Applied machine learning for the risk-stratification and clinical decision support of hospitalised patients with dengue in Vietnam

Machine learning for predicting severe dengue in Puerto Rico

BACKGROUND

Distinguishing 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.

METHODS

We 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.

RESULTS

Among 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/mm3), 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%.

CONCLUSIONS

ML 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.

Status of Digital Health Technology Adoption in 5 Vietnamese Hospitals: Cross-Sectional Assessment

BACKGROUND

Digital health technologies (DHTs) have been recognized as a key solution to help countries, especially those in the low- and middle-income group, to achieve the Sustainable Development Goals (SDGs) and the World Health Organization's (WHO) Triple Billion Targets. In hospital settings, DHTs need to be designed and implemented, considering the local context, to achieve usability and sustainability. As projects such as the Vietnam ICU Translational Applications Laboratory are seeking to integrate new digital technologies in the Vietnamese critical care settings, it is important to understand the current status of DHT adoption in Vietnamese hospitals.

OBJECTIVE

We aimed to explore the current digital maturity in 5 Vietnamese public hospitals to understand their readiness in implementing new DHTs.

METHODS

We assessed the adoption of some key DHTs and infrastructure in 5 top-tier public hospitals in Vietnam using a questionnaire adapted from the Vietnam Health Information Technology (HIT) Maturity Model. The questionnaire was answered by the heads of the hospitals' IT departments, with follow-up for clarifications and verifications on some answers. Descriptive statistics demonstrated on radar plots and tile graphs were used to visualize the data collected.

RESULTS

Hospital information systems (HIS), laboratory information systems (LIS), and radiology information systems-picture archiving and communication systems (RIS-PACS) were implemented in all 5 hospitals, albeit at varied digital maturity levels. At least 50% of the criteria for LIS in the Vietnam HIT Maturity Model were satisfied by the hospitals in the assessment. However, this threshold was only met by 80% and 60% of the hospitals with regard to HIS and RIS-PACS, respectively. Two hospitals were not using any electronic medical record (EMR) system or fulfilling any extra digital capability, such as implementing clinical data repositories (CDRs) and clinical decision support systems (CDSS). No hospital reported sharing clinical data with other organizations using Health Level Seven (HL7) standards, such as Continuity of Care Document (CCD) and Clinical Document Architecture (CDA), although 2 (40%) reported their systems adopted these standards. Of the 5 hospitals, 4 (80%) reported their RIS-PACS adopted the Digital Imaging and Communications in Medicine (DICOM) standard.

CONCLUSIONS

The 5 major Vietnamese public hospitals in this assessment have widely adopted information systems, such as HIS, LIS, and RIS-PACS, to support administrative and clinical tasks. Although the adoption of EMR systems is less common, their implementation revolves around data collection, management, and access to clinical data. Secondary use of clinical data for decision support through the implementation of CDRs and CDSS is limited, posing a potential barrier to the integration of external DHTs into the existing systems. However, the wide adoption of international standards, such as HL7 and DICOM, is a facilitator for the adoption of new DHTs in these hospitals.

Predicting deterioration in dengue using a low cost wearable for continuous clinical monitoring

Close vital signs monitoring is crucial for the clinical management of patients with dengue. We investigated performance of a non-invasive wearable utilising photoplethysmography (PPG), to provide real-time risk prediction in hospitalised individuals. We performed a prospective observational clinical study in Vietnam between January 2020 and October 2022: 153 patients were included in analyses, providing 1353 h of PPG data. Using a multi-modal transformer approach, 10-min PPG waveform segments and basic clinical data (age, sex, clinical features on admission) were used as features to continuously forecast clinical state 2 h ahead. Prediction of low-risk states (17,939/80,843; 22.1%), defined by NEWS2 and mSOFA < 6, was associated with an area under the precision-recall curve of 0.67 and an area under the receiver operator curve of 0.83. Implementation of such interventions could provide cost-effective triage and clinical care in dengue, offering opportunities for safe ambulatory patient management.

Proceedings of the dengue endgame summit: Imagining a world with dengue control

The first dengue "endgame" summit was held in Syracuse, NY over August 9 and 10, 2023. Organized and hosted by the Institute for Global Health and Translational Sciences at SUNY Upstate Medical University, the gathering brought together researchers, clinicians, drug and vaccine developers, government officials, and other key stakeholders in the dengue field for a highly collaborative and discussion-oriented event. The objective of the gathering was to discuss the current state of dengue around the world, what dengue "control" might look like, and what a potential roadmap might look like to achieve functional dengue control. Over the course of 7 sessions, speakers with a diverse array of expertise highlighted both current and historic challenges associated with dengue control, the state of dengue countermeasure development and deployment, as well as fundamental virologic, immunologic, and medical barriers to achieving dengue control. While sustained eradication of dengue was considered challenging, attendees were optimistic that significant reduction in the burden of dengue can be achieved by integration of vector control with effective application of therapeutics and vaccines.

