Sociodemographic variation in cumulative probability of recorded special educational needs and disability provision during primary school in England: A staggered cohort study of children with cerebral palsy

Public and patient involvement

This study is part of a wider programme of work called Health Outcomes for People in Education (HOPE)¹⁵. At the outset of HOPE, we consulted with stakeholders, including parent/carers of children who have requested SEND provision. These consultations revealed frustration with the variation in timing and provision of SEND services, particularly for parents and carers who disproportionately lacked resources to advocate for their child(ren). These discussions led us to design this study to investigate inequities in the timing of SEND provision.

Study design and setting

A series of cohorts, staggered by school Year at entry into education, were constructed using linked de-identified child-level administrative health and educational records from England. A staggered cohort design was chosen to reflect the real-world staggered entry of children into state-funded educational settings, improving inferences about the timing of SEND provision.

Data sources and linkage

This study used the ‘Education and Child Health Insights from Linked Data’ (ECHILD) database¹⁶, which comprises de-identified population-based linked administrative health and school records of children and young people in England. Health records in this study were derived from Hospital Episode Statistics Admitted Patient Care (HES APC), which contains records of all National Health Service (NHS)-funded inpatient hospital activity in England, including records for approximately 97% of births in England¹⁷. Educational records are derived from the National Pupil Database (NPD), which contains pupil- and school-level information¹⁸. The primary source of information in NPD are the termly School Censuses (collected in October, January and May each academic year), containing information on pupil enrolment, absences and exclusions and SEND provision from state-funded mainstream schools, special schools and some alternative provision. There is also an annual alternative provision census and pupil referral unit census (collected until 2012/13, after which it was merged with the School Census). Pre-school pupils (aged between 2 and 5 years) are included in the School Census if provision is within a nursery school or nursery class attached to a state-funded school (hereafter referred to as “nursery”). Individual records in HES and NPD are linked by NHS England using deterministic algorithms based on name, date of birth, assigned sex/parent-reported gender and postcode¹⁹.

Population and follow-up

Our population consisted of children with hospital-record defined cerebral palsy, born in an NHS hospital in England between 1 September 2003 and 31 August 2013, and recorded as entering a state-funded nursery or school in England between ages 2 and 4 years (see Figure S1). Births were identified in HES APC using a combination of diagnostic and procedure codes, healthcare resource group codes and administrative variables as outlined in Zylbersztejn et al.²⁰ We excluded multiple births (defined in HES APC data) to reduce the risk of including erroneous links²⁰. Cerebral palsy was defined by any hospital admission before January School Census of cohort entry (defined below) with an International Classification of Diseases 10^th Revision (ICD-10) code G80.*, including all subtypes, in any of the 20 diagnoses fields (Table S1)²¹.

We created three cohorts, corresponding to points of entry into state-funded nursery or primary school for most pupils: nursery Year 1 (N1; pupils expected to be aged 2 years by the start of the school year); nursery Year 2 (N2; 3 years by the start of the school year); and Reception (4 years by the start of the school year). In England, school attendance is compulsory from the age of 5 years meaning that, depending on their date of birth, children are required to enter school at some point in Reception. However, most children attend early years education before these dates, where SEND provision may also be assigned (and recorded). In this study, we only included school-based nursery settings, but children may also attend early years provision organised by the private, voluntary and independent sectors. School-based nurseries have a higher proportion of children from deprived backgrounds than other early year settings, reflecting purposeful development of maintained nursery schools in disadvantaged areas and eligibility policies²². Policies changed over the course of the study period, with government funding between 12–30 hours of early years provision a week during term time for children aged 3/4 years old, and additional entitlements for children aged 2/3 and older for families with low incomes and children with an EHCP²³.

For each pupil, the start of follow-up was set to the day before their first appearance in a January census (in N1, N2 or Reception). Pupils were followed up until the first chronological event of: first recorded SEND provision, January census in Year 6 (the last year of primary school, pupils aged 10/11 years), loss to follow-up, or end of the study (the January census in 2019). Loss to follow-up was defined by nonappearance in any NPD School Census following entry into the study. Apart from death, reasons for non-appearance were not recorded in this study, but may also include transfer to a non-state-funded school, home schooling, emigration and off-rolling (illegal exclusion)²⁴.

