高级检索
贫困所致传染病（英文）
期刊首页
过刊列表
高级检索
稿件发表
• Research Article •
Machine learning for predicting severe dengue in Puerto Rico
Zachary J. Madewell
Dania M. Rodriguez
Maile B. Thayer
Vanessa Rivera-Amill
Gabriela Paz-Bailey
Laura E. Adams
Joshua M. Wong
作者及单位信息
·
DOI： 10.1186/s40249-025-01273-0
0
0
0
0
0
0
PDF下载
APP内阅读
ABSTRACT

BackgroundDistinguishing between non-severe and severe dengue is crucial for timely intervention and reducing morbidity and mortality. World Health Organization (WHO)-recommended warning signs offer a practical approach for clinicians but have limited sensitivity and specificity. This study aims to evaluate machine learning (ML) model performance compared to WHO-recommended warning signs in predicting severe dengue among laboratory-confirmed cases in Puerto Rico.

MethodsWe analyzed data from Puerto Rico’s Sentinel Enhanced Dengue Surveillance System (May 2012-August 2024), using 40 clinical, demographic, and laboratory variables. Nine ML models, including Decision Trees, K-Nearest Neighbors, Naïve Bayes, Support Vector Machines, Artificial Neural Networks, AdaBoost, CatBoost, LightGBM, and XGBoost, were trained using fivefold cross-validation and evaluated with area under the receiver operating characteristic curve (AUC-ROC), sensitivity, and specificity. A subanalysis excluded hemoconcentration and leukopenia to assess performance in resource-limited settings. An AUC-ROC value of 0.5 indicates no discriminative power, while values closer to 1.0 reflect better performance.

ResultsAmong the 1708 laboratory-confirmed dengue cases, 24.3% were classified as severe. Gradient boosting algorithms achieved the highest predictive performance, with an AUC-ROC of 97.1% (95% CI: 96.0-98.3%) for CatBoost using the full 40-variable feature set. Feature importance analysis identified hemoconcentration (≥ 20% increase during illness or ≥ 20% above baseline for age and sex), leukopenia (white blood cell count < 4000/mm 3), and timing of presentation at 4-6 days post-symptom onset as key predictors. When excluding hemoconcentration and leukopenia, the CatBoost AUC-ROC was 96.7% (95% CI: 95.5-98.0%), demonstrating minimal reduction in performance. Individual warning signs like abdominal pain and restlessness had sensitivities of 79.0% and 64.6%, but lower specificities of 48.4% and 59.1%, respectively. Combining ≥ 3 warning signs improved specificity (80.9%) while maintaining moderate sensitivity (78.6%), resulting in an AUC-ROC of 74.0%.

ConclusionsML models, especially gradient boosting algorithms, outperformed traditional warning signs in predicting severe dengue. Integrating these models into clinical decision-support tools could help clinicians better identify high-risk patients, guiding timely interventions like hospitalization, closer monitoring, or the administration of intravenous fluids. The subanalysis excluding hemoconcentration confirmed the models' applicability in resource-limited settings, where access to laboratory data may be limited.

Dengue;Ensemble learning;Gradient boosting;Feature importance;Clinical decision support;Caribbean

Full list of author information is available at the end of the article

Conception and design of the study: ZJM, MBT, and JMW. Acquisition of data: VR. Analysis and interpretation of data: ZJM. Drafting the article: ZJM. Revising the article critically for important intellectual content: ZJM, DR, MBT, VR, GP, LEA, and JMW. All authors have made substantial contributions to the study. All authors read and approved the final manuscript.

引用本文

Zachary J. Madewell,Dania M. Rodriguez,Maile B. Thayer,et al. Machine learning for predicting severe dengue in Puerto Rico[J]. Infect Dis Poverty,2025,14(01)：50-66.

DOI:10.1186/s40249-025-01273-0

PERMISSIONS

Request permissions for this article from CCC.

