2014.ISDS.Abstracts.Final.pdf


ISDS Annual Conference Proceedings 2014. This is an Open Access article distributed under the terms of the Creative Commons Attribution-
Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, 
and reproduction in any medium, provided the original work is properly cited.

ISDS 2014 Conference Abstracts

Dry Climate as a Predictor of Chagas’ Disease Irregular 
Clusters: A Covariate Study

Luiz H. Duczmal*1, Gladston J. Moreira2, Luís Paquete3, David Menotti2, Ricardo 
Takahashi1 and Denise Burgarelli1

1Statistics, Universidade Federal de Minas Gerais, Belo Horizonte, Brazil; 2Universidade Federal de Ouro Preto, Ouro Preto, Brazil; 
3Universidade de Coimbra, Coimbra, Portugal

Objective
We employ climate information to assess the possible spatial 

dependence on the occurrence of Chagas’ disease irregular clusters 
in Central Brazil, using a variant of the Spatial Scan Statistic [1], the 
Geo-Dynamic Scan (GDScan) [4].

Introduction
Chagas’ disease, caused by the protozoan Trypanosoma cruzi, is 

spread mostly by Triatominae bugs. High carbon dioxide emission 
and strong infra-red (IR) radiation are indicative of their presence. 
Periods of low atmospheric water saturation favor their dispersal, 
when the bugs’ IR perception is high [2].

The Fast Subset Scan (FSScan) [3] is very efficient for the 
detection of the most likely geographic cluster. Covariate studies 
associating the presence of regular clusters with environmental factors 
are routinely done using the Circular Scan, the simplest version of the 
Spatial Scan statistic. However, if the study employs irregular clusters 
instead, accurate results depend on the generation of a rich family of 
variants of the primary cluster.

Methods
GDScan’s bi-objective optimization approach simultaneously 

minimizes the population and maximizes the number of cases of 
each candidate solution. The efficient solution set of the bi-objective 
problem is formed by several candidate solutions, as opposed to the 
(generally) unique solution provided by other methods, such as the 
Circular Scan or the FSScan. Each candidate solution z is a non-
dominated solution, in the sense that any other solution w with more 
cases than z and smaller population than z cannot exist.

The idea is to penalize regions which have humid climates, since 
the bugs prefer dry places. Figure 1 shows Minas Gerais state in 
central Brazil, indicating Chagas’ disease rates (Figure 1a) and 
dryness levels during the driest season, from June to August (Figure 
1b). The efficient solution set returned by GDScan is then evaluated 
according to dryness value. The cluster solutions located in drier 
regions, considered as more plausible, should receive the higher 
values. For clusters with very similar likelihood ratio scores, the most 
plausible solution is the one with maximum dryness.

Results
Figure 1c shows GDScan’s primary cluster variant with the lowest 

precipitation level, corresponding to the best solution found which 
incorporate the driest regions.

Figure 1d presents variants of the primary cluster provided by 
GDScan (open circles) and FSScan (crosses); the “better” solutions 
(more cases within less populated clusters) are located in the lower-left 
part of the graph. As we can see, primary cluster variants generated 
by GDScan are better situated in the graph than those generated by 
FSScan.

Conclusions
The cluster of Figure 1c is nearly identical with the most likely 

cluster found by both GDScan and FSScan (not shown). This is 

an indication of a strong correlation between dryness and Chagas’ 
disease incidence, supporting previous findings [2]. This information 
could be useful in Chagas’ disease surveillance and prevention.

GDScan is computationally more expensive; however, it finds 
more potentially useful variants of the primary cluster as compared 
to FSScan (e.g., those variants with more desirable covariate values).

Chagas’ disease rates (a), precipitation levels with darker shades indicating 
dryer areas (b), GDScan’s primary cluster variant with lowest precipitation 
level (c), and variants of the primary cluster, provided by GDScan (open 
circles) and FSScan (crosses) (d).

Keywords
irregular cluster; spatial scan statistic; Chagas’ disease; Geo Dynamic 
Scan; Fast Subset Scan

Acknowledgments

The authors were supported by CNPq, Fapemig and CAPES.

References

1. Kulldorff M (1997) A spatial scan statistic. Comm. Stat: Th. Meth.
26:1481-1496.

2. Catalá SS (2011) The infra-red (IR) landscape of triatoma infestans.
an hypothesis about the role of IR radiation as a cue for triatominae 
dispersal. Infection, Genetics and Evolution 11(8):1891-1898.

3. Neill DB (2012) Fast subset scan for spatial pattern detection. JRSSB.
74:337-360.

4. Moreira GJP, Paquete L, Duczmal LH, Menotti D, Takahashi RHC
(2014). Multi-objective dynamic programming for spatial cluster 
detection. Env. Ecol. Stat. (to appear)

*Luiz H. Duczmal

E-mail: duczmal@ufmg.br    

Online Journal of Public Health Informatics * ISSN 1947-2579 * http://ojphi.org * (1):e124, 201