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

BACKGROUND

Identifying patients at risk of dengue shock syndrome (DSS) is vital for effective healthcare delivery. This can be challenging in endemic settings because of high caseloads and limited resources. Machine learning models trained using clinical data could support decision-making in this context.

METHODS

We developed supervised machine learning prediction models using pooled data from adult and paediatric patients hospitalised with dengue. Individuals from 5 prospective clinical studies in Ho Chi Minh City, Vietnam conducted between 12th April 2001 and 30th January 2018 were included. The outcome was onset of dengue shock syndrome during hospitalisation. Data underwent random stratified splitting at 80:20 ratio with the former used only for model development. Ten-fold cross-validation was used for hyperparameter optimisation and confidence intervals derived from percentile bootstrapping. Optimised models were evaluated against the hold-out set.

FINDINGS

The final dataset included 4,131 patients (477 adults and 3,654 children). DSS was experienced by 222 (5.4%) of individuals. Predictors were age, sex, weight, day of illness at hospitalisation, indices of haematocrit and platelets over first 48 hours of admission and before the onset of DSS. An artificial neural network model (ANN) model had best performance with an area under receiver operator curve (AUROC) of 0.83 (95% confidence interval [CI], 0.76-0.85) in predicting DSS. When evaluated against the independent hold-out set this calibrated model exhibited an AUROC of 0.82, specificity of 0.84, sensitivity of 0.66, positive predictive value of 0.18 and negative predictive value of 0.98.

INTERPRETATION

The study demonstrates additional insights can be obtained from basic healthcare data, when applied through a machine learning framework. The high negative predictive value could support interventions such as early discharge or ambulatory patient management in this population. Work is underway to incorporate these findings into an electronic clinical decision support system to guide individual patient management.

Higher Plasma Viremia in the Febrile Phase Is Associated With Adverse Dengue Outcomes Irrespective of Infecting Serotype or Host Immune Status: An Analysis of 5642 Vietnamese Cases

BACKGROUND

One of the generally accepted constructs of dengue pathogenesis is that clinical disease severity is at least partially dependent upon plasma viremia, yet data on plasma viremia in primary versus secondary infections and in relation to clinically relevant endpoints remain limited and contradictory.

METHODS

Using a large database comprising detailed clinical and laboratory characterization of Vietnamese participants enrolled in a series of research studies executed over a 15-year period, we explored relationships between plasma viremia measured by reverse transcription-polymerase chain reaction and 3 clinically relevant endpoints-severe dengue, plasma leakage, and hospitalization-in the dengue-confirmed cases. All 4 dengue serotypes and both primary and secondary infections were well represented. In our logistic regression models we allowed for a nonlinear effect of viremia and for associations between viremia and outcome to differ by age, serotype, host immune status, and illness day at study enrollment.

RESULTS

Among 5642 dengue-confirmed cases we identified 259 (4.6%) severe dengue cases, 701 (12.4%) patients with plasma leakage, and 1441 of 4008 (40.0%) patients recruited in outpatient settings who were subsequently hospitalized. From the early febrile phase onwards, higher viremia increased the risk of developing all 3 endpoints, but effect sizes were modest (ORs ranging from 1.12-1.27 per 1-log increase) compared with the effects of a secondary immune response (ORs, 1.67-7.76). The associations were consistent across age, serotype, and immune status groups, and in the various sensitivity and subgroup analyses we undertook.

CONCLUSIONS

Higher plasma viremia is associated with increased dengue severity, regardless of serotype or immune status.

Using NS1 Flavivirus Protein Microarray to Infer Past Infecting Dengue Virus Serotype and Number of Past Dengue Virus Infections in Vietnamese Individuals

Background

In recent years, researchers have had an increased focus on multiplex microarray assays, in which antibodies are measured against multiple related antigens, for use in seroepidemiological studies to infer past transmission.

