Detecting Malaria Hotspots in Haiti, a Low-Transmission Setting

Abstract

In 2006, Haiti committed to malaria elimination when the transmission was thought to be low, but before robust national parasite prevalence estimates were available. In 2011, the first national population-based survey confirmed the national malaria parasite prevalence was < 1%. In both 2014 and 2015, Haiti reported approximately 17,000 malaria cases identified passively at health facilities. To detect malaria transmission hotspots for targeting interventions, the National Malaria Control Program (NMCP) piloted an enhanced geographic information surveillance system in three departments with relatively high-, medium-, and low-transmission areas. From October 2014-September 2015, NMCP staff abstracted health facility records of confirmed malaria cases from 59 health facilities and geo-located patients' households. Household locations were aggregated to 1-km2 grid cells to calculate cumulative incidence rates (CIRs) per 1,000 persons. Spatial clustering of CIRs were tested using Getis-Ord Gi* analysis. Space-time permutation models searched for clusters up to 6 km in distance using a 1-month malaria transmission window. Of the 2,462 confirmed cases identified from health facility records, 58% were geo-located. Getis-Ord Gi* analysis identified 43 1-km2 hotspots in coastal and inland areas that overlapped primarily with 13 space-time clusters (size: 0.26-2.97 km). This pilot describes the feasibility of detecting malaria hotspots in resource-poor settings. More data from multiple years and serological household surveys are needed to assess completeness and hotspot stability. The NMCP can use these pilot methods and results to target foci investigations and malaria interventions more accurately.

Figure 1.

Pilot geographic information system enhanced surveillance system workflow and health facility inclusion criteria. (A) Health facilities in the Department were covered using the following criteria. (B) This is the process GIS Assistants used to locate patients at their homes and collect GPS coordinates. This figure appears in color at www.ajtmh.org.

Figure 2.

Adjusted cumulative incidence rate of malaria cases per 1,000 persons in Haiti. (A) Departments were split into primary (1, 4, 7, and 9) and secondary (2 and 3, 5 and 6, 8, and 10) coverage areas and excluded areas in black. Malaria cases were geo-located at their residences within coverage areas. (B) Cumulative incidence rate (CIR) in the Grande Anse Department of geo-located cases. (C) CIR in the Sud Department of geo-located cases. (D) CIR in the Sud-Est Department of geo-located cases. This figure appears in color at www.ajtmh.org.

Figure 3.

Line-listed confirmed cases at health facilities by month of diagnosis. This figure appears in color at www.ajtmh.org.

Figure 4.

Getis-Ord Gi* cluster analysis of malaria cumulative incidence rates and 1-month Kulldorff space–time permutation clusters. Forty-three malaria incidence rate clusters were identified using Getis-Ord Gi* and 13 space–time clusters were identified from October 2014–September 2015. (A) Twenty Getis-Ord Gi* clusters and seven space–time clusters (P < 0.05) were detected in the Grande Anse Department. (B) Seventeen Getis-Ord GI* clusters (15 clusters at P < 0.05, two clusters at P < 0.10) and two space–time clusters (P < 0.05) were detected in the Sud Department. (C) Six Getis-Ord Gi* clusters and four space–time clusters (P < 0.05) were detected in the Sud-Est Department. This figure appears in color at www.ajtmh.org.