BACKGROUND: Case-control studies have been proposed as an appropriate tool for health impact evaluation of insecticide-treated nets (ITN) programmes. METHODS: A dispensary-based case-control study was carried out in one village in Tanzania. Each case of fever and parasitaemia in a child under 5 years was paired with one community and one dispensary control without fever and parasitaemia. Cases and controls were compared with regard to ITN ownership and other factors assessed by a questionnaire. A cross-sectional survey of factors associated with parasitaemia, including ITN use, was carried out during the study. Dispensary attendance rates of the study children were calculated using passive case detection data. RESULTS: Cases and dispensary controls had higher dispensary attendance rates compared to community controls and children with nets attended more for most of the illness events. A comparison of cases and community controls showed a strong and statistically significant association between untreated net use and being a case (odds ratio [OR] = 2.1, 95% CI : 1.3-3.4). For those with ITN there was a smaller and weaker association between risk of being a case and ITN use (OR =1.4, 95% CI : 0.9-2.2). Comparison of cases and dispensary controls showed no association between untreated or treated nets and the risk of being a case (for treated nets OR = 0.9, 95% CI : 0.5-1.4 and for untreated nets OR = 1.2, 95% CI : 0.7-2.0). These results are contrary to those from the cross-sectional assessment, where children with ITN had a lower prevalence of parasitaemia than those with no nets (OR = 0.5, 95% CI : 0.3-0.9), and also contrary to other assessments of the health impact of ITN in this population. CONCLUSIONS: The positive association between mild malaria and net ownership is counter-intuitive and best explained by attendance bias, since children with nets attended more frequently for all curative and preventive services at the dispensary than those without nets. Dispensary-based case-control studies may not be appropriate for assessing impact of treated nets on clinical malaria, while cross-sectional surveys might represent an attractive alternative.