Background: China has made progress in reducing maternal mortality ratio (MMR), yet county-level spatiotemporal heterogeneity persists. This study aims to identify spatiotemporal disparities in MMR and quantify the impacts of various administrative levels on these disparities.

Methods: We analyzed county-level MMR panel data from 1996 to 2015, employing the spatial Gini coefficient, Anselin Local Moran's I, and Getis-Ord Gi^* to assess spatiotemporal disparities related to spatial inequity and geographic clustering. Additionally, we applied a Bayesian multiscale spatiotemporally varying intercepts (BMSTVI) model to unveil the national temporal trend and multiple sub-national spatial patterns in maternal mortality risk. We further quantified the relative contributions of five sub-national administrative levels using the spatiotemporal variance partitioning index (STVPI).

Results: Results suggested that from 1996 to 2015, the proportion of MMR in counties achieving Sustainable Development Goals (SDGs) increased from 27.05% to 93.40%, yet spatiotemporal disparities remained. The spatial Gini coefficient and geographic clustering analyses indicated temporally varying but spatially stable inequities patterns, highlighting the Spatial Inequity Lock-in (SILI) effect. Hotspot analysis identified sensitive and exemplary counties, underscoring the need for targeted regional interventions. The BMSTVI model indicated a declining trend in MMR risk over 20 years, with the most substantial reduction from 2003 to 2012. While the geographic distribution of high-risk areas remained relatively stable, analyses at finer administrative levels enabled more precise identification of affected locations and improved intervention effectiveness. Finally, the STVPI revealed that spatial effects contributed 83.91% (95% CIs: 78.66%-89.47%) to MMR variations, far exceeding the 11.60% (95% CIs: 7.27%-16.55%) from temporal effects. The contribution from the administrative county-level was the highest (29.15%, 95% CIs: 19.69%-35.06%), followed by contributions from the seven geographical regions (14.10%, 95% CIs: 6.61%-34.06%), rural-urban differences (13.77%, 95% CIs: 4.93%-39.2%), provincial level (12.41%, 95% CIs: 8.06%-16.85%), and city level (11.21%, 95% CIs: 7.53%-13.84%).

Discussion: These findings underscore the crucial need for region-specific, time-sensitive policies to achieve maternal health equity across Chinese counties. This study provides a robust empirical foundation for a multi-tiered adaptive policy framework grounded in systematic spatiotemporal assessment across macro, meso, and micro scales to guide targeted maternal health interventions globally.

Keywords: China; MMR; SDG 3; county level; multiscale effects; small area; spatial inequity lock-in effect; spatiotemporal heterogeneity.

Copyright © 2025 Liao, Yuan, He, Xu, Tang, Xie, Tang, Tang, Zeng, Zhang and Song.

背景： 中国在降低孕产妇死亡率（MMR）方面取得了进展，但在县一级的时空异质性仍然存在。本研究旨在识别MMR的时空差异，并量化不同行政级别对这些差异的影响。

方法： 我们分析了1996年至2015年间县级MMR面板数据，使用空间基尼系数、安塞尔本地莫兰指数（Anselin Local Moran's I）和Getis-Ord Gi^*来评估与空间不平等和地区聚集有关的时空差异。此外，我们应用了贝叶斯多尺度时空变截距模型（BMSTVI），揭示了孕产妇死亡风险的国家时间趋势及多个次国家级的空间模式。进一步地，我们使用时空方差分解指数（STVPI）量化了五个次国家级行政级别对MMR变化的相对贡献。

结果： 结果显示，在1996年至2015年间，实现可持续发展目标（SDGs）的县中孕产妇死亡率的比例从27.05%增加到93.40%，但时空差异仍然存在。空间基尼系数和地理聚集分析表明存在随时间变化但在空间上稳定的不平等模式，突出了“空间不平等锁定”效应。热点分析识别了敏感和典型县，强调了区域针对性干预的必要性。BMSTVI模型显示，在20年内孕产妇死亡风险呈下降趋势，其中从2003年到2012年的减少幅度最大。虽然高风险地区的地理分布相对稳定，但在更细行政级别的分析中可以更准确地识别受影响地区并提高干预效果。最后，STVPI表明空间效应贡献了83.91%（95% CI: 78.66%-89.47%）的孕产妇死亡率变化，远超时间效应所占的11.60%（95% CI: 7.27%-16.55%）。行政县一级贡献最高（29.15%，95% CI: 19.69%-35.06%），其次是七个地理区域（14.10%，95% CI: 6.61%-34.06%）、城乡差异（13.77%，95% CI: 4.93%-39.2%）、省级水平（12.41%，95% CI: 8.06%-16.85%）和市级水平（11.21%，95% CI: 7.53%-13.84%）。

讨论： 这些发现强调了制定针对特定地区、时间敏感的政策以实现中国县一级孕产妇健康公平的重要性。本研究为一个基于系统时空评估的多层次适应性政策框架提供了坚实的实证基础，该框架涵盖宏观、中观和微观尺度，以指导全球针对性的孕产妇健康管理干预。

关键词： 中国；MMR；SDG 3；县一级；多尺度效应；小区域；空间不平等锁定效应；时空异质性。