Definitions and Severity/Risk Prediction Tools for Bronchopulmonary Dysplasia: A Scoping Review
Aditya Hemendra Bhatt, Somashekhar Marutirao Nimbalkar, Lalan K. BhartiABSTRACT
Background
Bronchopulmonary dysplasia (BPD) remains a major complication of extreme prematurity, but diagnosis and severity classification have become increasingly challenging in contemporary neonatal respiratory care. Changing respiratory support practices and survival patterns have exposed limitations in older oxygen‐centric definitions, while prognostic tools have emerged to estimate BPD severity or BPD or death risk earlier in the clinical course.
Objective
To map post‐2016 evidence on BPD definitions, severity classifications, and severity/risk prediction tools in preterm infants, and to summarise differences in construction, classifiability, healthcare‐context transportability, and prediction of clinically relevant outcomes.
Methods
This scoping review followed PRISMA‐ScR reporting guidance and the JBI Population‐Concept‐Context framework. PubMed/MEDLINE, Embase, Web of Science, and Google Scholar were searched from January 1, 2017 to February 28, 2026, with citation chaining of eligible articles. No language restriction was planned at search level, but only English full‐text articles were included because of feasibility. Two reviewers independently screened titles/abstracts and full texts; disagreements were resolved by consensus with a third author. Data were charted using a standardised framework and synthesised descriptively. Formal risk‐of‐bias appraisal was not performed because this was a scoping review. The review protocol was not registered, and no external funding was received.
Results
Eighteen studies met inclusion criteria after screening. Evidence clustered into conceptual revision of BPD definitions, comparative validation studies of newer definitions against short‐ and long‐term outcomes, and severity/risk prediction tools including online estimators, machine‐learning adaptations, and nomogram‐based models. Newer support‐mode‐based definitions generally improved classifiability compared with oxygen‐centric definitions and showed outcome gradients in representative cohorts; for example, one NRN definition correctly predicted death or serious respiratory morbidity in 81% of infants, while estimator performance varied by timing and cohort, with reported C‐statistic/AUC values ranging from approximately 0.67 to 0.91 in key prediction‐tool studies.
Conclusions
Post‐2016 BPD research shows a shift from oxygen‐centred criteria toward contemporary support‐based classification and dynamic risk estimation. Current evidence supports careful use‐case‐specific selection of definitions and cautious local validation of prognostic tools rather than a universal single best definition or estimator. Future work should harmonise outcome targets, strengthen external validation and calibration, and develop multidimensional, phenotype‐informed classification systems.