We carried out a retrospective longitudinal study using primary care electronic health records (EHRs) from all general practices (GP) in seven geographically contiguous areas in NEL: Barking & Dagenham, City & Hackney, Havering, Newham, Redbridge, Tower Hamlets, and Waltham Forest. The study protocol can be found in supplementary file 1 and the Reporting of studies Conducted using Observational Routinely-collected health Data (RECORD) Statement in supplementary file 2.

We defined the cohort as children registered with a NEL GP on their second birthday between 1^st January 2016 and 30^th October 2021. These children would have been eligible to receive their MMR1 vaccination between 12 and 24 months of age between 1^st January 2015 and 30^th October 2021. Children could enter the cohort at any point between birth and their second birthday.

Pseudonymised data were provided from the NEL Discovery Data Service which receives daily primary care EHR data from all GPs in NEL. Demographic and clinical data were extracted for children ever registered with a NEL GP and included children who may have died or left the area. Data were extracted on 23^rd November 2021 and included all clinical events up to 1^st November 2021. All data were extracted and managed according to UK National Health Service (NHS) information governance requirements [18].

Every addressable location in Great Britain is assigned a Unique Property Reference Number (UPRN). UPRNs identify a place of residence at a granular level, identifying individual properties, for example houses or flats within a block or building shell. UPRNs are allocated to GP-recorded addresses using the validated AddreSS MatchInG to Unique Property Reference Numbers (ASSIGN) algorithm [19], and pseudonymised into Residential Anonymised Linkage Fields (RALFs) within the Discovery Data Service, using a study-specific encryption key. We have previously described the process for identifying household members at a point in time [20, 21].

We identified 167,790 children born between 1^st January 2014 and 30^th October 2019 and with a single ‘regular’ (as opposed to temporary) NEL GP registration on their second birthday, with at least one RALF. We excluded 1,744 children with a poor-quality RALF match, and 4,315 with a RALF associated with a non-residential building. We further excluded 2 with unknown sex or implausible address dates. We retained only those living with at least one adult aged 18-100 years (figure 1). Our final study sample comprised 150,949 children (90.0%). Characteristics of those excluded from the study sample can be found in supplementary file 3 table 1.

We extracted sociodemographic, household and geographic data for each child, together with – for each child – all clinical events relating to MMR1 procedures. We derived a proxy date of birth by assigning day of birth to the first day of the week using the calendar week and year of birth.

We defined timely MMR1 vaccination as receipt of the MMR1 vaccination between 12 and 24 months of age. This is consistent with the Cover of Vaccination Evaluated Rapidly (COVER) measures in place during the study period [3]. We identified MMR1 vaccination using a pre-specified list of clinical Systematized Medical Nomenclature for Medicine (SNOMED) terms (supplementary file 3 table 2). In accordance with the UK Health Security Agency guidelines [2], MMR1 vaccinations given before 12 months of age were ignored, and in these analyses, follow-up was censored at 24 months, however some children may have received their vaccination after this age.

The main exposure was residential mobility during early childhood, defined as the number of GP-recorded addresses in the child’s EHR, as identified by unique residential RALFs between their earliest GP registration and either their MMR1 vaccination date or second birthday. We assumed that children with periods of time without a NEL GP registration were living elsewhere and added one to the count of RALFs for each period of greater than 30 days when a child was not registered with a NEL GP. Similarly, we added one to the count of RALFs for children who were aged greater than 60 days old when they first registered with a NEL GP. Residential mobility (number of RALFs) was categorised into one, two, three or more (see figure 2). 2,345 (1.6%) of 150,949 children had at least one unregistered period lasting more than 30 days (2,294 had one unregistered period, 51 had two or more). Unregistered periods of more than 30 days ranged from 31 to 657 days, with a mean length of 113.5 days (standard deviation: 100.7) and median length of 74 days (interquartile range: 44,146). We assume that 24,943 (16.5%) of 150,949 children moved into NEL during the period of follow-up – they were aged more than 60 days old at the time of their first NEL GP registration. The age at first GP registration ranged from 0 to 778 days, with a mean age of 226.4 days (standard deviation: 171.9) and a median age of 169 days (interquartile range: 78,336).

We included individual-, household-, and area-level characteri-stics as covariates. Individual-level characteristics included the child’s sex and ethnic background, classified using the NHS 16+1 categorisation: White British, White Irish, Other White, Chinese, White & Asian, White & Black African, White & Black Caribbean, Other Asian, Other Mixed, Other, Bangladeshi, Indian, Pakistani, Black African, Black Caribbean, and Other Black [22].

