Child health and social care outcomes in large households: a population-based linked-data study in Northwest London Authors: Sandeepa Arora, Nicholas Longford, Steven C Hope, Dougal Hargreaves Published: 2026 ABSTRACT AbstractIntroductionChildren in larger households in the UK are at increased risk of poverty due to policies such as the two-child benefit limit, but evidence on their health and social care outcomes is limited. We used linked administrative data to examine whether outcomes differ between children in smaller (1–2 children) and larger (≥3 children) households in Northwest London (NWL).MethodsWe conducted a retrospective, population-based study using the Discover-NWL database (2005–2024). Outcomes included dental extraction (at <12 years of age), antidepressant prescription, emergency hospital admission (respiratory and any cause), emergency department (ED) visit and Child in Need (CiN) status. Analyses were descriptive and stratified by Index of Multiple Deprivation quintiles; multilevel logistic regression adjusted for age, sex, ethnicity, deprivation and within-household clustering.ResultsData were available on 477 602 children aged 0–18 years. In unadjusted descriptive analyses, children in larger households had higher population-level rates of dental extraction, respiratory admissions and CiN status, and lower rates of ED visits and antidepressant prescribing.In adjusted multilevel models, larger household size was associated with higher odds of dental extraction (OR 1.07 per additional child, 95% CI 1.05 to 1.10) and CiN status (OR 2.90, 95% CI 2.77 to 3.04), and lower odds of antidepressant prescribing (OR 0.86, 95% CI 0.82 to 0.90) and ED visits (OR 0.94, 95% CI 0.93 to 0.95).ConclusionChildren in larger households experience different patterns of health and social care contact compared with those in smaller households. These population-level findings, supported by adjusted analyses, provide new evidence on household size as a marker of child health and social care inequalities and may inform future research and policy.Keywords: Sociodemographic Factors, statistics and numerical data, Public Health WHAT IS ALREADY KNOWN ON THIS TOPIC WHAT THIS STUDY ADDS Using linked routine data on nearly half a million children, we show that children in larger households (3+ children) in Northwest London had higher rates of dental extraction, respiratory admissions and Child in Need status, but lower rates of emergency department visits and antidepressant prescribing. Using linked routine data on nearly half a million children, we show that children in larger households (3+ children) in Northwest London had higher rates of dental extraction, respiratory admissions and Child in Need status, but lower rates of emergency department visits and antidepressant prescribing. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY Household size is a useful marker of child health inequalities. Replication in other regions and further research into correlates of poverty could inform welfare and service policies affecting larger families. Household size is a useful marker of child health inequalities. Replication in other regions and further research into correlates of poverty could inform welfare and service policies affecting larger families. Introduction Child poverty is a persistent determinant of health, shaping outcomes across the life course.1,2 In the UK, child poverty has increased from 3.6 million in 2010/2011 to 4.3 million in 2022/2023.3 Chronic respiratory disease, poor dental health, obesity, food insecurity and mental health problems are consistently linked with poverty, often mediated by poor diet and living conditions.4–8 Beyond health, rising child poverty contributes to higher rates of children entering the social care system.9–12 The combined health, social and financial burdens of poverty disproportionately affect larger households.13 Policy changes introduced since 2017, such as the two-child limit on benefit payments and the household benefit cap, are likely contributors to widening inequalities.10,12,14 Projections indicate that, under current policies, child poverty could reach 55% for households with three or more children and 77% for those with four or more by 2027/2028.11 The two-child limit restricts benefits for a third or later child, while the benefit cap sets a maximum on total household payments, which often hits larger families hardest. Qualitative research has linked these restrictions to poorer child health, but population-level quantitative evidence remains scant.10,12 Household composition is shaped by diverse cultural, religious and life-course pathways, reflecting heterogeneity in fertility patterns by education, ethnicity, migration history and family formation trajectories in the UK.15,16 Household size and composition can influence children’s health and other outcomes in many ways, and these influences should not all be reduced to the lens of child poverty. However, theoretical and empirical literature suggests that, under conditions of economic constraint, larger household size may be associated with an increased likelihood of exposure to poverty-related disadvantage, as limited household resources are shared among more members.17 Longitudinal evidence from the UK Millennium Cohort Study demonstrates that family resources mediate the relationship between socioeconomic circumstances and child outcomes, highlighting the importance of household context in shaping health and development.18 Broader UK policy research further indicates that material resources influence child health through multiple pathways, including housing quality, food security, parental stress and access to