follow these groups of labour migrants in the Danish registers; albeit only with health data from 1979 and onwards as the National Patient Registry was established then. This would make analyses from arrival and onwards of ‘old’ labour migrants more limited but still this group would be highly relevant to explore in relation to the HME in a future study. Not including these labour migrant groups naturally limits the external validity of the study.
Our analyses of the effect of duration of residence on disease occurrence among recently arrived refugees and family reunited immigrants imply that, when explaining migrants' advantageous health outcomes, the ruling theory of the HME should be used with caution, and other explanatory models should be included. On a policy level, our results imply that preventative and diagnostic health efforts in the immigration country should target relevant migrant groups from upon arrival and include non-communicable as well as communicable diseases.