评价本文
*以上评分为匿名评价

This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply 2025. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativeco mmons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Background
Dengue is a significant public health concern worldwide, with approximately 390 million infections annually, of which 96 million manifest clinically [ 1 , 2 , 3 ]. In Puerto Rico, 39,094 dengue cases were reported from 2010 to 2024, including 15,077 hospitalizations (38.6%) and 94 fatalities (0.2%) [ 4 ]. A surge in dengue cases on the island in 2024 prompted a public health emergency declaration by Puerto Rico’s Department of Health, highlighting the ongoing threat of dengue to the island [ 4 , 5 ]. These regular outbreaks strain healthcare resources and pose substantial morbidity and mortality risks. A critical aspect of managing dengue is distinguishing between non-severe and severe cases, as the latter require intensive medical intervention to prevent complications and fatalities. Early identification of patients at risk of severe dengue is important for timely intervention and improved patient outcomes. However, predicting which patients will progress to severe dengue remains a challenge, often leading to delayed treatment and increased healthcare burden.
The World Health Organization (WHO) has recommended identifying severe dengue through clinical assessment of warning signs such as persistent vomiting, abdominal pain, mucosal bleeding, restlessness, and hepatomegaly [ 6 ]. Although these warning signs offer a practical approach for clinicians, their specificity and sensitivity in accurately predicting severe dengue are limited. Studies have shown that relying solely on these warning signs can result in both false positives and negatives, potentially leading to over- or under-treatment of patients [ 7 , 8 , 9 , 10 , 11 ]. The substantial burden of dengue on the healthcare system, both in terms of economic cost and human suffering, underscores the need for innovative approaches to disease diagnosis. A more accurate and efficient method for risk stratification could lead to substantial improvements in patient care and resource allocation.
In recent years, machine learning (ML) has emerged as a powerful tool for analyzing complex datasets and uncovering patterns not easily discernible by traditional methods. In the context of dengue, ML models can analyze a multitude of factors beyond the established warning signs, including patient demographics, laboratory results, clinical symptoms, and epidemiological data, to enhance the prediction of severe disease [ 12 , 13 ]. While prior studies have demonstrated the utility of ML in predicting severe dengue, they were limited by smaller sample sizes, inclusion of a single serotype, and/or reliance on a narrow range of predictor variables [ 14 , 15 ]. This study expands on previous efforts by leveraging a larger dataset including all four dengue serotypes, incorporating a broader range of predictor variables, and comparing the performance of multiple ML models, including ensemble methods. Furthermore, we assess the applicability of these models in resource-constrained settings and their alignment with WHO dengue warning signs. By leveraging ML, we aim to improve the accuracy of severe dengue predictions, offering a more robust and data-driven approach to risk stratification. If successful, these models could transform dengue management by enabling early, accurate identification of high-risk patients, ultimately leading to improved patient outcomes and reduced mortality rates. Additionally, targeted intervention for high-risk patients can optimize resource allocation, ensuring critical care is available to those who need it most.
This project specifically leverages data from Puerto Rico’s Sentinel Enhanced Dengue Surveillance System (SEDSS), which has detailed clinical and laboratory information on dengue cases, allowing for the exploration of potential predictors of disease severity beyond the conventional warning signs. In addition to exploring ML approaches, this project also aims to evaluate the performance of WHO-recommended warning signs in predicting severe dengue among laboratory-confirmed cases, which include both molecular (RT-PCR) and serologic (immunoglobulin M [IgM]ELISA) testing. If ML models demonstrate improved prediction of severe dengue compared to traditional methods, this could highlight their potential to transform clinical decision-making and improve patient outcomes.
Methods
Study population
In this study, we used data from SEDSS, an ongoing facility-based study in Puerto Rico that tracks the frequency and causes of acute febrile illness [ 16 , 17 ]. Our study included data from SEDSS from May 2012 to August 2024. SEDSS has included five sites: (1) Centro Médico Episcopal San Lucas (CMESL) in Ponce, a tertiary acute care facility (2012-present), (2) Hospital Episcopal San Lucas (HESL)—Guayama, a secondary acute care hospital (2013-2015), (3) Hospital de La Universidad de Puerto Rico in Carolina, another secondary acute care teaching hospital (2013-2015), (4) Centro de Emergencia y Medicina Integrada (CEMI), an outpatient acute care clinic in Ponce (2016-present), and (5) Auxilio Mutuo Hospital, a tertiary care facility in the San Juan Metro Area (2018-present).
Study enrollment and data collection
SEDSS enrolls participants using convenience sampling. Potential participants are identified by triage nurses as any patient with an acute febrile illness (AFI) defined by the presence of fever (≥ 38.0 °C for temperatures measured orally, ≥ 37.5 °C for temperatures measured rectally, and ≥ 38.5 °C for temperatures measured axillarily for both children and adults) at the time of triage or chief complaint of having a fever within the past seven days. During the Zika virus epidemic in Puerto Rico (June 2016-June 2018), patients were eligible if they presented with either rash and conjunctivitis, rash and arthralgia, or fever [ 18 ]. Starting in April 2020, patients with cough or dyspnea within the last 14 days (with or without fever) were also eligible to better capture respiratory viruses [ 19 ]. No age groups were excluded, although infants were only eligible for enrollment if they presented to the hospital after their initial discharge after birth. After meeting the inclusion criteria and being informed about the study, participants provided written informed consent. In cases where patients were incapacitated at the time of triage due to acute illness, consent was sought after their stabilization.
SEDSS collects data via patient interviews and medical record reviews at enrollment and convalescence (~7-14 days later). The case investigation form (CIF) gathers information about patient demographics, comorbidities, and clinical features. The convalescent sample processing form (CSPF) echoes CIF data, adding the second specimen collection date and AFI severity indicators (hospitalizations, clinic visits). Inpatient medical data for participants with AFIs who were admitted to the hospital from CMESL, HESL-Guayama, and Auxilio Mutuo Hospital also were collected using a separate form (Hospital Admitted Abstraction Form) to collect key clinical indicators of disease severity and progression. For admitted patients, these data included information on extent and nature of hemorrhage, plasma leakage (e.g., ascites and pleural and cardiac effusions), hematologic indicators of increased intravascular permeability (e.g., hematocrit and serum albumin levels), additional blood pressure and heart rate measures to assess shock, and indicators of severe organ involvement (e.g., liver impairment, meningitis, and encephalitis) [ 17 ].
Dengue warning signs and severe dengue were defined by the World Health Organization [ 20 ], incorporating available clinical indicators from SEDSS intake and follow-up forms and abstracted inpatient medical records. Dengue warning signs were defined by abdominal pain or tenderness, persistent vomiting, plasma leakage (pleural or pericardial effusion or ascites), mucosal bleeding, restlessness, hemoconcentration (defined as either a hematocrit increase of ≥ 20% during illness or a hematocrit value ≥ 20% above baseline for age and sex), or hepatomegaly. Detailed definitions for these variables have been provided previously [ 21 ]. Severe dengue was defined as severe plasma leakage or shock, severe bleeding, or severe organ impairment [ 21 ]. The presence and overlap of warning signs among severe dengue cases were visualized using an Euler plot via the eulerr R package [ 22 ]. All analyses were done using R software, version 4.4.0 (R Foundation for Statistical Computing, Vienna, Austria) [ 23 ].
Sample collection and laboratory procedures
Blood, nasopharyngeal (NP), and oropharyngeal (OP) specimens were collected at enrollment from eligible participants. Additional blood samples (serum and whole blood) were also collected during the convalescent phase. Participation required providing at least one sample (blood or OP/NP swab). NP/OP swabs were collected for testing respiratory pathogens and are not the focus of this study. RT-PCR was conducted on specimens collected within the first 7 days post-onset to detect DENV, and ELISA for IgM antibodies was performed on specimens collected > 3 days post-onset to identify probable DENV infection [ 24 ]. The same standardized laboratory detection systems were used across all participating sites in SEDSS to ensure consistency in diagnostic methods.
Variables
