287 Am. J. Trop. Med. Hyg., 58(3), 1998, pp. 287–298 Copyright  1998 by The American Society of Tropical Medicine and Hygiene EXPLORATORY SPACE-TIME ANALYSIS OF REPORTED DENGUE CASES DURING AN OUTBREAK IN FLORIDA, PUERTO RICO, 1991–1992 AMY C. MORRISON, ARTHUR GETIS, MARILYN SANTIAGO, JOSE G. RIGAU-PEREZ, AND PAUL REITER Dengue Branch, Centers for Disease Control and Prevention, San Juan, Puerto Rico; Department of Geography, San Diego State University, San Diego, California; Water Resources Division, Caribbean District, United States Geological Survey, Guaynabo Abstract. The spatial and temporal distributions of dengue cases reported during a 1991–1992 outbreak in Florida, Puerto Rico (population = 8,689), were studied by using a Geographic Information System. A total of 377 dengue cases were identified from a laboratory-based dengue surveillance system and georeferenced by their residential addresses on digital zoning and U.S. Geological Survey topographic maps. Weekly case maps were generated for the period between June and December 1991, when 94.2% of the dengue cases were reported. The temporal evolution of the epidemic was rapid, affecting a wide geographic area within seven weeks of the first reported cases of the season. Dengue cases were reported in 217 houses; of these 56 (25.8%) had between two and six reported cases. K- function analysis was used to characterize the spatial clustering patterns for all reported dengue cases (laboratory- positive and indeterminate) and laboratory-positive cases alone, while the Barton and David and Knox tests were used to characterize spatio-temporal attributes of dengue cases reported during the 1991–1992 outbreak. For both sets of data significant case clustering was identified within individual households over short periods of time (three days or less), but in general, the cases had spatial pattern characteristics much like the population pattern as a whole. The rapid temporal and spatial progress of the disease within the community suggests that control measures should be applied to the entire municipality, rather than to the areas immediately surrounding houses of reported cases. The potential for incorporating Geographic Information System technologies into a dengue surveillance system and the limitations of using surveillance data for spatial studies are discussed. Dengue fever, a viral disease transmitted by the mos- quito Aedes (Stegomyia) aegypti (L.), can spread rapidly in explosive epidemics. 1 Since no vaccine or chemother- apy is available, prevention and control of the disease are dependent on vector control measures such as source re- duction (to eliminate larval habitats) or ultra-low volume spraying (to kill adult mosquitoes). During outbreaks, emergency measures are often centered on reported den- gue cases, a practice that assumes that female Ae. aegypti rarely travel further than 50–100 m during their lifetime. 2 Recent studies of oviposition, however, have indicated that dispersal is driven by the search for oviposition sites, 3, 4 and that in Puerto Rico much greater distances are in- volved. 5 The relative contribution of mosquitoes, persons, or both to the dispersal of dengue viruses within a com- munity is poorly understood. Several investigators have reported on clusters of cases inside the same or adjacent houses and descriptions of the focal nature of the disease are relatively common, 1, 6–8 but there are no previous stud- ies on the spatial-temporal patterns of the disease. The Geographic Information System (GIS), a computer system that can store, assemble, manipulate and analyze geo- graphically referenced information, offers a new approach to geographic studies of disease patterns. In a GIS environ- ment, geographically referenced point data sets can be ana- lyzed by considering the distance between each point and all other points, facilitating the use of a variety of methods that describe and analyze point patterns 9–11 and disease clus- tering in time and space. 12–14 We report here on the appli- cation of these methods to surveillance data obtained during an outbreak of dengue in the municipality of Florida, Puerto Rico in 1991–1992. In addition, we discuss the potential for incorporating GIS technologies into a dengue surveillance system and the limitations of using surveillance data for spa- tial studies. MATERIALS AND METHODS Study area. The municipality of Florida (area = 26 km 2 ; population = 8,689) is a small rural community 15 located in the hills of north central Puerto Rico (Figure 1). In 1991, this community had the highest reported incidence of dengue (15.7/1,000) in Puerto Rico. 16 Dengue-2 was the predomi- nant serotype. 16 The municipality consists of nine well-de- fined urbanizations (neighborhoods or housing develop- ments), two public housing projects, and seven rural neigh- borhoods. Some of these areas are separated by steep, eroded karst limestone hills ( 250 m), which probably act as a natural barrier for Ae. aegypti. The community was originally selected for an entomolog- ic study because it is spatially separated from other popu- lation centers. Dengue reporting in Florida during this period was probably better than in other areas on the island because it had a readily accessible government health center (Centro de Diagnostico y Tratamiento). Community awareness of dengue during the epidemic was high because of the large number of cases and the occurrence of four highly publicized cases of dengue hemorrhagic fever (DHF), one of which was fatal. 16 Data sources. Data surveillance database. Data were ob- tained from a laboratory-based surveillance system of the San Juan Laboratories, Dengue Branch, National Center for Infectious Diseases, Centers for Disease Control and Pre- vention (CDC). 17 Blood samples from clinically suspected dengue cases are submitted to CDC from government clin- ics, public and private hospitals, and physicians’ offices throughout Puerto Rico, along with a standardized dengue case information sheet (DCI). The DCI requests data on the home address, age, sex, and date of onset of symptoms of the patients. It also contains a checklist to indicate the symp- toms and signs reported by the patient or elicited in the eval- uation. The surveillance information used in this retrospec- tive study is presented by date of onset of symptoms or