Evaluating the effectiveness of IPTi on malaria using routine health information from sentinel health centres in southern Tanzania.

BACKGROUND: Intermittent preventive treatment of malaria in infants (IPTi) consists of the administration of a treatment dose of sulphadoxine-pyrimethamine (SP) at the time of routine vaccinations. The use of routine Health Management and Information Services (HMIS) data to investigate the effect of IPTi on malaria, anaemia, and all-cause attendance in children aged 2-11 months presenting to 11 health centres in southern Tanzania is described. METHODS: Clinical diagnosis of malaria was confirmed with a positive blood slide reading from a quality assurance laboratory. Anaemia was defined using two thresholds (mild [Hb<11 g/dL], severe [Hb<8 g/dL]). Incidence rates between IPTi and non-implementing health centres were calculated using Poisson regression, and all statistical testing was based on the t test due to the clustered nature of the data. RESULTS: Seventy two per cent of infants presenting in intervention areas received at least one dose of IPTi--22% received all three. During March 2006-April 2007, the incidence of all cause attendance was two attendances per person, per year (pppy), including 0.2 episodes pppy of malaria, 0.7 episodes of mild and 0.13 episodes of severe anaemia. Point estimates for the effect of IPTi on malaria varied between 18% and 52%, depending on the scope of the analysis, although adjustment for clustering rendered these not statistically significant. CONCLUSIONS: The point estimate of the effect of IPTi on malaria is consistent with that from a large pooled analysis of randomized control trials. As such, it is plausible that the difference seen in health centre data is due to IPTi, even thought the effect did not reach statistical significance. Findings draw attention to the challenges of robust inference of effects of interventions based on routine health centre data. Analysis of routine health information can reassure that interventions are being made available and having desired effects, but unanticipated effects should trigger data collection from representative samples of the target population.