Background
Laboratory capacity to confirm malaria cases in Tanzania is low and presumptive treatment of malaria is being practiced widely. In malaria endemic areas WHO now recommends systematic laboratory testing when suspecting malaria. Currently, the use of Rapid Diagnostic Tests (RDTs) is recommended for the diagnosis of malaria in lower level peripheral facilities, but not in health centres and hospitals. In this study, the following parameters were evaluated: (1) the quality of routine microscopy, and (2) the effects of RDT implementation on the positivity rate of malaria test results at three levels of the health system in Dar es Salaam, Tanzania.

Methods
During a baseline cross-sectional survey, routine blood slides were randomly picked from 12 urban public health facilities in Dar es Salaam, Tanzania. Sensitivity and specificity of routine slides were assessed against expert microscopy. In March 2007, following training of health workers, RDTs were introduced in nine public health facilities (three hospitals, three health centres and three dispensaries) in a near-to-programmatic way, while three control health facilities continued using microscopy. The monthly malaria positivity rates (PR) recorded in health statistics registers were collected before (routine microscopy) and after (routine RDTs) the intervention in all facilities.

Results
At baseline, 53% of blood slides were reported as positive by the routine laboratories, whereas only 2% were positive by expert microscopy. Sensitivity of routine microscopy was 71.4% and specificity was 47.3%. Positive and negative predictive values were 2.8% and 98.7%, respectively. Median parasitaemia was only three parasites per 200 white blood cells (WBC) by routine microscopy compared to 1226 parasites per 200 WBC by expert microscopy. Before RDT implementation, the mean test positivity rates using routine microscopy were 43% in hospitals, 62% in health centres and 58% in dispensaries. After RDT implementation, mean positivity rates using routine RDTs were 6%, 7% and 8%, respectively. The sensitivity and specificity of RDTs using expert microscopy as reference were 97.0% and 96.8%. The positivity rate of routine microscopy remained the same in the three control facilities: 71% before versus 72% after. Two cross-sectional health facility surveys confirmed that the parasite rate in febrile patients was low in Dar es Salaam during both the rainy season (13.6%) and the dry season (3.3%).

Conclusions
The quality of routine microscopy was poor in all health facilities, regardless of their level. Over-diagnosis was massive, with many false positive results reported as very low parasitaemia (1 to 5 parasites per 200 WBC). RDTs should replace microscopy as first-line diagnostic tool for malaria in all settings, especially in hospitals where the potential for saving lives is greatest.