Available online on 15.08.2026 at http://jddtonline.info

Journal of Drug Delivery and Therapeutics

Open Access to Pharmaceutical and Medical Research

Copyright  © 2026 The   Author(s): This is an open-access article distributed under the terms of the CC BY-NC 4.0 which permits unrestricted use, distribution, and reproduction in any medium for non-commercial use provided the original author and source are credited

Open Access Full Text Article                                                                Review Article

AI-Driven Predictive Models for Early Detection of Pediatric Sepsis: A Systematic Review and Meta-Analysis

Achsa Sharon Shibu ¹, Aadhira J ¹,  Mitra S ¹, Muhammd Marzuq U A ¹, Nickson P ¹, Dhivya K *²

¹ Pharm. D Intern, C.L. Baid Metha College of Pharmacy, Thoraipakkam, Chennai, Tamil Nadu, India

² Associate Professor, Department of Pharmacy Practice, C.L. Baid Metha College of Pharmacy, Thoraipakkam, Chennai, Tamil Nadu, India

 

 

INTRODUCTION

Sepsis is a life-threatening condition characterized by organ dysfunction resulting from a dysregulated host response to infection ¹. Sepsis continues to be a significant global health issue in pediatrics, particularly in pediatric intensive care units (PICUs), where it contributes substantially to morbidity and mortality. Early identification is critical; however, healthcare inequality, patient progress, and delays in recognition of septic infection in low- and middle-income countries can lead to worse outcomes ². The worldwide prevalence of pediatric sepsis is dependent upon local healthcare infrastructure and socioeconomic status. Children in high- and middle-income countries are more likely to have interventions available, while those in resource-poor settings experience higher rates of sepsis largely because of maternal infections, preterm births, and a lack of infection prevention measures ^3. 

Neonatal sepsis is still a major contributor to infant mortality and primarily due to perinatal risk factors such as maternal infection, premature birth, and lack of care around the time of birth. The most commonly implicated pathogens in neonatal sepsis involve gram-negative organisms such as Escherichia coli, Klebsiella pneumoniae, and Pseudomonas aeruginosa and gram-positive organisms such as Staphylococcus aureus and Streptococcus pneumoniae ⁴. Recently, viral pathogens responsible for sepsis have emerged as significant contributors of serious illness in patients with conditions such as influenza, Respiratory Syncytial Virus, and Severe Acute Respiratory Syndrome Coronavirus 2 infection, Fungal organisms can also be thought of in patients with sepsis but primarily pose a risk in immunocompromised children or those undergoing chemotherapy or organ transplantation ⁵.

Recent developments in Artificial intelligence (AI)-based models and biomarker investigations have drastically improved early diagnosis and risk assessment. Machine learning (ML) algorithms incorporated into health records and laboratory data streamline diagnostic accuracy. Emerging biomarkers, such as microRNAs (miRNAs), presepsin (sCD14-ST), and procalcitonin-to-albumin ratio, show promise in differentiating sepsis from other inflammatory conditions ^6,7. By integrating these biomarkers with AI-based models, further real-time risk assessment improvements are possible. However, AI-based sepsis-prediction models still face challenges. Biomarkers vary between populations, so larger studies are needed to make the results more reliable. Implementation in the clinical workflow comes with challenges, to ⁸. Ethical issues such as data privacy, algorithmic bias, and model accessibility also need resolving. AI decision support systems, rapid pathogen identification, and compliance to evidence-based recommendations may provide strategies to optimize antibiotics and improve patient outcomes. Preventive measures such as improved maternal healthcare, screening of neonates for sepsis, and vaccination programs are vital in reducing pediatric sepsis mortality. Vaccination against Streptococcus pneumoniae, Haemophilus influenzae type B and Neisseria meningitidis decreases incidence of bacterial sepsis. Improving infection control protocols in (Neonatal Intensive Care unit) NICUs and PICUs is necessary to minimize hospital-acquired infections and improve sepsis care ⁹.

AI has revolutionized pediatric sepsis prediction by utilizing ML and deep learning (DL) models to analyze (Electronic health records) EHRs, biomarkers, and vital signs, enabling early intervention before sepsis escalates. These models fall into four primary categories: traditional ML models, deep learning models, hybrid models, and biomarker-integrated models ¹⁰. Traditional ML algorithms like Random Forest (RF), Gradient Boosting Machines (GBM), and Support Vector Machines (SVM) effectively classify sepsis risk based on structured clinical data. Hybrid AI models combine ML and DL techniques to improve prediction accuracy while minimizing false positives. Ensemble learning models, integrating RF, neural networks, and XGBoost, offer enhanced specificity. 