Outcome

Recorded SEND provision was defined at the January census from school entry to Year 6 (ages 2/3–10/11 years) of the School Census, alternative provision census and pupil referral unit census. The School Census is also collected in November and March; however, we use the January census only as the information recorded in this census is used to determine funding (therefore, it is more likely to contain complete information). We defined two categories of SEND provision: SEND support (previously called School Action and School Action Plus); and Education, Health and Care Plans (EHCPs; including statements of SEN). We purposefully use the term recorded SEND provision as opposed to SEN provision (the label used in the NPD) to signify that such a recording is indicative of educational support for children with a learning difficulty and/or disability⁶. We use the label ‘recorded’ SEND provision to emphasise that a record of SEND provision in the educational data does not necessarily indicate that SEND provision was received, and there is no information about the precise elements of support received.

Sociodemographic groups

In this study, sociodemographic groups were included as indicators of social strata that, through structural inequities in power, money and resources, shape access (and barriers) to timely interventions, such as SEND provision^25,26. We present results by pupil gender, income deprivation affecting children index (IDACI) and means-tested free school meal (FSM) eligibility. Parent-reported gender (female or male, only two categories available) and IDACI were measured in the NPD at the pupil’s first appearance in a January census. IDACI is an area-level measure of socioeconomic circumstances, calculated as the proportion of children 0–15 years living in families in receipt of various benefits²⁷. IDACI scores are calculated at the Lower layer Super Output Area level (each containing approximately 400 to 1,200 households) of residential address. Owing to the small sample size, we dichotomised IDACI into pupils living in the most deprived 20% and least deprived 80% of IDACI scores for the main analyses. Means-tested FSM eligibility (yes or no), a household-level indicator of socioeconomic circumstances, was measured in the January census in Reception. To be eligible for FSMs through the means-tested route, families must receive certain benefits and, for those claiming Universal Credit or Child Tax Credit, have an annual pre-tax income under £7,400 or £16,190, respectively²⁸. We defined FSM eligibility at Reception as this is the first school Year where means-tested eligibility is applied to all pupils.

Additional sociodemographic groups known to be associated with the allocation of SEND provision (region of residence, racial-ethnic group and English as an additional language (EAL)) could not be included in the final analyses due to small sample sizes. We therefore present information about these groups to describe cohort composition only. These variables were measured in the NPD at the pupil’s first appearance in a January census. Where EAL and racial-ethnic group were not available at school entry, we imputed the nearest non-missing value between school entry and Year 1 (age 5/6).

Covariates

We included the following covariates from HES APC birth records: year of birth, defined according to the academic calendar (e.g., 2003/04 includes 1 September 2003 to 31 August 2004, inclusive); month of birth; and gestational age at birth, grouped into very preterm (22-<32 weeks), moderately preterm (32-<34), late preterm (34-<37), early term (37-<39), term (39-<42) and post-term (42-<45). We defined a series of comorbidities using diagnoses in hospital admission records from birth until study entry, based on the Hardelid chronic conditions list (Table S2)²⁹. We created a variable “time since school entry”, calculated as the day before a pupil’s first appearance in a January census minus their entry date into school. This was to account for differences in the date a child begins nursery/school and their first possibility of having SEND recorded (i.e., the January census). We also generated “age at first cerebral palsy recording”, calculated as the date of the pupil’s first recording of cerebral palsy in HES APC minus the date of their birth admission.

Statistical analysis

Case definition. We plotted the cumulative incidence of cerebral palsy as identified in HES APC records by year of birth and age of first recording (up to 5 years) to validate our study population against other data sources to assess biases in our case definition¹.

Study population. We described the study population characteristics, by staggered cohort, including missing data, in numbers and percentages. Next, we examined the changes in the three categories of recorded SEND provision (none, SEND support and EHCP) during nursery and primary school. We used Kaplan–Meier failure curves to plot the cumulative incidence of recorded SEND provision (any and EHCP) for each cohort.

Sociodemographic comparisons. For each cohort (first enrollment in N1, N2, Reception) and sociodemographic variable combination (gender, IDACI and FSM eligibility), we used inverse probability weighted (IPW) pooled logistic regression to estimate the cumulative probabilities of first recorded SEND provision over time (i.e. school Year)^30,31, and then to derive the differences in cumulative probabilities between their categories. The models were used to standardise these comparisons by predefined demographic and clinical factors known to be associated with the allocation of SEND provision: year of birth, month of birth, gestational age, age at first cerebral palsy recording in HES APC records, subtypes of cerebral palsy (denoted by asterisks in Table S1), chronic conditions and time since school entry^10,32. We present results in terms of standardised probability and absolute difference for each year after school entry, representing the cumulative probability and difference of first recorded SEND provision for each value of each exposure compared with its baseline. We estimated 95% confidence intervals for the cumulative probabilities and their differences using 1000 bootstrap samples. We considered pooling these results across cohorts using fixed effects models³³. To assess the appropriateness of the pooled logistic regression model, for each cohort, we also estimated the cumulative probability curves of first recorded SEND provision using unweighted pooled logistic regression and compared the estimates with those found using the Kaplan-Meier estimator.

