Al-Mamun Mohammad A, Jeun Ki Jin, Brothers Todd, Asare Ernest O, Shawwa Khaled, Ahmed Imtiaz
Department of Pharmaceutical Systems and Policy, West Virginia University, Morgantown, West Virginia, United States of America.
Department of Pharmacy Practice, University of Rhode Island, Kingston, Rhode Island, United States of America.
PLoS One. 2025 Jul 11;20(7):e0310749. doi: 10.1371/journal.pone.0310749. eCollection 2025.
Acute kidney injury (AKI) can lead to an approximate ninefold increased risk for developing chronic kidney disease (CKD). Despite this, many AKI survivors lack proper nephrology follow-up, highlighting the urgent need to identify patient profiles before onset CKD. Thus, we aimed to develop a patient profiling algorithm to identify clinical phenotypes from AKI to CKD progression.
This retrospective study utilized electronic health records data from 2010 to 2022. We classified AKI into three groups: Hospital Acquired AKI (HA-AKI), Community Acquired AKI (CA-AKI), and No-AKI. We developed a custom patient profiling algorithm by combining network-based community and variable clustering methods to examine risk factors among three groups. The top three clusters were presented using comorbidities and medical procedures network graphs, and matched between two methods to find similarities and dissimilarities.
Among 58,876 CKD patients, 10.2% (5,981) and 11.5% (6,762) had HA-AKI and CA-AKI, respectively. The No-AKI group had a higher comorbidity burden compared to AKI groups, with average comorbidities of 2.84 vs. 2.04. Commonly risk factors observed in both AKI cohorts included long-term opiate analgesic use, atelectasis, history of ischemic heart disease, and lactic acidosis. The comorbidity network in HA-AKI patients was more complex compared to CA-AKI and No-AKI groups with higher number of diagnosis (64 vs 62 vs 55). The HA-AKI cohort had several conditions with higher degree (mean number of edges connected to each diagnosis) and betweenness centrality (bridges connecting different diagnosis clusters) including high cholesterol (34, 91.10), chronic pain (33, 103.38), tricuspid insufficiency (38, 113.37), osteoarthritis (34, 56.14), and removal of GI tract components (37, 68.66) compared to the CA-AKI cohort.
Our proposed patient profiling algorithm successfully identifies AKI phenotypes toward CKD progression, offering a promising approach to identify early risk factors for CKD in improving targeted prevention strategies and reducing healthcare expenditures.
急性肾损伤（AKI）可使患慢性肾脏病（CKD）的风险增加约9倍。尽管如此，许多AKI幸存者缺乏适当的肾脏病学随访，这凸显了在CKD发病前识别患者特征的迫切需求。因此，我们旨在开发一种患者特征分析算法，以识别从AKI到CKD进展过程中的临床表型。
这项回顾性研究利用了2010年至2022年的电子健康记录数据。我们将AKI分为三组：医院获得性AKI（HA-AKI）、社区获得性AKI（CA-AKI）和非AKI。我们通过结合基于网络的社区和变量聚类方法开发了一种定制的患者特征分析算法，以检查三组之间的风险因素。使用合并症和医疗程序网络图展示前三个聚类，并在两种方法之间进行匹配以找出异同点。
在58876例CKD患者中，分别有10.2%（5981例）和11.5%（6762例）患有HA-AKI和CA-AKI。与AKI组相比，非AKI组的合并症负担更高，平均合并症数分别为2.84和2.04。在两个AKI队列中常见的风险因素包括长期使用阿片类镇痛药、肺不张、缺血性心脏病史和乳酸酸中毒。与CA-AKI和非AKI组相比，HA-AKI患者的合并症网络更复杂，诊断数量更多（分别为64、62和55）。与CA-AKI队列相比，HA-AKI队列有几种疾病的度数（连接每个诊断的边的平均数）和中介中心性（连接不同诊断聚类的桥梁）更高，包括高胆固醇（34，91.10）、慢性疼痛（33，103.38）、三尖瓣关闭不全（38，113.37）、骨关节炎（34，56.14）和胃肠道部分切除（37，68.66）。
我们提出的患者特征分析算法成功识别了AKI向CKD进展的表型，为识别CKD的早期风险因素提供了一种有前景的方法，有助于改进针对性预防策略并降低医疗费用。
