Comprehensive analysis of multivariable models for predicting severe dengue prognosis: systematic review and meta-analysis

Hyelan Lee, Seungjae Hyun, Sangshin Park

Research output: Contribution to journalReview articlepeer-review

1 Scopus citations

Abstract

Dengue fever has often been confused with other febrile diseases, with deterioration occurring in the later state. Many predictive models for disease progression have been developed, but there is no definite statistical model for clinical use yet. We retrieved relevant articles through Global Health, EMBASE, MEDLINE and CINAHL Plus. The Prediction Model Risk of Bias Assessment Tool was adopted to assess potential bias and applicability. Statistical analysis was performed using Meta-DiSc software (version 1.4). Of 3184 research studies, 22 were included for the systematic review, of which 17 were selected for further meta-analysis. The pooled data of predictive accuracy was as follows: the sensitivity was 0.88 (95% CI 0.86 to 0.89), the specificity was 0.60 (95% CI 0.59 to 0.60), the positive likelihood ratio was 2.83 (95% CI 2.38 to 3.37), the negative likelihood ratio was 0.20 (95% CI 0.14 to 0.0.29) and the diagnostic OR was 16.31 (95% CI 10.25 to 25.94). The area under the summary receiver operating characteristic curve value was 0.86 (SE=0.02) with 0.79 (SE=0.02) of the Cochran Q test value. The overall predictive power of models in this study was relatively high. With careful adaption and standardization, the implementation of predictive models for severe dengue could be practical in actual clinical settings.

Original languageEnglish
Pages (from-to)149-160
Number of pages12
JournalTransactions of the Royal Society of Tropical Medicine and Hygiene
Volume117
Issue number3
DOIs
StatePublished - 1 Mar 2023

Keywords

  • dengue
  • meta-analysis
  • prediction model
  • severe dengue
  • systematic review

Fingerprint

Dive into the research topics of 'Comprehensive analysis of multivariable models for predicting severe dengue prognosis: systematic review and meta-analysis'. Together they form a unique fingerprint.

Cite this