Evaluation of an operational malaria outbreak identification and response system in Mpumalanga Province, South Africa

Marlize Coleman¹ , Michael Coleman²^,3 , Aaron M Mabuza⁴ , Gerdalize Kok⁴ , Maureen Coetzee⁵^,6 and David N Durrheim⁷

¹School of Animal, Plant & Environmental Sciences, University of the Witwatersrand, Johannesburg, Gauteng, South Africa

²Liverpool School of Tropical Medicine, Pembroke Place, Liverpool, L3 5QA, UK

³Medical Research Council, Durban, South Africa

⁴Mpumalanga Department of Health, 66 Anderson Street, Nelspruit, 1200, South Africa

⁵Vector Control Reference Unit, National Institute for Communicable Diseases, National Health Laboratory Service, 1 Modderfontein Road, Sandringham, 2131 Johannesburg, South Africa

⁶SA Research Chair in Medical Entomology & Vector Control, School of Pathology, University of the Witwatersrand, Johannesburg, South Africa

⁷Hunter New England Population Health and Hunter Medical Research Institute, Locked Bag 10, Wallsend, 2287, Australia

author email corresponding author email

Malaria Journal 2008, 7:69doi:10.1186/1475-2875-7-69


Published:	27 April 2008

Abstract

Background and objective

To evaluate the performance of a novel malaria outbreak identification system in the epidemic prone rural area of Mpumalanga Province, South Africa, for timely identification of malaria outbreaks and guiding integrated public health responses.

Methods

Using five years of historical notification data, two binomial thresholds were determined for each primary health care facility in the highest malaria risk area of Mpumalanga province. Whenever the thresholds were exceeded at health facility level (tier 1), primary health care staff notified the malaria control programme, which then confirmed adequate stocks of malaria treatment to manage potential increased cases. The cases were followed up at household level to verify the likely source of infection. The binomial thresholds were reviewed at village/town level (tier 2) to determine whether additional response measures were required. In addition, an automated electronic outbreak identification system at town/village level (tier 2) was integrated into the case notification database (tier 3) to ensure that unexpected increases in case notification were not missed.

The performance of these binomial outbreak thresholds was evaluated against other currently recommended thresholds using retrospective data. The acceptability of the system at primary health care level was evaluated through structured interviews with health facility staff.

Results

Eighty four percent of health facilities reported outbreaks within 24 hours (n = 95), 92% (n = 104) within 48 hours and 100% (n = 113) within 72 hours. Appropriate response to all malaria outbreaks (n = 113, tier 1, n = 46, tier 2) were achieved within 24 hours. The system was positively viewed by all health facility staff. When compared to other epidemiological systems for a specified 12 month outbreak season (June 2003 to July 2004) the binomial exact thresholds produced one false weekly outbreak, the C-sum 12 weekly outbreaks and the mean + 2 SD nine false weekly outbreaks. Exceeding the binomial level 1 threshold triggered an alert four weeks prior to an outbreak, but exceeding the binomial level 2 threshold identified an outbreak as it occurred.

Conclusion

The malaria outbreak surveillance system using binomial thresholds achieved its primary goal of identifying outbreaks early facilitating appropriate local public health responses aimed at averting a possible large-scale epidemic in a low, and unstable, malaria transmission setting.