LETTERS

Detecting Epidemic The monitoring system detected Although the ability of the system 2 malaria outbreaks in Kabale, 1 each to generate timely epidemic warnings Malaria, in 2005 and 2006. During the most re- is encouraging, data from 1 sentinel To the Editor: In the fi eld of ma- cent outbreak, the fi rst warnings of ab- site highlighted a potential limita- laria epidemic early warning, there ex- normally high malaria incidence were tion of using routine data from clini- ists an unfortunate but frequently ac- communicated from the district team cal diagnoses of malaria as a basis for curate perception that health systems to the NMCP on June 5, 1 month be- epidemic detection. As elsewhere, in many affected countries learn of fore reports of the outbreak appeared routine outpatient data for Bufundi epidemics by way of the popular press in the press and >2 weeks before case Health Centre suggested the occur- rather than through formal disease sur- numbers began to peak (Figure, panel rence of a malaria outbreak starting in veillance systems. Malaria epidemics A). In the 6 weeks from May 29 to early June and peaking in early July are often easily recognized (albeit too July 9, Kabale’s 5 sentinel sites re- of 2006 (Figure, panel B). The slight late) by laypersons (1), but most rou- corded 4,637 clinical malaria cases, delay in the onset of the outbreak at tine disease surveillance systems lack 159% more than expected for this pe- this site was plausible, given its high the ability to provide accurate, timely riod. Although the sentinel network elevation (2,200 m) and geographic indications of aberrations in case num- consists of health centers with limited remoteness. Data for patients with bers. The World Health Organization inpatient facilities, available data on parasitologically confi rmed malaria, (WHO) has set specifi c targets for admissions showed a similar tempo- available through an ongoing malaria early detection and control of malaria ral pattern, with 616 patients admitted transmission study at Bufundi Health epidemics as part of a wider strategy to during the same 6-week period, 188% Centre (3), showed a different tempo- cut the global impact of malaria in half more than expected. ral pattern of incidence, however. As by 2010 (2). We describe experiences during a recent epidemic in southwest Uganda and examine the performance of a pilot early detection system. In 2002, the Ugandan Ministry of Health began developing and piloting a new district-level malaria monitor- ing system in Kabale and (3). Located in Uganda’s southwest- ern highlands, these districts have ex- perienced several serious malaria epi- demics in recent years, most notably during the El Niño year of 1998 (4,5). In this new system, data generated from representative health facilities are collated, entered on computer, and analyzed by district teams on a weekly basis. Incoming data on clinical ma- laria are compared with a baseline of historical illness data from which the effects of long-term temporal trends have been removed, and an objective anomaly measure, or “standardized departure,” is used to provide a simple, intuitive index of deviation from ex- pected weekly levels of incidence (3). Electronic reports are disseminated by Figure. A) Weekly observed and expected numbers of outpatient (OP) and inpatient email to the National Malaria Control (IP) cases of clinically diagnosed malaria from sentinel health centers in , southwestern Uganda, May–August 2006. B) Weekly numbers of clinically diagnosed Programme (NMCP) and others, in- malaria cases and the proportion of case-patients subsequently testing positive for cluding WHO and the United Nations Plasmodium falciparum infection by rapid diagnostic test at Bufundi, Kabale district, Children’s Fund. southwestern Uganda, May–August 2006.

Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 13, No. 5, May 2007 779 LETTERS part of this study, all samples from ma- measures to use and how best to target 7. The US President’s Malaria Initiative. laria case-patients identifi ed through them (8). Without these procedures, Lancet. 2006;368:1. 8. Abeku TA. Response to malaria epidem- clinical diagnosis were subject to a the value of early detection will be ics in Africa. Emerg Infect Dis [serial on Paracheck-Pf immunoassay test (Or- seriously undermined. the Internet]. 2007 May [4 Apr 2007]. chid Biomedical Systems, Verna, Goa, Available from http://www.cdc.gov/EID/ India). Results indicated that, at the content/13/5/681.htm Funding for this research was provided peak of the apparent malaria outbreak, by the Bill and Melinda Gates Foundation Address for correspondence: Jonathan Cox, the percentage of samples from clini- (through the Gates Malaria Partnership, London School of Hygiene & Tropical Medicine, cally diagnosed cases that produced a London School of Hygiene and Tropical Department of Infectious and Tropical Diseases, positive diagnostic test was as low as Medicine) and through the Department for Keppel Str, London WC1E 7HT, UK; email: 4% (Figure, panel B). These results International Development, UK. [email protected] are unlikely to refl ect poor diagnostic performance of the testing (6); febrile illness other than malaria was likely Jonathan Cox,* the cause of the outbreak. Tarekegn Abeku,* Recent experiences in Kabale James Beard,* also highlight the potentially unwieldy James Turyeimuka,† nature of indoor residual spraying Enoch Tumwesigye,† campaigns in the absence of spatial Michael Okia,‡ Extensively targeting. In Kabale, a district-wide and John Rwakimari‡ spraying campaign supported by the *London School of Hygiene and Tropi- Drug-resistant US President’s Malaria Initiative (7) cal Medicine, London, UK; †Kabale Dis- Tuberculosis, Italy was planned for the 2006 transmission trict Health Management Team, Kabale, and Germany season. However, shortages of trained Uganda; and ‡Ministry of Health, , personnel and other institutional de- Uganda To the Editor: Twenty-three lays meant that spraying could not countries have reported >1 case of begin until the third week of June, by extensively drug-resistant tubercu- References which time the epidemic had peaked losis (XDR TB) (1); however, in- (and densities of vector mosquitoes 1. Nájera JA, Kouznetsov RL, Delacollette formation about XDR TB is still in- had presumably begun to fall). By C. Malaria epidemics, detection and con- complete. In particular, the response trol, forecasting and prevention. Geneva: of XDR TB to treatment in countries July 17, <50% of the targeted struc- Division of Control of Tropical Diseases, tures had been sprayed. In the future, World Health Organization; 1998. with low incidence is not known. We careful targeting of spraying to areas 2. World Health Organization/Roll Back compared mortality rates from XDR of highest epidemic risk might lead to Malaria. Malaria early warning sys- TB with those from multidrug-resis- tems—concepts, indicators and partners. tant (MDR) TB. more timely completion of spraying A framework for fi eld research in Africa. activities. It might also be benefi cial Geneva: The Organization; 2001. WHO/ We analyzed data from all cul- to create special spray teams that can CDS/RBM/2001.32. ture-confi rmed TB cases diagnosed respond quickly to specifi c alerts. 3. Abeku TA, Hay SI, Ochola S, Langi P, during 2003–2006 by the TB clinical Beard B, de Vlas SJ, et al. Malaria epi- reference centers in Italy (Sondalo, Recent experiences in Kabale demic early warning and detection in have underlined the potential value African highlands. Trends Parasitol. Milan, Rome) and Germany (Borstel, of simple monitoring tools for early 2004;20:400–5. Grosshansdorf, Bad-Lippspringe) and detection of epidemics but have also 4. Lindblade KA, Walker ED, Onapa AW, reviewed original clinical records. Katungu J, Wilson ML. Highland malaria Drug susceptibility testing for fi rst- shown potential barriers to effective in Uganda: prospective analysis of an epi- epidemic control. Our fi ndings high- demic associated with El Niño. Trans R and second-line anti-TB drugs was light the need to build systems that Soc Trop Med Hyg. 1999;93:480–7. performed according to World Health improve routine collection of data 5. Kilian AHD, Langi P, Talisuna A, Organization (WHO) recommenda- Kabagambe G. Rainfall pattern, El Niño tions by quality-assured laboratories on parasitologically confi rmed cases and malaria in Uganda. Trans R Soc Trop of malaria and allow rapid investiga- Med Hyg. 1999;93:22–3. and retested at WHO Supranational tion of anomalies in incoming clinical 6. Mboera LE, Fanello CI, Malima RC, Tal- Reference Laboratories (Rome/Milan; data. It is equally important to develop bert A, Fogliati P, Bobbio F, et al. Compar- Borstel) (2–4). ison of the Paracheck-Pf test with micros- XDR TB was defi ned as resis- procedures that translate early warn- copy, for the confi rmation of Plasmodium ing information into timely decisions falciparum malaria in Tanzania. Ann Trop tance to at least rifampin and isoniazid concerning which epidemic control Med Parasitol. 2006;100:115–22. (MDR TB defi nition) in addition to

780 Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 13, No. 5, May 2007