Clinical predictors of HIV infection in febrile children attending health facilities in Tanzania.


Ressource 1Download: BIB_0161B772CE35.P001.pdf (484.05 [Ko])
State: Public
Version: After imprimatur
Serval ID
A Master's thesis.
Publication sub-type
Master (thesis) (master)
Clinical predictors of HIV infection in febrile children attending health facilities in Tanzania.
Institution details
Université de Lausanne, Faculté de biologie et médecine
Publication state
Issued date
Number of pages
Introduction: In Tanzania and most Sub-Saharan countries of Africa, the HIV infection remains an important cause of child mortality. A good opportunity to test children for HIV is when they attend health facilities for a fever episode. As HIV rapid tests are not always available in resource-limited countries, the use of a clinical algorithm could provide guidance to clinicians on which febrile children should be tested in priority during these consultations. The criteria used in the actual IMCI algorithm may not necessary be the best predictors of HIV infection in children as little is known on the subject. The aim of the present study was to identify the clinical and laboratory predictors of HIV in febrile children attending outpatient clinics in Tanzania and to calculate their diagnostic performance, as well as to assess the performance of the IMCI criteria and algorithm used in Tanzania at the time of this study and determine if a combination of new predictors could have better diagnostic performance.
Methods: A nested case-control analysis in a prospective observational study on etiologies of fever in children aged less than 10 years attending health facilities in rural and urban Tanzania was performed. Detailed history taking, clinical examination, laboratory investigations and final diagnoses (based on pre-defined criteria) were obtained. The sensitivity, specificity, positive and negative likelihood ratios were calculated in bivariate analysis for all significant predictors. Stepwise backward logistic regression models were used to identify independent predictors of HIV infection and calculate their adjusted Odds Ratios. New algorithms using the best clinical predictors found in the analyses were generated and compared with the IMCI HIV branch.
Results: Of 1004 consecutive febrile children (median age 18 months), 16 were HIV positive. Regarding history taking: difficulty to breath (LR+ 6.9), chronic condition (LR+ 4.7), recent travel history (LR+ 4.1), sick contact person (LR+ 2.8) were found to be independently predictive of HIV. Regarding clinical signs: lymphadenopathy at any site (LR+ 10.8), chest indrawing (LR+ 10.7), hepatomegaly (LR+ 9.5), abdominal tenderness (LR+ 6.2), low weight for age (LR+ 4.0) and fast breathing (LR+ 3.5) were also predictive. Laboratory predictors were: hemoglobin <8 mg/dl (LR+ 2.8), naso-pharygeal pneumococcus load ≥107 cfu/ml (LR+ 2.2) and naso-pharygeal Staphylococcus aureus carriage (LR+ 2.0). Children with HIV were significantly more at risk to have a severe illness of any type (LR+ 2.9), and a diagnosis of radiological end- point pneumonia (LR+ 14.8), acute otitis media (LR+ 5.6), or bronchiolitis (LR+ 3.7). On the contrary, the diagnosis of upper respiratory infection negatively predicted the HIV infection (LR+ 0.3). The HIV branch of the IMCI algorithm had a sensitivity of 93.8% and a specificity of 73.7% when using 1 criteria as cut off for testing, and a sensitivity of 56.3% and specificity of 97.2% when using 2. A new algorithm using lymphadenopathy at any site, radiological end-point pneumonia, chronic condition, abdominal tenderness, difficulty to breath, acute otitis media and oral thrush as criteria showed a sensitivity of 93.8%, and a specificity of 81.4% when using one criterion as cut-off..
Conclusion: Several clinical and laboratory predictors for HIV were found in febrile outpatient children, some of which are not included in the present IMCI criteria. These findings could serve to improve the actual IMCI algorithm and potentially increase the number of HIV positive children detected during consultations for acute care.
HIV, predictor, children, Tanzania, algorithm
Create date
31/08/2016 15:26
Last modification date
20/08/2019 13:23
Usage data