A total of 40 variables were selected based on the WHO’s criteria for severe dengue, physicians’ clinical experience, and a review of current literature to potentially differentiate between severe and non-severe dengue cases [ 20 , 21 , 25 , 26 , 27 , 28 ]. These features included age group, days post onset of symptoms, clinical signs and symptoms, laboratory findings, pre-existing health conditions, and dengue virus serotype. Clinical symptoms included report of fever, rash, headache, myalgia, abdominal pain, chills, itchy skin, eye pain, nasal discharge, cough, sore throat, persistent vomiting, diarrhea, arthralgia, arthritis, back pain, calf pain, nausea, no appetite, and restlessness. Clinical signs, as observed by healthcare providers, included objective fever at the time of enrollment, yellow skin (jaundice), observed bruising, conjunctivitis, hepatomegaly, mucosal bleeding, pale skin, and blue lips (cyanosis). Clinical laboratory findings comprised leukopenia (defined as white blood cell count < 4000/mm 3), the calculated value of hemoconcentration (an increase in the concentration of red blood cells due to plasma loss), and dengue immune status (primary or post-primary) as measured from results for anti-dengue virus immunoglobulin G (IgG) on or before day 5 of illness. Pre-existing health conditions like obesity (body mass index ≥ 30), gastritis, chronic arthritis, hypertension, chronic kidney disease, diabetes, thyroid disease, and high cholesterol were also considered. Additionally, dengue virus serotype was included, coded as "unknown" for probable cases identified through IgM ELISA, as serotype data was only available for confirmed reverse transcription polymerase chain reaction (RT-PCR) cases. Thrombocytopenia and clinical fluid accumulation were excluded from the analysis, as they generally manifest after the onset of severe disease or are components of its definition, making them less useful as predictive features. This comprehensive set of variables was intended to represent the multifaceted nature of factors influencing dengue infection severity.
Sensitivity analyses
To further explore the performance of the models in resource-constrained settings, where complete blood counts (CBCs), dengue immune status, and serotype information might not be readily available, we performed sub-analyses to evaluate model robustness and applicability. The sub-analyses included: (1) excluding results found on a CBC (i.e., leukopenia and hemoconcentration), (2) excluding IgG and serotype results, and (3) excluding leukopenia, hemoconcentration, IgG, and serotype results. These analyses were designed to assess how well the models could predict progression to severe dengue in various clinical scenarios, particularly where access to comprehensive clinical laboratory results or pathogen-specific testing might be limited.
To ensure the robustness of our models and account for potential confounding from co-circulating arboviruses, we performed a sensitivity analysis using CatBoost, the highest-performing individual ML model, excluding cases that tested positive for chikungunya virus (CHIKV) by either IgM or RT-PCR. This analysis aimed to confirm that the predictive features for severe dengue remain consistent even in the absence of CHIKV, given the potential overlap in clinical presentations between the two viruses. Additionally, we evaluated CatBoost performance separately for RT-PCR-confirmed and ELISA-probable cases to assess its applicability across sampling methods and collection timings.
Sampling
Our analysis included laboratory-confirmed dengue cases, confirmed by either molecular or serologic testing, focusing on differentiating between severe and non-severe cases. Due to an imbalance in the dataset, where non-severe cases were more prevalent, we used upsampling to balance the class distribution. Upsampling involved increasing the number of minority class samples (severe dengue cases) through duplication, which helps to prevent model bias towards the majority class and improves the model’s ability to accurately predict severe dengue cases [ 29 ]. Upsampling was done using the upSample function from the caret package in R [ 30 ]. To mitigate the risk of overfitting commonly associated with upsampling, we incorporated additional measures, including cross-validation and the use of regularization techniques within the ML algorithms. These strategies ensured that the models generalized well to unseen data by minimizing reliance on duplicated samples during training. Furthermore, model performance was assessed on a separate, unaltered test set comprising 30% of the original dataset, ensuring that evaluation metrics were not artificially inflated by the upsampling process. This partitioning, combined with robust evaluation, validated that upsampling improved model performance without compromising generalizability.
Machine learning models
An initial logistic regression (LR) model served as a baseline simple model to explore the relationship between potential predictors and the outcome of severe dengue. Stepwise selection, implemented using the stepAIC function from the MASS package in R [ 31 ], was used to iteratively add or remove variables to identify the optimal model with the lowest Akaike Information Criterion. This approach balances model complexity and goodness-of-fit by selecting variables that contribute significantly to the model. The final logistic regression model, derived from stepwise selection, was evaluated on both the training and testing sets.
In addition, we used nine ML methods to predict severe dengue and analyze feature importance. A comprehensive selection of models was used to leverage different strengths, enhance predictive performance, and provide a nuanced understanding of the factors contributing to severe dengue. The algorithms used include Decision Trees (DT), K-Nearest Neighbors (KNN), Naïve Bayes, Support Vector Machines (SVM), Artificial Neural Networks (ANN), Adaptive Boosting (AdaBoost), Categorical Boosting (CatBoost), Light Gradient Boosting Machine (LightGBM), and eXtreme Gradient Boosting (XGBoost). DTs create a tree-like structure to make predictions by recursively splitting the data based on feature values [ 32 ]. KNNs predict the class of a data point by considering the majority class of its closest neighbors in the feature space [ 33 ]. Naïve Bayes is a probabilistic classifier that applies Bayes’ theorem, assuming independence between predictors [ 34 ]. SVMs identify the optimal hyperplane to separate classes, making them effective for high-dimensional data [ 35 ]. ANNs are inspired by biological neural networks and consist of interconnected nodes that can capture complex patterns [ 36 ]. AdaBoost, CatBoost, LightGBM, and XGBoost are ensemble methods that combine multiple weak learners to improve predictive performance [ 37 , 38 ]. AdaBoost adjusts weights to focus on difficult-to-predict instances, CatBoost handles categorical features effectively, LightGBM is efficient with large datasets due to its leaf-wise tree growth, and XGBoost uses regularization techniques to prevent overfitting, enhancing accuracy and robustness [ 37 , 38 , 39 ].
The hyperparameters (model-specific settings, such as learning rate, maximum tree depth, or number of estimators) for each ML model were carefully tuned using a grid search strategy to optimize performance. The area under the receiver operating characteristic curve (AUC-ROC) was used as the optimization metric, ensuring a focus on maximizing classification performance. We used five-fold cross-validation during model training to enhance robustness and mitigate overfitting. Specific details of the grid search strategy and parameters included in each model are provided in Table S1 .
The following R packages were used for model implementation: rpart [ 40 ]for DT, e1071 [ 41 ]for Naïve Bayes and SVM, caret [ 30 ]for KNN, nnet [ 42 ]for ANN, keras [ 43 ]for DNN, ada [ 44 ]for AdaBoost, catboost [ 45 ]for CatBoost, lightgbm [ 46 ]for LightGBM, and xgboost [ 47 ]for XGBoost.
Ensemble model
To leverage the predictive power of multiple ML algorithms, we used an ensemble learning approach with a stacked generalization framework. This method combines the strengths of various individual models to improve overall predictive performance and robustness. We used predictions from LR and the nine different ML models as base learners in our ensemble. Specifically, we used a logistic regression model as the meta-learner to combine the outputs of the base models. This approach allows the meta-model to learn the optimal combination of base models’ predictions. To improve the performance of the meta-model, we again used stepwise selection with the stepAIC function from the MASS package [ 31 ].
Pearson correlation coefficients were calculated to measure the linear correlation between the predictions of the ML models. Comparing correlations provides insights into the degree of agreement between different ML approaches, which can highlight model consistency and reliability in capturing relevant patterns in the data. High correlations between models may indicate robustness in variable selection and prediction, whereas low correlations can signal differences in how models leverage the dataset. This helps assess whether ensemble approaches could enhance predictions by combining complementary model strengths. The results were visualized in a heatmap using ggplot2 [ 48 ].
Performance evaluation
Model performance for each ML model and the meta-model was evaluated on both the training and testing sets using AUC-ROC as the primary performance metric. AUC-ROC is an aggregate measure of performance across all possible classification thresholds, providing a comprehensive assessment of the model’s ability to distinguish between classes. We used the DeLong method to calculate the confidence intervals for the AUC-ROC to ensure accurate estimation of the model’s performance [ 49 ].
The performance of the ensemble model was further evaluated using several metrics, including accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), F1 score, and Cohen’s kappa. These metrics provide a comprehensive view of the model’s performance, capturing both the ability to correctly classify severe dengue cases and the overall agreement between predicted and actual classifications.
Feature importance