Models powered by AI enhance early detection, risk stratification and monitoring of pediatric sepsis. Machine learning models, specifically, artificial neural networks (ANNs) have been shown to outperform traditional approaches such as support vector machines (SVM) and logistic regression (LR), while producing very high accuracy levels. Predictive ML models that included common clinical variables (e.g., gestational age, birth weight, and inflammatory markers) had an increased accuracy when diagnosing neo-nates with infection ¹¹. Moreover, customized ML algorithms that incorporated electronic health record data were similarly shown to offer strong predictive performance relative to standard scoring systems such as the pediatric logistic organ dysfunction-2 (PELOD-2) score, and the systemic inflammatory response syndrome (SIRS) criteria ^12,13. Furthermore, deep learning models (e.g., long short-term memory (LSTM)) optimized the analysis of data in ICU settings, and AI models that combined clinical data with biomarkers improved the sensitivity and specificity of early detection of sepsis ^14,15.

AI has demonstrated significant potential in the early detection and prediction of sepsis. By utilizing advanced ML algorithms and integrating diverse data sources, AI models can substantially improve the accuracy and timeliness of sepsis diagnosis. As these technologies evolve, their clinical implementation could lead to earlier interventions, better patient outcomes, and lower healthcare costs ¹⁶. 

METHODOLOGY 

A systematic review and meta-analysis were conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.

Literature Search 

A comprehensive literature search was conducted across multiple electronic databases, including PubMed, Medline, Google scholar, Cochrane, and Scopus. The search strategy utilized both Medical Subject Headings (MeSH) terms and free-text keywords related to pediatric sepsis, artificial intelligence, machine learning, predictive analytics, vital signs, and laboratory results. An example search query might include terms such as: 

• “Pediatric sepsis” OR “childhood sepsis” 
• “Artificial intelligence” OR “AI” OR “machine learning” 
• “Predictive model” OR “predictive analytics” 

The search will focus on studies published in English from the inception of each database to the present. Additionally, references to the included studies will be manually screened to identify any other relevant articles. 

Inclusion Criteria

• Studies that focus on pediatric populations (≤18 years old). 
• Research that develops, validates, or tests AI-driven models for sepsis detection. 
• Studies using vital signs and/or laboratory results as input variables for the AI model. 
• Peer-reviewed articles or conference proceedings. 
• Studies that report performance metrics such as sensitivity, specificity, accuracy, or area under the receiver operating characteristic curve (AUROC). 

Exclusion Criteria

• Studies involving only adult populations. 
• Non-AI-based models or heuristic algorithms. 
• Studies lacking sufficient methodological detail or outcome data. 
• Reviews, editorials, commentaries, or opinion pieces without original data. 
• Articles not published in English. 

 

 

Data extraction

Data from eligible studies were collected using a standardized extraction form. Key information to be gathered includes: 

Study Characteristics

This includes details such as the author, year of publication, study design, and the setting of the study. 

Population Characteristics

Information on the age range of participants, sample size, and clinical setting (e.g., emergency department, ICU). 

Model Details 

The type of AI model used (e.g., random forest, neural network) and the methods used for training and validation. 

Outcome Measures

Performance metrics like sensitivity, specificity, accuracy, AUROC, and precision-recall curves were used, but our study mainly considered AUROC values.

Statistical Analysis

The AUROC serves as the primary performance metric as it quantifies how effectively the AI model separates septic from non-septic patients. This is especially relevant for imbalanced data where the positive (septic) cases are outnumbered compared to the negative (non-septic) cases. AUROC measures performance across all potential thresholds unlike sensitivity or specificity. This makes AUROC a more robust performance measure, since the score reflects the performance of the comparison across the relevant threshold space. AUROC adds the ability to compare different AI models easily with models having greater predictive accuracy having higher AUROC values. AUROC can help clinicians with earlier identification of sepsis cases while avoiding many false positives, which enhances patient treatment. The AI in this study were shown to perform very well with an average AUROC of 86.8%, which shows the encouraging performance from AI in the prediction of pediatric sepsis.

• AUROC values were converted into Cohen’s d to measure effect size.
• Upper and lower confidence intervals (CI) were calculated.
• A forest plot was generated to visualize the predictive performance of AI models. 

 

Figure 1: PRISMA

The Results should report actual screening numbers, such as records identified, duplicates removed, screened, full texts assessed, exclusions with reasons, and final studies included.

RESULTS 

Results of the study showed that a meta-analysis across 14 studies contained 96,764,476 patients. Table 1 including the study number, title, authors, study design, year, and country. This summary is helpful for presenting comparative information across studies with regard to research methods utilized, publication trends, and composition across geographical locations.