Statistical analyses were conducted using Stata v18³⁴ in the Office for National Statistics Secure Research Service (ONS SRS). The code used in this study is available from www.github.com/UCL-CHIG/HOPE-CP-SEND. Changes since the study protocol are described in the extended data (Box S1)³⁵. The extended data is available at: https://zenodo.org/records/15674559.

Sensitivity analyses

We repeated the main analyses on a population of children defined using a wider definition of cerebral palsy and paralytic syndromes, as has been applied elsewhere^1,8,21. For this definition, we included children with hospital-record-defined cerebral palsy (G80.*), hemiplegia (G81.*), paraplegia and tetraplegia (G82.0-Q82.5), diplegia (G83.0) and/or monoplegia (G83.1-Q83.3), extended data Table S1²¹.

Derivation of study population

There were 5,553,737 singleton births in NHS hospitals between 1 September 2003 and 31 August 2012, of which 9,225 (0.17%) had hospital record identified cerebral palsy before the age of 6 years (Figure 1). The proportion of children with cerebral palsy increased to 0.22% (12,013) when other paralytic syndromes were also included in the case definition, which compares to a period prevalence of 0.34% in young people aged 0–21 years in an external study that applied this wider definition across multiple linked heath datasets¹. Cumulative incidence curves demonstrate slightly lower proportions of recorded cerebral palsy among children born in later study years (extended data Figure S2; extended data Figure S3), mirroring decreases observed in registry data³⁶. From our study population of children with recorded cerebral palsy, we excluded 1,100 (11.9%) children who: died before expected entry into N1; did not link to the NPD; or did not appear in a school census at any entry points into our study. For the main analysis, we further excluded 2,446 (26.5%) children whose first hospital admission indicating cerebral palsy occurred after study entry. Our final cohort for the main analyses included 5,669 children: 896 (15.8%) entering school in N1, 2,073 (36.6%) in N2 and 2,700 (47.7%) in Reception.

Figure 1. Flow chart of study population selection.

*Definition used to compare prevalence of cerebral palsy in our study with national estimates from other data sources; **square brackets denote cohort derived using the extended definition of cerebral palsy and paralytic syndromes (see Table S1); HES = hospital episode statistics, N1 = nursery Year 1, N2 = nursery Year 2, NPD = national pupil database; R = Reception.

Study population

Of the 5669 children in the final study population, the majority were male (3381; 59.6%), White-British (4298; 75.8%) and had English recorded as their primary language (4984; 87.9%; Table 1). In comparison to pupils entering school in N1 and N2, those with first recorded entry in Reception were more likely to: live in areas in the least deprived IDACI quintiles; not be eligible for FSMs; live in the East, South East and South West of England; speak English as a first language; and be White-British. A total of 84.4% (4782) children had SEND provision recorded at entry into school, including 35.1% (1,988) with recorded SEND support and 49.3% (2794) with recorded EHCPs. The N1 cohort had the highest proportion of recorded SEND support at entry (403, 45.0%), whereas the Reception cohort had the highest proportion of recorded EHCPs (1494, 55.3%). Overall, 1,928 (34.0%) pupils’ recorded SEND provision status changed during follow-up, with the majority transitioning from recorded SEND support to EHCP (1,116, 57.9% of changes), followed by no SEND provision to SEND support (307; 15.9%) and no SEND provision to EHCP (217, 11.3%; extended data Table S3). Therefore, we modelled two separate outcomes: time to any recorded SEND provision and time to recorded EHCP.

Sociodemographic comparisons

During follow-up, 95.2% (5399/5669) of pupils had any recorded SEND provision and 72.3% (4098/5670) had an EHCP recorded at least once. Excluding pupils with any recorded SEND provision at entry into school, the median follow-up was 371 days (interquartile range, IQR, 364-1092) for any recorded SEND provision and 1092 days (IQR 371-2191) for recorded EHCP (extended data Table S4). For all cohorts, cumulative probabilities of first recorded SEND provision plateaued by 3–4 years of follow-up, while probabilities of first recorded EHCPs showed step changes beyond this time (extended data Figure S4). The cumulative probability curves derived using unweighted pooled logistic regression (extended data Figure S4) were similar to those estimated using the Kaplan-Meier method (extended data Figure S5). We did not pool cohort results due to heterogeneity across cohort populations and results.