Feature importance was calculated for each ML model to quantify the contribution of each variable to the model’s predictive accuracy. This approach enables the identification of the most influential features, which enhances our understanding of the factors driving the predictions for severe dengue. Feature importance was assessed for both the 40-variable feature set and a subset excluding CBCs, IgG, and serotype results. Different methods were applied across the ML algorithms to determine feature importance. For ensemble-based methods, including XGBoost, LightGBM, and CatBoost, feature importance was calculated using the Gain metric, which measures the contribution of each feature to the model’s decision-making process. Gain represents the improvement in the model’s accuracy brought by a feature, with higher values indicating greater importance. For XGBoost, LightGBM, and CatBoost, we used the xgb.importance, lgb.importance, and catboost.get_feature_importance functions from the xgboost [ 47 ], lightgbm [ 46 ], and catboost [ 45 ]packages, respectively.
Permutation importance was applied to assess feature importance for KNN, Naive Bayes, and ANN. This method involves randomly shuffling feature values and measuring the subsequent decline in model performance. A substantial decrease in accuracy indicates a highly influential feature. For DT, feature importance was determined by the reduction in impurity (Gini index or entropy) achieved by splitting data based on that feature. AdaBoost assigned importance to features based on their contribution to correcting errors in subsequent models, with higher weights indicating greater influence. For SVM, feature importance was derived from the absolute value of the model coefficients. The magnitude of these coefficients reflects the influence of each feature on the decision boundary, with larger coefficients indicating greater importance.
We also calculated and plotted SHapley Additive exPlanations (SHAP) values for the top three performing models based on AUC-ROC. SHAP values provide a nuanced measure of each feature’s contribution to the prediction of severe dengue cases, enabling a deeper understanding of model decision-making. Positive SHAP values indicate a higher likelihood of severe dengue, whereas negative values suggest a protective effect. The SHAP approach is particularly valuable as it allows for the decomposition of the prediction into individual feature contributions, offering a clear interpretation of how different variables influence the model’s predictions.
Post-hoc variable reduction analysis
To assess the predictive performance of a simplified variable set, we conducted a post-hoc analysis using CatBoost, the ML model that achieved the highest AUC. This analysis aimed to identify the minimum number of features needed to maintain high accuracy. We began with the top features identified through SHAP values in the original 40-variable analysis, adding one feature at a time, starting with the highest-ranking. At each step, we evaluated the AUC-ROC to determine the impact of including additional features. The goal was to develop a more streamlined model that remains feasible and interpretable, especially in clinical settings with limited diagnostic resource.
Diagnostic accuracy of warning signs
In addition to ML, we evaluated the diagnostic accuracy of individual warning signs for identifying severe dengue cases. The performance of each warning sign was assessed using sensitivity, specificity, PPV, NPV, and AUC-ROC. Sub-analyses assessed the performance of warning signs by dengue serotype and immune status. Dengue serotype was determined via RT-PCR, whereas immune status was classified based on IgG antibody results in the first 5 days after illness onset (primary: IgG; post-primary: positive IgG). Cases lacking serotype or immune status data were excluded from sub-analyses. This approach aimed to identify potential clinical differences in the presentation and predictive capacity of warning signs for severe dengue across serotypes and infection statuses.
Ethics statement
The Institutional Review Boards at the Centers for Disease Control and Prevention (CDC), Auxilio Mutuo, and Ponce Medical School Foundation approved the SEDSS study protocols 6214, and 120,308-VR/2311173707, respectively. Written consent to participate was obtained from all adult participants and emancipated minors. For minors aged 14 to 20 years, written consent was obtained, and for those aged 7 to 13 years, parental written consent and participant assent were obtained.
Results
Characteristics of dengue cases
From May 2012 to August 2024, there were 51,877 unique AFI visits from 41,647 participants enrolled in SEDSS, including 8,404 hospitalizations or transfers and 75 deaths. Of these visits, there were 50,189 AFI visits from 40,495 participants tested for DENV. From these, 1708 (3.4%) had dengue (1218 confirmed, 490 probable). The majority of the 1,206 serotyped dengue cases were DENV-1 ( n = 905, 75.0%), followed by DENV-3 ( n= 149, 12.4%), DENV-2 ( n= 102, 8.5%), and DENV-4 ( n= 50, 4.1%). Of 1708 dengue cases, 759 (44.4%) were hospitalized or transferred, and two (0.1%) died. Of the 730 participants assessed for immune status using DENV IgG, 577 (79.0%) were positive, indicating post-primary dengue, whereas the remaining 153 (21.0%) were negative, suggesting primary dengue infections.
Among the 1708 laboratory-confirmed dengue cases, 24.3% ( n= 415) were classified as severe dengue. Severe dengue was similar between ELISA-probable cases (27.4%) and RT-PCR-confirmed cases (23.1%) ( P= 0.072). The median duration from symptom onset to presentation at the emergency room was 3 days [interquartile range (IQR): 2, 5]for RT-PCR-confirmed cases and 4 days (IQR: 2, 5) for ELISA-probable cases. Compared to those without severe dengue, participants with severe dengue were more likely to present between 4 to 6 days post-symptom onset (52.4% vs 32.0%, P < 0.001) and be aged 10-19 years (51.6% vs 36.3%, P < 0.001) ( Table 1 ). Among dengue cases tested, a higher proportion of severe cases were post-primary DENV infections (85.7% vs 76.3%, P= 0.007). Participants with severe dengue had a higher prevalence of warning signs such as persistent vomiting (37.6% vs 20.4%), abdominal pain (79.0% vs 51.6%), restlessness (64.6% vs 40.9%), mucosal bleeding (22.9% vs 13.8%), and hemoconcentration (20.7% vs 3.3%) compared to lab-confirmed non-severe cases (all P < 0.001) ( Fig. 1 ). All 30 dengue cases with seizures were classified as severe dengue. Leukopenia (77.1% vs. 53.5%) was more prevalent among participants with severe dengue ( P < 0.001).
Euler plot of proportion of severe dengue cases with each warning sign, sentinel enhanced dengue surveillance system, Puerto Rico, 2012-2024
  Total N = 1708 n (column %) Severe dengue N = 415 n (column %) Not severe N = 1293 n (column %) P-value
Days post onset       < 0.001
0 128 (7.5) 18 (4.3) 110 (8.5)  
1-3 810 (47.4) 133 (32.0) 677 (52.4)  
4-6 695 (40.7) 243 (58.6) 452 (35.0)  
7+ 75 (4.4) 21 (5.1) 54 (4.2)  
Age group       < 0.001
<1 32 (1.9) 5 (1.2) 27 (2.1)  
1-4 109 (6.4) 11 (2.7) 98 (7.6)  
5-9 238 (13.9) 42 (10.1) 196 (15.2)  
10-19 684 (40.0) 214 (51.6) 470 (36.3)  
20-29 225 (13.2) 38 (9.2) 187 (14.5)  
30-39 118 (6.9) 27 (6.5) 91 (7.0)  
40-49 93 (5.4) 20 (4.8) 73 (5.6)  
50+ 209 (12.2) 58 (14.0) 151 (11.7)  
Female sex 802 (47.0) 185 (44.6) 617 (47.7) 0.290
DENV immune status       < 0.001
Post-primary 577 (33.8) 180 (43.4) 397 (30.7)  
Primary 153 (9.0) 30 (7.2) 123 (9.5)  
Not tested 978 (57.3) 205 (49.4) 773 (59.8)  
DENV Serotype       < 0.001
1 905 (53.0) 226 (54.5) 679 (52.5)  
2 102 (6.0) 15 (3.6) 87 (6.7)  
3 149 (8.7) 18 (4.3) 131 (10.1)  
4 50 (2.9) 19 (4.6) 31 (2.4)  
Unknown 502 (29.4) 137 (33.0) 365 (28.2)  
Comorbidities        
Chronic pulmonary disease or asthma 347 (20.3) 76 (18.3) 271 (21.0) 0.273
Cancer 27 (1.6) 6 (1.4) 21 (1.6) 0.978
Chronic kidney disease 12 (0.7) 5 (1.2) 7 (0.5) 0.285
Coronary heart disease 52 (3.0) 16 (3.9) 36 (2.8) 0.347
Diabetes 108 (6.3) 29 (7.0) 79 (6.1) 0.601
High cholesterol 84 (4.9) 24 (5.8) 60 (4.6) 0.420
Hypertension 156 (9.1) 41 (9.9) 115 (8.9) 0.611
Arthritis 23 (1.3) 5 (1.2) 18 (1.4) 0.965
Thyroid disease 74 (4.3) 16 (3.9) 58 (4.5) 0.682
Obesity 214 (12.5) 43 (10.4) 171 (13.2) 0.148
Gastritis 28 (1.6) 9 (2.2) 19 (1.5) 0.451
Warning signs        
Persistent vomiting 420 (24.6) 156 (37.6) 264 (20.4) < 0.001
Abdominal pain 995 (58.3) 328 (79.0) 667 (51.6) < 0.001
Restlessness 797 (46.7) 268 (64.6) 529 (40.9) < 0.001
Mucosal bleeding 274 (16.0) 95 (22.9) 179 (13.8) < 0.001
Hemoconcentration 129 (7.6) 86 (20.7) 43 (3.3) < 0.001
Hepatomegaly 47 (2.8) 19 (4.6) 28 (2.2) 0.015
Other clinical signs/symptoms        
Fever 1695 (99.2) 412 (99.3) 1283 (99.2) 1.000
Conjunctivitis 147 (8.6) 26 (6.3) 121 (9.4) 0.064
Chills 1362 (79.7) 369 (88.9) 993 (76.8) < 0.001
Nausea 1175 (68.8) 333 (80.2) 842 (65.1) < 0.001
No appetite 1330 (77.9) 363 (87.5) 967 (74.8) < 0.001
Rash 971 (56.9) 286 (68.9) 685 (53.0) < 0.001
Yellow skin 69 (4.0) 36 (8.7) 33 (2.6) < 0.001
Itchy skin 627 (36.7) 190 (45.8) 437 (33.8) < 0.001
Bruise 103 (6.0) 35 (8.4) 68 (5.3) 0.025
Headache 1432 (83.8) 372 (89.6) 1060 (82.0) < 0.001
Eye pain 1026 (60.1) 288 (69.4) 738 (57.1) < 0.001
Myalgia 1276 (74.7) 362 (87.2) 914 (70.7) < 0.001
Arthralgia 1062 (62.2) 301 (72.5) 761 (58.9) < 0.001
Back pain 912 (53.4) 261 (62.9) 651 (50.3) < 0.001
Calf pain 627 (36.7) 180 (43.4) 447 (34.6) < 0.001
Arthritis 251 (14.7) 84 (20.2) 167 (12.9) < 0.001
Nasal discharge 545 (31.9) 149 (35.9) 396 (30.6) 0.052
Sore throat 615 (36.0) 168 (40.5) 447 (34.6) 0.034
Cough 726 (42.5) 189 (45.5) 537 (41.5) 0.167
Diarrhea 736 (43.1) 228 (54.9) 508 (39.3) < 0.001
Seizure 30 (1.8) 30 (7.2) 0 (0.0) < 0.001
Pale skin 783 (45.8) 266 (64.1) 517 (40.0) < 0.001
Blue lips 72 (4.2) 35 (8.4) 37 (2.9) < 0.001
Laboratory        
Leukopenia 1012 (59.3) 320 (77.1) 692 (53.5) < 0.001
Demographic and clinical characteristics of participants with laboratory-confirmed dengue (RT-PCR and IgM ELISA) by severity, Sentinel Enhanced Dengue Surveillance System, Puerto Rico, 2012-2024

