Study population and case definition

The MOA database contains longitudinal demographic and reception data for all asylum seekers who arrived in asylum reception in The Netherlands during the period 2000–2008. The MOA database is derived from the main administrative system of the COA, which is the basis for the provision of housing and coverage of healthcare costs and included a dedicated part for medical purposes. It contains health care data of MOA and family practitioners for the period that the asylum seekers stayed in asylum reception [16]. The database also includes data for all newborns in asylum reception and their mothers.

MOA staff and family physicians used the problem-oriented record (POR) method for general medical recording [22]. In this method, main and chronic health problems are recorded on the problem list along with International Classification of Primary Care (ICPC) code, date of diagnosis, and a short open field description. MOA staff entered problem list data from paper medical records in the electronic medical record system. As of 2009 the organisation of health care for asylum seekers changed and the use of the nationwide database was discontinued.

We first selected all children born in asylum reception between 1 January 2000 and 31 December 2008. Next, we selected their mothers by using the social unit number, which allows linkage between family members. HIV case attribution started with identification of HIV status among women who delivered in reception on the basis of ICPC-code B90. The number of women with code B90 was 82. The consistency between the ICPC-code and the open field description was manually checked. For 7 women their medical record contained ICPC-code B90 but the open field texts referred to a negative HIV test without any further indication of HIV status. These women were not allocated case status. Next, an open field description search was done for the text ‘HIV’ or ‘AIDS’ for women without a B90 ICPC-code. For 6 women the open field description contained a clear description of HIV diagnosis. The total number of women with a record of HIV diagnosis was 81. For 75 of these women the date of HIV diagnosis was before the date of their last delivery in reception; they were allocated case status. For three women the date of HIV diagnosis was within two months after their last delivery; they were also allocated case status. For three other women, the time between delivery and recording of HIV status was longer. For two of these women, their new born was recorded to be HIV positive with a diagnosis date around the same date as their mother (4 and 7 months after delivery). As it is very likely that they were HIV positive at the time of delivery, these women were allocated case status. The third woman was diagnosed four years after delivery and was not allocated case status. For the newborn children, HIV case status allocation followed the same approach as for their mothers. We identified women who gave birth to more than one child during their stay in reception by linking all children born in reception to their mother and by counting the number of newborn children for each mother.

Independent variables

The variable ‘age group at delivery’ refers to the age at the time of the last delivery in reception. The variable ‘family composition’ distinguishes three groups of women: women who arrived in asylum reception with a partner, women who arrived without a partner, and unaccompanied minor asylum seekers (UMAs; children who have been separated from both parents and relatives, and who are not being cared for by an adult) [23]. The variable ‘migration phase at time of conception’ is derived by combining the date of arrival with the date of delivery. It distinguishes between women who got pregnant before arrival, within the first year after arrival or more than one year after arrival in asylum reception.

‘Country of origin’ is the country recorded by the immigration department; in general the country of nationality. Only countries with 30 or more women in the study population were analysed individually. Countries were grouped into ‘Regions of origin’ following the World Bank classification [24].

The variable ‘HIV testing policy at first antenatal consultation’ distinguishes between pregnancies for which the first antenatal consultation took place before or after the implementation of universal antenatal HIV testing in The Netherlands [21,25]. This distinction corresponds to delivery before and after 1 July 2004, as the first consultation preferably takes place before the 13th week of a pregnancy.

Analysis

The HIV prevalence rate was calculated as the number of women that were allocated case status, divided by the total number of women that delivered in reception. The mother-to-child-transmission (MTCT) rate was calculated as the number of newborns with HIV diagnosis divided by the number of women who were HIV positive at the time of delivery. HIV and MTCT prevalence rates are presented as percentages.

Binary logistic regression was used to calculate odds ratios (OR) with 95% confidence intervals (CI) that express how the likelihood of having HIV is associated with age group at delivery, family composition, country of origin, migration phase at time of conception and the HIV testing policy at first antenatal consultation. We used multivariate regression models to control for all other variables included in the analysis. Statistical Software SPSS (IBM Inc., Version 20.0, Somers, NY, USA) was used for all analyses.

Ethics

In line with Dutch legislation, the MOA privacy statement included a statement on the use of data for epidemiological purposes. Unique identifying numbers in the database were encrypted. Because we used only data collected for health care purposes, the medical ethics review committee (METC) of the Academic Medical Center at the University of Amsterdam stated that no METC approval was required (letter W12-276#12.17.0315).

Strengths and limitations

Strengths of this study are the large number of women in the study, the nationwide coverage, the availability and completeness of demographic data for all women who delivered in asylum reception in The Netherlands during the study period and their newborns, and the availability of data under a policy of universal antenatal HIV testing during the second half of the study.

