© International Epidemiological Association 1999
International Journal of Epidemiology 1999;28:982–989
Printed in Great Britain
American cutaneous leishmaniasis
in Southeast Brazil: space–time clustering
George Luiz Lins Machado-Coelho,a Renato Assunção,b Wilson Mayrinkc and Waleska Teixeira Caiaffad
Background American cutaneous leishmaniasis (ACL) is endemic in many rural areas of Brazil
where different transmission patterns of the disease have been described. This
ecological study was carried out in a municipality located in Southeast Brazil and
aimed to investigate the space-temporal patterns of the disease and environmental
risk factors from 1966 to 1996.
Methods
Incident ACL cases were defined by clinical diagnosis, confirmed by a positive
skin test and/or parasitological examination. Age-adjusted morbidity rate of ACL
was calculated by year for this municipality and their different census enumeration
districts. The homogeneity χ2 test, Moran and empirical Bayes index and Knox
procedure were employed for testing the significance of clusters in time, space
and in time-space, respectively. A Poisson regression model was used to identify
environmental factors related to rate variability.
Results
A total of 1712 new ACL cases were reported with a yearly incidence rate of
48/100 000 inhabitants. Higher incidence rates were detected in 1968, 1974, and
1988 (100, 160, and 190 cases/100 000, respectively) with evidence of spatial clustering from 1986 to 1993. Significant space–time clustering with epidemic peaks
followed by low incidence in subsequent periods was observed. The incidence
rates of ACL were independently associated with rural areas; areas lacking
sanitary installations and with higher proportion of exposed garbage (P , 0.01).
Conclusions This study suggests that ACL rates vary across space and time. Rural areas and
some environmental factors could explain part of this variation. Environmental
modifications in the vicinity of households over time and accumulation of susceptible individuals are discussed as possible factors responsible for variability.
Keywords
American cutaneous leishmaniasis, cluster analysis, temporal-space patterns,
temporal series, geography, herd immunity
Accepted
22 March 1999
American cutaneous leishmaniasis (ACL) is a zoonosis caused
by Leishmania species of subgenera Viannia and Leishmania. It is
transmitted by phlebotomine sandflies and affects primarily
sylvan mammalian hosts, mainly small rodents from tropical
forests. It is an important public health problem, with clinical
manifestations varying from simple cutaneous to mucocutaneous and disseminate lesions. It is highly endemic in several
Brazilian regions. About 20 000 new cases have been notified
annually over the last 5 years, and 10% of these cases are from
the Southeast Region.1
a Department of Pharmacy, Federal University of Ouro Preto (UFOP), Brazil.
b Department of Statistic, c Department of Parasitology, d Department of Social
and Preventive Medicine, Federal University of Minas Gerais (UFMG),
Brazil.
Reprints requests to: Dr George LL Machado-Coelho, Rua Costa Sena, 171,
Ouro Preto, Minas Gerais, Brazil. 35400–000. E-mail: [email protected]
Until the 1970s, human infection was only related to contact
between humans and forests.2–5 However, in the last two
decades, the disease has been described among individuals living
in settled agricultural areas, where deforestation took place a
long time ago.6 Recently, it has been suggested7–9 that there has
been a modification in the transmission patterns, with the
disease occurring in periurban populations. These new findings
have lead to the ecological idea that environmental factors have
changed sandflies’ behaviour. They are now living inside houses
and outdoors close to the household.10 Little attention has been
paid to the space and temporal dynamics of the transmission
patterns of this disease and to the correlates between environmental factors and ACL.
In this paper we present the results of a continuous collection
of ACL cases through a surveillance system carried out in
Caratinga municipality, Southeast Brazil, where different transmission patterns have been hypothesized in the last three
decades. The aims of this study were to estimate the incidence
982
AMERICAN CUTANEOUS LEISHMANIASIS: SPACE–TIME CLUSTERING
983
Figure 1 Caratinga municipality subdivided by rural and urban census enumeration districts, Minas
Gerais State, Southeast Brazil
rates of ACL from 1966 to 1996, to test for space and temporal
patterns in the rates and to correlate them to socio-geographic
factors.
Material and Methods
Study area
Caratinga is an endemic municipality for the disease. It is an area
of 2234 km2 located in Minas Gerais State, Southeast Brazil,
and lying within the boundaries of 19°19’ and 20°01’S, and
41°46’ and 42°31’W. Caratinga Town, the administrative centre
of the municipality, is situated at 575 m above sea level. The
municipality has 12 districts, subdivided according to Brazilian
Census Bureau into 130 census enumeration districts (CED).
