Search

BackgroundAt least a third of dengue patients develop plasma leakage with increased risk of life-threatening complications. Predicting plasma leakage using laboratory parameters obtained in early infection as means of triaging patients for hospital admission is important for resource-limited settings.MethodsA Sri Lankan cohort including 4,768 instances of clinical data from N = 877 patients (60.3% patients with confirmed dengue infection) recorded in the first 96 hours of fever was considered. After excluding incomplete instances, the dataset was randomly split into a development and a test set with 374 (70%) and 172 (30%) patients, respectively. From the development set, five most informative features were selected using the minimum description length (MDL) algorithm. Random forest and light gradient boosting machine (LightGBM) were used to develop a classification model using the development set based on nested cross validation. An ensemble of the learners via average stacking was used as the final model to predict plasma leakage.ResultsLymphocyte count, haemoglobin, haematocrit, age, and aspartate aminotransferase were the most informative features to predict plasma leakage. The final model achieved the area under the receiver operating characteristics curve, AUC = 0.80 with positive predictive value, PPV = 76.9%, negative predictive value, NPV = 72.5%, specificity = 87.9%, and sensitivity = 54.8% on the test set.ConclusionThe early predictors of plasma leakage identified in this study are similar to those identified in several prior studies that used non-machine learning based methods. However, our observations strengthen the evidence base for these predictors by showing their relevance even when individual data points, missing data and non-linear associations were considered. Testing the model on different populations using these low-cost observations would identify further strengths and limitations of the presented model.

Leptospirosis is endemic in Sri Lanka. There is a need for updated seroprevalence studies in endemic areas, to improve the understanding of disease dynamics, risk factors, control methods, and for clinical diagnosis. The cut-off titres for the microscopic agglutination test (MAT) for diagnosis of acute leptospirosis depend on community seroprevalence, and can vary based on locality and serovar. This study aimed to identify the seroprevalence, geographical determinants, and associations of seropositivity of leptospirosis in the district of Colombo in Sri Lanka, and to determine diagnostic cut-off titres for MAT in the community studied. This study utilized a stratified cluster sampling model in the Colombo district of Sri Lanka, to sample individuals living in urban and semi-urban areas. Serovar specific MAT titres were measured on recruited individuals using a panel of saprophytic (Leptospira biflexa) and 11 pathogenic Leptospira spp. serovars. Associations between environmental risk factors and MAT positivity were examined, with location mapping using GIS software. A total of 810 individuals were included. The mean age was 51.71 years (SD 14.02) with male predominance (60%). A total of 429 (53%) tested positive at a titer of 1/40 or more for the saprophytic Leptospira biflexa serovar Patoc. Pathogenic serovar MAT was positive at a titer of 1/40 or more for at least one serovar in 269 (33.2%) individuals. From the perspective of screening for clinical disease, serovar-specific cut-off titres of 1/80 for Leptospira spp. serovars Hebdomadis, Icterohaemorrhagiae, Pomona, Ratnapura and Patoc, 1/160 for serovars Pyrogenes and Cynopteri, and 1/40 for other serovars were determined, based on the 75th quartile MAT titre for each serovar. Serovar Pyrogenes (15.9%) had the highest seroprevalence, with serovars Ratnapura, Bankinang and Australis accounting for 9.9%, 9.6% and 9.3% respectively. When the proposed new cut-offs were applied, Bankinang(9.6%) Australis(9.3%), Pyrogenes(6.9%) and Ratnapura(6.9%) were the most prevalent serovars. No significant differences in seroprevalence or serovar patterns were noted between urban and semi-urban settings. Individuals seropositive for Australis, Ratnapura and Icterohaemorrhagiae were clustered around main water bodies as well as around smaller tributaries and paddy fields. Those positive for the serovar Pyrogenes were clustered around inland tributaries, smaller water sources and paddy fields. Associations of MAT positivity included high risk occupational exposure, environmental exposure including exposure to floods, bathing in rivers and lakes, using well-water for bathing, contact with stagnant water, propensity to skin injuries, presence of rats in the vicinity, and proximity to water sources. For pathogenic serovars, high-risk occupational exposure remained statistically significant following adjustment for other factors (adjusted OR = 2.408, CI 1.711 to 3.388; p

Leptospirosis is a zoonotic infection with significant morbidity and mortality. The clinical presentation of leptospirosis is known to mimic the clinical profile of other prevalent tropical fevers. Laboratory confirmation of leptospirosis is based on the reference standard microscopic agglutination test (MAT), direct demonstration of the organism, and isolation by culture and DNA detection by polymerase chain reaction (PCR) amplification. However these methods of confirmation are not widely available in resource limited settings where the infection is prevalent, and reliance is placed on clinical features for provisional diagnosis. In this prospective study, we attempted to develop a model for diagnosis of leptospirosis, based on clinical features and standard laboratory test results.The diagnostic score was developed based on data from a prospective multicentre study in two hospitals in the Western Province of Sri Lanka. All patients presenting to these hospitals with a suspected diagnosis of leptospirosis, based on the WHO surveillance criteria, were recruited. Confirmed disease was defined as positive genus specific MAT (Leptospira biflexa). A derivation cohort and a validation cohort were randomly selected from available data. Clinical and laboratory manifestations associated with confirmed leptospirosis in the derivation cohort were selected for construction of a multivariate regression model with correlation matrices, and adjusted odds ratios were extracted for significant variables. The odds ratios thus derived were subsequently utilized in the criteria model, and sensitivity and specificity examined with ROC curves.A total of 592 patients were included in the final analysis with 450 (180 confirmed leptospirosis) in the derivation cohort and 142 (52 confirmed leptospirosis) in the validation cohort. The variables in the final model were: history of exposure to a possible source of leptospirosis (adjusted OR = 2.827; 95% CI = 1.517-5.435; p = 0.001) serum creatinine > 150 micromol/l (adjusted OR = 2.735; 95% CI = 1.374-4.901; p = 0.001), neutrophil differential percentage > 80.0% of total white blood cell count (adjusted OR 2.163; 95% CI = 1.309-3.847; p = 0.032), serum bilirubin > 30 micromol/l (adjusted OR = 1.717; 95% CI 0.938-3.456; p = 0.049) and platelet count < 85,000/mm3 (adjusted OR = 2.350; 95% CI = 1.481-4.513; p = 0.006). Hosmer-Lemeshow test for goodness of fit was 0.931. The Nagelkerke R2 was 0.622. The area under the curve (AUC) was noted as 0.762. A score value of 14 reflected a sensitivity of 0.803, specificity of 0.602, a PPV of 0.54, NPV of 0.84, a positive LR of 2.01 and a negative LR of 0.32.The above diagnostic model for diagnosis of leptospirosis is suggested for use in clinical settings. It should be further validated in clinical practice.