services.19 In the context of contemporary UK welfare policies that scale financial support by household size, household composition may shape children’s exposure to poverty-related disadvantage through differences in how resources are shared within households. This study therefore examines whether children living in larger households experience different health and social care outcomes, and how these patterns vary by area-level deprivation under current UK welfare policy. Advances in data linkage methodologies, including the use of de-identified healthcare records and pseudonymised address identifiers, such as the Unique Property Reference Number (UPRN), enable closer examination of household-level influences on child health and social care outcomes.20–22 Linkage of de-identified records to household identifiers allows assessment of whether children in larger households experience different outcomes from those in smaller households, and whether these patterns vary by area-level deprivation.21,23,24 Such linkage provides valuable population coverage, although interpretation must take account of the limits of routinely collected data, such as under-recording of some conditions and variation in coding practices. Evidence on how household size relates to child health and social care outcomes remains weak. Focusing on outcomes that are strongly associated with disadvantage and are consistently captured in administrative records, we examine unplanned hospital activity, dental extractions, antidepressant prescribing, emergency admission for a respiratory condition and Children in Need (CiN) status. In the absence of direct measures of household income or material deprivation in administrative data, household size and area-level deprivation are used as pragmatic, policy-relevant indicators of socioeconomic disadvantage, consistent with established epidemiological and policy approaches.17–19,25 Child poverty is often defined in terms of financial resources, although other dimensions, such as housing quality and stability, access to services and social support are also relevant. While these measures do not capture the full complexity of disadvantage, they provide the best available population-level indicators for examining inequalities under current policy conditions. This study aims to describe patterns of health and social care outcomes among children by household size, and to assess how they vary across levels of area deprivation. We expected to observe systematic differences between children in larger households (three or more children) and smaller households (one or two children), and that these differences would vary by Index of Multiple Deprivation (IMD) quintile. Methods Data We used the Discover-Northwest London (NWL) database, a population-based database covering NWL. It draws on records from more than 400 provider organisations, including over 350 general practitioner (GP) practices and all acute providers attended by NWL patients.26 The NWL population has an age–sex distribution and prevalence of comorbidities comparable to the UK population overall, but it is ethnically more diverse.26–28 Data access was via a remote secure server in a Trusted Research Environment, hosted by Imperial College Health Partners, London, UK.26 Access to Discover-NWL data is available only via approved application and analysis within the Trust Research Environment and is subject to information governance approvals. Record linkage is performed by the Discover team using the National Health Service number as the unique identifier. Primary care records are available from 2005 onwards, while linked primary–secondary care records are available from 2015 to 2024. Household membership was defined using the UPRN, which assigns individuals to a single residential address at the time of data extraction. This identifier was available only at the time of data extraction and was therefore used to assign children to households at a single time point; changes in UPRN over time were not available within the dataset. The NWL region includes eight London boroughs: Brent, Ealing, Hammersmith and Fulham, Harrow, Hillingdon, Hounslow, Kensington and Chelsea and Westminster. Social care data are available only for Harrow (years 2018/2019–2022/2023) and for Kensington and Chelsea and Westminster (2022/2023). As of 26 September 2024, the database contained records for 2.6 million NWL residents registered with a GP, representing about 96% of the total population. Of the 499 494 individuals aged 0–18 years, our analysis focused on 477 602 (95.6%) with a valid UPRN that corresponded to a single household. Children without a valid UPRN were excluded, mainly due to population movement such as temporary registrations, or because they were living in institutions such as hotels, boarding schools, university campuses or other settings that cannot be assigned to a single household. Details of included and excluded housing types are provided in online supplemental materials 1. Throughout this study, the ages of children refer to their ages on 26 September 2024, the date of data extraction. The study design was a retrospective cohort, with UPRN, population denominator and sociodemographic characteristics defined at this single time point, and outcomes measured as cumulative events during 2005–2024. Variables Outcomes and stratification Household size was based on UPRN-defined household membership at data extraction and treated as fixed for analysis. Health and social care outcomes are summarised in table 1. These outcomes were selected based on their public health relevance and availability in the Discover database. Table 1Health and social care outcomes included in the