To permit comparisons, the CED were grouped in 21 urban and
53 rural CED (Figure 1).
Subjects
Incident ACL cases were patients attending for the first time a
reference clinic in Caratinga during 1966–1996, and reported to
a specific passive surveillance system for leishmaniasis. Patients
with one or more active skin lesions, clearly demarcated, were
diagnosed as cutaneous leishmaniasis. The presence of either
concomitant/or not sequellae of mucous membranes, such as
oral or naso-pharyngeal lesions defined the mucocutaneous
form. All clinical diagnoses were confirmed by positive skin test
(Montenegro) and/or parasitological examination.
Population data
Demographic data for the study area were obtained through
the official files supplied by ‘Instituto Brasileiro de Geografia e
Estatística’, the official Brazilian Census Agency. The 1960,
1970, 1980 and 1991 census data for Caratinga stratified by age
were used to calculate growth rates and to estimate the population size from 1966 to 1996.11–14 Regarding information on
CED, only the 1980 and 1991 census data were available.
Morbidity rates
Direct standardized rates which took into account differences in
the age population distribution among CED were calculated.
The reference population was the 1980 local population in each
age category.
Time homogeneity and time trend
Annual homogeneity of municipal level rates of ACL were
tested using a χ2 test in order to compare the total number of
984
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Table 1 Demographic and clinical characteristics of American cutaneous leishmaniasis cases according to the districts in Caratinga, Southeast
Brazil, 1966–1996
Age groupa (years)
Rurala
Districts
No.
cases (n)
Caratinga
345
22.9
47.8
13.0
44.6
42.3
95.6
Dom Lara
53
98.1
58.5
17.0
37.7
45.3
96.2
98.2
Entre Folhas
zone (%)
Gender
Male (%)
0–9 (%)
10–19 (%)
20+ (%)
Clinical diagnosis
Cutaneous form (%)
55
81.8
45.5
12.7
27.3
60.0
Imbé
177
97.7
50.8
24.3
19.2
56.5
93.7
Ipaba
18
50.0
38.9
27.8
50.0
22.2
100.0
Santa Barbara
Santa Rita
Santo A. do Manhuaçu
141
94.3
60.3
20.7
25.0
54.3
98.6
43
34.9
46.5
20.9
44.2
34.9
95.3
177
93.8
61.6
22.0
23.2
54.8
94.3
52
96.2
53.8
19.2
17.3
63.5
96.2
São J do Jacutinga
117
92.3
56.4
19.7
32.5
47.9
94.9
Sapucaia
104
97.1
54.8
16.5
21.4
62.1
99.0
Ubaporanga
317
92.4
55.2
27.3
25.1
47.6
96.2
Vargem Alegre
113
87.6
61.9
10.6
20.4
69.0
96.5
1712
77.3
54.2
19.6
29.2
51.3
96.0
São Cândido
Total
a P , 0.01.
ACL cases observed with those expected under the assumption
of constant rate across time.
The time trend of ACL rates were also tested using a log-linear
model on the assumption of a Poisson distribution for cases using
the population as offset.15 We assumed that the annual count
of ACL cases in this region followed a Poisson distribution with
mean equal to the population size times the underlying timevarying rate. The logarithm of annual rates follow a linear trend
on time with parameters a and b estimated by maximum likelihood. Since the log-linear model is specific for time trend, it is
more powerful for detecting this departure from the constant
rate hypothesis than the χ2 test for homogeneity.
Space clustering
Initially, we tested whether the underlying rates among the
CED were spatially heterogeneous or not. Since the disease is
rare, it was convenient to aggregate the counts of ACL cases
with the aim of obtaining more stable rates. To investigate time
variation on the endemic pattern, three time periods for which
CED population data were available, were used: 1980–1985,
1986–1993, and 1994–1996. A powerful test for rates heterogeneity proposed by Oden16 was used.
Acceptance of the hypothesis of heterogeneous rates implies
that there was some spatial variation of the underlying rates
in the whole region but the scale of this variation could be smaller
than the typical CED area. If the latter was true, we would have
different rates among the areas but, if these rates were mapped
by CED, they would not show a spatial pattern. The index proposed by Moran was used to test for spatial autocorrelation at
the CED level. Moran’s index tests if rates from CED neighbours
tend to be more similar than rates from CED chosen arbitrarily.17
We considered as neighbours those CED sharing boundaries.