 

 

Table 1: Study Characteristics

Study No	Title	Authors	Study design	Year	Country / Place
	Pediatric septic shock estimation using deep learning and electronic medical records 17	Ji Weon Lee et al	Retrospective observational study	2010-2023	Seoul
	Pediatric Severe Sepsis Prediction Using Machine Learning 18	Sidney Le et al	Retrospective set	2011-2016	California San Francisco
	Machine learning models for early sepsis recognition in the neonatal intensive care unit using readily available electronic health record data 19	Aaron J et al	Retrospective case-control design	2019	Spain
	A Continuous Late-Onset Sepsis Prediction Algorithm for Preterm Infants Using Multi-Channel Physiological Signals From a Patient Monitor 20	Zheng Peng et al	Quantitative study	2016-2018	Netherlands
	Cardiorespiratory signature of neonatal sepsis: development and validation of prediction models in 3 NICUs 21	Sherry L et al	Multi-center cohort design	2023	United States
	Vital Sign-Based Detection of Sepsis in Neonates Using Machine Learning 22	Antoine Honore et al	Single-center retrospective cohort study	2023	Sweden
	Using Machine Learning to Predict Invasive Bacterial Infections in Young Febrile Infants Visiting the Emergency Department 23	I-Min Chiu et al	Retrospective study	2011-2018	Taiwan
	Machine Learning for Early Warning of Septic Shock in Children With Hematological Malignancies Accompanied by Fever or Neutropenia: A Single Center Retrospective Study 24	Hansong Wang et al	Retrospective study	2021	China
	Effective diagnosis of sepsis in critically ill children using probabilistic graphical model 25	Tuong Minh Nguyen et al	Retrospective study	2010-2019	China
	Prediction of Late-Onset Sepsis in Preterm Infants Using Monitoring Signals and Machine Learning 26	Cabrera-Quiros, Laura et al	Retrospective study	2021	United States
	Medical decision support using machine learning for early detection of late-onset neonatal sepsis 27	Subramani Mani et al	Retrospective cohort study	2014-2017	United States
	Development and clinical impact assessment of a machine-learning model for early prediction of late-onset sepsis 28	Merel A.M. van den Berg et al	Retrospective cohort study	2023	Netherlands
	Identification of Pediatric Sepsis Subphenotypes for Enhanced Machine Learning Predictive Performance: A Latent Profile Analysis 29	Tom Velez, Tony Wang et al	Retrospective study	2019	Washington, D.C.United States
	Deep Learning Model to Predict Serious Infection Among Children with Central Venous Lines 30	Azade Tabaie et al	Retrospective cohort study	2013-2018	United States

 

 

Table 2 provides a summary of the AI models employed in each study and the respective interpretations. This summary allows for comparisons between different AI methodologies (e.g., random forest, logistic regression, neural networks), to identify the most efficacious methods for sepsis prediction. 

 

 

Table 2: AI Model Used in the Included Studies

Study No	AI MODEL USED	CONCLUSION
1.	Artificial Neural Network	The deep learning model presented contributes to the early diagnosis of septic shock in children, decreasing diagnostic burden, providing certainty in clinical decision making, and giving an opportunity to safely initiate treatment rapidly, but must be validated externally in prospective studies to ensure clinical fidelity.
2.	Decision Trees	This algorithm uses artificial intelligence to continuously monitor electronic health records (EHRs) for possible signs of severe sepsis in pediatric inpatients. It analyzes real-time data, including vital signs, labs, and other clinical parameters, to identify subtle changes before symptoms present. Timely detection of possible sepsis allows for interventions for treatment through antibiotics, fluid resuscitation, and hospital resources, leading to better patient outcomes. Automating sepsis recognition decreases the manual workload, decreasing the time to a medical evaluation and intervention for established sepsis while minimizing false alarms.
3.	Logistic Regression, Decision Trees, Random Forests, Gradient Boosted Trees (XGBoost), Support Vector Machines (SVM), Neural Networks, k-Nearest Neighbors (k-NN), And Naïve Bayes.	ML algorithms can predict sepsis in NICU infants up to hours before clinical symptoms appear, potentially allowing for earlier intervention. More datasets and additional input features, such as continuous vital signs and biomarker trends, could enhance prediction. Nonetheless, substantial research and real-world prospective studies will be needed to validate models across populations. Future research should focus on model generalizability, reducing false positives, and integrating into the clinical workflow for optimizing neonatal care.
4.	Extreme Gradient Boosting (XGB), k-Nearest Neighbors (k-NN), Logistic Regression (LR), And Support Vector Machine (SVM).	This study shows that LOS in NICUs can be predicted using physiologic signals routinely measured in clinical practice. In addition to variables derived from cardiorespiratory waveforms, motion-based features added to predictive accuracy alongside ECG and chest impedance. The visualizations of the variables contributed significantly to the interpretability of the feature selection while further enhancing usability in a clinical practice. Next steps would be to validate the signal and model parameters with larger studies across multiple centers.
5.	Logistic Regression, a Neural Network, An Extreme Gradient Boosting Classifier Or Xgboost, And Random Forest	A cardiorespiratory early warning score predicts late-onset sepsis within 24 hours, outperforming heart rate or demographics alone.
6.	Naïve Bayes Classifier.	This algorithm can predict sepsis in NICU patients 24 hours before clinical suspicion on vital signs and demographics alone. We emphasize the need for standardized definitions of sepsis and to test the method prospectively. Clinical decision support systems have the potential of improving patient care, appropriateness of resources, and patient outcomes.
7.	Logistic Regression (LR), Support Vector Machine (SVM), And Extreme Gradient Boosting (XGBoost)	This study evaluated LR, SVM, and XGBoost for predicting IBIs in febrile infants. All outperformed traditional scoring, with SVM achieving optimal results using fewer features.
8.	Xgboost Algorithm	The SSEW model provides an accurate estimate of pediatric hematology-oncology patients' risk of septic shock up to 24 hours prior to its development, and enables intervention as needed. The model uses real-time clinical data to assess risk, identify at-risk patients, and help facilitate optimizing treatment pathways. However, the model warrants exploration of its performance and reliability in prospective studies in broader patient populations. Future studies should investigate additional accuracy and false alarm reduction, and working toward a model that would work in a clinical workflow would be ideal.
9.	Tree Augmented Naive Bayes (TAN)	PGM is a reliable diagnostic tool for pediatric sepsis, with lab tests aiding prediction and vital signs helping rule it out. Further validation with diverse datasets is needed.
10.	Logistic regressor, Naive Bayes, Nearest Mean Classifier	Routine monitoring data can predict sepsis, with ECG, respiration, and motion features identifying late-onset sepsis hours before clinical deterioration.
11.	Support Vector Machines (SVM), Artificial Neural Networks (ANN), Decision Trees, and Bayesian Networks	Routinely collected EHR data can effectively predict late-onset neonatal sepsis before clinical diagnosis. AI-driven decision support systems have the potential to enhance early detection, allowing for timely interventions and improved neonatal outcomes.
12.	Generalized Additive Models, Logistic Regression, And Xgboost	An ML algorithm was trained for early LOS detection, evaluated on prediction horizons and clinical impact. This study provides the most extensive retrospective simulation for LOS prediction to date.
13.	Latent Profile Analysis	LPA identified four pediatric sepsis subphenotypes, improving prediction for high-mortality cases. Larger studies are needed for validation.
14.	Bidirectional Long Short-Term Memory (LSTM) Network With Focal Loss and An Attention Mechanism to Predict the Onset of Psi	A deep learning model can predict CLABSI onset in hospitalized children 48 hours before specimen collection using EHR data.