Any recorded SEND provision. In all 3 cohorts, males had a higher cumulative probability of any recorded SEND provision than females, although the difference in the N1 and N2 cohorts did not appear until pupils reached Reception (Figure 2; extended data Table S5). The N2 cohort had the greatest estimated difference in cumulative probabilities between genders (2.13% higher for males at Year 3, 95% CI 1.44, 2.83; Table S6). Pupils in the 20% most and 80% least deprived IDACI groups had similar estimated probabilities of any recorded SEND provision (Figure 3), apart from the N1 cohort, where pupils in the most deprived 20% group had a 1.05% lower cumulative probability of any recorded SEND provision at Year 1 compared with the least deprived 80% (95% CI -1.81, -0.29; Table S6). There was no difference in the cumulative probability of any recorded SEND provision by FSM eligibility for the N1 cohort, some difference during follow-up for the N2 cohort (1.42% higher at Year 3 for pupils who were FSM eligible, 95% CI 0.83, 2.01) and a greater difference for the Reception cohort (Figure 4). Children eligible to receive FSMs who entered state-funded school in Reception had a higher estimated cumulative probability of any recorded SEND provision (3.47% higher than the non-FSM eligible group at Year 3, 95% 2.78, 4.15; Table S6). This difference was present throughout follow-up period.

Figure 2. Predicted cumulative probability (95% confidence intervals) and absolute difference (95% confidence intervals) of first recorded SEND provision (any), by gender (stratified by cohort): estimated using weighted pooled logistic regression with 1000 bootstraps.

Displayed predictions are restricted by limited non-events in some follow-up years. N1 = nursery Year 1; N2 = nursery Year 2; R = Reception; SEND = special educational needs and disability.

Figure 3. Predicted cumulative probability (95% confidence intervals) and absolute difference (95% confidence intervals) of first recorded SEND provision (any), by IDACI group (stratified and cohort: estimated using weighted pooled logistic regression with 1000 bootstraps.

Displayed predictions are restricted by limited non-events in some follow-up years. IDACI = income deprivation affecting children index; N1 = nursery Year 1; N2 = nursery Year 2; R = Reception; SEND = special educational needs and disability.

Figure 4. Predicted cumulative probability (95% confidence intervals) and absolute difference (95% confidence intervals) of first recorded SEND provision (any), by FSM eligibility (stratified and cohort: estimated using weighted pooled logistic regression with 1000 bootstraps.

Displayed predictions are restricted by limited non-events in some follow-up years. FSM = free school meal; N1 = nursery Year 1; N2 = nursery Year 2; R = Reception; SEND = special educational needs and disability.

Recorded EHCP. Males had a higher cumulative probability of recorded EHCP across cohorts, although confidence intervals for probabilities overlapped with females for most years of follow-up (Figure 5; extended data Table S7). The greatest gender difference was in the N2 cohort, where males had a 3.51% increased cumulative probability compared with females at Year 6 (95% CI 1.48, 5.53; extended data Table S8). There were differences in the cumulative probability of EHCPs by IDACI group for the N1 and N2 cohorts (Figure 6). The difference in the N1 cohort appeared at Reception, where pupils living in the most deprived 20% IDACI areas had an 8.94% lower probability of recorded EHCP (95% -13.01, -4.88; extended data Table S8). Pupils in the N2 cohort living in the most deprived 20% IDACI areas had a 3.48% lower cumulative probability of recorded EHCP in Year 6 (95% CI -5.51, -1.45; extended data Table S8). Cumulative probability plots showed consistent differences during follow-up for this cohort. We found no differences in the cumulative probabilities of recorded EHCP for children who were or were not FSM eligible (Figure 7).

Figure 5. Predicted cumulative probability (95% confidence intervals) and absolute difference (95% confidence intervals) of first recorded EHCP, by gender (stratified by cohort): estimated using weighted pooled logistic regression with 1000 bootstraps.

Displayed predictions are restricted by limited non-events in some follow-up years. EHCP = education, health and care plan; N1 = nursery Year 1; N2 = nursery Year 2; R = Reception.

