Fig 1.
Selection workflow of dengue patients with warning signs that required hospitalizations and patients with no warning signs and no hospitalization required for both discovery and validation cohort.
Table 1.
Demographic descriptions of dengue RT-PCR positive patients classified according to designated clinical outcomes.
Fig 2.
Laboratory characteristics (A-Viral copy number at Day 1–3; B-Platelet count; C-Lymphocytes count) of hospitalized dengue patients with warning signs (WS + Hosp. Group) compared to non-hospitalized patients with no warning signs (Non-WS + Non-Hosp. Group). P-value (P) is shown only for statistically significant comparisons on Day 1–3 and Day 4–7 between the two groups.
Fig 3.
Sensitivity and specificity plots of Model 13 (A) and Model 14 (B) with the varying probability cut-offs using the discovery cohort. The y-axis shows the true positive rate (sensitivity in blue) and the true negative rate (specificity in red) of the model’s capability at different probability cutoff on the x-axis. The probability cutoff range (x-axis) allows the investigators to choose how sensitive and specific they want the model to be at different setting depending on the aim of the models.
Table 2.
Early prognostic models of warning signs and hospitalization from the discovery cohort.
Fig 4.
Laboratory characteristics of patients in the validation cohort who are predicted at Day 1–3 p.f. to develop warning signs and require hospitalization at post Day 1–3 p.f. using Model 13 (A, B and C) and Model 14 (D, E and F). Viral Ct level at Day 1–3 (A, D), platelet (B, E) and lymphocyte (C, F) levels of the patients who were predicted into either the “Non-WS + Non-Hosp” group or “WS + Hosp” group. WS- Warning Signs; Hosp-Hospitalisation. P-value (P) is shown only for statistically significant comparisons on Day 1–3 and Day 4–7 between the two predicted groups.
Table 3.
Early prognostic performance of the top selected models with an independent validation cohort across a range of probability cutoff.