 

 

Table 3 provides a summary of the AUROC values reported in each study, which describe the predictive performance of each of the AI models to discriminate septic from non-septic patients. AUROC values are the primary metric used to report model performance, with higher AUROC values reflecting better predictive performance.

The AUROC had a mean value of 86.8%, implying the AI models' overall high predictive efficiency towards early pediatric sepsis detection.

 

Table 3: AUROC Value

 

 

Figure 2: FOREST PLOTTING

 

 

DISCUSSION:

Artificial Intelligence has revolutionized healthcare by improving diagnostic accuracy, clinical workflows, and patient outcomes. Early sepsis detection is crucial in pediatrics, where rapid assessment reduces morbidity and mortality risks ³¹. AI predictive models outperform traditional scoring systems, enabling earlier diagnosis and treatment initiation ³³. AI is also transforming radiology, pathology, genomics, and clinical decision-making. Deep learning detects abnormalities in MRI and CT scans, while natural language processing (NLP) extracts insights from electronic health records (EHR) for continuous diagnosis and personalized treatment [33].AI chatbots and virtual assistants improve patient access to medical advice, reducing the burden on healthcare professionals ³⁴ .AI plays a key role in the pharmaceutical industry, driving advancements in drug discovery, pharmacovigilance, and precision medicine. Machine learning (ML) models predict drug interactions, optimize dosing strategies, and enhance medication safety ³⁵. Automation in life sciences and pharmaceutical manufacturing improves efficiency, reliability, and consistency while reducing human error ³⁶. These innovations enhance healthcare system performance and product quality.

In pediatric intensive care units (PICUs), AI models analyze real-time patient data, including vital signs and biomarkers, to detect sepsis before symptoms appear ³⁷. AI models outperform clinical scores such as Pediatric Logistic Organ Dysfunction-2 (PELOD-2) and Systemic Inflammatory Response Syndrome (SIRS) criteria by analyzing vast datasets for improved risk assessment ^38,39

A number of AI-based techniques have been effectively implemented for pediatric sepsis identification, including: Machine Learning (ML) techniques such as Random Forest, Gradient Boosting, and Support Vector Machines analyze structured clinical data to assess sepsis risk ⁴⁰ . Deep Learning (DL) models like Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and transformer architectures detect physiological changes for early sepsis identification ⁴¹. Explainable AI (XAI) methods such as SHAP (Shapley Additive Explanations) and LIME (Local Interpretable Model-Agnostic Explanations) provide transparency and information about the AI-model predictions so that clinicians can understand the rationale for AI-generated suggestions ⁴².

Although it holds significant promise, AI implementation in identification of pediatric sepsis faces a number of obstacles. Data security is a concern whenever AI models utilize large amounts of patient data for training purposes ⁴³. Another challenge is the "black box" nature of deep learning models, making it difficult for clinicians to interpret predictions ⁴⁴. Developing AI in clinical settings also requires upgrades to both infrastructure and training of clinicians while ensuring all models are validated against diverse patient populations ⁴⁵. While AI models can be very sensitive, but can have lower specificity producing false-alarms and unnecessary interventions increasing days in the hospital and healthcare utilization ⁴⁶.