The empirical Bayes index (EBI) proposed by Assunção and
Reis18 is an adjustment of Moran’s index which takes into
account the different population size used to calculate the rates.
It is more robust to eventual outliers due to small population
size and it is more powerful than Moran’s index.
Space and time clustering
The space-time interaction of cases of leishmaniasis was
determined through the statistic of Knox.19 This test assumes
that the number of ACL cases living near to each other and
occurring close in time is higher than it could be due to chance
alone. Note that the absence of spatial autocorrelation for each
time period does not imply, as a consequence, an absence of
space-time interaction.15 Amercian cutaneous leishmaniasis
cases were considered close in space if they belonged to the
same CED and were close in time if they belonged to the same
season within a year.
To test for the presence of time periodicity each year was
divided into two periods, rainy and dry, according to the pattern
of rainfall in Caratinga,20 based on the information that the
sandfly population size varies between seasons.10 The rainy
season extends from October to March, and the dry season
starts in April and ends in September.
Poisson regression analysis
In order to explain rate variability, a generalized linear model14
was fitted to the ACL case counts assuming a Poisson distribution with the population size as offset and with the logarithm of
the underlying rates depending on the following covariates:
population density, urban/rural indicator, household number,
average of household inhabitants, proportion of houses without
sanitary disposal, without water installation and with exposed
garbage. Information on these covariates was available only for
the 1991 census year.
Results
A total of 1712 new cases of ACL were reported in Caratinga
from 1966 to 1996, with an average crude incidence rate of
48/100 000 inhabitants.
Table 1 shows selected demographic and clinical variables by
area of residence. About 77% of all cases were from rural areas,
except for Caratinga, Ipaba and Santa Rita districts where
AMERICAN CUTANEOUS LEISHMANIASIS: SPACE–TIME CLUSTERING
985
Figure 2 The age adjusted incidence rate (per 100 000) by year of diagnosis and time trend of American
cutaneous leishmaniasis (ACL) in Caratinga, Southeast Brazil, 1966–1996
Table 2 Time trenda for annual counts of American cutaneous
leishmaniasis (ACL) in Caratinga, Southeast Brazil, 1966–1996
Coefficientsa
Intercept
Linear trend
Value
SE
t-value
P-value
–8.50
0.22
–38.73
,0.0001
0.01
0.003
4.07
0.0003
a Log-linear model using Poisson distribution.
77.1%, 50% and 65% of cases, respectively, were from urban
areas (P , 0.01). Gender distribution was found similar across
the area strata (P = 0.06). The median age was 20.0 years and
the average age was 25.5 (range 24.7–26.4 years). The age distribution of ACL cases according to the districts was significantly
different (P , 0.01), thus justifying age standardization.
According to the clinical diagnosis, the majority of cases (96%)
presented with the cutaneous form.
The incidence rates of ACL by year and the time trends are
shown in Figure 2. There is strong evidence of annual variation
during the three decades according to the χ2 homogeneity test
(χ2 = 1411.1, d.f. = 30, P , 0.01). The trend test gives evidence
of an increased rate of reported cases through the years (Table 2).
According to the fitted generalized linear model, ACL morbidity
increases at a rate of 5 cases/100 000 inhabitants a year and the
underlying rates varied from 41.4 in 1966 to 57.2 in 1996.
In order to investigate if the trend of the disease was different
among age categories, separated log-linear Poisson regression
models were used. We tested for interactions between the trend
of the incidence rates in the age classes, 0–9, 10–19, >20 years,
stratified for rural and urban areas. Trend in the incident rates
was not different for any age category, regardless of the area of
residence. So, pooling the age for rural areas, the model showed
increased rates of ACL cases over time, varying from 43.81 in
1970 to 169.08/100 000 in 1996 (P , 0.01). On the other hand,
the rates in urban areas presented a steady decrease over time
varying from 61.12 in 1970 to 2.71/100 000 in 1996 (P , 0.01)
(Figure 3).
In Figure 4, the results of the ACL spatial distribution are
presented for each CED from Caratinga. There was a wide variation on incidence rates over the three time periods. The results
of Oden’s test (Table 3) provide strong evidence for rejecting
the rate homogeneity hypothesis in each of the three periods.
Moran’s index did not show a spatial autocorrelation, but the
EBI test found evidence for this autocorrelation from 1986 to
1993. The EBI test was preferable to the Moran’s test because
the population in each area was different.