studyCategoryOutcomeDefinitionData coverage*Dental cariesDental extractionExtraction in hospital, ages 0–12 years†NWL, Apr 2005 to Jan 2024Respiratory healthEmergency respiratory admissionEmergency admission with primary ICD-10 diagnosis J00–J99†‡NWL, Apr 2005 to Jan 2024Mental healthSSRI prescriptionPrescription of any SSRI§NWL, Apr 2005 to Jan 2024Social services involvementCiNChild documented as ‘in need’ under statutory guidance¶Harrow (2018–2022); Kensington and Chelsea and Westminster (2022)Acute health problems (general)Emergency admission (any cause)All-cause emergency hospital admissionNWL, Apr 2005 to Jan 2024Acute health problems (general)ED visitVisit to an ED for any causeNWL, Apr 2005 to Jan 2024*Coverage refers to availability of data in NWL Discover database†Extractions at older ages may reflect orthodontic procedures rather than dental caries.‡ICD-10 refers to the International Classification of Diseases, 10th Revision.§SSRI are the main class of antidepressants recommended for children and adolescents in UK clinical guidance. The full list of SSRI included is provided in online supplemental materials 2.¶Defined according to Department for Education Children in Need Census statutory guidance.CiN, Child in Need; ED, emergency department; NWL, Northwest London; SSRI, selective serotonin reuptake inhibitors. Health and social care outcomes included in the study Coverage refers to availability of data in NWL Discover database Extractions at older ages may reflect orthodontic procedures rather than dental caries. ICD-10 refers to the International Classification of Diseases, 10th Revision. SSRI are the main class of antidepressants recommended for children and adolescents in UK clinical guidance. The full list of SSRI included is provided in online supplemental materials 2. Defined according to Department for Education Children in Need Census statutory guidance. CiN, Child in Need; ED, emergency department; NWL, Northwest London; SSRI, selective serotonin reuptake inhibitors. The outcomes examined reflect recorded healthcare and social care contacts within routine administrative systems and therefore represent patterns of service utilisation and statutory involvement rather than direct measures of underlying morbidity or social need. Analyses of CiN outcomes were restricted to the boroughs and years for which data were available and should be interpreted as exploratory. Area-level deprivation Area-level deprivation was measured using the 2019 English IMD, the official measure of small-area socioeconomic deprivation in England. IMD is a composite index based on seven domains: income, employment, education, health, crime, barriers to housing and services and living environment. It was linked at the lower-layer super output area level using residential postcode and categorised into national quintiles (IMDq 1=most deprived; IMDq 5=least deprived). IMD is widely used in epidemiological and policy research and is strongly associated with child health outcomes, healthcare utilisation, and children’s social care involvement in England.29–31 Covariates and stratification Sociodemographic variables were summarised by household size, sex (male, female), age group (0–4, 5–9, 10–14, 15–18 years), ethnicity (white, Asian, black, mixed, other) and IMD quintile. We described the number and proportion of children with each outcome by household size. In addition to child-level analyses, we conducted household-level analyses, identifying households with at least one positive case to estimate the overall household burden (online supplemental materials 3, online supplemental table SM3). Analysis Descriptive analyses Associations between household size (1–2 versus 3+ children) and outcomes were assessed using χ2 tests. χ2 statistics, df and 95% CIs are reported as indicators of precision, not for formal inference, given the near-complete population coverage and large sample size. No correction for multiple testing was applied, as analyses are descriptive and exploratory. Analyses were stratified by deprivation, comparing more deprived households (IMDq 1–2) with less deprived households (IMDq 3–5). Multilevel regression analysis As an additional analysis, we fitted multilevel logistic regression models for each outcome with household as a random effect, using maximum likelihood.32 The logit of each outcome was modelled as a quadratic function of age, IMD quintile and household size, with adjustment for sex and ethnicity. These models provided adjusted associations and estimates of within-household clustering (online supplemental materials 5, online supplemental tables SM5a and b). Covariate selection was informed by prior evidence and data availability within linked administrative records. Direct measures of household income, parental education, housing quality and parental physical and mental health were not available, so residual confounding cannot be ruled out. These models were used to assess whether descriptive patterns were robust to adjustment for demographic and socioeconomic factors and within-household clustering, rather than to estimate causal effects of household size. Outcomes were defined as binary indicators of whether a child ever experienced the event during the observation period. Logistic regression was used to compare outcome rates across background characteristics such as sex and ethnicity and to accommodate outcomes with different structures, including non-recurrent outcomes such as CiN status. Follow-up time varied between individuals due to age, and recurrent events were not modelled explicitly. Multicollinearity was assessed by examining correlations between covariates and the stability of model