A simple nomogram to predict dengue shock syndrome: A study of 4522 south east Asian children

Dengue shock syndrome (DSS) substantially worsens the prognosis of children with dengue infection. This study aimed to develop a simple clinical tool to predict the risk of DSS. A cohort of 2221 Thai children with a confirmed dengue infection who were admitted to King Chulalongkorn Memorial Hospital between 1987 and 2007 was conducted. Another data set from a previous publication comprising 2,301 Vietnamese children with dengue infection was employed to create a pooled data set, which was randomly split into training (n = 3182), testing (n = 697) and validating (n = 643) datasets. Logistic regression was compared to alternative machine learning algorithms to derive the most predictive model for DSS. 4522 children, including 899 DSS cases (758 Thai and 143 Vietnamese children) with a mean age of 9.8 ± 3.4 years, were analyzed. Among the 12 candidate clinical parameters, the Bayesian Model Averaging algorithm retained the most predictive subset of five covariates, including body weight, history of vomiting, liver size, hematocrit levels, and platelet counts. At an Area Under the Curve (AUC) value of 0.85 (95% CI: 0.81-0.90) in testing data set, logistic regression outperformed random forest, XGBoost and support vector machine algorithms, with AUC values being 0.82 (0.77-0.88), 0.82 (0.76-0.88), and 0.848 (0.81-0.89), respectively. At its optimal threshold, this model had a sensitivity of 0.71 (0.62-0.80), a specificity of 0.84 (0.81-0.88), and an accuracy of 0.82 (0.78-0.85) on validating data set with consistent performance across subgroup analyses by age and gender. A logistic regression-based nomogram was developed to facilitate the application of this model. This work introduces a simple and robust clinical model for DSS prediction that is well-tailored for children in resource-limited settings.

A pilot study investigating severe community-acquired febrile illness through implementation of an innovative microbiological and nucleic acid amplification testing strategy in Timor-Leste (ISIN-MANAS-TL)

Objectives

Acute febrile illness (AFI) causes significant health-seeking, morbidity, and mortality in Southeast Asia. This pilot study aimed to describe presentation, etiology, treatment, and outcomes of patients with AFI at one hospital in Timor-Leste and assessing the feasibility of conducting larger studies in this setting.

Methods

Patients attending Hospital Nacional Guido Valadares with tympanic or axillary temperature ≥37.5°C in whom a blood culture was taken as part of routine clinical care were eligible. Participants were followed up daily for 10 days and again after 30 days. Whole blood was analyzed using a real-time quantitative polymerase chain reaction assay detecting dengue virus serotypes 1-4 and other arthropod-borne infections.

Results

A total of 82 participants were recruited. Polymerase chain reaction testing was positive for dengue in 14 of 82 (17.1%) participants and blood culture identified a bacterial pathogen in three of 82 (3.7%) participants. Follow-up was completed by 75 of 82 (91.5%) participants. High rates of hospital admission (58 of 82, 70.7%), broad-spectrum antimicrobial treatment (34 of 82, 41.5%), and mortality (9 of 82, 11.0%) were observed.

Conclusions

Patients with AFI experience poor clinical outcomes. Prospective observational and interventional studies assessing interventions, such as enhanced diagnostic testing, clinical decision support tools, or antimicrobial stewardship interventions, are required and would be feasible to conduct in this setting.

Mapping patient pathways and understanding clinical decision-making in dengue management to inform the development of digital health tools

BACKGROUND

Dengue is a common viral illness and severe disease results in life-threatening complications. Healthcare services in low- and middle-income countries treat the majority of dengue cases worldwide. However, the clinical decision-making processes which result in effective treatment are poorly characterised within this setting. In order to improve clinical care through interventions relating to digital clinical decision-support systems (CDSS), we set out to establish a framework for clinical decision-making in dengue management to inform implementation.

METHODS

We utilised process mapping and task analysis methods to characterise existing dengue management at the Hospital for Tropical Diseases, Ho Chi Minh City, Vietnam. This is a tertiary referral hospital which manages approximately 30,000 patients with dengue each year, accepting referrals from Ho Chi Minh city and the surrounding catchment area. Initial findings were expanded through semi-structured interviews with clinicians in order to understand clinical reasoning and cognitive factors in detail. A grounded theory was used for coding and emergent themes were developed through iterative discussions with clinician-researchers.

RESULTS

Key clinical decision-making points were identified: (i) at the initial patient evaluation for dengue diagnosis to decide on hospital admission and the provision of fluid/blood product therapy, (ii) in those patients who develop severe disease or other complications, (iii) at the point of recurrent shock in balancing the need for fluid therapy with complications of volume overload. From interviews the following themes were identified: prioritising clinical diagnosis and evaluation over existing diagnostics, the role of dengue guidelines published by the Ministry of Health, the impact of seasonality and caseload on decision-making strategies, and the potential role of digital decision-support and disease scoring tools.

CONCLUSIONS

The study highlights the contemporary priorities in delivering clinical care to patients with dengue in an endemic setting. Key decision-making processes and the sources of information that were of the greatest utility were identified. These findings serve as a foundation for future clinical interventions and improvements in healthcare. Understanding the decision-making process in greater detail also allows for development and implementation of CDSS which are suited to the local context.