P-values were calculated using either the chi-square test or Fisher’s exact test, as appropriate, based on the sample sizes in each category

Performance of warning signs for predicting severe dengue
Among warning signs, abdominal pain and restlessness had the highest sensitivities for predicting severe dengue at 79.0% and 64.6%, respectively, but the lowest specificities of 48.4% and 59.1% ( Table 2 ). In contrast, hepatomegaly and hemoconcentration demonstrated the highest specificities at 97.8% and 96.7%, respectively, but were less sensitive at 4.6% and 20.7%, respectively. The presence of any warning sign yielded the highest sensitivity (92.8%) but a low specificity (29.2%), with an AUC-ROC of 61.1%. Combining three or more warning signs increased the specificity to 65.1% while maintaining moderate sensitivity (87.2%), resulting in the highest AUC-ROC (71.3%) among the combinations tested.
Warning sign True positive n (%) True negative n (%) False positive n (%) False negative n (%) Sensitivity % (95% CI) Specificity % (95% CI) Positive predictive value % (95% CI) Negative predictive value % (95% CI) AUC-ROC % (95% CI)
Persistent vomiting 156 (9.1) 1029 (60.2) 264 (15.5) 259 (15.2) 37.6 (32.9, 42.4) 79.6 (77.3, 81.7) 37.1 (32.5, 42.0) 79.9 (77.6, 82.0) 58.5 (56.0, 61.1)
Abdominal pain 328 (19.2) 626 (36.7) 667 (39.1) 87 (5.1) 79.0 (74.8, 82.9) 48.4 (45.7, 51.2) 33.0 (30.0, 36.0) 87.8 (85.2, 90.1) 60.4 (58.5, 62.3)
Restlessness 268 (15.7) 764 (44.7) 529 (31.0) 147 (8.6) 64.6 (59.8, 69.2) 59.1 (56.4, 61.8) 33.6 (30.3, 37.0) 83.9 (81.3, 86.2) 58.7 (56.7, 60.8)
Mucosal bleeding 95 (5.6) 1114 (65.2) 179 (10.5) 320 (18.7) 22.9 (18.9, 27.2) 86.2 (84.2, 88.0) 34.7 (29.0, 40.6) 77.7 (75.4, 79.8) 56.2 (53.2, 59.2)
Hemoconcentration 86 (5.0) 1250 (73.2) 43 (2.5) 329 (19.3) 20.7 (16.9, 24.9) 96.7 (95.5, 97.6) 66.7 (57.8, 74.7) 79.2 (77.1, 81.1) 72.9 (68.7, 77.1)
Hepatomegaly 19 (1.1) 1265 (74.1) 28 (1.6) 396 (23.2) 4.6 (2.8, 7.1) 97.8 (96.9, 98.6) 40.4 (26.4, 55.7) 76.2 (74.0, 78.2) 58.3 (51.1, 65.5)
Any warning sign 388 (22.7) 347 (20.3) 946 (55.4) 27 (1.6) 93.5 (90.7, 95.7) 26.8 (24.4, 29.3) 29.1 (26.7, 31.6) 92.8 (89.7, 95.2) 60.9 (59.1, 62.7)
Only one warning sign 78 (9.0) 347 (40.1) 413 (47.7) 27 (3.1) 74.3 (64.8, 82.3) 45.7 (42.1, 49.3) 15.9 (12.8, 19.4) 92.8 (89.7, 95.2) 54.3 (52.2, 56.4)
Only two warning signs 126 (14.9) 347 (41.0) 347 (41.0) 27 (3.2) 82.4 (75.4, 88.0) 50.0 (46.2, 53.8) 26.6 (22.7, 30.9) 92.8 (89.7, 95.2) 59.7 (57.3, 62.1)
Three or more warning signs 184 (24.7) 347 (46.6) 186 (25.0) 27 (3.6) 87.2 (81.9, 91.4) 65.1 (60.9, 69.2) 49.7 (44.5, 54.9) 92.8 (89.7, 95.2) 71.3 (68.4, 74.1)
Performance of warning signs for predicting severe dengue ( n= 1708), Sentinel Enhanced Dengue Surveillance System, Puerto Rico, 2012-2024

