Evaluation of linkage between health, education and social care administrative data for 21 million children in England
This study evaluates the linkage of health, education, and social care data for 21 million children in England, revealing that while over 90% of recent cohorts are successfully linked, significant under-representation persists among children from deprived areas, London residents, and minoritised ethnic groups, highlighting the critical need for researchers to account for these demographic biases in their analyses.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
Imagine trying to understand the health of a nation's children by looking at a single, massive map. In England, this map is being drawn not by surveying every family, but by stitching together three different sets of official records: the medical notes from hospitals and clinics, the attendance and achievement logs from schools, and the files kept by social care services. For decades, these records have lived in separate buildings, guarded by different rules. But researchers have been working to link them, creating a powerful tool called ECHILD. This system allows scientists to see how a child's health, their time in the classroom, and their family's social circumstances intertwine over a lifetime. The hope is that by seeing the whole picture, we can spot problems earlier and design better support for the millions of young people growing up in England. However, for this map to be useful, it must be accurate. If the map misses entire neighborhoods or specific groups of people, the conclusions drawn from it could be misleading, potentially leaving the most vulnerable children without the help they need.
A team of researchers recently set out to check the quality of this massive digital map. They wanted to know exactly how many children were successfully connected across these different records and, just as importantly, who was missing. They looked at data covering 21 million children born between 1984 and 2023. The process involved matching a child's school record to their medical record using secure, anonymous identifiers. The results showed that the system is remarkably effective at its core task. For children born in recent years, more than 90 percent of those with a hospital birth record were successfully linked to a school record. Similarly, about 88 percent of children who were in school in the early 2000s were found to have a corresponding medical record. In total, the system now holds linked information on 21 million individuals, creating a resource that spans over 350 million years of observation.
Yet, the study revealed that this map is not perfectly complete, and the missing pieces are not random. The researchers found that the likelihood of a child being included in the linked data depends heavily on who they are and where they live. Children living in London, those from more deprived areas, and those from ethnic minority backgrounds were less likely to have their records successfully linked compared to their peers. For instance, children recorded as having a Black or Asian ethnicity, or those living in the most economically disadvantaged neighborhoods, appeared in the linked data less often than White children or those in wealthier areas. This pattern held true across different age groups and time periods. The reasons for these gaps are varied. Some children may have opted out of having their health data used for research, while others might have attended private schools that are not fully captured in the national database, or they may have moved to England after their early childhood without a prior medical record in the system.
The researchers also investigated whether the system was accidentally connecting the wrong people, such as linking one child's school record to another child's medical file. They found that false matches were extremely rare, occurring in less than one percent of cases. When they did happen, they were more common among children with missing or unclear information on their records, such as those without a fixed address or with incomplete demographic details. This suggests that the errors are usually due to poor data quality rather than a flaw in the matching technology itself. The study also highlighted a specific historical quirk: children born between 1997 and 2002 were less likely to be linked because, during those years, babies were not automatically given a unique national health number at birth, leading to confusion in the records that made linking difficult.
The significance of these findings lies in what they tell us about the limits of using big data to understand society. While the ECHILD system is a powerful tool for generating evidence to improve child health, it does not represent every single child equally. The groups that are most likely to be under-represented are often the very groups that face the greatest health and social challenges. If researchers use this data without accounting for these gaps, they might mistakenly believe that certain health issues are less common in deprived areas or among minority groups, simply because those children are missing from the dataset. The authors emphasize that understanding who is missing is just as critical as understanding who is present. By knowing the boundaries of their map, scientists can adjust their methods to ensure that the stories they tell about children's health are fair and accurate, rather than inadvertently silencing the voices of those who need to be heard the most.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.