DECISION-AWARE TWO-STAGE CLASSIFICATION FOR EARLY SEPSIS PREDICTION IN IMBALANCED ICU DATA

Ali Aslan; Emin Demir

TR EN

DENGESİZ YOĞUN BAKIM ÜNİTESİ VERİLERİNDE ERKEN SEPİS TAHMİNİ İÇİN KARAR DUYARLI İKİ AŞAMALI SINIFLANDIRMA

Öz

Yoğun bakım ünitelerinde sepsisın erken teşhisi mortaliteyi azaltmak için kritik öneme sahip olmakla birlikte, klinik veri setlerindeki sınıf dengesizliği sepsis sinyallerinin baskın sağlıklı hasta gürültüsü içinde kaybolmasına neden olmaktadır. Bu çalışma, geleneksel modellerin sıklıkla gözden kaçırdığı belirsizlik bölgelerindeki zor vakaları tespit etmek amacıyla karar duyarlı iki aşamalı bir sınıflandırma çerçevesi önermektedir. Küresel bir XGBoost sınıflandırıcısının ardından veri odaklı belirsizlik aralığını [0.037–0.48] hedefleyen maliyet duyarlı bir iyileştirme modeli uygulanmaktadır. Veri seti, PhysioNet 2019 kapsamında 14.057 hastadan elde edilen 40.336 saatlik YBÜ kaydını kapsamakta olup test setinde 234 sepsis vakası yer almaktadır. Entegre sistem sepsis yakalama oranını %51'den %97'ye yükselterek yanlış negatif sayısını 115'ten 6'ya düşürmüştür. Hassasiyet-Geri Çağırma AUC değeri 0.268'den 0.515'e yükselerek gerçek bir ayrım gücü iyileşmesini doğrulamıştır. SHAP analizi önemli bir strateji değişikliğini ortaya koymuştur: küresel model makro göstergelere odaklanırken, iyileştirme modeli hemodinamik ortalamalar ve volatilite ölçütlerini ön plana çıkarmıştır. Bulgular, belirsiz vakalar için özerk bir ikinci görüş mekanizmasının klinik yanlış negatifleri belirgin biçimde azalttığını göstermektedir. Bu yaklaşım, gerçek zamanlı klinik karar destek sistemlerine entegrasyon için şeffaf ve verimli bir yol sunmaktadır.

Anahtar Kelimeler

DECISION-AWARE TWO-STAGE CLASSIFICATION FOR EARLY SEPSIS PREDICTION IN IMBALANCED ICU DATA

Öz

Early sepsis detection is vital for reducing ICU mortality, yet clinical class imbalance masks critical sepsis signals within the dominant healthy population noise. This study proposes a decision-aware two-stage framework to identify hard cases in uncertainty zones that traditional models frequently miss. A global XGBoost classifier is followed by a cost-sensitive refinement model targeting samples within a data-driven uncertainty interval [0.037–0.48]. The dataset comprises 40,336 hourly ICU records from 14,057 patients (PhysioNet/Computing in Cardiology Challenge 2019), with 234 sepsis cases in the test set (8.3% prevalence). The integrated system improves sepsis recall from 51% to 97%, reducing false negatives from 115 to 6. Precision-Recall AUC improved from 0.268 to 0.515 against a random baseline of 0.083, confirming genuine discriminative improvement. SHAP analysis revealed a strategic feature shift: the global model relied on macro-indicators such as last recorded temperature and mean respiratory rate, while the refinement model re-prioritised haemodynamic and respiratory mean values alongside volatility measures, demonstrating capacity to detect physiological instability in insidious sepsis cases. Providing an autonomous second opinion for ambiguous cases significantly reduces clinical false negatives. This approach offers a transparent and computationally efficient pathway for integrating advanced analytics into real-time clinical decision support systems.