AUC-ROC Area under receiver operating characteristic curve, CI Confidence interval

Performance of warning signs for predicting severe dengue demonstrated some variability across serotypes and immune status, though the interpretation is constrained by limited sample sizes and overlapping confidence intervals ( Tables S2 - S3 ).
Performance evaluation of machine learning models
The ensemble model demonstrated a strong correlation between predictions from the CatBoost, XGBoost, LightGBM, and AdaBoost models, with Pearson correlation coefficients of 0.91, 0.89, 0.89, and 0.84, respectively, indicating that these gradient boosting models had substantial influence on the ensemble’s predictions ( Fig. 2 ). This high correlation suggests that the models may be selecting and emphasizing similar variables in their predictive processes. In contrast, weaker correlations were observed between the ensemble model and simpler models like KNN (0.48), Naïve Bayes (0.48), and DT (0.51), indicating different prediction patterns and potential differences in variable selection. Additionally, high inter-model correlations among gradient boosting models, particularly between LightGBM and XGBoost (0.98), further support the idea that these models capture similar patterns in the data and rely on comparable sets of variables.
Pearson’s correlation of predictions between machine learning models, Sentinel Enhanced Dengue Surveillance System, Puerto Rico, 2012-2024. Pearson correlation coefficients measure the linear agreement between the predictions of different machine learning models. Higher values indicate similar prediction patterns across models, suggesting that models are identifying similar cases as severe dengue. Darker colors represent higher correlations
AUC values for the 40-variable feature set across various ML models demonstrated varying levels of predictive performance. Gradient boosting algorithms achieved the highest AUC values of 97.1% for CatBoost, 95.5% for XGBoost, and 94.5% for LightGBM, indicating strong discriminatory power ( Fig. 3 ). ANN showed moderate performance (AUC = 88.4%), whereas LR and SVM had lower discrimination (AUC = 79.4% and 78.9%, respectively). KNN, Naïve Bayes, and DT had the lowest AUC values of 74.1%, 75.9%, and 76.2%, respectively, indicating limited predictive ability. The ensemble meta-model provided a slight improvement over CatBoost with an AUC of 97.7%.
Forest plot of AUC values for Decision Trees (DT), K-Nearest Neighbors (KNN), Naïve Bayes, Support Vector Machines (SVM), Artificial Neural Networks (ANN), Adaptive Boosting (AdaBoost), Categorical Boosting (CatBoost), Light Gradient Boosting Machine (LightGBM), eXtreme Gradient Boosting (XGBoost), and ensemble models for a 40-variable feature set and subsets excluding CBCs, IgG, and serotype results, Sentinel Enhanced Dengue Surveillance System, Puerto Rico, 2012-2024. DeLong method was used to obtain the 95% confidence intervals for the AUC-ROC. CBC Complete blood count, IgG Immunoglobulin G, AUC-ROC Area under the receiver operating characteristic curve.
Exclusion of immune status and serotype data minimally affected model performance across all ML algorithms ( Fig. 3 ). Conversely, removing leukopenia and hemoconcentration significantly reduced predictive power for Naïve Bayes, LR, SVM, and ANN (AUC decreased by 3.2%-5.2%). CatBoost, XGBoost, LightGBM, and the ensemble model consistently maintained high performance, showing minimal to no change in AUC, regardless of the inclusion or exclusion of leukopenia and hemoconcentration. Excluding CHIKV-positive cases in the sensitivity analysis resulted in minimal changes to the AUC-ROC scores for CatBoost, confirming that the model’s predictive performance for severe dengue remains robust even in the presence of co-circulating arboviruses ( Table S4 ). CatBoost performance was consistent across RT-PCR-confirmed (AUC = 94.8%) and ELISA-probable (AUC = 95.3%) cases ( Table S5 ).
The ensemble model with 40 variables achieved the highest overall AUC of 97.7% with corresponding sensitivity and specificity of 95.6% and 93.3%, respectively ( Table S6 ). The F1 score was 94.5% and Kappa was 88.9%, indicating a high level of agreement and balanced performance between precision and recall in the model’s classification of severe dengue cases.
Feature importance
For the 40-variable feature set, SHAP values identified hemoconcentration, days post symptom onset, and leukopenia as most important features for CatBoost, XGBoost, and LightGBM ( Fig. 4 , Figures S1 - S2 ). Similarly, LR highlighted these variables as having the highest adjusted odds ratios for severe dengue [hemoconcentration: adjusted odd ratios (a OR)= 7.02; leukopenia: a OR = 2.24; days post onset 4-6 days: a OR = 1.96]( Table S7 ). Additionally, these models highlighted pale skin, age group, and the clinical warning signs of restlessness, abdominal pain, and persistent vomiting as key predictors of severe dengue progression. Hemoconcentration also stood out as a top feature for Naïve Bayes and SVM ( Figures S3 - S4 ). AdaBoost, which focuses on correcting errors from previous classifiers, assigned greater importance to chronic conditions such as high cholesterol, chronic arthritis, and hypertension. Although hepatomegaly is a recognized warning sign, it had a lower importance score in our analysis, suggesting it may play a more limited role in predicting severe dengue in this context.
SHapley Additive exPlanations (SHAP) values for the 40 Features in CatBoost, Sentinel Enhanced Dengue Surveillance System, Puerto Rico, 2012-2024. SHAP values measure each feature’s contribution to the prediction of severe dengue in the CatBoost model. Positive SHAP values indicate a higher likelihood of severe dengue, while negative values suggest a lower likelihood (or protective effect). Each dot represents a single case, with its horizontal position showing the SHAP value, reflecting the strength and direction of the feature’s impact. The color of the dots indicates the actual feature value for each case. For most features, values are binary (0 or 1), representing presence or absence (e.g., rash or no rash). For age group, the scale ranges from 0 to 7, with 0 indicating the youngest age group (< 1 year) and 7 indicating the oldest age group (≥ 50 years). An example interpretation: if 'persistent vomiting’ has a positive SHAP value and the dot is green (value= 1), it indicates that the presence of persistent vomiting strongly increases the likelihood of severe dengue for that case. The mean SHAP values shown on the right represent the average absolute impact of each feature across all cases, indicating the overall importance of that feature in the model’s predictions
Post-hoc variable reduction analysis
To explore a more streamlined predictive model, we conducted a post-hoc variable reduction analysis using the ML model with the highest AUC, CatBoost. Starting with the top feature identified by SHAP values (days post onset), we sequentially added variables, assessing AUC-ROC at each step. The AUC improved consistently with each additional variable, though the gains diminished over time. By including just 20 variables—compared to the original 40-variable set—the model achieved an AUC of 96.5% ( Fig. 5 ). The optimal reduced feature set included days post onset, hemoconcentration, leukopenia, restlessness, pale skin, abdominal pain, age group, diarrhea, rash, persistent vomiting, cough, calf pain, sore throat, arthralgia, itchy skin, eye pain, back pain, mucosal bleeding, myalgia, and arthritis. This reduced model offers a more practical and interpretable approach while maintaining high predictive accuracy, making it feasible for use in clinical settings, especially where diagnostic resources are limited.
Iterative improvement in area under the curve (AUC) with additional variables in CatBoost model for severe dengue prediction, Sentinel Enhanced Dengue Surveillance System, Puerto Rico, 2012-2024. This figure shows the change in AUC as top-performing variables are sequentially added to the CatBoost model. Starting with the highest-impact feature, "Days post onset," each subsequent model includes one additional variable in the order of their mean SHAP values. The combinations of variables and their AUC, along with 95% confidence intervals, are shown to demonstrate the predictive gain with each added variable
Discussion
Our study underscores the potential of ML models, particularly gradient boosting algorithms, to outperform traditional warning signs in predicting severe dengue. This improved predictive ability could transform clinical decision-making, enabling earlier and more accurate identification of high-risk patients, thereby improving outcomes in dengue-endemic regions like Puerto Rico.
Hemoconcentration, days post symptom onset, and leukopenia emerged as the most important features across multiple ML models, aligning with their known relevance in dengue prognosis [ 50 , 51 , 52 ]. Hemoconcentration, which reflects plasma leakage through an increased red blood cell concentration, was consistently highlighted as a top predictor by CatBoost, XGBoost, LightGBM, and several other algorithms. Days post symptom onset is a crucial temporal marker, likely capturing the dynamic nature of disease as cases often progress to the critical phase of dengue (when severe disease occurs) 3-7 days after symptom onset. Future studies could explore how days post-onset and other predictors evolve over time, particularly focusing on the timing of key markers like hemoconcentration and leukopenia. Understanding these dynamics could refine predictive models and better inform intervention timing. Leukopenia, or low white blood cell count, often reflects the body’s response to viral infections, including dengue. In addition, pale skin, age group, and clinical warning signs such as restlessness, abdominal pain, and persistent vomiting were identified as key predictors of severe dengue progression. The variability in feature importance across models emphasizes the complexity of severe dengue prediction, highlighting the need for tailored approaches that account for both individual patient characteristics and disease progression.
Our analysis highlights the strong predictive performance of gradient boosting algorithms—CatBoost, XGBoost, and LightGBM—with AUC values above 94%, reflecting their ability to capture complex, non-linear patterns in clinical data [ 39 , 53 , 54 , 55 ]. Despite this high predictive accuracy, the interpretability of these models remains a limitation in clinical settings, where transparency in decision-making is critical for trust and practical use [ 56 , 57 , 58 ]. Compared to simpler models like LR, which offers straightforward interpretations of how each variable influences severe dengue risk, gradient boosting algorithms can be challenging to interpret. This trade-off between high performance and interpretability suggests that ML models may be most useful as supplementary tools for alerting clinicians to high-risk cases, rather than as standalone decision aids. For example, predictive ML models could be incorporated into clinical decision support systems used in emergency departments or outpatient clinics, enabling real-time risk stratification for severe dengue. This integration could help healthcare providers prioritize patients for hospitalization, allocate medical resources more efficiently, and guide timely interventions to prevent complications and fatalities. These tools could enable early and accurate identification of high-risk patients, improve patient outcomes, and optimize resource allocation.
In contrast, simpler models like Naïve Bayes, Decision Trees, and KNN showed weaker correlations and lower AUC values, reflecting their limitations in capturing data complexity. The ensemble model, achieving the highest AUC of 97.7%, demonstrates the added value of combining multiple ML algorithms to enhance predictive accuracy, with high sensitivity and specificity, making it particularly useful in resource-limited settings [ 59 , 60 ]. Additionally, the exclusion of immune status and serotype data had minimal impact on model performance, indicating these variables are not essential for accurate prediction in this context. The high NPV across models suggests that ML tools can still be valuable for identifying low-risk patients who may not require intensive monitoring. In these cases, the ML model’s recommendation could serve as an early discharge or outpatient management decision-support mechanism, further optimizing healthcare resource allocation.
The post-hoc variable reduction analysis demonstrated that a streamlined set of 20 variables achieved strong predictive accuracy (AUC of 96.5%), close to the full 40-variable model. This reduced set offers a balance between interpretability and performance, making it more practical for clinical application, particularly in settings with limited diagnostic resources. By focusing on essential predictors—such as hemoconcentration, days post onset, leukopenia, and key symptoms—this approach prioritizes feasibility and interpretability, even if it means a modest sacrifice in predictive power. For added clinical utility, LR could complement ML approaches by further refining and validating the reduced variable set with interpretable odds ratios, enabling clinicians to apply these findings more confidently in practice [ 61 ].
Traditional warning signs showed both strengths and limitations. Abdominal pain and restlessness were the most sensitive indicators, consistent with other studies [ 7 , 10 ], yet their low specificities limit their utility. Conversely, markers like hepatomegaly and hemoconcentration had high specificity but low sensitivity. Combining multiple warning signs improved specificity while maintaining high sensitivity, yielding the highest AUC among tested combinations. In contrast, gradient boosting ML algorithms offered a more balanced approach with high sensitivity and specificity, crucial for accurate risk stratification in clinical settings. Our findings align with a recent study where an 8-gene XGBoost model outperformed clinical warning signs, significantly improving negative predictive power and demonstrating strong generalizability across patient cohorts [ 57 ]. Although our models focus on demographic and clinical features, the integration of gene expression data presents an intriguing avenue for future exploration [ 57 , 58 ].
This study has several limitations. First, these ML models need to be re-fitted to different variables and populations to ensure accuracy across various settings. Second, the relatively small dataset increases the risk of overfitting, potentially affecting the models’ robustness and generalizability. Third, the prevalence of DENV-1 cases from the 2012-2013 outbreak may limit the models’ applicability to other periods, regions, populations, age groups, or serotypes. Fourth, the models were developed using data from the SEDSS, where inclusion criteria required febrile illness, potentially limiting generalizability to broader populations. Fifth, due to limited sample sizes for serotypes and immune statuses, we could not conduct ML analyses for these subgroups; future studies with larger datasets are needed to validate findings. Sixth, the lack of time-course data for the onset and progression of key clinical and laboratory features limited our ability to evaluate their predictive utility over time. Seventh, there is a potential limitation related to the inclusion of false negatives—SEDSS cases who may have later presented to a non-SEDSS facility with severe disease—although this is considered unlikely given typical healthcare-seeking behaviors. Eighth, the SEDSS data’s robustness may not accurately reflect real-world conditions, where datasets are often sparse, contain free-text fields, or have incomplete information, potentially affecting model performance. Ninth, dengue and severe dengue are often underdiagnosed and underreported, which could impact model results; however, our findings likely represent a conservative estimate when accounting for underreporting. Finally, implementing ML models in clinical practice may require computational resources and infrastructure not available in all settings, particularly in low-resource environments where dengue is endemic. Despite these limitations, this study builds on prior research [ 14 , 15 ]by employing a larger and more representative dataset that includes all four dengue serotypes, leveraging nine machine learning models, and integrating ensemble approaches to enhance predictive accuracy. Unlike previous studies, we assessed model performance in resource-constrained scenarios, compared predictions against WHO warning signs, and evaluated prediction agreement across models. These contributions advance the applicability and robustness of ML tools for severe dengue prediction, addressing key gaps in earlier research and providing a foundation for broader clinical and public health implementation.
Conclusions
This study found that ML models, particularly gradient boosting algorithms, outperformed traditional warning signs in predicting severe dengue by integrating multiple variables and capturing complex interactions. These models demonstrated high specificity and sensitivity, making them valuable tools for early identification of patients at highest risk for progression to severe disease. Implementing these models in clinical decision-making could help identify patients at highest risk for progression to severe dengue, reducing unnecessary hospitalizations and easing healthcare burdens. Although resource constraints may limit direct ML implementation in some settings, platforms like SEDSS can still leverage ML techniques to identify key predictors of severe disease. This approach can optimize patient care by prioritizing the most critical predictors, even in low-resource environments where advanced ML algorithms may not be feasible.
附件材料