estimates when variables were included jointly. Correlations were weak to moderate, and no evidence of instability was observed. Model fit was assessed by comparing observed and fitted outcome rates across household size and IMD quintile combinations (online supplemental materials 5, online supplemental figure SM5). Residual diagnostics were not used because, for binary outcomes in multilevel logistic regression, residuals are less informative for overall model fit; calibration of observed and fitted rates provides a more relevant assessment. Sensitivity analysis Sensitivity analyses were conducted to examine the implications of the analytical assumptions related to household stability over time. These included age-stratified analyses, given that household composition is more stable within narrower age bands, and analyses restricted to a fixed follow-up window (2021–2023) to address unequal time at risk. Additional stratified analyses by ethnicity were undertaken to examine consistency across population subgroups. Full results are presented in online supplemental materials 6 and online supplemental tables SM6a–d. All data preparation and management were carried out in Snowflake SQL. Statistical analyses were performed using R V.4.0.4. Results Health and social care outcomes Descriptive summaries for the outcomes are presented below. Healthcare outcomes included hospital-based dental extractions, emergency admissions for respiratory conditions, antidepressant prescriptions (selective serotonin reuptake inhibitors (SSRI)), emergency hospital admissions for any cause and emergency department (ED) visits across NWL between 2005 and 2024. Social care outcomes were represented by CiN status, available only for Harrow (2018–2022) and the Bi-borough (Kensington and Chelsea and Westminster, 2022) and analysed on an exploratory basis. Demographic characteristics Table 2 presents unadjusted descriptive analyses characterising population-level demographic patterns by household size; interpretation is supported by multilevel adjusted analyses reported in the online supplemental materials 5. Most children lived in households with one (24.3%) or two (36.2%) children. Two-child households were most common across all ethnic groups, particularly among white (38.7%), mixed (37.8%) and Asian (37.5%) children. Black children were somewhat less likely to live in two-child households (28.2%) and more likely to live in larger households. Children in the most deprived quintile (IMDq 1) were more likely to live in three-child households (21.6%) compared with those in the least deprived areas (IMDq 5, 17.9%). Gender distribution was similar across household sizes. Table 2Demographic characteristics of children by household size (counts and percentages in parentheses underneath)Number of childrenCharacteristic12345+TotalEthnicity Asian31 430 (21.8)54 121 (37.5)29 939 (20.8)14 618 (10.1)14 054 (9.7)144 162 Black9899 (20.7)13 517 (28.2)10 529 (22.0)6721 (14.0)7259 (15.1)47 925 Mixed9794 (29.6)12 513 (37.8)6197 (18.7)2551 (7.7)2032 (6.1)33 087 Other15 575 (22.6)22 704 (32.9)14 852 (21.5)7854 (11.4)7994 (11.6)68 979 White44 789 (27.4)63 348 (38.7)30 511 (18.6)13 028 (8.0)11 933 (7.3)163 609 Missing/unknown4532 (22.8)6645 (33.5)4089 (20.6)2232 (11.3)2342 (11.8)19 840IMD quintile 1 (most deprived)15 708 (24.0)21 488 (32.8)14 187 (21.6)7467 (11.4)6694 (10.2)65 544 237 854 (24.0)53 514 (33.9)32 697 (20.7)16 911 (10.7)17 037 (10.8)158 013 332 777 (24.3)48 749 (36.1)27 254 (20.2)13 278 (9.8)12 979 (9.6)135 037 421 085 (25.1)33 118 (39.5)15 698 (18.7)7024 (8.4)6975 (8.3)83 900 5 (least deprived)8595 (24.5)15 979 (45.5)6281 (17.9)2324 (6.6)1929 (5.5)35 108Sex Female56 245 (24.1)84 436 (36.3)46 650 (20.0)23 167 (9.9)22 414 (9.6)232 912 Male59 768 (24.4)88 405 (36.1)49 463 (20.2)23 835 (9.7)23 199 (9.5)244 670Age group (years) 0–432 770 (30.0)37 831 (34.7)19 436 (17.8)9690 (8.9)9331 (8.6)109 058 5–922 896 (18.2)48 427 (38.6)27 831 (22.2)13 544 (10.8)12 808 (10.2)125 506 10–1426 830 (19.8)50 811 (37.4)29 468 (21.7)14 557 (10.7)14 076 (10.4)135 742 15–1833 523 (31.2)35 779 (33.3)19 382 (18.1)9213 (8.6)9399 (8.8)107 296Percentages are row percentages, calculated as the number in each child-category divided by the total within each background variable group. 20 individuals with gender recorded as “Other/Unknown” are excluded.CYP, children and young people; IMD, Index of Multiple Deprivation. Demographic characteristics of children by household size (counts and percentages in parentheses underneath) Percentages are row percentages, calculated as the number in each child-category divided by the total within each background variable group. 20 individuals with gender recorded as “Other/Unknown” are excluded. CYP, children and young people; IMD, Index of Multiple Deprivation. Child-level outcomes by household size Table 3 (with corresponding rates plotted in figure 1) shows child-level outcomes by household size. The proportion of children who had at least one dental extraction was lower in single-child households (1.6%) and higher in households with five or more children (2.4%). Similarly, among children living in boroughs with available social care data, the proportion with CiN was 3.2% in single-child households and 6.8% in households with five or more children. Table 3Distribution of health and social care outcomes by household size with percentage (%)Household size (number of children)Children N (%)ED visitEmergency inpatient admissionDental extractionCiN¶SSRI PrescriptionRespiratory admission1116 019(24.3)74 094(63.9)18 593(16.0)1889(1.6)1071(3.2)770(0.7)4248(3.7)2172 848(36.2)115 018(66.5)30 549(17.7)3373(2.0)1900(4.1)698(0.4)8020(4.6)396 117(20.1)62 034(64.5)16 