There are multiple important advantages to the use of AI for the detection of pediatric sepsis: Early Detection: AI has the potential to recognize sepsis several hours before clinical symptoms arise so that treatments may begin earlier and possibly improving chance of survival ⁴⁷. Automation and Efficiency: AI has the possibility to rapidly and effectively analyze large volumes of patient data, and trigger real-time alerts for those responsible for patient care ⁴⁸. Personalized Treatment Plans: AI generates individualized treatment plans based on patient specific risk factors, thereby improving clinical outcomes ⁴⁹.

Although AI has many advantages, it still has limitations such as Lack of diverse pediatric data, which makes it difficult for AI models to generalize across populations, Variations in hospital protocols require continuous AI model updates and validation ⁵⁰. AI models demand high computational resources, posing challenges in resource-limited settings ⁵¹ and AI biases may exacerbate disparities in sepsis diagnosis, leading to inequitable healthcare outcomes ⁵².

AI models for pediatric sepsis detection have evolved from classical models to machine learning, hybrid models, and biomarker integration. Machine learning improves interpretability, deep learning analyzes complex datasets, hybrid models enhance predictive accuracy, and biomarker integration supports precision medicine. Challenges such as limited data availability, clinical validation barriers, and computational costs persist. Future research will refine AI models, integrate AI into clinical workflows, and standardize data to enhance early sepsis detection and optimize pediatric patient care ⁵³.

CONCLUSION

AI-driven models aid early pediatric sepsis detection with an AUROC of 86.8%. The integration of real-time vital signs, biomarkers, and waveform data enhance predictive accuracy for timely, reliable diagnoses. While ML and DL improve predictions, challenges remain in validation, clinical adoption, and interpretation. Future research will develop resource-efficient, user-friendly AI algorithms to enhance patient outcomes. Efforts will also focus on standardizing AI in pediatric sepsis diagnosis and expanding its application to cancer patient care.

Author Contributions:

Achsa Sharon Shibu: Literature search, study screening and selection, data collection, data extraction, data analysis, interpretation of findings, preparation of tables and figures, manuscript writing, and review and editing.

Aadhira J: Conceptualization, literature search, study screening and selection, data collection, data extraction, data analysis, statistical analysis and meta-analysis, PRISMA flow diagram preparation, interpretation of results, and manuscript writing.

Mitra S.: Data collection, data extraction, data analysis, interpretation of findings, manuscript writing, and review and editing.

Muhammad Marzuq U A: Literature search, study screening, data collection, data extraction, data analysis, manuscript writing, and review and editing.

Nickson P: Literature search, study screening, data collection, data extraction, data analysis, manuscript writing, and review and editing.

Dhivya K: data analysis, and manuscript writing.

Acknowledgment: We would like to thank our management of C.L. Baid Metha College of Pharmacy for providing us with all the essential facilities in bringing out this article.

Funding Source – “Authors received no external funding for this research." 

Competing Interests/Conflict of Interest –"The authors declare no conflict of interest." 

Ethical Approval – "Ethical approval was not required for this study because it is based on previously published literature.

REFERENCES

1.     Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The third international consensus definitions for sepsis and septic shock (SEpsis-3). JAMA. 2016 Feb 23;315(8):801. Available from: https://doi.org/10.1001/jama.2016.0287 

2. Aslan AT, Permana B, Harris PNA, Naidoo KD, Pienaar MA, Irwin AD. The opportunities and challenges for artificial intelligence to improve sepsis outcomes in the Paediatric Intensive Care Unit. Current Infectious Disease Reports. 2023 Oct 31;25(11):243–53. Available from: https://doi.org/10.1007/s11908-023-00818-4 

3. Guo L, Han W, Su Y, Wang N, Chen X, Ma J, et al. Perinatal risk factors for neonatal early-onset sepsis: a meta-analysis of observational studies. The Journal of Maternal-Fetal & Neonatal Medicine. 2023 Sep 24;36(2):2259049. Available from: https://doi.org/10.1080/14767058.2023.2259049 

4. El-Aziz RMA, Rayan A. Early detection of sepsis using machine learning algorithms. Alexandria Engineering Journal. 2024 Oct 19;111:47–56. Available from: https://doi.org/10.1016/j.aej.2024.10.005 

5.     Wang H, Zhang R, Xu J, Zhang M, Ren X, Wu Y. Development of a prognosis prediction model for pediatric sepsis based on the NLPR. Journal of Inflammation Research. 2024 Oct 1;Volume 17:7777–91. Available from: https://doi.org/10.2147/jir.s479660 

6. Huang C, Chen J, Zhan X, Li L, An S, Cai G, et al. Clinical value of laboratory biomarkers for the diagnosis and early identification of Culture-Positive sepsis in neonates. Journal of Inflammation Research. 2023 Nov 1;Volume 16:5111–24. Available from: https://doi.org/10.2147/jir.s419221 

