The study group of interest in this study consisted of mothers categorised as asylum seekers or refugees who gave birth in Wales. While immigration records are not currently held by the SAIL Databank, the WLGP [23] contains Read codes for particular patient characteristics, including specific codes for physicians to record their patient’s asylum seeker or refugee status: 13ZB (refugee) and 13ZN (asylum seeker) [24]. We used these Read codes to select patients recorded as refugees or asylum seekers by their general practitioner (GP) in Wales. A similar approach has previously been used successfully by researchers to select a cohort of migrants (all types) through GP record codes related to visa status, language, country of birth, and origin [25].

For comparative analyses, we selected UK-born mothers who were not categorised as refugees or asylum seekers using the mother’s country of birth as recorded in the ADBE and GP records.

We included the mothers for all births recorded in the MIDS, which covers the period from April 2014 to present. We excluded mothers aged under 18 as adolescent pregnancies are themselves linked to adverse outcomes [26] and teenage mothers are subject to particular safeguarding practices within maternity care in Wales [27], which may affect their use of services and outcomes. We also excluded mothers lacking a valid ALF and those lacking valid values for refugee status or country of birth.

Non-refugee migrant mothers (i.e. individuals whose recorded country of birth is outside the UK and who were not recorded as refugees or asylum seekers by their GP) were excluded from this analysis as this group may contain refugees or asylum seekers who were not recorded as such by their GP, introducing bias in the analyses.

Following the exclusion of duplicate records and ineligible individuals, we built three nested study cohorts to accommodate different denominators for various types of perinatal indicators:

We conducted statistical analysis using IBM SPSS Statistics software. Adopting an exploratory data analysis approach [28], we first examined the quality of the available data, producing frequency tables and data visualisations to identify patterns and possible outliers.

We examined outliers on a case-by-case basis and removed values determined as invalid in definitions of maternal health indicators by the United Kingdom (UK) Department of Health and Social Care [29]. We also created derived variables (e.g. BMI derived from height and weight) and categorised some values for ease of analysis (e.g. grouped ethnicity categories from detailed categories).

To describe the characteristics of the cohort of mothers categorised as refugees or asylum seekers, we selected a range of variables that are associated with maternal health outcomes in the literature (e.g. maternal age [30], parity [31], obesity [32]) and were available in the linked dataset. We produced cross-tabulations comparing frequencies among the study group and the comparator group of UK-born mothers and presented counts and valid percentages (i.e. excluding missing data). We performed chi-squared tests to identify any statistically significant differences (p < 0.05) in findings. While there is no publicly available data on numbers of mothers seeking sanctuary in Wales, we compared our results to external data where possible to contextualise our findings and assess their validity.

This study was part of a broader mixed-method doctoral research project which included public engagement activities to inform data collection and analysis. As part of engagement initiatives, the lead author consulted with maternity care providers and third-sector organisations providing services to mothers seeking sanctuary, as well as experts by experience, i.e. refugee mothers living in Wales, to better understand the lived experience of seeking sanctuary and guide research activities.

The SAIL Databank provided WLGP records dating from April 2014 until the date of provision (November 2023) for all mothers present in the MIDS. We observed a total of 933 recorded Read codes indicating asylum seeker or refugee status related to 729 individuals, indicating that some individuals had these codes recorded multiple times. Indeed, while the majority of individuals only had one GP record pertaining to refugee status in the dataset, 15.6% of individuals had two records containing refugee or asylum seeker status codes, and 6.2% had three or more. These repeated records could reflect discontinuity of care (i.e. individuals changing GP and disclosing their immigration status to multiple healthcare professionals) or the same GP recording their status multiple times. The majority (81%) of these Read codes indicated asylum seeker status (Read code 13ZN) rather than refugee status (Read code 13ZB).

After excluding ineligible records, the study group included 602 mothers categorised as refugees or asylum seekers who experienced 777 births during the study period, resulting in 791 babies. The comparator group comprised 146 665 non-refugee UK-born mothers, 195 680 birth events, and 198 538 resulting babies (see Figure 1).

We examined some key characteristics of the cohort members to provide a brief demographic profile. The proportion of missing data for each variable ranged from 0% (maternal age) to 28.3% (ethnic group) (see Table 1).

The majority of mothers seeking sanctuary in Wales were born in the Middle East and North Africa (36.7%) or Sub-Saharan Africa (32.4%). Smaller proportions of the group originate from Europe and Central Asia (10.2%), South Asia (9.1%), East Asia and Pacific (9.5%), and Latin America or the Caribbean (2.1%). These results appear broadly consistent with Home Office data concerning the nationalities of migrants who submitted an asylum application in the UK in recent years showing large numbers of applicants from these regions [33]. This coherence with external country of birth data supports the validity of the method used to select the refugee and asylum seeker group within the larger dataset.

While data regarding the ethnicity of the mothers is fairly incomplete (28.3% missing data) and should be interpreted with caution, the two groups appear to have very different profiles (see Table 2). Among mothers seeking sanctuary, 29% are Black (or mixed Black), 24% are Asian (or mixed Asian), and an additional 32% were of another non-White ethnicity. Comparatively, the vast majority (96.9%) of UK-born mothers were of White ethnicity.

Examining all births recorded in the dataset (see Table 3), we observed that asylum seekers and refugees were more likely than UK-born mothers to live in a deprived area as measured by the Welsh Indicator of Multiple Deprivation (WIMD) (56.0% vs 24.7% living in the most deprived quintile). Mothers categorised as refugees or asylum seekers were also more likely to be aged 35 or over at the time of birth (19.9% vs 17.7%), and less likely to be nulliparous (29.7% vs 41.0%), be obese (19.0% vs 29.9%), or smoke at the time of birth (2.2% vs 16.1%). The figures for UK-born mothers appear broadly consistent with national statistics: in 2024, 17.5% of births in Wales were to mothers aged 35 or over, 32.1% to mothers with a BMI over 30, 12.2% to mothers who smoked at the time of birth [34], and 44.7% to nulliparous mothers [35]. These external data, however, refer to all births in Wales, including both UK-born and migrant mothers, only in the year 2024 and thus are not directly comparable to the present study findings.

The use of GP records to select refugees and asylum seekers in datasets has considerable limitations. This method likely does not provide complete coverage of the target population of refugees and asylum seekers as it only identifies those whose immigration status was recorded in the dataset, excluding any refugee or asylum seeker who didn’t disclose their immigration status to their GP or whose GP didn’t record their status through the appropriate Read codes. The data recorded may also be inexact due to human error when inputting the data into the information system. In addition, due to a history of data sharing between the NHS and the Home Office for immigration control [44], vulnerable migrants may be more reticent to access healthcare services or share their immigration status.

Additionally, while the majority of Read codes indicated asylum seeker status rather than refugee, this does not reflect the proportions of asylum seekers and refugees in the UK: in 2023, there were far fewer asylum seekers (approximately 138,000) than refugees (approximately 449,000) [45]. Asylum seekers may be more likely to disclose their immigration status as it is linked to more restrictions (specific accommodation, no right to work, etc.). In addition, GPs may not be aware of the difference between the two statuses or may not update records once an individual obtains refugee status. The two Read codes may not accurately represent differences between asylum seeker and refugee status.