Significant evidence of space-time clustering in rural areas for
all decades and only for the 1970s and 1980s in urban areas was
found using the Knox analysis (Table 4). The pattern of this
space-time clustering was composed of epidemic peaks followed
by small incidence in the subsequent periods (data not shown).
Poisson regression analysis found a significant effect in the
overall rate of ACL in this region. The incidence rates were
independently associated with rural areas, areas without sanitary installations and areas with higher proportion of exposed
garbage as compared to urban areas, areas with a higher proportion of sanitary disposal and less exposed garbage (Table 5).
The model was able to explain 20% of the variation of the
incident rates observed among CED. Due to strong correlation,
all others covariates were not included in the final model.
Because of a very small population and very high incidence
rate, one CED was excluded from the model.
Discussion
This study shows an increase in the incidence rates of reported
ACL cases, with a significant positive trend over time, in a rural
area located in Southeast Brazil over the last 30 years. American
cutaneous leishmaniasis occurred in outbreaks, clustering
in space and time. Three independent ecological factors were
986
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Figure 3 Time trend and incidence rate (per 100 000) of American cutaneous leishmaniasis (ACL) according to residence
area, age and year of diagnosis, Caratinga, Southeast Brazil, 1970–1996
associated with incidence rate variability: rural area, area with a
higher proportion of households lacking sanitary disposal, and
with a higher proportion of households with exposed garbage.
Four possible explanations can be proposed: (1) variation of
exposure related to environmental modifications, such as urbanization of endemic rural areas; (2) variation in sandfly habits
and distribution as a consequence of the first argument; (3)
variation in the number of mammalian hosts and, (4) variation
in the number of susceptible individuals.
With respect to explanation (1), some studies in Brazil have
shown an association between the environmental modifications
of urbanization and the emergence or re-emergence of ACL
cases.7–9 Although information on urbanization in this area is
lacking, given that the incidence rates have been decreasing
over time in urban areas, contrasting with the steady increase
on incidence rates in rural areas, we postulate that environmental modifications could be occurring in both directions. For
urban areas these modifications would result in a decreased
exposure to several factors related to transmission. On the other
hand, for rural areas, a continuous or maybe increased exposure of human beings within the sylvan environment could be
able to maintain the transmission of the disease. To corroborate
this argument the adjusted Poisson model showed a significant
association with rural area after taking into account other social
or geographic factors.
With respect to explanation (2), some articles have argued
that the increase in ACL cases in children aged 0–9 years living
either in urban or rural areas may be due to increased occurrence of sandflies in and around households, possibly resulting
in a new transmission pattern for the disease.9,10,21 The absolute
number of sandflies captured in this area has varied, as well as
the relative proportion of different species. Also, data collected
in this area have shown that most of the sandflies were captured in woodland (59%), around domestic sites (33.4%) and
from the margins of rice fields (7.4%).6 Consequently, using the
time increase on incidence rates regardless of age group as a
surrogate marker for the transmission pattern in rural areas, the
data do not support a new transmission pattern in this area due
to increased sandfly occurrence in and around households over
the last 30 years.
With respect to explanation (3), we observed that the risk was
larger among the rural CED with a large number of houses
AMERICAN CUTANEOUS LEISHMANIASIS: SPACE–TIME CLUSTERING
987
Figure 4 American cutaneous leishmaniasis (ACL) age-adjusted incidence rate by tree periods according to
census enumeration districts, Caratinga, Southeast Brazil, 1980–1996
Table 3 Comparison of estimates using Moran’s, empirical Bayes and Oden’s indexes to evaluated the heterogeneity and autocorrelation spatial of
the incidence rates of American cutaneous leishmaniasis in Caratinga, Southeast Brazil, 1980–1996
Moran’s index
Empirical Bayes index
Oden’s index
Z-value
P-value
EBI
P-value
Z-value
1980–1985
–0.28
0.78
10.43
0.18
43.94
,0.01
1986–1993
1.37
0.17
30.24
0.02
364.35
,0.01
1994–1996
–0.66
0.50
18.14
0.31
78.41
,0.01
1980–1996
0.80
0.42
28.59
0.04
459.36
,0.01
Period
P-value
Table 4 Space-time interaction of American cutaneous leishmaniasis cases according to the residence area (urban/rural) in decades, Caratinga,
Southeast Brazil, 1966–1996
Urban areas
Rural areas
Total
Z-value
P-value
Z-value
P-value
Z-value
60
1.34
0.090
2.27
0.012
3.39
,0.001
70
2.36
0.009
10.03
,0.001
8.58
,0.001
80
3.25
0.001
10.04
,0.001
11.92
,0.001
90
0.70
0.242
4.37
,0.001
4.31
,0.001
Total
6.99
,0.001
26.97
,0.001
37.88
0.000
Decade
P-value
988
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Table 5 American cutaneous leishmaniasis incidence rate according to selected environmental risks per census enumeration districts, Caratinga,
Southeast Brazil, 1991
Coefficientsa
Value
SE
t-value
Intercept
–6.92
0.247
–28.05
,0.001
% houses without sanitary disposal
0.010
0.003
3.56
,0.001
0.024
0.003
8.54
,0.001
–1.651
0.239
–6.90
,0.001
% houses with exposed garbage
Urbanization indicator
P-value
a Poisson multiple regression.