755(17.4)2256(2.3)1371(5.8)299(0.3)4549(4.7)447 004(9.8)28 534(60.7)8071(17.2)1160(2.5)712(6.4)100(0.2)2272(4.8)5+45 614(9.6)25 124(55.1)6857(15.0)1104(2.4)722(6.8)96(0.2)1973(4.3)All477 602(100.0)304 804 (63.8)80 825(16.9)9782(2.1)5776(4.6)1963(0.4)21 062(4.4)Notes: ¶CiN denominators are restricted to boroughs with available data: Harrow (2018–2022), Kensington and Chelsea and Westminster (2022). Percentages are column percentages (children with outcome/total in household size group).CiN, Child in Need; ED, emergency department; SSRI, selective serotonin reuptake inhibitor. Distribution of health and social care outcomes by household size with percentage (%) Notes: ¶CiN denominators are restricted to boroughs with available data: Harrow (2018–2022), Kensington and Chelsea and Westminster (2022). Percentages are column percentages (children with outcome/total in household size group). CiN, Child in Need; ED, emergency department; SSRI, selective serotonin reuptake inhibitor. Children’s health and social care outcomes by household size with 95% CIs in NWL. ED, emergency department; NWL, Northwest London; SSRI, selective serotonin reuptake inhibitors. By contrast, SSRI prescriptions were more common in smaller households, recorded for 0.7% of children in single-child households and 0.2% in households with five or more children. Respiratory admissions increased modestly with household size (3.7% in single-child households vs 4.3–4.8% in multi-child households). Emergency inpatient admissions and ED visits showed smaller relative differences, with the highest proportions in two-child households and the lowest in households with five or more children. The results in table 3 are presented graphically in online supplemental materials 4 and online supplemental figure SM4. Comparisons of smaller and larger households Table 4 compares outcomes between children in smaller (1–2 children) and larger (3+ children) households. Unadjusted rates of dental extraction were higher in larger households (2.4% vs 1.8%), as were respiratory admissions (4.7% vs 4.2%), and, among those living in boroughs with available social care data, CiN status (6.2% vs 3.7%). Conversely, ED visits (62.3% vs 65.5%) and SSRI prescriptions (0.3% vs 0.5%) were less common among children in larger households. Emergency inpatient admissions were very similar across groups (16.8% vs 17.0%). These differences were generally modest in magnitude. Table 4Proportion of children with health and social care outcomes comparing smaller (1–2 children) and larger (3+ children) households, stratified by deprivationOutcomeHousehold groupMost deprived %(95% CI)Less deprived %(95% CI)All children %(95% CI)χ²(p value)ED visitSmaller (1–2)65.5 (65.4 to 65.6)67.7 (67.4 to 68.0)63.7 (63.5 to 63.9)1254 (<0.001)Larger (3+)61.3 (61.1 to 61.5)63.7 (63.4 to 64.0)58.9 (58.6 to 59.2)Emergency inpatient admissionSmaller (1–2)17.0 (16.9 to 17.1)17.8 (17.6 to 18.0)16.4 (16.2 to 16.6)5 (0.043)Larger (3+)16.8 (16.7 to 16.9)17.7 (17.5 to 18.0)15.9 (15.7 to 16.1)Dental extractionSmaller (1–2)1.8 (1.8 to 1.8)2.2 (2.1 to 2.3)1.5 (1.4 to 1.6)279 (<0.001)Larger (3+)2.4 (2.3 to 2.5)2.8 (2.7 to 2.9)2.0 (1.9 to 2.1)CiNSmaller (1–2)3.7 (3.6 to 3.8)2.8 (2.7 to 2.9)4.1 (4.0 to 4.2)2060 (<0.001)Larger (3+)6.2 (6.1 to 6.3)5.0 (4.9 to 5.1)6.7 (6.5 to 6.9)SSRI prescriptionSmaller (1–2)0.5 (0.5 to 0.5)0.5 (0.5 to 0.5)0.5 (0.5 to 0.5)180 (<0.001)Larger (3+)0.3 (0.3 to 0.3)0.2 (0.2 to 0.2)0.3 (0.3 to 0.3)Respiratory admissionSmaller (1–2)4.2 (4.1 to 4.3)4.6 (4.5 to 4.7)4.0 (3.9 to 4.1)96 (<0.001)Larger (3+)4.7 (4.6 to 4.8)5.0 (4.9 to 5.1)4.3 (4.2 to 4.4)Smaller households (1–2 children) and larger households (3+ children). IMD refers to the Index of Multiple Deprivation, grouped as most deprived (quintiles 1–2) and less deprived (quintiles 3–5). Values are percentages with 95% CIs. χ² denotes the chi-squared test statistic and p values indicate statistical significance for comparisons between smaller and larger households within each outcome.CiN, Child in Need; ED, emergency department; SSRI, selective serotonin reuptake inhibitor. Proportion of children with health and social care outcomes comparing smaller (1–2 children) and larger (3+ children) households, stratified by deprivation Smaller households (1–2 children) and larger households (3+ children). IMD refers to the Index of Multiple Deprivation, grouped as most deprived (quintiles 1–2) and less deprived (quintiles 3–5). Values are percentages with 95% CIs. χ² denotes the chi-squared test statistic and p values indicate statistical significance for comparisons between smaller and larger households within each outcome. CiN, Child in Need; ED, emergency department; SSRI, selective serotonin reuptake inhibitor. Additional multilevel models including interactions between household size and IMD quintile showed that socioeconomic gradients in outcomes persisted across household sizes, with higher rates of dental extraction and social care involvement observed in larger households within each deprivation stratum (online supplemental materials 5c; online supplemental figure SM5). Stratification by deprivation As shown in table 4 and online supplemental materials 4, online supplemental figure SM4, differences between larger and smaller households were broadly consistent across deprivation groups. For example, dental extractions were more common in larger households in both more deprived (2.8% vs 2.2%) and less deprived (2.0% vs 1.5%) areas. Similarly, CiN status was more common in larger households across deprivation strata, 6.2% versus 3.7% in more deprived and 5.0% versus 2.8% in less deprived households. The only exception was emergency inpatient