This study also has certain limitations. First of all, despite the large study population, the numbers of HIV cases in subgroups (e.g. by country of origin) are small. Therefore, the prevalence rates estimated for these specific groups lack precision. Furthermore, the absence of data for pregnancies that were not carried to term may have influenced HIV prevalence rates, as HIV may be associated with spontaneous and induced abortions [26,27].

HIV prevalence rates among women attending antenatal care are generally considered to be a proxy for the HIV prevalence in the general population. However, it is unclear whether this also applies to the asylum population [28,29]. Many factors may influence the variations in HIV prevalence rates between pregnant women, other women and men [30]. Women may be particularly vulnerable to HIV infection in times of war and during flight [12,31]. In addition, asylum seekers are relatively young and more often male and single than the general population. Due to lack of detailed knowledge on such factors, it is unknown whether the HIV prevalence rates found among the asylum seekers who gave birth in The Netherlands are higher or lower than in the overall asylum-seeking population.

In addition, the recording of HIV diagnoses on the medical problem list may have been incomplete because results from antenatal screening may not always have been communicated to MOA or the family physician. This may have resulted in underestimation of HIV prevalence rates.

Finally, the change in the HIV testing policy in July 2004 from selective to opting-out screening may have influenced the results. In the first part of the study period, HIV cases may have been missed and as a consequence the HIV prevalence rates may be underestimated. However, the accounts of health professionals suggest that the uptake of HIV testing among pregnant asylum seekers, especially from sub-Sahara Africa, was already very high before 2004. In addition, the multivariate model showed no substantial or significant increase in HIV prevalence rates after the introduction of universal testing compared to the previous period of selective testing.

With respect to the generalisability of our findings to other host countries, it has to be taken into account that the composition of the asylum population may strongly differ between countries. In addition, the HIV prevalence may be influenced by differences between host countries in reception conditions, HIV screening policies and the availability of preventive interventions for asylum seekers [32,33].

Interpretation of results

Comparisons to other countries can be made only with regards to HIV prevalence in the general asylum population, as antenatal surveillance data are not available for asylum seekers in other host countries. The antenatal HIV prevalence in this study (1.6%) is low compared to the prevalence rate found in asylum-seeking men and women in a small study in the UK (3.9%) and comparable to the prevalence rate observed in Italy (1.5%) [13]. The HIV prevalence among sub-Saharan African women who delivered in reception in our study (3.4%) was higher than among the sub-Saharan African asylum-seeking women who were tested for HIV at arrival in the United States (1.4%) [34]. The difference might be related to the composition of the study population (women who gave birth in The Netherlands compared to all women at arrival in the US) or to a different distribution of the countries of origin. However, the US study does not provide data on country of origin.

The HIV prevalence rates that we observed for sub-Saharan women are similar or higher than the rates observed among sub-Saharan African women living in their countries of origin or in refugee camps (S1 Table). Differences may be explained by various factors, e.g. declining HIV prevalence rates in the countries of origin and selection on HIV risk during the migration processes. The higher HIV rates found among Rwandan asylum seekers compared to rates in Rwanda, for example, may reflect the strong decrease in HIV prevalence observed in Rwanda between 1998 and 2003 [35]. In addition, urban-rural differences in HIV prevalence may have played a role: in 2002 the antenatal HIV prevalence was 13% in Kigali whereas it was 3% in rural settings [35].

For regions of origin other than sub-Saharan Africa, the antenatal HIV prevalence rate was at the same level as in the Dutch population. The number of women in the study was too small to obtain precise estimates. However, for the countries of origin from which most of the asylum seekers originated (Afghanistan, Iran, Iraq, and Syria), the HIV prevalence estimates for the population aged 15–49 years in 2006 were 0.2% or less, which is similar to or lower than our estimates for those in The Netherlands (0.2%) [36]. Although the estimated HIV prevalence in the Russian Federation was higher (1.1%), no cases were observed in Russian women who delivered in reception in our study [36].

The higher HIV prevalence rates among UMAs and women who arrived without partner may reflect the particular vulnerability of these girls and women, both in their countries of origin, during the flight and in the host country [12,37]. Our finding that HIV prevalence rates among women who got pregnant after arrival in The Netherlands were as high as among women who were already pregnant at arrival, may reflect the absence of an active HIV screening policy for all men and women. An active HIV screening policy would have made more women aware of their HIV infection, so that they could choose not to have unprotected sex or to postpone becoming pregnant. The high HIV prevalence rates among UMAs who got pregnant within one year after arrival, may reflect the particular vulnerability of these girls shortly after arrival [12,38]. We documented a number of cases showing transmission of HIV to asylum-seeking women and girls during their stay in asylum reception in the Netherlands. This adds to the evidence pointing towards HIV transmission after migration to Europe and underlines the importance of preventive interventions [38].

The decrease in the MTCT rate from 9.8% before the implementation of universal HIV testing to 0% after implementation in 2004, illustrates the potential effectiveness of this screening policy with respect to the transmission from mother to child in the asylum population. This is in line with the effectiveness shown for the general population [21].