without sanitary installations and with exposed garbage. These
findings suggest that human activity could attract the zoonotic
hosts to the domestic environment, as observed by other
authors.10,22,23
Arguments (2) and (3) support the idea of formation of an
ACL artificial biocenosis around the household, contributing to
the argument of transmission in this area.
With respect to explanation (4), some authors21 have suggested that, in Brazilian regions, ACL has explosive outbreaks
with very low incident case rates in the subsequent period.
There is a strong evidence that, following clinically apparent
infection, treated individuals acquire lifelong resistance,24
increasing the herd immunity. In this region, which has a very
low rate of migration into the area, three epidemic peaks were
observed in the last three decades. These peaks could be
explained by the birth of new susceptible cohorts of individuals
with a subsequent accumulation of susceptible individuals
reaching a threshold sufficient to initiate an epidemic. When
the susceptibles pool was depleted, the epidemics declined until
a new large susceptible pool was formed again.
Also, considering that the population growth rate was constant
over time, variation in the susceptible pool due to migration and
death rates does not seem a plausible explanation for ACL case
variation in this region.
These ecological associations could be spurious if the
variation in incidence was due to changes in the surveillance
system during the epidemic period, with diagnosis, notification
and flow of information improvement, as well as changes in
popular knowledge of the disease, resulting in more medical
care seeking behaviour. Also, changes in the disease classification and diagnostic criteria were not observed in this area
during the study. In Caratinga, the continuous existence of
collaboration between the Federal University of Minas Gerais
and the public health service created a consistent surveillance
system over the last 30 years. Surveillance bias would not be
expected to produce a distorted trend in the observed ACL rates
in this setting.
Despite the impossibility of accurately measuring relevant
exposures at the individual level in this large number of subjects, this study suggests an apparent cluster of ACL in small
rural areas over time.25 Differences in the communities that
might explain a large amount of the incidence rate variability
could be an increased sandfly population in the vicinity of
dwellings due to low sanitation including higher proportion of
exposed garbage, associated with cyclic provision of susceptible
individuals. This suggests the creation of a very susceptible biocenosis that would be responsible for occasional peaks of incidence within small regions.
Further studies, combining individual-level analysis in each
group, followed by an ecological analysis of all the group using
the results from the individual level will be able to clarify the
findings described here.
Acknowledgements
The authors would like to acknowledge Mr Jair Cecílio de
Paula, Ms Jane de Paula Medina, Ms Edna Cristina Rocha
Muniz, Professor Paul Williams, Dr Ana Clara Moura Mourão
and Dr Odair Genaro for their help during this study. This study
was supported by the CNPq, PRPG/UFMG, FAPEMIG, FNS and
FBB.
References
1 Pereira GFM, Fonseca HHR. Leishmaniose tegumentar americana:
epidemiologia e controle. Rev Soc Bras Med Trop 1994;27:45–50.
2 Lainson R. The american leishmaniases: some observations on their
ecology and epidemiology. Trans R Soc Trop Med Hyg 1982;77:569–96.
3 Montoya J, Jaramillo C, Palma G, Gomez T, Segura I, Travi B. Report
of an epidemic outbreak of tegumentary leishmaniasis in a coffeegrowing area of Colombia. Mem Inst Oswaldo Cruz 1990;85:119–21.
4 Rangel EF, Azevedo AC, Andrade CA, Souza NA, Wermelinger ED.
Studies on sandfly fauna (Diptera: Psychodidae) in a foci of cutaneous
leishmaniasis in Mesquita, Rio de Janeiro State, Brazil. Mem Inst
Oswaldo Cruz 1990;85:39–45.
5 Walsh JF, Molyneux DH, Birley MH. Deforestation: effects on vector-
borne disease. Parasitology 1993;106:S55–75.