Household-level characteristics included the number of children living in the household and household composition. Both were defined on the date of the child’s second birthday. The number of children in the household was defined as the total number of people (including the cohort child) aged under 18 years sharing the same RALF as the cohort child on the cohort child’s second birthday. We categorised this into: one, two, three, four or more. We categorised household composition using a modified Harper and Mayhew method [23] by counting the number of household members in three age brackets: 0-18 years (children), 19-64 years (working age adults) and 65 or older (older adults) and grouping into: working-aged adults with children; a single working age adult with children; other.

Area-level characteristics included local authority of the child’s GP and area-level deprivation. We used the 2019 Income Deprivation Affecting Children Index (IDACI) at decile level [24], linked using the 2011 Lower layer Super Output Area (LSOA) of the child’s home address. An LSOA has an average population of 1,500 people or 650 households. The IDACI score measures the proportion of children under 16 in low-income households for an area. IDACI deciles were collapsed into quintiles from most to least deprived.

We explored variation in the proportion of children receiving a MMR1 vaccination by residential mobility and other covariates, and described the differences in the proportion of children receiving the MMR1 vaccination by 24 months of age. We also explored variation in residential mobility by the covariates. We conducted binary logistic regression to estimate the crude and adjusted odds (odds ratio [OR] and 95% confidence interval [CI]) of a timely MMR1 vaccination by residential mobility, using a stepwise approach to adjust for individual-, household-, and area-level covariates. We present OR (95% CI) of a timely MMR1 vaccination in four models: 1) Univariable; 2) Adjustment for individual-level factors (sex and ethnic background); 3) Adjustment for individual- and household-level factors (number of children in the household and household composition); 4) Adjustment for individual-, household-, and area-level factors (IDACI quintile and local authority).

All analyses were conducted using Stata (MP/17.0).

We repeated all analyses without adding one more to the count of addresses for children who were aged greater than 60 days old when they first registered with a NEL GP (figure 2 example 9) or for each period of greater than 30 days when a child was not registered with a NEL GP (figure 2 examples 6 and 7). We also repeated all analyses on a sub-sample of 109,711 children born between 1^st January 2014 and 20_th March 2018 and therefore eligible to receive their MMR1 vaccination between 12 and 24 months of age between 1^st January 2015 and 20^th March 2020. The receipt of MMR1 vaccination among these children is expected to be unaffected by the 2019 Coronavirus (COVID-19) pandemic. Results of both sensitivity analyses are presented in the supplementary material.

Patients and the public were not involved in the design of this research.

In sensitivity analyses not adjusting for periods without a GP registration, 94.4% of children had only one GP-recorded address before their MMR1 vaccination or second birthday, 5.4% had two, and 0.2% three or more (supplementary file 3 table 5). The proportion of children receiving their MMR1 vaccination by 24 months of age varied by residential mobility (supplementary file 3 table 6). Children who had two, or three or more GP-recorded addresses were less likely to have received their MMR1 vaccination by their second birthday, compared with those with one (81.3%; 95% CI: 80.5,82.1; 74.5%; 68.9,79.4; 85.0%; 84.8,85.2; respectively). After adjustment for individual-, household-, and area-level covariates, children with two, or three or GP-recorded addresses were less likely than those with only one to receive their MMR1 vaccination by 24 months of age more (OR: 0.74; 95% CI: 0.69,0.79; 0.45; 0.33,0.62) (figure 4 and supplementary file 3 table 7).

Results were similar in sensitivity analyses restricted to children eligible for their MMR1 vaccination before the COVID-19 pandemic (supplementary file 3 tables 8-10). After adjustment for individual-, household-, and area-level covariates, children who had two, or three or more GP-recorded addresses were less likely than those with only one to receive their MMR1 vaccination by 24 months of age (OR: 0.45; 95% CI: 0.43,0.47; 0.33; 0.29,0.38) (figure 4 and supplementary file 3 table 10).

To our knowledge, this is the first study in an urban, mobile, and ethnically diverse population to examine MMR1 receipt by residential mobility. After adjustment for covariates known to be associated with MMR1 vaccination and residential mobility, we found a reduced likelihood of MMR1 vaccination among children with increased residential mobility. Compared with children with just one address, children with two GP-recorded addresses were 54% less likely, and those with three or more, 68% less likely, to receive their MMR1 vaccination by 24 months of age.