Methods

We assess the performance of a flavivirus microarray assay for determining past dengue virus (DENV) infection history in a dengue-endemic setting, Vietnam. We tested the microarray on samples from 1 and 6 months postinfection from DENV-infected patients (infecting serotype was determined using reverse-transcription polymerase chain reaction during acute, past primary, and secondary infection assessed using plaque reduction neutralization tests 6 months postinfection).

Results

Binomial models developed to discriminate past primary from secondary infection using the protein microarray (PMA) titers had high area under the curve (0.90–0.97) and accuracy (0.84–0.86). Multinomial models developed to identify most recent past infecting serotype using PMA titers performed well in those with past primary infection (average test set: κ = 0.85, accuracy of 0.92) but not those with past secondary infection (κ = 0.24, accuracy of 0.45).

Conclusions

Our results suggest that the microarray will be useful in seroepidemiological studies aimed at classifying the past infection history of individuals (past primary vs secondary and serotype of past primary infections) and thus inferring past transmission intensity of DENV in dengue-endemic settings. Future work to validate these models should be undertaken in different transmission settings and with samples later after infection.

The value of daily platelet counts for predicting dengue shock syndrome: Results from a prospective observational study of 2301 Vietnamese children with dengue

BACKGROUND

Dengue is the most important mosquito-borne viral infection to affect humans. Although it usually manifests as a self-limited febrile illness, complications may occur as the fever subsides. A systemic vascular leak syndrome that sometimes progresses to life-threatening hypovolaemic shock is the most serious complication seen in children, typically accompanied by haemoconcentration and thrombocytopenia. Robust evidence on risk factors, especially features present early in the illness course, for progression to dengue shock syndrome (DSS) is lacking. Moreover, the potential value of incorporating serial haematocrit and platelet measurements in prediction models has never been assessed.

METHODOLOGY/PRINCIPAL FINDINGS

We analyzed data from a prospective observational study of Vietnamese children aged 5-15 years admitted with clinically suspected dengue to the Hospital for Tropical Diseases in Ho Chi Minh City between 2001 and 2009. The analysis population comprised all children with laboratory-confirmed dengue enrolled between days 1-4 of illness. Logistic regression was the main statistical model for all univariate and multivariable analyses. The prognostic value of daily haematocrit levels and platelet counts were assessed using graphs and separate regression models fitted on each day of illness. Among the 2301 children included in the analysis, 143 (6%) progressed to DSS. Significant baseline risk factors for DSS included a history of vomiting, higher temperature, a palpable liver, and a lower platelet count. Prediction models that included serial daily platelet counts demonstrated better ability to discriminate patients who developed DSS from others, than models based on enrolment information only. However inclusion of daily haematocrit values did not improve prediction of DSS.

CONCLUSIONS/SIGNIFICANCE

Daily monitoring of platelet counts is important to help identify patients at high risk of DSS. Development of dynamic prediction models that incorporate signs, symptoms, and daily laboratory measurements, could improve DSS prediction and thereby reduce the burden on health services in endemic areas.

A Prognostic Model for Development of Profound Shock among Children Presenting with Dengue Shock Syndrome

PURPOSE

To identify risk factors and develop a prediction model for the development of profound and recurrent shock amongst children presenting with dengue shock syndrome (DSS).

METHODS

We analyzed data from a prospective cohort of children with DSS recruited at the Paediatric Intensive Care Unit of the Hospital for Tropical Disease in Ho Chi Minh City, Vietnam. The primary endpoint was "profound DSS", defined as ≥2 recurrent shock episodes (for subjects presenting in compensated shock), or ≥1 recurrent shock episodes (for subjects presenting initially with decompensated/hypotensive shock), and/or requirement for inotropic support. Recurrent shock was evaluated as a secondary endpoint. Risk factors were pre-defined clinical and laboratory variables collected at the time of presentation with shock. Prognostic model development was based on logistic regression and compared to several alternative approaches.

