Applications of spatio temporal analytical methods in surveillance of ross river virus disease

APPLICATIONS OF SPATIO-TEMPORAL ANALYTICAL METHODS IN SURVEILLANCE OF ROSS RIVER VIRUS DISEASE BY WENBIAO HU BMed A thesis submitted for the Degree of Doctor of Philosophy in the Cen

APPLICATIONS OF SPATIO-TEMPORAL

ANALYTICAL METHODS IN SURVEILLANCE OF

ROSS RIVER VIRUS DISEASE

BY

WENBIAO HU BMed

A thesis submitted for the Degree of Doctor of Philosophy in the Centre

for Health Research, Queensland University of Technology

MAY 2005

For my wife, Xiaodong and our son Junqian

KEYWORDS

Classification and regression trees, cluster analysis, generalised linear model, geographic information system, interpolation, polynomial distributed lag model, principal components analysis, Ross River virus disease, seasonal auto-regressive integrated moving average, socio-ecological factors, time series analysis

SUMMARY

The incidence of many arboviral diseases is largely associated with social and environmental conditions Ross River virus (RRV) is the most prevalent arboviral disease in Australia It has long been recognised that the transmission pattern of RRV

is sensitive to socio-ecological factors including climate variation, population movement, mosquito-density and vegetation types This study aimed to assess the relationships between socio-environmental variability and the transmission of RRV using spatio-temporal analytic methods

Computerised data files of daily RRV disease cases and daily climatic variables in Brisbane, Queensland during 1985-2001 were obtained from the Queensland Department of Health and the Australian Bureau of Meteorology, respectively Available information on other socio-ecological factors was also collected from relevant government agencies as follows: 1) socio-demographic data from the Australia Bureau of Statistics; 2) information on vegetation (littoral wetlands, ephemeral wetlands, open freshwater, riparian vegetation, melaleuca open forests, wet eucalypt, open forests and other bushland) from Brisbane City Council; 3) tidal activities from the Queensland Department of Transport; and 4) mosquito-density from Brisbane City Council

Principal components analysis (PCA) was used as an exploratory technique for discovering spatial and temporal pattern of RRV distribution The PCA results show that the first principal component accounted for approximately 57% of the information, which contained the four seasonal rates and loaded highest and positively

characterised by three groups with high, medium and low incidence of disease, and it suggests that there are at least three different disease ecologies The variation in spatio-temporal patterns of RRV indicates a complex ecology that is unlikely to be explained by a single dominant transmission route across these three groupings Therefore, there is need to explore socio-economic and environmental determinants of RRV disease at the statistical local area (SLA) level

Spatial distribution analysis and multiple negative binomial regression models were employed to identify the socio-economic and environmental determinants of RRV disease at both the city and local (ie, SLA) levels The results show that RRV activity was primarily concentrated in the northeast, northwest and southeast areas in Brisbane The negative binomial regression models reveal that RRV incidence for the whole of the Brisbane area was significantly associated with Southern Oscillation Index (SOI) at a lag of 3 months (Relative Risk (RR): 1.12; 95% confidence interval (CI): 1.06 - 1.17), the proportion of people with lower levels of education (RR: 1.02; 95% CI: 1.01 - 1.03), the proportion of labour workers (RR: 0.97; 95% CI: 0.95 – 1.00) and vegetation density (RR: 1.02; 95% CI: 1.00 – 1.04) However, RRV incidence for high risk areas (ie, SLAs with higher incidence of RRV) was significantly associated with mosquito density (RR: 1.01; 95% CI: 1.00 - 1.01), SOI at

a lag of 3 months (RR: 1.48; 95% CI: 1.23 - 1.78), human population density (RR: 3.77; 95% CI: 1.35 - 10.51), the proportion of indigenous population (RR: 0.56; 95% CI: 0.37 - 0.87) and the proportion of overseas visitors (RR: 0.57; 95% CI: 0.35 – 0.92) It is acknowledged that some of these risk factors, while statistically significant, are small in magnitude However, given the high incidence of RRV, they may still be important in practice The results of this study suggest that the spatial pattern of RRV

disease in Brisbane is determined by a combination of ecological, socio-economic and environmental factors

The possibility of developing an epidemic forecasting system for RRV disease was explored using the multivariate Seasonal Auto-regressive Integrated Moving Average (SARIMA) technique The results of this study suggest that climatic variability, particularly precipitation, may have played a significant role in the transmission of RRV disease in Brisbane This finding cannot entirely be explained by confounding factors such as other socio-ecological conditions because they have been unlikely to change dramatically on a monthly time scale in this city over the past two decades SARIMA models show that monthly precipitation at a lag 2 months (β=0.004, p=0.031) was statistically significantly associated with RRV disease It suggests that that there may be 50 more cases a year for an increase of 100 mm precipitation on average in Brisbane The predictive values in the model were generally consistent with actual values (root-mean-square error (RMSE): 1.96) Therefore, this model may have applications as a decision support tool in disease control and risk-management planning programs in Brisbane

The Polynomial distributed lag (PDL) time series regression models were performed

to examine the associations between rainfall, mosquito density and the occurrence of RRV after adjusting for season and auto-correlation The PDL model was used because rainfall and mosquito density can affect not merely RRV occurring in the same month, but in several subsequent months The rationale for the use of the PDL technique is that it increases the precision of the estimates We developed an epidemic forecasting model to predict incidence of RRV disease The results show that 95%

density and rainfall, respectively The predictive values in the model were generally consistent with actual values (RMSE: 1.25) The model diagnosis reveals that the residuals were randomly distributed with no significant auto-correlation The results

of this study suggest that PDL models may be better than SARIMA models (R-square increased and RMSE decreased) The findings of this study may facilitate the development of early warning systems for the control and prevention of this wide-spread disease