Anahtar Kelimeler

Kaynakça

Abbas, Ghulam Husain, Palash Sen, Oviya Anjali Giri, and Nawaid Hussain Khan. 2025. ‘Artificial Intelligence-Based Predictive Modeling for Early Detection of Sepsis in Hospitalized Patients: A Systematic Review and Meta-Analysis’. Critical Care Explorations 7(12):e1360. doi:10.1097/CCE.0000000000001360.
AbuHaweeleh, Mohannad N., Adiba Tabassum Chowdhury, Mehrin Newaz, Purnata Saha, Khandaker Reajul Islam, Jaya Kumar, Muhammad E. H. Chowdhury, and Shona Pedersen. 2025. ‘Sepsis Mortality Prediction Using Machine Learning and Deep Learning - a Systematic Review’. BMC Medical Informatics and Decision Making 26(1):16. doi:10.1186/s12911-025-03286-z.
Akyüz, Melike, Yunus Doğan, Atakan Koçyiğit, and Ayşe Pınar Miran. 2024. ‘Prediction of Sepsis for the Intensive Care Unit Patients with Stream Mining and Machine Learning’. Pamukkale Üniversitesi Mühendislik Bilimleri Dergisi 30(3):354–65. doi: 10.5505/pajes.2023.84899
Al-Mualemi, Bilal Yaseen, and Lu Lu. 2021. ‘A Deep Learning-Based Sepsis Estimation Scheme’. IEEE Access 9:5442–52. doi:10.1109/ACCESS.2020.3043732.
Chanci, Daniela, Jocelyn R. Grunwell, Alireza Rafiei, Stephanie R. Brown, Michael J. Ripple, Natalie R. Bishop, Prakadeshwari Rajapreyar, Lisa M. Lima, and Rishikesan Kamaleswaran. 2025. ‘Machine Learning Model for Daily Prediction of Pediatric Sepsis Using Phoenix Criteria’. Pediatric Research 1–7. doi:10.1038/s41390-025-04221-8.
Chen, Tianqi, and Carlos Guestrin. 2016. ‘XGBoost: A Scalable Tree Boosting System’. Pp. 785–94 in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’16. New York, NY, USA: Association for Computing Machinery. Doi: 10.1145/2939672.2939785
Delahanty, Ryan J., JoAnn Alvarez, Lisa M. Flynn, Robert L. Sherwin, and Spencer S. Jones. 2019. ‘Development and Evaluation of a Machine Learning Model for the Early Identification of Patients at Risk for Sepsis’. Annals of Emergency Medicine 73(4):334–44. doi:10.1016/j.annemergmed.2018.11.036.
D’Onofrio, Valentino, Dries Heylen, Murih Pusparum, Inge Grondman, Johan Vanwalleghem, Agnes Meersman, Reinoud Cartuyvels, Peter Messiaen, Leo A. B. Joosten, Mihai G. Netea, Dirk Valkenborg, Gökhan Ertaylan, and Inge C. Gyssens. 2022. ‘A Prospective Observational Cohort Study to Identify Inflammatory Biomarkers for the Diagnosis and Prognosis of Patients with Sepsis’. Journal of Intensive Care 10(1):13. doi:10.1186/s40560-022-00602-x.

