Urban structure alone consistently recovered the relative spatial distribution of dengue susceptibility in unseen years, with year-specific Spearman correlations ranging from 0.49 to 0.76. Structural predictors achieved stronger mean prospective spatial ranking than recent epidemiological history, while combining both information sources produced little additional gain. Temporal generalization collapsed when year-specific spatial correspondence was disrupted (null mean Spearman = 0.0048; empirical $p=0.002$), but was reproduced when coherent place-level susceptibility trajectories were preserved. Cross-classification models retained 93.79\% of within-classification ranking performance after matching training-set size, although the corresponding transportability gap was not unusual relative to random partitions of epidemic years ($p=0.8024$). These findings support a persistent place-based spatial hierarchy of dengue susceptibility that remains informative across markedly different epidemic conditions. Measured urban structure provides an informative predictive representation of this hierarchy, but does not constitute a complete or causal explanation of its persistence. Persistent susceptibility mapping may therefore complement short-term epidemiological surveillance by supporting longer-term, place-based urban health planning.

翻译：暂无翻译