Figure 6. Predicted cumulative probability (95% confidence intervals) and absolute difference (95% confidence intervals) of first recorded EHCP, by IDACI group (stratified and cohort: estimated using weighted pooled logistic regression with 1000 bootstraps.

Displayed predictions are restricted by limited non-events in some follow-up years. EHCP = education, health and care plan; IDACI = income deprivation affecting children index; N1 = nursery Year 1; N2 = nursery Year 2; R = Reception.

Figure 7. Predicted cumulative probability (95% confidence intervals) and absolute difference (95% confidence intervals) of first recorded EHCP, by FSM eligibility (stratified and cohort: estimated using weighted pooled logistic regression with 1000 bootstraps.

Displayed predictions are restricted by limited non-events in some follow-up years. EHCP = education, health and care plan; FSM = free school meal; N1 = nursery Year 1; N2 = nursery Year 2; R = Reception.

Sensitivity analysis

Extending our case definition of children with cerebral palsy to include other paralytic syndromes did not alter the results (extended data Figures S6-S11).

Summary

In this England-wide study of 5,669 children with cerebral palsy, we found that 95% of children had any SEND provision recorded during nursery or primary school, 84% of whom had SEND provision recorded on entry into school. Males consistently showed a higher cumulative probability of any recorded SEND provision compared to females across all cohorts, with differences emerging by Reception. The association between FSM eligibility and SEND provision varied by cohort, with the strongest association observed in the Reception cohort, where FSM eligible children who entered state-funded schools in Reception had a higher probability of recorded SEND provision. Different patterns were found for EHCPs, which 72% of pupils had recorded at least once during follow-up. We found that pupils entering state-funded education in nurseries who lived in the 20% most income-deprived areas had lower cumulative probabilities of EHCPs than those living in the 80% least deprived areas.

Strengths and limitations

To the best of our knowledge, this is one of the first studies to evaluate temporal equity in SEND provision in England. We advanced the current literature in two ways: by looking at SEND provision across primary school, rather than SEND provision at a single point in time; and separately studying EHCPs, which have a different pathway to assignment¹⁵. By focussing on children with cerebral palsy and standardising by health indicators, we reduced heterogeneity in the need for SEND provision in our population. Therefore, we can be more confident that differences are due to disparities in the allocation of SEND provision rather than differences in the underlying need for SEND provision. However, given the broad spectrum of cerebral palsy, its multiple aetiologies and associated conditions², it is unlikely that we fully accounted for differences between pupils, such as in cerebral palsy severity or Gross Motor Function Classification System functioning. Our health indicators were based on diagnoses recorded in hospital admissions data, meaning that we did not capture health or developmental indicators solely recorded in primary or community care. Furthermore, to be included in this study, children were required to have a recording of cerebral palsy in admission records prior to school entry. Given that early identification of cerebral palsy is thought to be socially patterned¹¹, it is likely that our results underestimate the inequities experienced by children with cerebral palsy across their lifespan.

A limitation of our study design is that the results may not be generalisable to a wider population of children requiring SEND provision due to the comparatively clearer aetiology present in this group. Most children in this study had SEND provision recorded at school entry, reflecting the high developmental needs in this population². In addition, due to the relatively low numbers of children with cerebral palsy, we could only study a few of the planned indicators of social position³⁵, and we were unable to examine intersections of disadvantage. There was likely some measurement error in the exposures assessed. Both IDACI and FSM eligibility are derived using benefits data, which underrepresents socioeconomic circumstances in some minoritised ethnic groups due to lower rates of state benefit claims³⁷. We defined FSM eligibility at Reception, which may lead to some measurement error in the N1 and N2 cohorts. Parental education is arguably a better measure of parental resources, as it is thought to reflect additional dimensions of disadvantage³⁸, but this was unmeasured in our administrative data sources. Another important limitation of this study is that a recording of SEND provision in the NPD does not indicate the receipt or quality of such provision. Parents of children with cerebral palsy have indicated a lack of joined up working between education and health services, even where an EHCP has been awarded⁵. In the 2023/24 academic year, His Majesty’s Courts and Tribunals Service reported 21,106 registered appeals in relation to SEND provision (not specific to cerebral palsy), a 55% increase on the previous year³⁹. However, whilst 27% of these appeals were against not securing an EHCP assessment, a much higher proportion (56%) related to the content of EHCP plans that were already in place, highlighting that securing an EHCP is a necessary but insufficient step towards appropriate provision.