7. Li T, Li X, Liu X, Zhu Z, Zhang M, Xu Z, et al. Association of Procalcitonin to Albumin Ratio with the Presence and Severity of Sepsis in Neonates. Journal of Inflammation Research. 2022 Apr 1;Volume 15:2313–21.  https://doi.org/10.2147/jir.s358067 

8. Xia Y, Long H, Lai Q, Zhou Y. Machine Learning Predictive Model for Septic Shock in Acute Pancreatitis with Sepsis. Journal of Inflammation Research. 2024 Mar 1;Volume 17:1443–52. Available from: https://doi.org/10.2147/jir.s441591 

9. Yang J, Hao S, Huang J, Chen T, Liu R, Zhang P, et al. The application of artificial intelligence in the management of sepsis. Medical Review. 2023 Oct 1;3(5):369–80. Available from: https://doi.org/10.1515/mr-2023-0039 

10. Di Sarno L, Caroselli A, Tonin G, Graglia B, Pansini V, Causio FA, et al. Artificial intelligence in Pediatric Emergency Medicine: applications, challenges, and future perspectives. Biomedicines. 2024 May 30;12(6):1220. Available from: https://doi.org/10.3390/biomedicines12061220 

11. Wang B, Wang QM, Li DX. An analysis of predictive factors for severe neonatal infection and the construction of a prediction model. Infection and Drug Resistance. 2023 Jun 1;Volume 16:3561–74.  http://dx.doi.org/10.2147/idr.s408126 

12. Zhu Y, Li X, Guo P, Chen Y, Li J, Tao T. <p>The accuracy assessment of presepsin (sCD14-ST) for mortality prediction in adult patients with sepsis and a head-to-head comparison to PCT: a meta-analysis</p> Therapeutics and Clinical Risk Management. 2019 Jun 1;Volume 15:741–53.   https://doi.org/10.2147/tcrm.s198735 

13. Yuan KC, Tsai LW, Lee KH, Cheng YW, Hsu SC, Lo YS, et al. The development an artificial intelligence algorithm for early sepsis diagnosis in the intensive care unit. International Journal of Medical Informatics. 2020 May 21;141:104176. Available from: https://doi.org/10.1016/j.ijmedinf.2020.104176 

14. Zhao Y, Zhu R, Hu X. Diagnostic capacity of miRNAs in neonatal sepsis: a systematic review and meta-analysis. The Journal of Maternal-Fetal & Neonatal Medicine. 2024 Jan 2;37(1):2345850.  https://doi.org/10.1080/14767058.2024.2345850 

15. Llitjos JF, Carrol ED, Osuchowski MF, Bonneville M, Scicluna BP, Payen D, et al. Enhancing sepsis biomarker development: key considerations from public and private perspectives. Critical Care. 2024 Jul 13;28(1):238. Available from: https://doi.org/10.1186/s13054-024-05032-9 

16. Li F, Wang S, Gao Z, Qing M, Pan S, Liu Y, et al. Harnessing artificial intelligence in sepsis care: advances in early detection, personalized treatment, and real-time monitoring. Frontiers in Medicine. 2025 Jan 6;11:1510792. Available from: https://doi.org/10.3389/fmed.2024.1510792 

17. Lee JW, Lee B, Park JD. Pediatric septic shock estimation using deep learning and electronic medical records. Acute and Critical Care. 2024 Aug 1;39(3):400–7. Available from: https://doi.org/10.4266/acc.2024.00031 

18. Le S, Hoffman J, Barton C, Fitzgerald JC, Allen A, Pellegrini E, et al. Pediatric Severe sepsis prediction using Machine learning. Frontiers in Pediatrics. 2019 Oct 11;7:413. Available from: https://doi.org/10.3389/fped.2019.00413 

19. Mani S, Ozdas A, Aliferis C, Varol HA, Chen Q, Carnevale R, et al. Medical decision support using machine learning for early detection of late-onset neonatal sepsis. Journal of the American Medical Informatics Association. 2013 Sep 17;21(2):326–36. Available from: https://doi.org/10.1136/amiajnl-2013-001854 

20. Peng Z, Varisco G, Long X, Liang RH, Kommers D, Cottaar W, et al. A Continuous Late-Onset sepsis prediction algorithm for preterm infants using Multi-Channel physiological signals from a patient monitor. IEEE Journal of Biomedical and Health Informatics. 2022 3;27(1):550–61.   https://doi.org/10.1109/jbhi.2022.3216055 

21. Kausch SL, Brandberg JG, Qiu J, Panda A, Binai A, Isler J, et al. Cardiorespiratory signature of neonatal sepsis: development and validation of prediction models in 3 NICUs. Pediatric Research. 2023 Jan 2;93(7):1913–21. Available from: https://doi.org/10.1038/s41390-022-02444-7 