RESULTS

The analysis population included 1207 children of whom 222 (18%) progressed to "profound DSS" and 433 (36%) had recurrent shock. Independent risk factors for both endpoints included younger age, earlier presentation, higher pulse rate, higher temperature, higher haematocrit and, for females, worse hemodynamic status at presentation. The final prognostic model for "profound DSS" showed acceptable discrimination (AUC=0.69 for internal validation) and calibration and is presented as a simple score-chart.

CONCLUSIONS

Several risk factors for development of profound or recurrent shock among children presenting with DSS were identified. The score-chart derived from the prognostic models should improve triage and management of children presenting with DSS in dengue-endemic areas.

Clinical characteristics of Dengue shock syndrome in Vietnamese children: a 10-year prospective study in a single hospital

BACKGROUND

Dengue shock syndrome (DSS) is a severe manifestation of dengue virus infection that particularly affects children and young adults. Despite its increasing global importance, there are no prospective studies describing the clinical characteristics, management, or outcomes of DSS.

METHODS

We describe the findings at onset of shock and the clinical evolution until discharge or death, from a comprehensive prospective dataset of 1719 Vietnamese children with laboratory-confirmed DSS managed on a single intensive care unit between 1999 and 2009.

RESULTS

The median age of patients was 10 years. Most cases had secondary immune responses, with only 6 clear primary infections, and all 4 dengue virus serotypes were represented during the 10-year study. Shock occurred commonly between days 4 and 6 of illness. Clinical signs and symptoms were generally consistent with empirical descriptions of DSS, although at presentation 153 (9%) were still febrile and almost one-third had no bleeding. Overall, 31 (2%) patients developed severe bleeding, primarily from the gastrointestinal tract, 26 of whom required blood transfusion. Only 8 patients died, although 123 of 1719 (7%) patients had unrecordable blood pressure at presentation and 417 of the remaining 1596 (26%) were hypotensive for age. The majority recovered well with standard crystalloid resuscitation or following a single colloid infusion. All cases were classified as severe dengue, while only 70% eventually fulfilled all 4 criteria for the 1997 World Health Organization classification of dengue hemorrhagic fever.

CONCLUSIONS

With prompt intervention and assiduous clinical care by experienced staff, the outcome of this potentially fatal condition can be excellent.

Assessment of microalbuminuria for early diagnosis and risk prediction in dengue infections

BACKGROUND

Dengue is the most important arboviral infection of humans. Following an initial febrile period, a small proportion of infected patients develop a vasculopathy, with children at particular risk for severe vascular leakage and shock. Differentiation between dengue and other common childhood illnesses is difficult during the early febrile phase, and risk prediction for development of shock is poor. The presence of microalbuminuria is recognized as a useful early predictor for subsequent complications in a number of other disorders with vascular involvement. Significant proteinuria occurs in association with dengue shock syndrome and it is possible that early-phase microalbuminuria may be helpful both for diagnosis of dengue and for identification of patients likely to develop severe disease.

METHODOLOGY/PRINCIPAL FINDINGS

We measured formal urine albumin to creatinine ratios (UACRs) in daily samples obtained from a large cohort of children with suspected dengue recruited at two outpatient clinics in Ho Chi Minh City, Vietnam. Although UACRs were increased in the 465 confirmed dengue patients, with a significant time trend showing peak values around the critical period for dengue-associated plasma leakage, urine albumin excretion was also increased in the comparison group of 391 patients with other febrile illnesses (OFI). The dengue patients generally had higher UACRs than the OFI patients, but microalbuminuria, using the conventional cutoff of 30 mg albumin/g creatinine discriminated poorly between the two diagnostic groups in the early febrile phase. Secondly UACRs did not prove useful in predicting either development of warning signs for severe dengue or need for hospitalization.

CONCLUSION/SIGNIFICANCE

Low-level albuminuria is common, even in relatively mild dengue infections, but is also present in many OFIs. Simple point-of-care UACR tests are unlikely to be useful for early diagnosis or risk prediction in dengue endemic areas.