Frequency and correlates of malaria overdiagnosis and treatment in western Kenya

Abstract

Background: Despite decreasing malaria burden in sub-Saharan Africa, fever in children is frequently attributed to malaria and empirically treated with antimalarial drugs. The 2010 WHO guidelines recommend laboratory confirmation of malaria before initiating therapy, but the uptake of these guidelines is slow. We determined frequency and predictors of malaria overtreatment in febrile children at two western Kenya rural hospitals in regions with varied malaria endemicity. Methods: We enrolled consecutive febrile children aged 6 months to 15 years presenting to Kisii Provincial (malaria endemicity: hypoendemic) or Homa Bay District (malaria endemicity: holoendemic) hospitals between March, 2012, and November, 2013. Detailed demographic, clinical, and laboratory data were collected, including physician diagnosis at admission and corresponding treatments. All children were screened for HIV with antibody or PCR testing, for malaria using smear microscopy and Paracheck Pf rapid diagnostic tests, and for invasive bacterial infections using blood culture. We determined the frequency of clinician-prescribed antimalarial treatment among those with negative smear and rapid diagnostic test results and evaluated independent predictors of overtreatment using multivariate logistic regression. Because the sites had marked differences in underlying malaria endemicities that might impact clinician suspicion of parasitaemia, analyses were stratified and compared by site. Findings: Among the 1362 children enrolled (685 in Kisii and 677 in Homa Bay), 46 (6·7%) and 310 (45·8%) had laboratory-confirmed malaria in Kisii and Homa Bay, respectively (p<0·001). Among children without malaria, 7·0% and 57·2% were empirically prescribed antimalarials at Kisii and Homa Bay, respectively (p<0·001). Significant predictors of malaria overtreatment in Homa Bay included presence of more than one danger sign according to Integrated Management of Childhood Illness guidelines (adjusted odds ratio 8·47, 95% CI 4·81–14·89), fever lasting more than 7 days (4·94, 1·90–12·86), and fever greater than 390C (3·07, 1·58–5·96). The sensitivity of clinical malaria diagnosis (vs gold-standard laboratory-based diagnosis) was lower in Kisii than in Homa Bay (73·0% vs 95·3%, p<0·001), whereas specificity was higher in Kisii than in Homa Bay (93·2% vs 33·7%, p<0·001). Interpretation: Malaria was treated, despite negative laboratory testing, more frequently in Homa Bay than in Kisii, probably owing to underlying higher malaria prevalence. Despite availability of diagnostic testing, malaria prevalence appears to influence the clinical management of children in Kenya and may result in missed opportunities to accurately diagnose alternative causes of fever. Strengthening health system use and adherence to existing treatment guidelines appears necessary, particularly in areas of high malaria endemicity. Funding: US National Institutes of Health.