Fawcett, Tom. 2006. ‘An Introduction to ROC Analysis’. Pattern Recognition Letters 27(8):861–74. doi:10.1016/j.patrec.2005.10.010.
Friedman, Jerome H. 2001. ‘Greedy Function Approximation: A Gradient Boosting Machine.’ The Annals of Statistics 29(5):1189–1232. doi:10.1214/aos/1013203451.
Gao, Jiayi, Yuying Lu, Negin Ashrafi, Ian Domingo, Kamiar Alaei, and Maryam Pishgar. 2024. ‘Prediction of Sepsis Mortality in ICU Patients Using Machine Learning Methods’. BMC Medical Informatics and Decision Making 24(1):228. doi:10.1186/s12911-024-02630-z.
Giannini, Heather M., Jennifer C. Ginestra, Corey Chivers, Michael Draugelis, Asaf Hanish, William D. Schweickert, Barry D. Fuchs, Laurie Meadows, Michael Lynch, Patrick J. Donnelly, Kimberly Pavan, Neil O. Fishman, C. William III Hanson, and Craig A. Umscheid. 2019. ‘A Machine Learning Algorithm to Predict Severe Sepsis and Septic Shock: Development, Implementation, and Impact on Clinical Practice*’. Critical Care Medicine 47(11):1485. doi:10.1097/CCM.0000000000003891.
Goh, Kim Huat, Le Wang, Adrian Yong Kwang Yeow, Hermione Poh, Ke Li, Joannas Jie Lin Yeow, and Gamaliel Yu Heng Tan. 2021. ‘Artificial Intelligence in Sepsis Early Prediction and Diagnosis Using Unstructured Data in Healthcare’. Nature Communications 12(1):711. doi:10.1038/s41467-021-20910-4.
Gray, Authia P., Erin Chung, Rebecca L. Hsu, Daniel T. Araki, Anna Gershberg Hayoon, Nicole Davis Weaver, Lucien R. Swetschinski, Eve E. Wool, Chieh Han, Tomislav Mestrovic, Kevin S. Ikuta, Nasir Abbas, Madineh Abbasi, Samar Abd ElHafeez, Wakgari Mosisa Abdisa, Meriem Abdoun, Auwal Abdullahi, Mesfin Abebe, Asrat Agalu Abejew, Alemwork Abie, Hassan Abolhassani, Hana J. Abukhadijah, Meshack Achore, Lisa C. Adams, Kamoru Ademola Adedokun, Ridwan Olamilekan Adesola, Usha Adiga, Qorinah … and Mohsen Naghavi. 2025. ‘Global, Regional, and National Sepsis Incidence and Mortality, 1990–2021: A Systematic Analysis’. The Lancet Global Health 13(12):e2013–26. doi:10.1016/S2214-109X(25)00356-0.
He, Zhengling, Lidong Du, Pengfei Zhang, Rongjian Zhao, Xianxiang Chen, and Zhen Fang. 2020. ‘Early Sepsis Prediction Using Ensemble Learning With Deep Features and Artificial Features Extracted From Clinical Electronic Health Records’. Critical Care Medicine 48(12):e1337. doi:10.1097/CCM.0000000000004644.
La Via, Luigi, Giuseppe Sangiorgio, Stefania Stefani, Andrea Marino, Giuseppe Nunnari, Salvatore Cocuzza, Ignazio La Mantia, Bruno Cacopardo, Stefano Stracquadanio, Serena Spampinato, Salvatore Lavalle, and Antonino Maniaci. 2024. ‘The Global Burden of Sepsis and Septic Shock’. Epidemiologia 5(3):456–78. doi:10.3390/epidemiologia5030032.
Li, Xiang, Xiao Xu, Fei Xie, Xian Xu, Yuyao Sun, Xiaoshuang Liu, Xiaoyu Jia, Yanni Kang, Lixin Xie, Fei Wang, and Guotong Xie. 2020. ‘A Time-Phased Machine Learning Model for Real-Time Prediction of Sepsis in Critical Care’. Critical Care Medicine 48(10):e884. doi:10.1097/CCM.0000000000004494.
Ma, Xuejie, Yaoqiong Mai, Yin Ma, and Xiaowei Ma. 2025. ‘Constructing an Early Warning Model for Elderly Sepsis Patients Based on Machine Learning’. Scientific Reports 15(1):10580. doi:10.1038/s41598-025-95604-8.
Moor, Michael, Nicolas Bennett, Drago Plečko, Max Horn, Bastian Rieck, Nicolai Meinshausen, Peter Bühlmann, and Karsten Borgwardt. 2023. ‘Predicting Sepsis Using Deep Learning across International Sites: A Retrospective Development and Validation Study’. eClinicalMedicine 62:102124. doi:10.1016/j.eclinm.2023.102124.