22.  Honoré A, Forsberg D, Adolphson K, Chatterjee S, Jost K, Herlenius E. Vital sign‐based detection of sepsis in neonates using machine learning. Acta Paediatrica. 2023 Jan 6;112(4):686–96. Available from: https://doi.org/10.1111/apa.16660 

23. Chiu IM, Cheng CY, Zeng WH, Huang YH, Lin CHR. Using machine learning to predict invasive bacterial infections in young febrile infants visiting the emergency department. Journal of Clinical Medicine. 2021 Apr 26;10(9):1875. Available from: https://doi.org/10.3390/jcm10091875 

24. Xiang L, Wang H, Fan S, Zhang W, Lu H, Dong B, et al. Machine learning for early warning of septic shock in children with hematological malignancies accompanied by fever or neutropenia: a Single Center retrospective study. Frontiers in Oncology. 2021 Jun 15;11:678743.  https://doi.org/10.3389/fonc.2021.678743 

25. Nguyen TM, Poh KL, Chong SL, Lee JH. Effective diagnosis of sepsis in critically ill children using probabilistic graphical model. Translational Pediatrics. 2023 Apr 1;12(4):538–51. Available from: http://dx.doi.org/10.21037/tp-22-510 

26. Cabrera-Quiros L, Kommers D, Wolvers MK, Oosterwijk L, Arents N, Van Der Sluijs-Bens J, et al. Prediction of Late-Onset sepsis in preterm infants using monitoring signals and machine learning. Critical Care Explorations. 2021 Jan 1;3(1):e0302. Available from: https://doi.org/10.1097/cce.0000000000000302 

27. Masino AJ, Harris MC, Forsyth D, Ostapenko S, Srinivasan L, Bonafide CP, et al. Machine learning models for early sepsis recognition in the neonatal intensive care unit using readily available electronic health record data. PLoS ONE. 2019 Feb 22;14(2):e0212665. https://doi.org/10.1371/journal.pone.0212665 

28. Van Den Berg M, Medina O, Loohuis I, Van Der Flier M, Dudink J, Benders M, et al. Development and clinical impact assessment of a machine-learning model for early prediction of late-onset sepsis. Computers in Biology and Medicine. 2023 Jun 15;163:107156.  https://doi.org/10.1016/j.compbiomed.2023.107156 

29. Velez T, Wang T, Koutroulis I, Chamberlain J, Uppal A, Yohannes S, et al. Identification of Pediatric sepsis subphenotypes for Enhanced Machine Learning Predictive Performance: A Latent Profile analysis. arXiv (Cornell University). 2019 Aug 23; Available from: http://arxiv.org/abs/1908.09038 

30. Tabaie A, Orenstein EW, Nemati S, Basu RK, Clifford GD, Kamaleswaran R. Deep learning model to predict serious infection among children with central venous lines. Frontiers in Pediatrics. 2021 15;9:726870.   https://doi.org/10.3389/fped.2021.726870 

31. Fu S, Li F, Qian SY. Artificial intelligence in pediatric intensive care units: current applications in sepsis management. World Journal of Pediatrics. 2026 May 1;22(5):501–10. Available from: https://doi.org/10.1007/s12519-026-01052-3 

32. Branda F, Scarpa F. Implications of artificial intelligence in addressing antimicrobial resistance: innovations, global challenges, and healthcare’s future. Antibiotics. 2024 May 29;13(6):502. https://doi.org/10.3390/antibiotics13060502 

33. Nagori A, Gautam A, Wiens MO, Nguyen V, Mugisha NK, Kabakyenga J, et al. Contextual phenotyping of pediatric sepsis cohort using large language models. arXiv (Cornell University). 2025 May 14; Available from: http://arxiv.org/abs/2505.09805 

34. Lauritsen SM, Kalør ME, Kongsgaard EL, Lauritsen KM, Jørgensen MJ, Lange J, et al. Early detection of sepsis utilizing deep learning on electronic health record event sequences. arXiv (Cornell University). 2019 Jun 7; http://arxiv.org/abs/1906.02956 

35. Mitra A, Ashraf K. Sepsis prediction and vital signs ranking in intensive care unit patients. arXiv (Cornell University). 2018 Dec 17; Available from: http://arxiv.org/abs/1812.06686 

36.  Tiewei L, Xiaojuan L, Xinrui L, Zhiwei Z, Min Z, Zhe X, et al. Association of procalcitonin-to-albumin ratio with the presence and severity of sepsis in neonates. J Inflamm Res. 2022;2313-21. DOI: 10.2147/JIR.S358067 

37.  Bedoya AD, Futoma J, Clement ME, Corey K, Brajer N, Lin A, et al. Machine learning for early detection of sepsis: an internal and temporal validation study. JAMIA Open. 2020 Apr 11;3(2):252–60. Available from: https://doi.org/10.1093/jamiaopen/ooaa006 

