Han Shasha, Li Sairan, Yang Yunhaonan, Liu Lihong, Ma Libing, Leng Zhiwei, Mair Frances S, Butler Christopher R, Nunes Bruno Pereira, Miranda J Jaime, Yang Weizhong, Shao Ruitai, Wang Chen
School of Population Medicine and Public Health, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing, China.
State Key Laboratory of Respiratory Health and Multimorbidity, Beijing, China.
Commun Med (Lond). 2024 Jul 11;4(1):139. doi: 10.1038/s43856-024-00563-2.
Current clustering of multimorbidity based on the frequency of common disease combinations is inadequate. We estimated the causal relationships among prevalent diseases and mapped out the clusters of multimorbidity progression among them.
In this cohort study, we examined the progression of multimorbidity among 190 diseases among over 500,000 UK Biobank participants over 12.7 years of follow-up. Using a machine learning method for causal inference, we analyzed patterns of how diseases influenced and were influenced by others in females and males. We used clustering analysis and visualization algorithms to identify multimorbidity progress constellations.
We show the top influential and influenced diseases largely overlap between sexes in chronic diseases, with sex-specific ones tending to be acute diseases. Patterns of diseases that influence and are influenced by other diseases also emerged (clustering significance P > 0.87), with the top influential diseases affecting many clusters and the top influenced diseases concentrating on a few, suggesting that complex mechanisms are at play for the diseases that increase the development of other diseases while share underlying causes exist among the diseases whose development are increased by others. Bi-directional multimorbidity progress presents substantial clustering tendencies both within and across International Classification Disease chapters, compared to uni-directional ones, which can inform future studies for developing cross-specialty strategies for multimorbidity. Finally, we identify 10 multimorbidity progress constellations for females and 9 for males (clustering stability, adjusted Rand index >0.75), showing interesting differences between sexes.
Our findings could inform the future development of targeted interventions and provide an essential foundation for future studies seeking to improve the prevention and management of multimorbidity.
目前基于常见疾病组合频率对多种疾病共存进行的聚类分析并不充分。我们估计了常见疾病之间的因果关系，并绘制出了它们之间多种疾病共存进展的聚类情况。
在这项队列研究中，我们在超过12.7年的随访期内，对超过50万名英国生物银行参与者的190种疾病的多种疾病共存进展情况进行了研究。我们使用一种用于因果推断的机器学习方法，分析了疾病在女性和男性中相互影响的模式。我们使用聚类分析和可视化算法来识别多种疾病共存进展星座图。
我们发现，在慢性病中，最具影响力和受影响的疾病在很大程度上在性别之间存在重叠，而特定性别的疾病往往是急性病。也出现了疾病相互影响的模式（聚类显著性P>0.87），最具影响力的疾病影响多个聚类，而受影响最大的疾病集中在少数几个聚类中，这表明在增加其他疾病发生风险的疾病中存在复杂的机制，而其发生风险因其他疾病而增加的疾病之间存在共同的潜在病因。与单向的多种疾病共存进展相比，双向的多种疾病共存进展在国际疾病分类章节内和章节间都呈现出显著的聚类趋势，这可为未来制定跨专业的多种疾病共存策略的研究提供参考。最后，我们确定了女性的10种多种疾病共存进展星座图和男性的9种（聚类稳定性，调整兰德指数>0.75），显示出有趣的性别差异。
我们的研究结果可为未来有针对性干预措施的开发提供参考，并为未来旨在改善多种疾病共存的预防和管理的研究提供重要基础。