We used routine primary care EHRs available for an entire population of children registered with all NHS general practices in one region of London. These results highlight inequalities in timely vaccination coverage, leaving children who experience residential mobility more susceptible to measles infection.

Coding of routine childhood vaccinations by primary care teams in NEL is facilitated by data entry templates with standardised coding enabling high quality recording of childhood vaccinations at the point of care. We developed and validated a SNOMED code specification which we mapped to Read code [25] and local system codes to ensure a maximally sensitive search strategy for identifying MMR1 vaccinations. We used robust statistical methods, including retrospective longitudinal analyses, to investigate the association between residential mobility and receipt of MMR1 vaccination. We confirmed our results were not impacted by the COVID-19 pandemic in our sensitivity analyses.

We used a robust methodology to identify household members on the second birthday. The ASSIGN algorithm has been shown to match 98.6% of primary care patient addresses to UPRNs [19]. We observed a marginally lower proportion of children receiving the MMR1 vaccination between 12 and 24 months of age among those excluded from the study sample, compared with those included. The explanation for this remains unclear. We excluded children living in non-residential households. It is possible that we included household members who no longer live at their registered address due to a time lag between a patients’ GP registrations, and a period of time where a patient has moved on from an area but remains registered with a GP. Hence, we may have overestimated the true number of children in the household. On the other hand, it is also possible we excluded children where it appeared they did not live with any other adults/household members in cases where other household members were not registered with a GP at the time of the child’s second birthday.

Our cohort definition included children registered with a NEL GP on their second birthday. By definition, this excluded children who may have been registered with and vaccinated by a NEL GP between 12 and 24 months of age, but who subsequently moved out of NEL and were not captured in our study sample, consequently excluding children experiencing residential mobility from our sample. Whilst not perfect, this approach enabled comparison of the characteristics of children registered with a NEL GP on their second birthday who did or did not receive MMR1 between 12 and 24 months of age. Additionally, we excluded children with multiple GP registrations, and those not living with any adults. It is possible these children have a different experience of MMR1 vaccination, compared with children included in our study sample.

We made some assumptions about residential mobility during periods of time when children were not registered with a NEL GP. We added one to the count of addresses for children aged greater than 60 days old when their first GP registration started, assuming these children had moved into NEL from elsewhere. Similarly, we added one to the count of addresses for each period of more than 30 days when a child was not registered with a NEL GP between their first registration and second birthday/date of MMR1 vaccination. We did not have information about the child’s place(s) of residence during these time periods and have assumed they lived elsewhere. We conducted sensitivity analyses without this adjustment, as we recognise it is possible unregistered periods are a result of administrative delays in registration. Whilst the magnitude of the effect of residential mobility was not as great, the direction of the effect remained the same, where children with two or more GP-recorded addresses were less likely to receive their MMR1 vaccination by 24 months of age.

EHRs are a source of routinely-collected administrative data, and do not provide information about the motivation for moving home. Linkage to other data sources, such as the Census, could provide additional information, for example about property tenure, which could aid interpretation of residential mobility. This was not in the scope of this study.

Whilst our study has focused on receipt of the MMR1 vaccination by 24 months of age, it is important to recognise that some children may receive a delayed vaccination after 24 months of age, and a second dose, now due at approximately 18 months of age, is essential for full protection [2]. Additional research investigating receipt of delayed vaccinations and the second MMR vaccination by age five years would further our understanding and improve identification of children with increased measles susceptibility.

Our finding supports research in Canada which found the likelihood of being incompletely immunised by age seven was greater in children who had moved residence two times or more, compared to those who had moved one time or less. It is hypothesised that increased residential mobility may inhibit the development of long-term relationships with GPs and other healthcare workers, as well as bringing additional logistical barriers to vaccination which may arise due the chaos and competing priorities of residential instability [11].

Our estimates of residential mobility are considerably lower than those estimated from a nationally representative sample of children participating in the Millennium Cohort study [12, 26], although this difference is likely explained by the use of a longer follow-up period (up to age five years) in these studies.