Persson, Inger, Andreas Östling, Martin Arlbrandt, Joakim Söderberg, and David Becedas. 2021. ‘A Machine Learning Sepsis Prediction Algorithm for Intended Intensive Care Unit Use (NAVOY Sepsis): Proof-of-Concept Study’. JMIR Formative Research 5(9):e28000. doi:10.2196/28000.
Rafiei, Alireza, Alireza Rezaee, Farshid Hajati, Soheila Gheisari, and Mojtaba Golzan. 2021. ‘SSP: Early Prediction of Sepsis Using Fully Connected LSTM-CNN Model’. Computers in Biology and Medicine 128:104110. doi:10.1016/j.compbiomed.2020.104110.
Reyna, Matthew A., Christopher S. Josef, Russell Jeter, Supreeth P. Shashikumar, M. Brandon Westover, Shamim Nemati, Gari D. Clifford, and Ashish Sharma. 2020. ‘Early Prediction of Sepsis From Clinical Data: The PhysioNet/Computing in Cardiology Challenge 2019’. Critical Care Medicine 48(2):210–17. doi:10.1097/CCM.0000000000004145.
Rosnati, Margherita, and Vincent Fortuin. 2021. ‘MGP-AttTCN: An Interpretable Machine Learning Model for the Prediction of Sepsis’. PLOS ONE 16(5):e0251248. doi:10.1371/journal.pone.0251248.
Santos, Miriam Seoane, Pedro Henriques Abreu, Nathalie Japkowicz, Alberto Fernández, Carlos Soares, Szymon Wilk, and João Santos. 2022. ‘On the Joint-Effect of Class Imbalance and Overlap: A Critical Review’. Artificial Intelligence Review 55(8):6207–75. doi:10.1007/s10462-022-10150-3.
Shashikumar, Supreeth P., Gabriel Wardi, Atul Malhotra, and Shamim Nemati. 2021. ‘Artificial Intelligence Sepsis Prediction Algorithm Learns to Say “I Don’t Know”’. Npj Digital Medicine 4(1):134. doi:10.1038/s41746-021-00504-6.
Srdić, Tijana, Siniša Đurašević, Iva Lakić, Aleksandra Ružičić, Predrag Vujović, Tanja Jevđović, Tamara Dakić, Jelena Đorđević, Tomislav Tosti, Sofija Glumac, Zoran Todorović, and Nebojša Jasnić. 2024. ‘From Molecular Mechanisms to Clinical Therapy: Understanding Sepsis-Induced Multiple Organ Dysfunction’. International Journal of Molecular Sciences 25(14):7770. doi:10.3390/ijms25147770.
Vuttipittayamongkol, Pattaramon, and Eyad Elyan. 2020. ‘Overlap-Based Undersampling Method for Classification of Imbalanced Medical Datasets’. Pp. 358–69 in Artificial Intelligence Applications and Innovations, edited by I. Maglogiannis, L. Iliadis, and E. Pimenidis. Cham: Springer International Publishing. doi: 10.1007/978-3-030-49186-4_30
Wang, Hao, Destiny Pounds, Wenhui Zhang, Alaa Y. Mokbel, Md Niamul Kabir, Xin Yao Lin, April Highlander, and Iman Dehzangi. 2025. ‘Early Sepsis Prediction Using Publicly Available Data: High-Performance AI/ML Models with First-Hour Clinical Information’. Diagnostics 15(21). doi:10.3390/diagnostics15212727.
Yang, Meicheng, Chengyu Liu, Xingyao Wang, Yuwen Li, Hongxiang Gao, Xing Liu, and Jianqing Li. 2020. ‘An Explainable Artificial Intelligence Predictor for Early Detection of Sepsis’. Critical Care Medicine 48(11):e1091. doi:10.1097/CCM.0000000000004550.
Zabihi, Morteza, Serkan Kiranyaz, and Moncef Gabbouj. 2019. ‘Sepsis Prediction in Intensive Care Unit Using Ensemble of XGboost Models’. P. Page 1-Page 4 in 2019 Computing in Cardiology (CinC). doi: 10.22489/CinC.2019.238
Zhang, Dongdong, Changchang Yin, Katherine M. Hunold, Xiaoqian Jiang, Jeffrey M. Caterino, and Ping Zhang. 2021. ‘An Interpretable Deep-Learning Model for Early Prediction of Sepsis in the Emergency Department’. Patterns 2(2). doi:10.1016/j.patter.2020.100196.