admissions, which were slightly less common in larger households in the least deprived areas (15.9% vs 16.4%). Online supplemental figure SM4 in online supplemental materials 4 presents full results stratified by deprivation quintile. Household-level analyses, estimating the proportion of households with at least one affected child, are presented in online supplemental materials 3. Multivariable logistic regression models with random coefficients for household-level clustering, adjusting for age, sex, ethnicity, deprivation and household size, are reported in online supplemental materials 5. The results they yielded are broadly consistent with the descriptive results. For example, the models confirmed that dental extractions and CiN status increased with household size, with predicted rates rising from 2.1% to 2.7% and 0.02% to 1.1%, respectively, between single-child and 5+-child households. In contrast, the descriptive variation in emergency and respiratory admissions by household size was not apparent after adjustment, suggesting these differences were largely attributable to demographic and socioeconomic factors. In adjusted analyses, the rates of CiN were much smaller for white children than for black and Asian children. The rates of dental extraction and being prescribed SSRI were higher for white children than for the other ethnic groups. The large estimated household-level variance for CiN status (28.43) indicated substantial within-household clustering, meaning that siblings of children with CiN status were frequently also recorded as CiN. The variances for the healthcare conditions were much smaller (estimates in the range 0.67–1.66) but far from trivial. Discussion In this population-based study in NWL, children in larger households (with three or more children) were more likely than those in smaller households (with one or two children) to have had dental extractions, emergency respiratory admissions and CiN status. These differences were observed among children living in both more deprived and less deprived areas. Children in larger households were less likely to have been prescribed antidepressants (SSRI) or to have attended the ED. The direction and magnitude of associations varied across outcomes, indicating modest and heterogeneous patterns rather than a consistent relationship between household size and health and social care outcomes. Associations with household size were generally weak, particularly for outcomes such as emergency and respiratory admissions, where differences between groups were limited in absolute terms. Although some of these differences were statistically significant, they correspond to relatively small changes in individual risk. Given the very large sample size, even small differences are likely to reach statistical significance, and p values should therefore be interpreted with caution. These findings should therefore be interpreted with attention to effect size rather than statistical significance alone. Nonetheless, even small differences in rates may be meaningful at the population level when large numbers of children are affected. We fitted multilevel models to adjust for age, sex, ethnicity, deprivation and within-household clustering, including interaction terms (online supplemental materials 5). The adjusted analyses were broadly consistent with the descriptive findings, with most estimated differences between household sizes remaining small. Household size should therefore be interpreted as a descriptive, policy-relevant marker rather than an independent causal determinant of outcomes. Household-level variance was greatest for CiN status, indicating strong clustering of social care involvement among siblings, while clustering was more modest for health service use outcomes such as emergency admissions and dental extractions. Previous studies have examined household size in relation to child health and well-being, often within child poverty and policies such as the two-child benefit cap.33 Our findings are consistent with the Marmot Review 10 Years On, which showed persistent health inequalities linked to child poverty in England.30 To our knowledge, no other UK study has used routine population data to examine household size, deprivation and health and social care outcomes for children. This study also adds to work on household-level linkage, building on research by Santos et al, who validated the use of the UPRN linkage in routine health data and demonstrated the reliability of this method.34 Our results show higher recorded levels of health and social care involvement among children in larger households for some outcomes, but not for all, highlighting heterogeneity in how household size interacts with deprivation. For example, dental extractions were most common among children in large households in deprived areas, while the lowest rates were seen in small households in the least deprived areas. Bywaters and colleagues showed that children in poorer areas are more likely to be involved with social care,31 which is broadly consistent with our findings on CiN. However, the higher observed CiN rates in less deprived groups may reflect difference in thresholds for intervention, detection and service response rather than lower underlying need. Previous research has shown that, after accounting for neighbourhood deprivation, social care involvement can vary substantially by local authority context, reflecting differences in referral practices, thresholds and service capacity.35 Multilevel models including household size×IMD quintile interactions suggested that these differences reflect effect modification rather than stratification alone (online supplemental materials 5, online