Diagnosing dengue fever
Motivation
Dengue fever is a common viral illness spread by mosquitoes in tropical and subtropical parts of the world. More than 5 million people are infected each year. While most people recover within one or two weeks, some cases become severe, requiring hospitalization to treat shock and internal bleeding. Roughly 5,000 people die of dengue fever each year.
While there is no medication that can cure dengue, prompt diagnosis is still helpful so patients can get supportive treatment and be closely monitored for signs of severe dengue. This data comes from a study in Vietnam which tested whether simple lab tests and symptom data can be used to accurately diagnose dengue fever, rather than relying on more expensive dengue-specific lab tests. If this were possible, doctors could diagnose patients more quickly and cheaply, before more extensive lab test results are available.
Data
The data contains symptom and test results for 5,726 patients, all children 1–15 years old in several hospitals in Vietnam. Children were eligible to enroll if they had been sick less than 72 hours and had a fever, and the attending physician thought dengue fever was a possible diagnosis. Patients whose parents consented were enrolled in the study. On enrollment, blood samples were drawn for normal blood tests (such as white blood counts) and dengue fever tests.
If the patient was hospitalized, a second blood sample was drawn for additional blood tests to see if they had dengue fever antibodies. Of those patients who were not hospitalized, 10% were randomly selected to get a second blood sample.
The definitive diagnosis of dengue fever was made by using three tests: RT-PCR, serology, and ELISA assay.
Data preview
dengue.csv
Variable descriptions
| Variable | Description |
|---|---|
| DateEnrol | Date patient was enrolled in the study. (Time is provided, but always midnight or noon; probably because Excel automatically converted the date to a time.) |
| SiteNo | Which hospital (of several hospitals participating in the study) the patient was from |
| StudyNo | Identifying number for the patient |
| Sex | Sex of the patient (1 = male, 2 = female) |
| Age | Age of the patient (years) |
| DayDisease | How many days the patient had been sick, when they were enrolled into the study |
| Vomiting | Whether the patient presented with vomiting (1 = yes, 2 = no) |
| Abdo | Whether the patient had abdominal pain (1 = yes, 2 = no) |
| Muco | Whether the patient had mucosal bleeding (1 = yes, 2 = no) |
| Skin | Whether the patient had skin bleeding (1 = yes, 2 = no) |
| Temp | Patient’s body temperature (Celsius) |
| BMI | Patient’s Body Mass Index |
| Height | Patient’s height (cm) |
| Weight | Patient’s weight (kg) |
| Flush | Whether the patient had redness of the skin caused by dilation of the capillaries |
| Hepatomegaly | Whether the patient had an enlarged liver |
| Rash | Whether the patient had a rash (“generalized macular, blanching, non-petechial rash”) |
| Injection | Whether the patient had conjunctival injection, i.e., red eyes due to prominent blood vessels in the conjuctiva |
| WBC | White blood cell count (103/mm3) |
| HCT | Hematocrit percentage |
| PLT | Platelet count (103/mm3) |
| NEUCount | Neutrophil count |
| LYMCount | Lymphocyte count |
| MONOCount | Monocyte count |
| ALB | Blood albumin concentration (g/L) |
| AST | Blood aspartat aminotransferase concentration (log2-transformed) |
| ALT | Unknown |
| CK | Blood creatine kinase concentration (log2-transformed) |
| Lab_Confirmed_Dengue | Whether the patient had laboratory-confirmed dengue fever (1 = yes, 2 = no) according to the results of RT-PCR, ELISA assay, or IgM seroconversion |
| Serotype | Which dengue fever serotype was detected (1-4), or “<LOD” if dengue levels were below the level of detection |
| NS1_PLATELIA | Result of the Platelia Dengue NS1 Ag ELISA assay, a dengue blood test (1 = positive, 2 = negative) |
| NS1_TRIP | Unclear; may be result of the BioRad NS1 Ag STRIP, a rapid antibody test for dengue fever (1 = positive, 2 = negative) |
| SeroConversion | For patients with two blood samples, whether they showed seroconversion, the change in antibodies created after a dengue infection (coding not specified) |
| Serology | For patients who showed seroconversion, whether the serological response was primary or secondary (0 = inconclusive, 1 = primary, 2 = secondary) |
| Pairedsample | Whether the patient had two different blood samples available (one at enrollment and one at early convalescence) for seroconversion testing (1 = yes, 0 = no; unknown what 2 means) |
Questions
- Compare the demographics of the patients across hospitals. Are the patients similar between hospitals, or do hospitals serve patients of different ages or sexes? Do patients at some hospitals present with more severe symptoms?
- Build a classifier to predict whether the patient had laboratory-confirmed dengue (
Lab_Confirmed_Dengue) from the blood test results, symptoms, and demographics that would be available for patients prior to diagnosis. Evaluate its performance. Is it accurate for diagnosing sensitivity? What are its false positive and false negative rates, and what is the positive predictive value?
References
Tuan NM, Nhan HT, Chau NVV, Hung NT, Tuan HM, Tram TV, et al. (2015) Sensitivity and Specificity of a Novel Classifier for the Early Diagnosis of Dengue. PLoS Neglected Tropical Diseases 9(4): e0003638. https://doi.org/10.1371/journal.pntd.0003638. Data is included in the Supporting Information, S1 Dataset, https://doi.org/10.1371/journal.pntd.0003638.s009.