The team, led by Luiz Jupiter Carneiro de Souza of the Oswaldo Cruz Foundation in Brasília together with colleagues at institutions including Rio de Janeiro State University, the University of Sorocaba, and the Federal University of Tocantins, has published a detailed protocol for a scoping review in the journal Frontiers in Drug Safety and Regulation. A scoping review does not pool results into a single statistical estimate the way a meta-analysis would. Instead, it systematically maps the existing landscape of research, identifying what has been studied, how, and where, and exposing the gaps that remain. The protocol, registered on the Open Science Framework, lays out a rigorous roadmap for charting how public health databases, from electronic health records to insurance claims and disease registries, have been deployed to detect signs of autism before or in support of a formal clinical diagnosis, in both children and adults, anywhere in the world.

The technical logic behind the approach is compelling. Electronic health records capture longitudinal clinical information generated during everyday care, administrative or claims databases track billing and healthcare utilization, and disease registries and surveillance systems systematically collect data on specific conditions. Because these systems record symptoms, healthcare utilization patterns, and developmental indicators long before anyone writes down a diagnosis, they hold the raw material for algorithms that can identify autism-related patterns pre-diagnostically. The evidence base already offers striking proof of concept. One study using electronic medical records found that 81 percent of cases flagged by an identification algorithm were later confirmed as, or judged probable or possible, autism diagnoses. Large pediatric cohort analyses built on health system data have produced predictive algorithms estimating an autism prevalence of 1.32 percent, demonstrating the potential of these sources for epidemiological surveillance and service planning.

Longitudinal analyses reinforce the promise further. Research drawing on large integrated health systems has traced distinct trajectories of physical and mental health conditions among autistic youth compared with non-autistic peers, revealing higher prevalence of obesity, neurological conditions, anxiety, and attention-deficit/hyperactivity disorder across developmental stages. Such findings show that routinely collected data can do more than flag a single condition; they can build risk profiles over time, supporting systematic monitoring of autistic populations and informing how health services plan and allocate care.

Yet the same data sources come with serious caveats. Ethical concerns around privacy and autonomy loom large whenever population-scale health records are analyzed. Data structures vary enormously across systems, coding practices differ between institutions and countries, and technical limitations in data integration and infrastructure management can undermine the quality, comparability, and scalability of any database-driven approach. A recent scoping review that synthesized evidence from 1,702 studies and 163 cohorts using health databases in autism research confirmed both the breadth of applications, including cohort construction and epidemiological analysis, and the persistent weaknesses: missing data, difficulties in case identification, and inconsistencies in data linkage across systems.

That earlier review also exposed a glaring geographical imbalance. The overwhelming majority of database-driven autism studies originate from high-income countries, a pattern the new protocol’s authors attribute to structural inequalities in digital health infrastructure, interoperability, data governance frameworks, and research investment. Crucially, the previous review focused on the characteristics of study populations rather than on the databases themselves, leaving unmapped the question of which public health databases exist, how they are structured, and how they are operationalized for early identification in different health system contexts. The new review is designed to fill precisely that hole.

Methodologically, the protocol follows the framework of the Joanna Briggs Institute and will be reported against the PRISMA-ScR checklist for scoping reviews. A comprehensive search strategy, developed and validated with an experienced information specialist librarian, will span MEDLINE via PubMed, EMBASE, Scopus, PsycINFO, Web of Science, and LILACS via the Virtual Health Library, the latter chosen to capture Latin American evidence alongside global literature. Gray literature from ProQuest Dissertations and Theses will be searched to reduce publication bias and surface methodological innovations not yet in peer-reviewed journals. Reference lists of included studies, expert consultation, and citations from prior reviews will supplement the searches. The review will run between May and September 2026, with all searches performed within a predefined window to ensure consistency and reproducibility.

Eligibility is deliberately broad on participants, including studies of individuals diagnosed with or suspected of having autism regardless of age, gender, or race, and even studies that avoid diagnostic codes altogether if they use alternative indicators derived from routinely collected data, such as symptom patterns or healthcare utilization, with the explicit aim of identifying or predicting autism. Genetic and biomarker-based databases are excluded, a boundary the authors justify by noting that such data are typically generated in controlled research or specialized clinical settings rather than being routinely available at population level, so their inclusion would not align with the goal of mapping scalable, health system-based strategies. Eligible designs include predictive modelling studies, validation studies of identification algorithms, screening approaches, and database-driven detection strategies, while purely epidemiological descriptions without an identification component are excluded.

Execution will be equally disciplined. Two trained reviewers will independently screen titles and abstracts using Rayyan, a web-based application for study selection, after automatic duplicate removal, and full-text screening will proceed in Microsoft Excel with exclusion reasons systematically recorded. Data extraction on a standardized, pilot-tested form will capture the type and characteristics of each database, its data sources and structure, the methods used for autism identification, the stage of identification from screening through confirmed diagnosis, outcomes such as accuracy and feasibility, and the geographical and healthcare system context. Reviewers must reach at least 80 percent agreement in calibration exercises before each stage. Analysis will combine structured narrative synthesis with descriptive statistics, categorizing databases by type, geographic origin, and healthcare system level, stratifying studies by World Bank country income classifications to probe global disparities, and grouping studies by primary purpose, whether early identification, predictive modelling, algorithm validation, or surveillance-related detection.

The stakes extend well beyond autism research. By mapping which databases are used, where, and for what, the review aims to illuminate broader inequalities in digital health infrastructure and research capacity, and to inform future efforts toward standardization, interoperability, and equitable development of database-driven identification strategies across diverse healthcare contexts. The authors acknowledge limitations, including heterogeneity in how studies report on databases and the relative novelty of early identification practices, which complicates search strategy design; they plan to mitigate these through robust procedures and a team trained in applying the eligibility criteria. Because the review uses only published scientific articles, ethical approval is not required, and findings will be disseminated through conference presentations and peer-reviewed publications. If the map they produce is as revealing as the protocol suggests, it could mark a turning point in how health systems worldwide spot autism sooner, using the data they already collect every day.

Subject of Research: Use of public health databases for the early identification of autism spectrum disorder

Article Title: Use of public health databases for the early identification of Autism spectrum disorder: a scoping review protocol

Article References: Souza, L. J. C. D., Nagem Lopes, L. P., Del Grossi Moura, M., Sasaki Zeredo, K. M., Carrillo, J. F. S., Lucas Júnior, J., Figueiró, M. F., Fernandes Neto, J. A. S., Lopes, L. C., & Rodrigues, W. (2026). Use of public health databases for the early identification of Autism spectrum disorder: a scoping review protocol. Frontiers in Drug Safety and Regulation, 6, Article 1816967. https://doi.org/10.3389/fdsfr.2026.1816967

Image Credits: AI Generated

DOI: 10.3389/fdsfr.2026.1816967

Keywords: autism spectrum disorder, public health databases, electronic health records, early identification, scoping review, predictive modelling, epidemiological surveillance, health data, diagnostic delay, global health disparities, Joanna Briggs Institute, PRISMA-ScR