supplemental table SM5c). Poverty is one factor, but household size may also be associated with differences in exposure and family context that correlate with observed outcomes. Different outcomes are likely shaped by different mechanisms, including shared exposure, family context and access to services. Parents may approach care differently with later children, and siblings increase exposure to common infections. In smaller households, firstborn children may attract more attention, and they are known to differ from later-born siblings in both positive ways (such as higher rates of breastfeeding and educational attainment) and negative ways (such as higher risk of preterm birth, more health service use and more prescriptions).36 Other work has found that larger families face disadvantages in education when combined with low income37; our study suggests that a similar pattern may be present for health and social care outcomes. Our finding that SSRI prescribing was less common among children in larger households is consistent with earlier reports of lower prescribing in more deprived groups, particularly in London.38 In that national study, the association reversed after adjustment for region and other covariates, underlining the complexity of antidepressant prescribing patterns. Taken together, these findings suggest that the lower prescribing observed in our descriptive analysis may reflect health system and contextual influences, rather than differences in underlying need. It is not possible to determine which differences by household size reflect expected variation and which represent inequity. For outcomes such as dental extractions, which are largely preventable through diet, hygiene and access to dental care, higher rates in larger households suggest avoidable harm. Given the population-level scope of the study, we did not prespecify thresholds for clinical or policy significance. The aim was to describe differences by household size rather than to test specific mechanisms, but even small differences may matter at population level and for service planning. Future research should test formal hypotheses and define policy-relevant thresholds. Strengths and limitations A major strength of this study was the use of linked, population-level administrative data covering nearly the entire recent NWL population. This provides near-complete enumeration of the region, minimises selection bias and gives access to recent data with minimal time lag compared with surveys. Linkage across healthcare and social care records also allowed us to describe outcomes across multiple domains. Household size was defined at a single time point (September 2024), and changes in composition over time could not be captured. This may lead to misclassification, for example, if siblings were born, left home or reached 19 years of age during the study period. Younger single-child households may later become multi-child households, whereas households with older children may be complete in terms of childbearing but may have ‘lost’ children who reached 19 years of age. As UPRN was available only at data extraction, household size was treated as fixed, and the analysis was designed to describe differences across household groups rather than estimate causal effects of household size itself. Findings should therefore be interpreted as descriptive patterns within a broader social and policy context in which household size and disadvantage are correlated. Age-stratified and restricted follow-up sensitivity analyses showed similar patterns to the main results (online supplemental materials 6). Exclusions based on UPRN availability and household definition may disproportionately affect highly mobile or marginal families, who may therefore be under-represented in the analysis, and this should be considered when interpreting the findings. Age distribution differed across household types, with a higher proportion of children aged 0–4 and 15–18 in single-child households. This may have influenced outcomes such as hospital admissions, ED visits and SSRI prescriptions, which are more common at these ages, and may also affect cumulative incidence through uneven exposure times, as younger children contribute shorter periods at risk. Although these differences cannot be fully controlled for, adjusted models showed similar patterns (online supplemental materials 5). Outcomes were modelled as binary indicators of whether a child ever experienced an event during the study period, which does not account for differences in follow-up time or repeated events. Alternative modelling approaches such as count or time-to-event models may be useful in future work but were not applied here because outcomes differed in how they were defined and recorded across data sources, making a consistent approach across all outcomes difficult. Another limitation is that we measure health status through provision of healthcare. This inevitably overlooks differences in accessibility and willingness to seek healthcare or social services, as well as how health problems are recognised or sometimes not acknowledged by parents or carers. These are important dimensions of child and family well-being, but they can be meaningfully assessed only by direct engagement with households, which was beyond the scope of this study. Generalisation beyond NWL should be cautious, as the population may not represent other UK regions. Another limitation is the absence of direct household-level measures of poverty. Instead, we used household size and area-level deprivation (IMD) as practical proxies, consistent with epidemiological approaches but not capturing the full complexity of disadvantage.39,40 Residual