Predicting Sepsis-Induced Acute Kidney Injury (AKI) Using Dynamic Graph Neural Networks on ICU Time-Series Data

Authors

  • Ayobami Emmanuel Mesioye * Department of Cybersecurity, McPherson University, Seriki Sotayo, Ogun State. https://orcid.org/0000-0003-0406-5226
  • Johnson Bisi Oluwagbemi Department of Computer Science, McPherson University, Seriki Sotayo, Ogun State. https://orcid.org/0009-0006-6882-9562
  • Racheal Ogundein Oladoyin Department of Public Health, McPherson University University, Seriki Sotayo, Ogun State.

https://doi.org/10.48314/isti.v2i3.44

Abstract

In Intensive Care Units (ICUs), sepsis-induced Acute Kidney Injury (AKI) is a life-threatening complication associated with high mortality and long-term morbidity. Sequential Organ Failure Assessment (SOFA) score, like other traditional screening tools is often calculated infrequently and fails to capture the complex, time-varying interplay between physiological systems that precedes organ failure. This study introduces a Dynamic Graph Neural Network (DGNN) approach to model the evolving relationships among physiological variables for the early and accurate prediction of sepsis-induced AKI. Unlike traditional time-series models, the DGNN represents each physiological variable such as Heart Rate (HR), lactate, creatinine, vasopressor dose as a node. This dynamically learn the weighted edges that capture the evolving patient pathophysiology at each time step. The DGNN was trained using multivariate time-series extracted from a cohort of adult septic IC patients within the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. The model was tasked with predicting the onset of AKI (Kidney Disease Improving Global Outcomes (KDIGO) criteria) within 6, 12, 24, and 36-hour prediction windows, benchmarked against the static SOFA score and an established deep learning model, the Long Short-Time Memory (LSTM) network. The experimental results show the supremacy in the performance of DGNN across all time horizons. For critical 12-hour prediction window, the model achieved an Area Under the Receiver Operating Characteristic (AUROCs) of 0.89, significantly outperforming LSTM (0.82) and baseline SOFA (0.71). Furthermore, DGNN maintains a robust Area Under the Receiver (AUC) of 0.80 for the 36-hour window, thus provides a much earlier warning than the current methods. Analysis of the learned graph edges revealed clinically relevant insights, such as the increasing influence of vasopressor dose and rising lactate on renal function markers preceding AKI onset. This model offers a more robust and accurate early warning system, reflecting the systemic nature of sepsis and holding significant potential to facilitate timely interventions and improve patient outcomes.

Keywords:

D sepsis, Acute kidney injury, Medical information mart for intensive care IV, Time-series prediction, Critical care

