Abstract Background Rapid urbanization, environmental degradation, and socioeconomic inequalities are reshaping exposure patterns and vulnerability profiles for respiratory diseases worldwide. Although extensive literature documents individual urban and environmental determinants, decision-makers still lack integrated, transparent, and reproducible frameworks to synthesize heterogeneous evidence and prioritize risk factors in a policy-relevant manner. Methods A PRISMA-aligned review (2014–2024) synthesized 43 studies, extracting and NLP-harmonizing 211 risk-factor entries into 29 standardized factors across seven categories using rule-based cleaning and semantic similarity checks. Factors were scored through a multi-criteria framework (Frequency, Impact, Applicability, Contextuality), with within-study normalization ensuring weights summed to 100%, followed by cross-study aggregation, LOCO robustness analyses, and descriptive profiling. Quality-control checks addressed duplicate factor labeling and consistency across studies, while a cluster-robust OLS model was applied as an ex-post validation of internal coherence, not for causal inference. Results Pollution-related exposures, urban form and green-space conditions, and climate-related determinants emerged as the most consistently prioritized configurations, while living conditions showed lower weights. Rankings were highly stable across robustness checks (Spearman ρ ≈ 0.96–0.99). Leave-one-criterion-out (LOCO) analyses further confirmed this stability, showing only limited rank shifts when each scoring dimension was removed sequentially. Top-k overlap remained high across LOCO scenarios, indicating that prioritization was not driven by any single criterion. Ex-post cluster-robust OLS validation indicated that Impact and Contextuality were the main drivers of prioritization, whereas Frequency was marginal and Applicability was not statistically significant. Discussion This study presents a reproducible prioritization framework that emphasizes severity and contextual relevance rather than causal estimation. The resulting patterns align with established urban health frameworks highlighting the combined roles of the built environment, environmental hazards, and social inequalities. Conclusions This study presents a reproducible decision-support framework for prioritizing urban health determinants, integrating systematic review, NLP-assisted harmonization, multi-criteria weighting, and econometric validation, with potential extensions toward predictive modeling, economic evaluation, and dynamic policy monitoring.
Erraji et al. (2026) studied this question.