Body Mass Index (BMI) remains a widely used indicator in public health for monitoring population-level weight trends and assessing risk factors associated with chronic disease. However, BMI data recorded in Electronic Health Records (EHRs) is often fragmented and incomplete. Up to 74% of potential adult records can be missing, creating significant blind spots that limit our ability to effectively understand obesity and plan public health strategies.
The study, led by Prof. Ashley Akbari’s team based in Population Data Science at Swansea University and directed by Fatemeh Torabi, addressed this challenge in their work, “Bridging the gap.” By developing a new methodology to link and clean data from multiple health sources, they advanced BMI data coverage across 20 years by retrieving and standardising over 53 million BMI records, utilising the Secure Anonymised Information Linkage (SAIL) Databank, the national trusted research environment (TRE) for Wales.
Prof. Ashley Akbari highlighted the significance of the study:
“This study exemplifies how the SAIL Databank enables population-scale research that is both methodologically rigorous and societally relevant. By securely linking and curating complex data, SAIL provides a trusted foundation for studies that inform public health policy and accelerate progress towards data-driven, patient-centred research.”
Fatemeh Torabi, senior author of the study, reflected on the methodological depth and human-centred motivation underpinning the research:
“At its core, this work is about restoring completeness and integrity to the data that represent people’s health journeys. From a statistical perspective, integrating BMI data across diverse healthcare systems is an intricate problem involving uncertainty, bias, and heterogeneity. Our methodological approach sought to reconcile these complexities without losing sight of the individuals behind the data. By improving the reliability of these records, we can generate insights that are both scientifically robust and meaningful to public health.”
We spoke with the lead author Dr. Jeanne Childs about the publication to learn more about their study.
1. Your paper highlights the “gap” in population-level BMI data. What is the biggest barrier researchers face when trying to use existing electronic health records (EHRs) to understand weight trends?
Dr. Childs the lead author answered: one of the key barriers is that BMI-related components—such as BMI value, category, height, and weight—are not routinely or consistently recorded in primary or secondary care unless there is a specific clinical need. As a result, there is a significant amount of missing or incomplete data, limiting our ability to generate accurate, population-wide insights into weight trends and related health outcomes.
2. You retrieved 53.4 million BMI records. What is the most innovative feature of your methodology that allowed you to achieve this scale and improve data quality so significantly?
The most innovative aspect of our approach was the integration and harmonisation of multiple data sources within the SAIL Databank. By linking four key population-level sources—general practice events, hospital admissions, maternity records, and children’s community health records, we were able to compile a large-scale, high-quality research ready data asset (RRDA) of BMI measurements across diverse settings and demographics. This multi-source linkage was critical in enhancing both the completeness and reliability of the data.
3. With this new, harmonised data asset now available, what is the next step for your team in building on this new data asset?
The next step is to make this harmonised BMI resource accessible to other research teams interested in studying population health in Wales. By providing a ready-to-use dataset, we aim to support and accelerate further research without requiring teams to independently extract, link, and clean raw BMI data. This will foster collaborative research, enable more consistent use of BMI data across studies, and ultimately enhance public health insights and policy development.
Using anonymised, individual-level linked data from the SAIL Databank, this new methodology achieved specific, measurable improvements in data coverage and is now a blueprint that can be adopted by other secure research environments (TREs) globally that hold similar linkable data.
Read the full publication in BMJ Open.