The operational scope was adults and older adolescents aged 17–19 years, with post-acute recovery defined as seven days to five years after injury. Sources describing paediatric, adult, mixed-age or unstated age populations were retained as explicit metadata categories. Acute neurosurgery, resuscitation fluids, osmotherapy, acute neuroprotection and intensive-care interventions were excluded from the post-acute rehabilitation synthesis.

Researchers searched English-language PubMed records from database inception through 21 August 2026 using seven complementary retrieval strands covering rehabilitation and function, cognitive rehabilitation, physical symptoms, sleep and vestibular care, mental health and behavioural recovery, participation and social recovery, care delivery, implementation and access, and outcome measurement or prognosis. The broad probes returned 202,254 overlapping hits, while the high-level retrieval paths returned 10,129 overlapping hits. These were overlapping retrieval counts rather than denominators for independent studies.

Deduplication produced 2,546 unique PubMed identifiers. EFetch supplied parseable XML metadata for 2,529 records, while 17 requested identifiers remained in a missing-record audit. A prespecified rule-based score combined evidence type, recency, domain relevance, recovery and population signals and a direct-TBI title signal, with penalties for animal, paediatric-only, acute or surgical and biomarker-only records. The score organized the parsed corpus into a priority subset of 220 records, and a separate deterministic calculation reproduced the stored score for all 2,529 records without mismatch.

The illustrative registry comprised 50 PubMed-indexed sources published from 2020 through 2026: 10 guidelines or consensus documents, 16 intervention syntheses, 10 implementation or service-evidence sources, 13 measurement or prognostic-evidence sources and one protocol or methods source. The protocol was retained as a methods reference and was not interpreted as completed effect evidence. No numerical quota was imposed by recovery domain. Each source received one title/abstract-audited primary recovery domain, one nominal source-role category and a source-specific selection rationale.

Source-level validation was conducted without prospective clinical validation. Two reviewers independently assessed 200 records sampled equally across four priority-score strata without access to the automated classifications. Their pre-consensus agreement was 190 of 200, or 95.0%, and all 10 disagreements were resolved by consensus. Specialists in neurosurgery, neurology, neurorehabilitation and psychiatry then reviewed all 50 curated sources and confirmed the prespecified source-role, primary-domain and clinical-context classifications.

The registry contained 12 sources in mental health and behaviour, 10 in multidisciplinary rehabilitation, nine in symptoms, sleep and physical recovery, seven in cognitive rehabilitation, seven in implementation and access, four in participation and social recovery and one in outcome measurement and prognosis. These counts describe the composition of the curated registry, not the importance or strength of evidence for any domain.

Reporting of clinical context was uneven across the 50 sources. Age language was absent from 23, while 20 contained adult language, five contained mixed adult and paediatric or adolescent language and two contained paediatric or adolescent language. General or pre-existing comorbidity language appeared in one source. Injury severity was stated in 20 sources: 13 referenced mild TBI or concussion and seven moderate-to-severe TBI; severity was unstated in 30. Recovery-time language appeared in 15 sources. Care setting was not stated in 38; seven mentioned rehabilitation or transition settings, two digital or remote settings, two vocational or education settings and one outpatient or community setting.

No molecular or biochemical marker, cell-type or brain-region signal appeared in the curated title and abstract metadata. The authors described these zero counts as reporting results, not evidence that the corresponding biology is clinically unimportant. The analyses were descriptive and did not calculate treatment effects, certainty ratings, risk-of-bias assessments, causal estimates, patient-level predictions, net benefit or treatment-effect direction.

Important limitations constrain interpretation. The search used PubMed and English-language records, and the 50-source registry was curated rather than derived from a completed multi-database, dual-reviewer full-text systematic review. Title and abstract metadata cannot establish risk of bias, overlap between primary studies, treatment-effect direction, effects of prior treatment, outcome harmonization, certainty of evidence, net benefit or patient-level prognosis. TBI-RECOVER is therefore a transparent structure for locating and describing high-level post-acute TBI recovery evidence, not a clinical decision algorithm, treatment recommendation or validated prognostic system.

Subject of Research: Architecture of a curated evidence registry for post-acute traumatic brain injury recovery

Article Title: Post-acute traumatic brain injury recovery: architecture of an illustrative curated evidence registry

Article References: Cao, F., Xu, L., Wang, W., Yu, T., & Xu, G. (2026). Post-acute traumatic brain injury recovery: architecture of an illustrative curated evidence registry. Frontiers in Neurology, 17, Article 1979438. https://doi.org/10.3389/fneur.2026.1979438

Image Credits: AI Generated

DOI: 10.3389/fneur.2026.1979438

Keywords: traumatic brain injury, post-acute recovery, evidence registry, rehabilitation, clinical decision context, PubMed, care settings, outcome measurement, implementation, Post-acute, traumatic, brain