Ayrıntılar

Birincil Dil

İngilizce

Konular

Makine Öğrenme (Diğer)

Bölüm

Araştırma Makalesi

Yazarlar

Ali Aslan
0000-0002-9674-5618
Türkiye

Emin Demir ^*
0009-0007-9657-2079
Türkiye

Yayımlanma Tarihi

3 Haziran 2026

Gönderilme Tarihi

28 Şubat 2026

Kabul Tarihi

20 Nisan 2026

Yayımlandığı Sayı

Yıl 2026 Cilt: 29 Sayı: 2

IZ

https://izlik.org/JA88KT22NL

Kaynak Göster

RIS / Bibtex

APA

Aslan, A., & Demir, E. (2026). DECISION-AWARE TWO-STAGE CLASSIFICATION FOR EARLY SEPSIS PREDICTION IN IMBALANCED ICU DATA. Kahramanmaraş Sütçü İmam Üniversitesi Mühendislik Bilimleri Dergisi, 29(2), 948-963. https://izlik.org/JA88KT22NL

AMA

1.Aslan A, Demir E. DECISION-AWARE TWO-STAGE CLASSIFICATION FOR EARLY SEPSIS PREDICTION IN IMBALANCED ICU DATA. Kahramanmaraş Sütçü İmam Üniversitesi Mühendislik Bilimleri Dergisi. 2026;29(2):948-963. https://izlik.org/JA88KT22NL

Chicago

Aslan, Ali, ve Emin Demir. 2026. “DECISION-AWARE TWO-STAGE CLASSIFICATION FOR EARLY SEPSIS PREDICTION IN IMBALANCED ICU DATA”. Kahramanmaraş Sütçü İmam Üniversitesi Mühendislik Bilimleri Dergisi 29 (2): 948-63. https://izlik.org/JA88KT22NL.

EndNote

Aslan A, Demir E (01 Haziran 2026) DECISION-AWARE TWO-STAGE CLASSIFICATION FOR EARLY SEPSIS PREDICTION IN IMBALANCED ICU DATA. Kahramanmaraş Sütçü İmam Üniversitesi Mühendislik Bilimleri Dergisi 29 2 948–963.

IEEE

[1]A. Aslan ve E. Demir, “DECISION-AWARE TWO-STAGE CLASSIFICATION FOR EARLY SEPSIS PREDICTION IN IMBALANCED ICU DATA”, Kahramanmaraş Sütçü İmam Üniversitesi Mühendislik Bilimleri Dergisi, c. 29, sy 2, ss. 948–963, Haz. 2026, [çevrimiçi]. Erişim adresi: https://izlik.org/JA88KT22NL

ISNAD

Aslan, Ali - Demir, Emin. “DECISION-AWARE TWO-STAGE CLASSIFICATION FOR EARLY SEPSIS PREDICTION IN IMBALANCED ICU DATA”. Kahramanmaraş Sütçü İmam Üniversitesi Mühendislik Bilimleri Dergisi 29/2 (01 Haziran 2026): 948-963. https://izlik.org/JA88KT22NL.

JAMA

1.Aslan A, Demir E. DECISION-AWARE TWO-STAGE CLASSIFICATION FOR EARLY SEPSIS PREDICTION IN IMBALANCED ICU DATA. Kahramanmaraş Sütçü İmam Üniversitesi Mühendislik Bilimleri Dergisi. 2026;29:948–963.

MLA

Aslan, Ali, ve Emin Demir. “DECISION-AWARE TWO-STAGE CLASSIFICATION FOR EARLY SEPSIS PREDICTION IN IMBALANCED ICU DATA”. Kahramanmaraş Sütçü İmam Üniversitesi Mühendislik Bilimleri Dergisi, c. 29, sy 2, Haziran 2026, ss. 948-63, https://izlik.org/JA88KT22NL.

Vancouver

1.Ali Aslan, Emin Demir. DECISION-AWARE TWO-STAGE CLASSIFICATION FOR EARLY SEPSIS PREDICTION IN IMBALANCED ICU DATA. Kahramanmaraş Sütçü İmam Üniversitesi Mühendislik Bilimleri Dergisi [Internet]. 01 Haziran 2026;29(2):948-63. Erişim adresi: https://izlik.org/JA88KT22NL