38. Moor M, Bennet N, Plecko D, Horn M, Rieck B, Meinshausen N, et al. Predicting sepsis in multi-site, multi-national intensive care cohorts using deep learning. arXiv (Cornell University). 2021 Jul 12; Available from: http://arxiv.org/abs/2107.05230 

39. Meeus M, Beirnaert C, Mahieu L, Laukens K, Meysman P, Mulder A, et al. Clinical decision support for improved neonatal care: The development of a machine learning model for the prediction of late-onset sepsis and necrotizing enterocolitis. The Journal of Pediatrics. 2023 Dec 6;266:113869. Available from: https://doi.org/10.1016/j.jpeds.2023.113869 

40.  Islam KR, Prithula J, Kumar J, Tan TL, Reaz MBI, Sumon MdSI, et al. Machine learning-based early prediction of sepsis using electronic health records: A systematic review. J Clin Med. 2023;12(17):5658. DOI: 10.3390/jcm12175658 

41.  Guo L, Han W, Su Y, Wang N, Chen X, Ma J, et al. Perinatal risk factors for neonatal early-onset sepsis: A meta-analysis of observational studies. J Matern-Fetal Neonatal Med. 2023;36(2):2259049. DOI: 10.1080/14767058.2023.2259049 

42. Xiao Y, Zhang G. Predictive value of a diagnostic Five-Gene biomarker for pediatric sepsis. Journal of Inflammation Research. 2024;17:2063–71. : https://doi.org/10.2147/jir.s447588 

43. Finzel B. Current methods in explainable artificial intelligence and future prospects for integrative physiology. Pflügers Archiv - European Journal of Physiology. 2025 Feb 25;477(4):513–29. Available from: https://doi.org/10.1007/s00424-025-03067-7 

44.  Li Z, Huang B, Yi W, Wang F, Wei S, Yan H, et al. Identification of potential early diagnostic biomarkers of sepsis. Journal of Inflammation Research. 2021 Mar 1;Volume 14:621–31. Available from: https://doi.org/10.2147/jir.s298604 

45. Larsen GY, Brilli R, Macias CG, Niedner M, Auletta JJ, Balamuth F, et al. Development of a quality improvement learning collaborative to improve pediatric sepsis outcomes. PEDIATRICS. 2020 Dec 16;147(1). Available from: https://doi.org/10.1542/peds.2020-1434 

46. Górriz JM, Álvarez-Illán I, Álvarez-Marquina A, Arco JE, Atzmueller M, Ballarini F, et al. Computational approaches to Explainable Artificial Intelligence: Advances in theory, applications and trends. Information Fusion. 2023 Jul 29;100:101945. Available from: https://doi.org/10.1016/j.inffus.2023.101945 

47. Scipion CEA, Manchester MA, Federman A, Wang Y, Arias JJ. Barriers to and facilitators of clinician acceptance and use of artificial intelligence in healthcare settings: a scoping review. BMJ Open. 2025 Apr 1;15(4):e092624. Available from: https://doi.org/10.1136/bmjopen-2024-092624 

48. Shashikumar SP, Mohammadi S, Krishnamoorthy R, Patel A, Wardi G, Ahn JC, et al. Development and prospective implementation of a large language model based system for early sepsis prediction. Npj Digital Medicine. 2025 May 16;8(1):290. Available from: https://doi.org/10.1038/s41746-025-01689-w 

49. O’Reilly D, McGrath J, Martin-Loeches I. Optimizing artificial intelligence in sepsis management: Opportunities in the present and looking closely to the future. Journal of Intensive Medicine. 2023 Nov 29;4(1):34–45. Available from: https://doi.org/10.1016/j.jointm.2023.10.001 

50. Gomes S, Dhanoa H, Assheton P, Carr E, Roland D, Deep A. Predicting sepsis treatment decisions in the paediatric emergency department using machine learning: the AiSEPTRON study. BMJ Paediatrics Open. 2025 May 1;9(1):e003273. Available from: https://doi.org/10.1136/bmjpo-2024-003273 

51. Culliton P, Levinson M, Ehresman A, Wherry J, Steingrub JS, Gallant S I. Predicting Severe Sepsis Using Text from the Electronic Health Record. arXiv (Cornell University). 2017 Nov 30; Available from: http://arxiv.org/abs/1711.11536 

52. Song W, Jung SY, Baek H, Choi CW, Jung YH, Yoo S. A Predictive model based on machine learning for the early detection of Late-Onset neonatal sepsis: Development and Observational study. JMIR Medical Informatics. 2020 Jun 7;8(7):e15965. Available from: https://doi.org/10.2196/15965 

53. Szakmany T, Fitzgerald E, Garlant HN, Whitehouse T, Molnar T, Shah S, et al. The ‘analysis of gene expression and biomarkers for point-of-care decision support in Sepsis‘ study; temporal clinical parameter analysis and validation of early diagnostic biomarker signatures for severe inflammation andsepsis-SIRS discrimination. Frontiers in Immunology. 2024 Jan 25;14:1308530. Available from: https://doi.org/10.3389/fimmu.2023.1308530