References

  1. [1] Kellum, J. A., Romagnani, P., Ashuntantang, G., Ronco, C., Zarbock, A., & Anders, H. J. (2021). Acute kidney injury. Nature reviews disease primers, 7(1), 52. https://doi.org/10.1038/s41572-021-00284-z
  2. [2] Han, J., Meng, S., & Sun, J. (2025). Multimodal prediction of sepsis-induced acute kidney injury: Integrating CT imaging, clinical data, radiomics, deep features, and nomogram-based risk assessment. Journal of radiation research and applied sciences, 18(3), 101709. https://doi.org/10.1016/j.jrras.2025.101709
  3. [3] Cheungpasitporn, W., Thongprayoon, C., & Kashani, K. B. (2024). Artificial intelligence and machine learning’s role in sepsis-associated acute kidney injury. Kidney research and clinical practice, 43(4), 417. https://doi.org/10.23876/j.krcp.23.298
  4. [4] Vagliano, I., Chesnaye, N. C., Leopold, J. H., Jager, K. J., Abu Hanna, A., & Schut, M. C. (2022). Machine learning models for predicting acute kidney injury: A systematic review and critical appraisal. Clinical kidney journal, 15(12), 2266–2280. https://doi.org/10.1093/ckj/sfac181
  5. [5] Takkavatakarn, K., & Hofer, I. S. (2023). Artificial intelligence and machine learning in perioperative acute kidney injury. Advances in kidney disease and health, 30(1), 53–60. https://doi.org/10.1053/j.akdh.2022.10.001
  6. [6] Lin, Y., Shi, T., & Kong, G. (2025). Acute kidney injury prognosis prediction using machine learning methods: A systematic review. Kidney medicine, 7(1), 100936. https://doi.org/10.1016/j.xkme.2024.100936
  7. [7] Xiao, Z., Huang, Q., Yang, Y., Liu, M., Chen, Q., & Huang, J. (2022). Emerging early diagnostic methods for acute kidney injury. Theranostics, 12(6), 2963. https://doi.org/10.7150/thno.71064
  8. [8] Fan, C., Ding, X., & Song, Y. (2021). A new prediction model for acute kidney injury in patients with sepsis. Annals of palliative medicine, 10(2), 1771778–1772778. https://doi.org/10.21037/apm-20-1117%0A
  9. [9] Shi, J., Han, H., Chen, S., Liu, W., & Li, Y. (2024). Machine learning for prediction of acute kidney injury in patients diagnosed with sepsis in critical care. Plos one, 19(4), e0301014. https://doi.org/10.1371/journal.pone.0301014
  10. [10] Ge, X., Chen, W., Shi, J., Zhang, J., Tai, H., Zhang, Y., & Han, H. (2025). Prediction of moderate-to-severe sepsis-associated acute kidney injury using a dual-timepoint machine learning model: Development, multiregional validation, and clinical deployment study. Journal of medical internet research, 27, e73840. https://doi.org/10.2196/73840
  11. [11] Zhang, L., Li, M., Wang, C., Zhang, C., & Wu, H. (2025). Prediction of acute kidney injury in intensive care unit patients based on interpretable machine learning. Digital health, 11, 20552076241311172. https://doi.org/10.1177/20552076241311173
  12. [12] Li, J., Zhu, M., & Yan, L. (2024). Predictive models of sepsis-associated acute kidney injury based on machine learning: A scoping review. Renal failure, 46(2), 2380748. https://doi.org/10.1080/0886022X.2024.2380748
  13. [13] Luo, X. Q., Yan, P., Zhang, N. Y., Luo, B., Wang, M., & Deng, Y. H. (2021). Machine learning for early discrimination between transient and persistent acute kidney injury in critically ill patients with sepsis. Scientific reports, 11(1), 20269. https://doi.org/10.1038/s41598-021-99840-6
  14. [14] Zhang, L., Wang, Z., Zhou, Z., Li, S., Huang, T., Yin, H., & Lyu, J. (2022). Developing an ensemble machine learning model for early prediction of sepsis-associated acute kidney injury. Iscience, 25(9), 1-14. https://doi.org/10.1016/j.isci.2022.104932
  15. [15] Lee, H. C., Yoon, S. Bin, Yang, S. M., Kim, W. H., Ryu, H. G., Jung, C. W., & Lee, K. H. (2018). Prediction of acute kidney injury after liver transplantation: Machine learning approaches vs. logistic regression model. Journal of clinical medicine, 7(11), 428. https://doi.org/10.3390/jcm7110428
  16. [16] Song, Z., Yang, Z., Hou, M., & Shi, X. (2022). Machine learning in predicting cardiac surgery associated acute kidney injury: A systemic review and meta analysis. Frontiers in cardiovascular medicine, 9, 951881. https://doi.org/10.3389/fcvm.2022.951881
  17. [17] Yun, G., Yi, J., Han, S., Seong, J., Menadjiev, E., Han, H., & Kim, S. (2025). Validation of an acute kidney injury prediction model as a clinical decision support system. Korean journal of nephrology. https://doi.org/10.23876/j.krcp.24.163
  18. [18] Luo, X. Q., Yan, P., Duan, S. B., Kang, Y. X., Deng, Y. H., Liu, Q., & Wu, X. (2022). Development and validation of machine learning models for real-time mortality prediction in critically ill patients with sepsis-associated acute kidney injury. Frontiers in medicine, 9, 853102. https://doi.org/10.3389/fmed.2022.853102
  19. [19] Quan, Z., Han, Z., Zeng, S., Wen, L., Wang, J., Li, Y., & Wang, H. (2025). Stage prediction of acute kidney injury in sepsis patients using explainable machine learning approaches. Frontiers in medicine, 12, 1667488. https://doi.org/10.3389/fmed.2025.1667488
  20. [20] Chaudhary, K., Vaid, A., Duffy, Á., Paranjpe, I., Jaladanki, S., & Paranjpe, M. (2020). Utilization of deep learning for subphenotype identification in sepsis-associated acute kidney injury. Clinical journal of the american society of nephrology, 15(11), 1557–1565. https://doi.org/10.2215/CJN.09330819
  21. [21] Zhou, Y., Feng, J., Mei, S., Zhong, H., Tang, R., Xing, S., & He, Z. (2023). Machine learning models for predicting acute kidney injury in patients with sepsis-associated acute respiratory distress syndrome. Shock, 59(3), 352–359. https://doi.org/10.1097/SHK.0000000000002065

Downloads

Published

2025-09-08

Issue

Vol. 2 No. 3 (2025): Information Sciences and Technological Innovations

Section

Articles

How to Cite

Emmanuel Mesioye, . A., Bisi Oluwagbemi , J., & Ogundein Oladoyin, R. (2025). Predicting Sepsis-Induced Acute Kidney Injury (AKI) Using Dynamic Graph Neural Networks on ICU Time-Series Data. Information Sciences and Technological Innovations, 2(3), 172-186. https://doi.org/10.48314/isti.v2i3.44

Similar Articles

11-16 of 16

You may also start an advanced similarity search for this article.