confounding is therefore likely, as we were unable to account for other determinants of child health and social care use, including household income, housing conditions, food insecurity, and parental physical and mental health. IMD is an area-level measure and does not capture socioeconomic variation within neighbourhoods; however, it is the official and widely used measure of small-area deprivation in England and is commonly used in population studies of child health and social care. Finally, the outcome measures themselves have limitations. ED visits and hospital admissions reflect not only health status but also healthcare seeking and access. Antidepressant prescribing is influenced by service availability, clinical practice and cultural factors, and therefore may not directly reflect mental health prevalence. Social care involvement (CiN) was available only for three boroughs and for only a few recent years. These findings should therefore be interpreted with caution. They may not be representative of NWL as a whole, and selection bias is likely; the CiN analyses are intended to be exploratory and not generalisable. Despite these limitations, using multiple routinely recorded outcomes provides a broad perspective on child health and social care needs without the expense of data being collected for a specific study. Recent policy developments should be considered when interpreting these findings. Projections of rising child poverty in larger families were based on policy assumptions that predated the UK Government’s Child Poverty Strategy published in December 2025,41 following the Child Poverty Taskforce and 2025 Budget measures. While these reforms signal renewed policy attention their effects are not yet observable in population-level administrative data. The patterns described here provide important context for understanding inequalities under earlier policy environment and for assessing future policy impact. Implications and future work In this study, UPRN was used solely as an administrative linkage tool to identify households and define household size. The contribution of this study lies in the substantive population patterns identified. These patterns were described through analysis of linked health and social care data, without implying causal attribution or prescribing targeted policy interventions at the household level. Children in larger households, particularly in more deprived areas, had higher recorded levels of health and social care involvement for some outcomes compared with children in smaller households. This suggests that household size warrants consideration when examining and addressing inequalities in children’s health and social care. The study also demonstrates the feasibility of UPRN linkage for examining such patterns. Analyses were carried out in a single region (NWL) and would benefit from replication in other regional and national datasets across England and Wales. Future research should investigate the pathways linking household size to outcomes and how interventions might best be targeted through policy and service delivery. Building on UPRN linkage, further integration of administrative data across health, social care and education systems could improve understanding of childhood inequalities and guide more timely preventive interventions. Recognising household size as a marker of disadvantage could help policymakers better identify families at risk and target preventive support. Conclusion This study examined whether health and social care outcomes differ between larger (three or more children) and smaller households. The results show a consistent pattern of higher rates of dental extraction, respiratory hospital admissions and CiN status in larger households, with lower rates of ED visits and antidepressant prescribing. All-cause emergency admission rates showed only modest differences by household size. While these findings are descriptive, they indicate population-level patterns in child health and social care that are relevant to inequalities research, policy and service planning. Supplementary material Acknowledgements The authors are grateful for support provided by the Northwest London Discover team that supports and maintains the de-identified database and the platform access. The data is provided by patients and collected by the National Health Service as part of their care and support. Footnotes Funding: The research was conducted using existing institutional infrastructure at Imperial College London and within the Discover-NWL programme. Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not applicable. Data availability free text: Data underlying this study were accessed through Discover-Northwest London (NWL) within the National Health Service Trust Research Environment. The data are held by a third party and are not publicly available due to information governance and confidentiality requirements. Access to Discover-NWL data may be granted to approved researchers following application to the NWL Discover Data Access Committee, with all analyses conducted within the Trust Research Environment in accordance with data protection and information governance requirements. Patient and public involvement: Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research. Correction notice: This article has been corrected since it was first published. The author's name has been updated from "Hope C Steven" to "Steven C Hope." Data availability statement Data may be obtained from a third party and are not publicly available. References Associated Data Supplementary Materials Data Availability Statement Data may be obtained from a third party and are not publicly available.