Supplementary material 1.

文档类型: DOCX
参考文献
1.
Bhatt S , Gething PW , Brady OJ , Messina JP , Farlow AW , Moyes CL ,et al. The global distribution and burden of dengue. Nature. 2013;496(7446):504-7.
返回引文位置Google Scholar
百度学术
万方数据
2.
Global, regional, and national age-sex-specific mortality for 282 causes of death in 195 countries and territories, 1980-2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet. 2018;392(10159):1736-88.
返回引文位置
3.
Madewell ZJ . Arboviruses and their vectors. South Med J. 2020;113(10):520-3.
返回引文位置Google Scholar
百度学术
万方数据
4.
Rodriguez DM , Madewell ZJ , Torres JM , Rivera A , Wong JM , Santiago GA ,et al. Epidemiology of dengue - Puerto Rico, 2010-2024. MMWR Morb Mortal Wkly Rep. 2024;73(49):1112-7.
返回引文位置Google Scholar
百度学术
万方数据
5.
Thayer MB , Marzan-Rodriguez M , Torres Aponte J , Rivera A , Rodriguez DM , Madewell ZJ ,et al. Dengue epidemic alert thresholds: A tool for surveillan ce and epidemic detection . Medrxiv. 2024;11:9.
返回引文位置Google Scholar
百度学术
万方数据
6.
WHO Guidelines Approved by the Guidelines Review Committee. Dengue guidelines for diagnosis, treatment, prevention and control. Geneva:World Health Organization; 2009.
返回引文位置
7.
Ahmad MH , Ibrahim MI , Mohamed Z , Ismail N , Abdullah MA , Shueb RH ,et al. The sensitivity, specificity and accuracy of warning signs in predicting severe dengue, the severe dengue prevalence and its associated factors. Int J Environ Res Public Health. 2018;15(9):2018.
返回引文位置Google Scholar
百度学术
万方数据
8.
Narvaez F , Gutierrez G , Pérez MA , Elizondo D , Nuñez A , Balmaseda A ,et al. Evaluation of the traditional and revised WHO classifications of dengue disease severity. PLoS Negl Trop Dis. 2011;5(11):e1397.
返回引文位置Google Scholar
百度学术
万方数据
9.
Macedo GA , Gonin MLC , Pone SM , Cruz OG , Nobre FF , Brasil P . Sensitivity and specificity of the World Health Organization dengue classification schemes for severe dengue assessment in children in Rio de Janeiro. PLoS ONE. 2014;9(4):e96314.
返回引文位置Google Scholar
百度学术
万方数据
10.
Jayaratne SD , Atukorale V , Gomes L , Chang T , Wijesinghe T , Fernando S ,et al. Evaluation of the WHO revised criteria for classification of clinical disease severity in acute adult dengue infection. BMC Res Notes. 2012;5:645.
返回引文位置Google Scholar
百度学术
万方数据
11.
Thein TL , Gan VC , Lye DC , Yung CF , Leo YS . Utilities and limitations of the World Health Organization 2009 warning signs for adult dengue severity. PLoS Negl Trop Dis. 2013;7(1):e2023.
返回引文位置Google Scholar
百度学术
万方数据
12.
Hoyos W , Aguilar J , Toro M . Dengue models based on machine l earning techniques: a systematic literature review . Artif Intell Med. 2021;119:102157.
返回引文位置Google Scholar
百度学术
万方数据
13.
Gupta G , Khan S , Guleria V , Almjally A , Alabduallah BI , Siddiqui T ,et al. DDPM a dengue disease prediction and diagnosis model using sentiment analysis and machine learning algorithms. Diagnostics. 2023. https://doi.org/10.3390/diagnostics13061093 .
返回引文位置Google Scholar
百度学术
万方数据
14.
Huang SW , Tsai HP , Hung SJ , Ko WC , Wang JR . Assessing the risk of dengue severity using demographic information and labo ratory test results with machine learning . PLoS Negl Trop Dis. 2020;14(12):e0008960.
返回引文位置Google Scholar
百度学术
万方数据
15.
Thanh NT , Luan VT , Viet DC , Tung TH , Thien V . A machine learning-based risk score for prediction of mechanical ventilation in children with dengue shock syndrome: a retrospective cohort study. PLoS ONE. 2024;19(12):e0315281.
返回引文位置Google Scholar
百度学术
万方数据
16.
Tomashek KM , Rivera A , Torres-Velasquez B , Hunsperger EA , Munoz-Jordan JL , Sharp TM ,et al. Enhanced surveillance for fatal dengue-like acute febrile illness in Puerto Rico, 2010-2012. PLoS Negl Tr op Dis . 2016;10(10):e0005025.
返回引文位置Google Scholar
百度学术
万方数据
17.
Madewell ZJ , Hernandez-Romieu AC , Wong JM , Zambrano LD , Volkman HR , Perez-Padilla J ,et al. Sentinel enhanced dengue surveillance system - Puerto Rico, 2012-2022. MMWR Surveill Summ. 2024;73(3):1-29.
返回引文位置Google Scholar
百度学术
万方数据
18.
Read JS , Torres-Velasquez B , Lorenzi O , Rivera Sanchez A , Torres-Torres S , Rivera LV ,et al. Symptomatic Zika Virus Infection in Infants, Children, and adolescents living in Puerto Rico. JAMA Pe diatr . 2018;172(7):686-93.
返回引文位置Google Scholar
百度学术
万方数据
19.
Wong JM , Volkman HR , Adams LE , García CO , Martinez-Quiñones A , Perez-Padilla J ,et al. Clinical features of COVID-19, dengue, and influenza among adults presenting to emergency departments and urgent care clinics—Puerto Rico, 2012-2021. Am J Trop Med Hyg. 2023;108(1):107.
返回引文位置Google Scholar
百度学术
万方数据
20.
World Health Organization. 2012. Handbook for clinical management of dengue. Available from:https://www.who.int/publications/i/item/9789241504713. Accessed 22 Jan 2024
返回引文位置Google Scholar
百度学术
万方数据
21.
Paz-Bailey G , Sánchez-González L , Torres-Velasquez B , Jones ES , Perez-Padilla J , Sharp TM ,et al. Predominance of severe plasma leakage in pediatric patients with severe dengue in puerto rico. J Infect Dis. 2022;226(11):1949-58.
返回引文位置Google Scholar
百度学术
万方数据
22.
Larsson J , Godfrey AJR , Gustafsson P , Eberly DH , Huber E , Privé F . eulerr: Area-Proportional Euler and Venn Diagrams with Ellipses. 2024. https://cran.r-project.org/web/packages/eulerr/index.htmlAccessed 15 Aug 2024
返回引文位置Google Scholar
百度学术
万方数据
23.
R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Co mputing . 2024
返回引文位置
24.
Tomashek KM , Lorenzi OD , Andújar-Pérez DA , Torres-Velásquez BC , Hunsperger EA , Munoz-Jordan JL ,et al. Clinical and epidemiologic characteristics of dengue and other etiologic agents among patients with acute febrile illness, Puerto Rico, 2012-2015. PLoS Negl Trop Dis. 2017;11(9):e0005859.
返回引文位置Google Scholar
百度学术
万方数据
25.
Tsheten T , Clements ACA , Gray DJ , Adhikary RK , Furuya-Kanamori L , Wangdi K . Clinical predictors of severe dengue: a systematic review and meta-analysis. Infect Dis Poverty. 2021;10(1):123.
返回引文位置Google Scholar
百度学术
万方数据
26.
Sangkaew S , Ming D , Boonyasiri A , Honeyford K , Kalayanarooj S , Yacoub S ,et al. Risk predictors of progression to severe disease during the febrile phase of dengue: a systematic review and meta-analysis. Lancet Infect Dis. 2021;21(7):1014-26.
返回引文位置Google Scholar
百度学术
万方数据
27.
Khan MAS , Al Mosabbir A , Raheem E , Ahmed A , Rouf RR , Hasan M ,et al. Clinical spectrum and predictors of severity of dengue among children in 2019 outbreak: a multicenter hospital-based study in Bangl adesh . BMC Pediatr. 2021;21(1):478.
返回引文位置Google Scholar
百度学术
万方数据
28.
Yuan K , Chen Y , Zhong M , Lin Y , Liu L . Risk and predictive factors for severe dengue infection: a systematic review and meta-analysis. PLoS ONE. 2022;17(4):e0267186.
返回引文位置Google Scholar
百度学术
万方数据
29.
Werner de Vargas V , Schneider Aranda JA , Dos Santos CR , da Silva Pereira PR , Victória Barbosa JL . Imbalanced data preprocessing techniques for machine learning: a systematic mapping study. Knowl Inf Syst. 2023;65(1):31-57.
返回引文位置Google Scholar
百度学术
万方数据
30.
Kuhn M . caret: Classification and Regression Training 2023. https://cran.r-project.org/web/packages/caret/index.htmlAccessed 17 Jun 2024
返回引文位置Google Scholar
百度学术
万方数据
31.
Ripley B , Venables B , Bates DM , Hornik K , Gebhardt A , Firth D . MASS: Support Functions and Datasets for Venables and Ripley’s MASS 2024. https://cran.r-project.org/web/packages/MASS/index.htmlAccessed 7 Aug 2024
返回引文位置Google Scholar
百度学术
万方数据
32.
Kotsiantis SB . Decision trees: a recent overview. Artif Intell Rev. 2013;39:261-83.
返回引文位置Google Scholar
百度学术
万方数据
33.
Peterson LE . K-nearest neighbor. Scholarpedia. 2009;4(2):1883.
返回引文位置Google Scholar
百度学术
万方数据
34.
Webb GI , Keogh E , Miikkulainen R . Naïve bayes. Encycl Machine Learn. 2010;15(1):713-4.
返回引文位置Google Scholar
百度学术
万方数据
35.
Hearst MA , Dumais ST , Osuna E , Platt J , Scholkopf B . Support vector machines. IEEE Intell Syst Appl. 1998;13(4):18-28.
返回引文位置Google Scholar
百度学术
万方数据
36.
Aggarwal CC . Neural networks and deep learning. Cham:Springer; 2018.
返回引文位置
37.
Schapire RE . The boosting approach to machine learning: an overview. In:Nonlinear estimation and classification. Cham:Springer; 2003. p. 149-71.
返回引文位置Google Scholar
百度学术
万方数据
38.
Bentéjac C , Csörgő A , Martínez-Muñoz G . A comparative analysis of gradient boosting algorithms. Artif Intell Rev. 2021;54:1937-67.
返回引文位置Google Scholar
百度学术
万方数据
39.
Khan AA , Chaudhari O , Chandra R . A review of ensemble learning and data augmentation models for class imbalanced problems: combination, implementation an d evaluation . Expert Syst Appl. 2024;244:122778.
返回引文位置Google Scholar
百度学术
万方数据
40.
Therneau T , Atkinson B , Ripley B . rpart: Recursive Partitioning and Regression Trees 2023. https://cran.r-project.org/web/packages/rpart/index.htmlAccessed 7 Aug 2024
返回引文位置Google Scholar
百度学术
万方数据
41.
Meyer D . e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly: E1071), TU Wien. 2023. https://cran.r-project.org/web/packages/e1071/index.htmlAccessed 20 June 2024
返回引文位置Google Scholar
百度学术
万方数据
42.
Ripley B . nnet: Feed-Forward Neural Networks and Multinomial Log-Linear Models 2023. https://cran.r-project.org/web/packages/nnet/index.htmlAccessed 18 June 2024
返回引文位置Google Scholar
百度学术
万方数据
43.
Kalinowski T . keras: R Interface to ’Keras’ 2024. https://cran.r-project.org/web/packages/keras/index.htmlAccessed 7 Aug 2024
返回引文位置Google Scholar
百度学术
万方数据
44.
Culp M . ada: The R Package Ada for Stochastic Boosting 2016. https://cran.r-project.org/web/packages/ada/index.htmlAccessed 18 June 2024
返回引文位置Google Scholar
百度学术
万方数据
45.
Prokhorenkova L , Gusev G , Vorobev A , Dorogush AV , Gulin A . CatBoost unbiased boosting with categorical features. Advances in neural information processing systems. 2018;2018(31):1.
返回引文位置Google Scholar
百度学术
万方数据
46.
Shi Y . lightgbm: Light Gradient Boosting Machine 2024:. [cited 2024 June 17 ] . Available from https://cran.r-project.org/web/packages/lightgbm/index.html
返回引文位置Google Scholar
百度学术
万方数据
47.
Chen T . xgboost: Extreme Gradient Boosting 2024:. [cited 2024 June 17 ] . Available from https://cran.r-project.org/web/packages/xgboost/index.html
返回引文位置Google Scholar
百度学术
万方数据
48.
Wickham H . ggplot2. Wiley Interdiscip Rev Comput Statis. 2011;3(2):180-5.
返回引文位置Google Scholar
百度学术
万方数据
49.
DeLong ER , DeLong DM , Clarke-Pearson DL . Comparing the areas under two or more correlated receiver operating characteristic curves: a non-parametric approach. Biometrics. 1988;44(3):837-45.
返回引文位置Google Scholar
百度学术
万方数据
50.
Lee IK , Huang CH , Huang WC , Chen YC , Tsai CY , Chang K ,et al. Prognostic factors in adult patients with dengue: developing risk scoring models and emphasizing factors associated with death ≤7 Days after illness onset and ≤3 days after presentation. J Clin Med. 2018. https://doi.org/10.3390/jcm7110396 .
返回引文位置Google Scholar
百度学术
万方数据
51.
Wakimoto MD , Camacho LA , Guaraldo L , Damasceno LS , Brasil P . Dengue in children: a systematic review of clinical and laboratory factors associated with severity. Expert Rev Anti Infect Ther. 2015;13(12):1441-56.
返回引文位置Google Scholar
百度学术
万方数据
52.
Sami CA , Tasnim R , Hassan SS , Khan AH , Yasmin R , Monir-Uz-Zaman M ,et al. Clinical profile and early severity predictors of dengue fever: current trends for the deadliest dengue infection in Bangladesh in 2022. IJID Reg. 2023;9:42-8.
返回引文位置Google Scholar
百度学术
万方数据
53.
Ming DK , Hernandez B , Sangkaew S , Vuong NL , Lam PK , Nguyet NM ,et al. Applied machine learning for the risk-stratification and clinical decision support of hospitalised patients with dengue in Vietnam. PLOS Digital Health. 2022;1(1): e0 000005 .
返回引文位置Google Scholar
百度学术
万方数据
54.
Ong SQ , Isawasan P , Ngesom AMM , Shahar H , AmM L , Nair G . Predicting dengue transmission rates by comparing different machine learning models with vector indices and meteorological data. Sci Rep. 2023. https://doi.org/10.1038/s41598-023-46342-2 .
返回引文位置Google Scholar
百度学术
万方数据
55.
Madewell ZJ , Rodriguez DM , Thayer MB , Rivera-Amill V , Torres Aponte J , Marzan-Rodriguez M ,et al. Machine learning for improved dengue diagnosis Puerto Rico. medRxiv. 2024;403:667.
返回引文位置Google Scholar
百度学术
万方数据
56.
Chaw JK , Chaw SH , Quah CH , Sahrani S , Ang MC , Zhao Y ,et al. A predictive analytics model using machine learning algorithms to estimate the risk of shock development among dengue patients. Healthc Anal. 2024;5:100290.
返回引文位置Google Scholar
百度学术
万方数据
57.
Liu YE , Saul S , Rao AM , Robinson ML , Agudelo Rojas OL , Sanz AM ,et al. An 8-gene machine learning model improves clinical prediction of severe dengue progression. Genome Med. 2022;14(1):33.
返回引文位置Google Scholar
百度学术
万方数据
58.
Hung SJ , Tsai HP , Wang YF , Ko WC , Wang JR , Huang SW . Assessment of the risk of severe dengue using intrahost viral population in dengue virus serotype 2 patients via machine learning. Front Cell Infect Microbiol. 2022;12:831281.
返回引文位置Google Scholar
百度学术
万方数据
59.
Cramer EY , Ray EL , Lopez VK , Bracher J , Brennen A , Castro Rivadeneira AJ ,et al. Evaluation of individual and ensemble probabilistic forecasts of COVID-19 mortality in the United States. Proc Natl Acad Sci U S A. 2022;119(15):e2113561119.
返回引文位置Google Scholar
百度学术
万方数据
60.
Dean NE , Pastore YPA , Madewell ZJ , Cummings DAT , Hitchings MDT , Joshi K ,et al. Ensemble forecast modeling for the design of COVID-19 vaccine efficacy trials. Vaccine. 2020;38(46):7213-6.
返回引文位置Google Scholar
百度学术
万方数据
61.
Ho TS , Weng TC , Wang JD , Han HC , Cheng HC , Yang CC ,et al. Comparing machine learning with case-control models to identify confirmed dengue cases. PLoS Negl Trop Dis. 2020;14(11):e0008843.
返回引文位置Google Scholar
百度学术
万方数据
备注信息
A
Zachary J. Madewell vogdef.cabdc0kco
B