Our findings are not consistent with evidence from the same national UK cohort, which found no difference in parentally-reported MMR1 immunisation status at three years of age among children who moved between birth and three years of age compared to those who did not move [11]. Similarly, in Wales it was reported that moving home frequently did not increase the odds of not being immunised when compared to not moving home [10]. As well as looking at coverage, they also reported that residential mobility did not impact on the timeliness of the receipt of vaccinations by 13 months of age. It is possible that we identify a different association between residential mobility and MMR1 vaccination due to differences in our study populations and definitions of timeliness. In NEL, the population is highly mobile, ethnically diverse and disadvantaged, with historically low uptake of MMR1 vaccination.

More broadly, our findings support other studies exploring the relationship between residential mobility and adverse health outcomes. For example, a study in New Zealand reported higher risk of potentially avoidable hospitalisations among children with higher residential mobility in the first 24 months of life [7], and in Wales, researchers found increased incidence of hospitalisations for infectious diseases and asthma among those with two or more moves [8].

Our research has contributed to a growing body of evidence around the home environment and subsequent health behaviours and outcomes. Managing the health care needs of children in mobile population groups is a challenge, especially when moving between different healthcare providers. It is possible that residential mobility leads to a lack of continuity in primary care, and is a potentially contributing factor to incomplete immunisation status [27]. However, not all address changes will result in a change of GP and we did not look at change in registered GP with residential mobility so cannot make this inference. It is possible that residential mobility, and the consequential administrative process of changing GP, may result in delayed vaccination. We were unable to explore change in registered GP with residential mobility.

As well as the direct impact of residential mobility, it is also likely that residential mobility is associated with unobserved characteristics which lead to lower vaccination rates. Residential mobility may arise as a consequence of opposing socioeconomic circumstances. For some, residential mobility may indicate financial insecurity as families navigate short-term rental accommodation and perhaps more precarious housing, yet for others, residential mobility may signify improved financial security as families move towards home ownership [26]. It remains that residential mobility is driven by a range of factors including changes in employment, partnerships, family size and composition, and housing tenure [28]. Whatever the motivation for moving home, we estimate a population attributable risk percentage of 18.4% (supplementary file 3 table 11). This means that if no children experienced a change in address, 18.4% more children would be expected to receive a MMR1 vaccination by 24 months of age, assuming a causal association.

Measles vaccination may currently receive less priority in a health care system facing multiple challenges and clinical priorities [29]. The need for targeted public health interventions around routine childhood vaccinations has been recognised internationally [30, 31]. In England, a 2023-2024 national catch-up campaign by the UK Health Security Agency vaccinated more than 50,000 children aged 15 months to five years against MMR1 [32]. There is strong evidence to support the effectiveness of primary care led quality improvement programmes to improve vaccine uptake [33]. National measures to tackle these inequalities include NHS England’s Quality and Outcomes Frameworks to incentivise timely routine childhood vaccinations in primary care [34].

While there is technical guidance for the registration of new patients [35], good practice guidance on what happens once the patient is registered has not been developed. Consequently, the process for documenting and managing the registration process, and any clinical review of new patients, varies from GP to GP. Where residential mobility results in a change of GP practice, at the very least, a desktop patient review should be conducted to check vaccination status. Where GPs have the capacity to offer in-person new patient health checks, a focus on vaccinations would identify children overdue for any routine primary vaccinations. In practices where new patient health checks are not offered, opportunistic conversations about vaccinations with caregivers during children’s consultations for other issues might remind and/or encourage caregivers to make vaccination appointments. It is however important to recognise that opportunistic conversations should only be instigated by staff with appropriate training and should align with the GP practice strategy. It’s possible that published good practice guidance on the GP registration process could begin to standardise this process to ensure that the vaccination status of all children is checked as early as possible after registration, to improve the timeliness of vaccination receipt.

As well as new registration checks, an emphasis on making appointments as accessible as possible could facilitate vaccination uptake. We found that residential mobility in NEL is more common in the most deprived areas, and among households with a single adult or many children. These families may face several barriers to accessing vaccination appointments, including juggling employment and childcare, with limited time to make and attend appointments for multiple children. Increasing the availability of flexible, mobile services could facilitate vaccination uptake among those most likely to experience residential mobility. A pilot was launched in January 2026 to trial provision of vaccinations during routine health visitor visits - this model of care may be a step towards flexible, mobile services that are required by families facing barriers to accessing routine vaccination appointments [36]. Qualitative research focusing on understanding the implications of residential mobility for children and their families could highlight how services might be adapted to better suit the needs of those experiencing residential mobility.