6 Mayrink W, Williams P, Coelho MV et al. Epidemiology of dermal
leishmaniasis in the Rio Doce valley, state of Minas Gerais, Brazil. Ann
Trop Med Parasitol 1979;73:123–37.
7 Oliveira-Neto MP, Pirmez C, Rangel E, Schubach A, Grimaldi JRG. An
outbreak of american cutaneous leishmaniasis (Leishmania braziliensis
braziliensis) in a periurban area of Rio de Janeiro City, Brazil: clinical
and epidemiological studies. Mem Inst Oswaldo Cruz 1988;83:427–35.
8 Passos VMA, Falcão AL, Katz N. Urban american cutaneous
leishmaniasis in the metropolitan region of Belo Horizonte, Minas
Gerais state, Brazil. Mem Inst Oswaldo Cruz 1990;85:243–44.
9 Passos VM, Falcão AL, Marzochi MCA et al. Epidemiological aspects
of American cutaneous leishmaniasis in a periurban area of the
metropolitan region of Belo Horizonte, Minas Gerais, Brazil. Mem Inst
Oswaldo Cruz 1993;88:103–10.
10 Teodoro U, Filho VLS, Lima EM et al. Observations on the behavior of
phlebotomines in forest and domiciliary environments in an area of
American cutaneous leishmaniasis in southern Brazil. Rev Saude
Publica 1993;27:242–49.
11 Instituto Brasileiro de Geografia e Estatística. Censo demográfico de 1960,
Minas Gerais, VII Recenseamento Geral do Brasil, Série Regional, Vol. I, Tom.
IX. Rio de Janeiro: IBGE, 1963.
AMERICAN CUTANEOUS LEISHMANIASIS: SPACE–TIME CLUSTERING
989
12 Instituto Brasileiro de Geografia e Estatística. Censo demográfico, Minas
20 Nimer E, Brandão AMPM. Balanço hídrico e clima da região dos cerrados.
Gerais, VIII Recenseamento Geral—1970, Série Regional, Vol. 1, Tom. XIV,
Part 3, Resultados segundo as Microrregiões, os Municípios e os Distritos. Rio
de Janeiro: IBGE, 1973.
21 Gomes AC, Yamamoto YI, Capinzaiki NA, Amaral NMM, Guimarães
Rio de Janeiro: FIBGE, 1989.
Brasil—1980, Vol. 1, Tom. 3, No. 14. Censo demográfico: dados distritais,
Minas Gerais. Rio de Janeiro: IBGE, 1982.
AJG. Aspectos ecológicos da leishmaniose tegumentar americana:
Prevalência/Incidência da infecção humana nos municípios de Pedro
de Toledo e Miracatu, São Paulo, Brasil. Rev Inst Med Trop São Paulo
1992;34:149–58.
14 Instituto Brasileiro de Geografia e Estatística. Censo demográfico, 1991,
22 Aragão MB. Sobre o comportamento de alguns insetos hematófagos.
Resultados do universo relativos às características da população e dos
domicílios, No. 18, Minas Gerais. Rio de Janeiro: IBGE, 1994.
23 Gomes AC, Rabelo EX, Santos JL, Galati EA. Aspectos ecológicos da
13 Instituto Brasileiro de Geografia e Estatística. IX Recenseamento Geral do
15 Estève J, Benhamou E, Raymond L. Statistical Methods in Cancer
Arq Biol Tecnol 1975;18:3–23.
Research. Vol IV. Descriptive Epidemiology. (IARC Scientific Publications
No. 128). Lyon: International Agency for Research on Cancer, 1994.
leishmaniose tegumentar americana. 3. Observações naturais sobre o
ritmo diário de atividade de Ps. intermedia em ambiente florestal e
extraflorestal. Rev Saude Publica 1983;17:23–30.
16 Oden N. Adjusting Moran’s I for population density. Stat Med 1995;
24 Weigle K, Saraiva NG. Natural history, clinical evolution, and the
14:17–26.
17 Moran PAP. Notes on continuous stochastic phenomena. Biometrika
1950;37:17–23.
18 Assunção RM, Reis EA. A new proposal to adjust Moran’s I for
population density. Stat Med 1998;9:18.
19 Knox EG. The detection of space-time interaction. Appl Stat 1964;13:
25–29.
host-parasite interaction in new world cutaneous leishmaniasis.
Clin Dermatol 1996;14:433–50.
25 Susser M. The logic in ecological: II. The logic of design. Am J Public
Health 1994;84:830–35.