Full list of author information is available at the end of the article

C

Conception and design of the study: ZJM, MBT, and JMW. Acquisition of data: VR. Analysis and interpretation of data: ZJM. Drafting the article: ZJM. Revising the article critically for important intellectual content: ZJM, DR, MBT, VR, GP, LEA, and JMW. All authors have made substantial contributions to the study. All authors read and approved the final manuscript.

D
10.1186/s40249-025-01273-0.T003
AdaBoost Adaptive boosting
aOR Adjusted odds ratio
AFI Acute febrile illness
ANN Artificial neural networks
AUC-ROC Area under the receiver operating characteristic curve
CatBoost Categorical boosting
CBC Complete blood count
CHIKV Chikungunya virus
CSPF Convalescent sample processing form
CEMI Centro de Emergencia y Medicina Integrada
CIF Case investigation form
CMESL Centro Médico Episcopal San Lucas
DENV Dengue virus
DT Decision trees
ELISA Enzyme-linked immunosorbent assay
HESL Hospital Episcopal San Lucas
IgG Immunoglobulin G
IgM Immunoglobulin M
IQR Interquartile range
KNN K-nearest neighbors
LightGBM Light gradient boosting machine
LR Logistic regression
ML Machine learning
NP Nasopharyngeal
NPV Negative predictive value
OP Oropharyngeal
PPV Positive predictive value
RT-PCR Reverse transcription polymerase chain reaction
SEDSS Sentinel enhanced dengue surveillance system
SHAP SHapley Additive exPlanations
SVM Support vector machines
XGBoost EXtreme gradient boosting
WHO World Health Organization
E
Not applicable.
F
The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the US Centers for Disease Control and Prevention.
G
Data cannot be shared publicly because data cannot be deidentified at the granular level of analyses performed. Data are available from the CDC management team (contact: dengue@cdc.gov) for researchers who meet the criteria for access to confidential data.
H
Declarations
I
The Institutional Review Boards at the Centers for Disease Control and Prevention (CDC), Auxilio Mutuo, and Ponce Medical School Foundation approved the SEDSS study protocols 6214, and 120308-VR/2311173707, respectively. Written consent to participate was obtained from all adult participants and emancipated minors. For minors aged 14 to 20 years, written consent was obtained, and for those aged 7 to 13 years, parental written consent and participant assent were obtained.
J
Not applicable.
K
The authors declare no conflict of interests.
评论 （0条）
注册
登录
时间排序
暂无评论，发表第一条评论抢沙发
MedAI助手（体验版）
文档即答
智问智答
机器翻译
回答内容由人工智能生成，我社无法保证其准确性和完整性，该生成内容不代表我们的态度或观点，仅供参考。
生成快照
文献快照

你好，我可以帮助您更好的了解本文，请向我提问您关注的问题。

0/2000

《中华医学会杂志社用户协议》 | 《隐私政策》

《SparkDesk 用户协议》 | 《SparkDesk 隐私政策》

网信算备340104764864601230055号 | 网信算备340104726288401230013号

技术支持：

历史对话
本文全部
还没有聊天记录
设置
模式
纯净模式沉浸模式
字号
小中大