Results in context and implications

In parallel to findings from whole-pupil population studies of SEND provision^10,12, we found that males have a higher probability of any recorded SEND provision. Our work further shows that this difference is consistently present by Reception for all cohorts among children with cerebral palsy. We did not find the same pattern for recorded EHCPs, indicating that the gender difference in all SEND provision is driven by SEND support (i.e., support organised by the school). The observed gender difference in SEND support could be attributed to several factors including gender-bias in teacher perceptions of pupil behaviour as hypothesised in previous studies^10,12 Our results add to evidence from Sweden, which has also indicated a gender-bias in a range of interventions for children with cerebral palsy, including physiotherapy and botulinum toxin, suggesting a broader pattern of gender-based disparities in the identification and provision of support for children with cerebral palsy^40,41. An alternative explanation is that the higher prevalence reflects a higher (unmeasured) need in the male population. While the higher incidence of cerebral palsy in males is well documented, there is mixed evidence regarding differences in the severity of the condition and associated comorbidities^42,43. Future research could explore primary SEND need to further disentangle differences in SEND provision among girls and boys with cerebral palsy.

We found that children with cerebral palsy living in the 20% most deprived areas (neighbourhood indicator of socioeconomic deprivation) had a lower prevalence of recorded EHCPs than those living in the other 80%. While this result contrasts with the observed rates of EHCPs reported in the entire pupil population, it agrees with research on pupils with any recorded SEND provision during primary school (used to define a population in need of SEND provision in one study)⁴⁴. This pattern points to inequities in local area resourcing, with wealth advantages in accessing early identification and intervention. We found this pattern among children entering school in nursery, but not Reception, which may reflect differences in populations entering state-funded school at this later stage. This discrepancy could be attributed to capacity constraints – children with higher needs attending school prior to the mandatory age may be dependent on available ‘space’ or the ability of the school to accommodate their needs. This may result in some children with higher needs delaying entry until Reception (or mandatory school age). Additionally, the ability to secure early placement, particularly in special schools or schools with appropriate support, may be influenced by parents' capacity to advocate effectively for their children, advantaging families with more resources or those who are more adept at navigating the education system. Simultaneously, families from less deprived areas may have better access to early intervention services and support, potentially leading to earlier identification of needs and EHCP applications before school entry, which in turn influences need trajectories.

FSM eligibility (a household indicator of socioeconomic deprivation) was associated with an increased probability of any recorded SEND provision, but only for those entering state-funded education in Reception. This was driven by SEND support as we observed no differences for EHCPs. No difference was found in the school-based nursery cohorts, which may reflect the differences in populations accessing these establishments. An association between FSM eligibility and SEND provision has long been recognised, with patterns thought to mostly reflect the higher levels of underlying need among pupils experiencing socioeconomic deprivation (a well-established structural determinant of health)²⁵. Amongst children with cerebral palsy, downstream risk factors from socioeconomic deprivation, including comorbidities, inaccessibility of appropriate healthcare, and other early interventions create barriers to optimal outcomes⁴⁵. These findings underscore the importance of further research to disentangle the complex interplay between socioeconomic deprivation, need, and provision of support for children with cerebral palsy. Evidence suggests that FSM eligible children are less likely to have a recorded EHCP if they live in a more deprived area, highlighting the importance of exploring the intersectional effects of deprivation⁴⁶.

Future research would benefit from extending this study to a wider population of pupils, which would also enable investigations into disadvantage among ethnically minoritised pupils, who face additional structural barriers to accessing appropriate services^47,48. Identifying a larger number of children with a need for SEND provision would enable a more nuanced analysis of inequities in the assignment of SEND provision. This may require the use of assessment-based measures that take developmental age into account and could make use of novel population cohort study-administrative data linkages⁴⁹. Another important component of further research is quantifying the effect of delays in SEND provision on longer-term educational and health outcomes. There is currently a lack of robust evidence on the effectiveness of SEND provision, particularly from trial data. Overall, our findings add weight to calls for a more equitable approach to who receives SEND provision⁵⁰, which is not driven by local area resources or parental advocacy skills. For children with cerebral palsy in particular, families have requested a more joined up referral and advice pathway from the early years, ideally with a key worker¹¹. These considerations are particularly pertinent given the accelerated pressure reported in the SEND system in the last decade, which is now described as having reached a crisis point⁵¹. Ultimately, without extra investment and system change, the SEND system will continue to be unsustainable⁵⁰.