Further analyses were conducted using classification trees to identify major mosquito species of Ross River virus (RRV) transmission and explore the threshold of mosquito

density for RRV disease in Brisbane, Australia The results show that Ochlerotatus vigilax (RR: 1.028; 95% CI: 1.001 – 1.057) and Culex annulirostris (RR: 1.013, 95%

CI: 1.003 – 1.023) were significantly associated with RRV disease cycles at a lag of 1 month The presence of RRV was associated with average monthly mosquito density

of 72 Ochlerotatus vigilax and 52 Culex annulirostris per light trap These results may

also have applications as a decision support tool in disease control and management planning programs

risk-As RRV has significant impact on population health, industry, and tourism, it is important to develop an epidemic forecast system for this disease The results of this study show the disease surveillance data can be integrated with social, biological and environmental databases These data can provide additional input into the development of epidemic forecasting models These attempts may have significant implications in environmental health decision-making and practices, and may help health authorities determine public health priorities more wisely and use resources more effectively and efficiently

TABLE OF CONTENTS

KEYWORDS III SUMMARY IV

TABLE OF CONTENTS 1

LIST OF TABLES 5

LIST OF FIGURES 7

DEFINITION OF TERMS 10

ABBREVIATIONS 12

STATEMENT OF ORIGIANL AUTHORSHIP 13

ACKNOWLEDGEMENTS 14

PUBLICATIONS BY THE CANDIDATE (2001 - 2004) 16

CHAPTER 1: INTRODUCTION AND BACKGROUND 20

1.1 INTRODUCTION 20

1.2 AIMS AND HYPOTHESES 24

1.3 SIGNIFICANCE OF THE THESIS 25

1.4 CONTENTS AND STRUCTURE OF THE THESIS 26

CHAPTER 2: APPLICATIONS OF GIS AND SPATIAL ANALYSIS IN MOSQUITO-BORNE DISEASE RESEARCH: A REVIEW OF RELATED LITERATURE 29

2.1 SYSTEMATIC REVIEW 29 2.2 CRITICAL APPRAISAL OF KEY SPATIO-TEMPORAL ANALYTIC

2.3 APPLICATIONS OF GIS AND SPATIO-TEMPORAL ANALYTIC METHODS

IN RRV RESEARCH 51

2.4 KNOWLEDGE GAPS IN THIS AREA 56

CHAPTER 3: STUDY DESIGN AND METHOD 58

3.1 STUDY SITE AND STUDY POPULATION 58

3.2 STUDY DESIGN 61

3.3 DATA COLLECTION AND MANAGEMENT 61

3.4 DATA LINKAGES 63

3.5 DATA ANALYSIS 63

3.6 THE LIMITATIONS OF THE STUDY 69

CHAPTER 4: SPATIAL AND TEMPORAL PATTERNS OF ROSS RIVER VIRUS IN BRISBANE, AUSTRALIA 72

ABSTRACT 73

4.1 INTRODUCTION 74

4.2 MATERIAL AND METHODS 76

4.3 RESULTS 78

4.4 DISCUSSION 84

REFERENCES 89

CHAPTER 5: SPATIAL ANALYSIS OF SOCIAL AND ENVIRONMENTAL FACTORS ASSOCIATED WITH ROSS RIVER VIRUS IN BRISBANE, AUSTRALIA 93

ABSTRACT 94

5.1 INTRODUCTION 95

5.2 MATERIALS AND METHODS 96

5.3 RESULTS 99

5.4 DISCUSSION 107

REFERENCES 114

CHAPTER 6: DEVELOPMENT OF A PREDICTIVE MODEL FOR ROSS RIVER VIRUS DISEASE IN BRISBANE, AUSTRALIA 119

ABSTRACT 120

6.1 INTRODUCTION 121

6.2 MATERIALS AND METHODS 123

6.3 RESULTS 128

6.4 DISCUSSION 141

ACKNOWLEDGEMENTS 146

REFERENCES 147

CHAPTER 7: RAINFALL, MOSQUITO DENSITY AND THE TRANSMISSION OF ROSS RIVER VIRUS: AN EPIDEMIC FORECASTING MODEL 153

ABSTRACT 154

7.1 INTRODUCTION 155

7.2 METHODS 157

7.3 RESULTS 159

7.4 DISCUSSION 167

ACKNOWLEDGEMENTS 170

REFERENCES 172

CHAPTER 8: MOSQUITO SPECIES AND THE TRANSMISSION OF ROSS RIVER VIRUS IN BRISBANE, AUSTRALIA 176

ABSTRACT 177

8.1 INTRODUCTION 178

8.2 MATERIALS AND METHODS 179

8.3 RESULTS 181

8.4 DISCUSSION 188

ACKNOWLEDGEMENTS 190

REFERENCES 191

CHAPTER 9: GENERAL DISCUSSION 194

9.1 INTRODUCTION 194

9.2 SUBSTANTIVE DISCUSSION 194

9.3 THE IMPLICATIONS OF THE STUDY 201

9.4 THE STRENGTHS AND LIMITATIONS OF THE STUDY 203

9.5 RECOMMENDATIONS 205

APPENDIX 211

DATA COLLECTION 211

REFERENCES 225

LIST OF TABLES

T ABLE 2 1 T HE CODING CATEGORIES FOR THE LITERATURE REVIEW 32

T ABLE 2 2 A RTICLE NUMBERS BY JOURNAL BASED ON GENERAL HEALTH DOMAIN

(F IRST 50 JOURNALS ) 34

T ABLE 2 3 A RTICLE NUMBERS BY JOURNAL BASED ON MBD 35

T ABLE 2 4 C HARACTERISTICS OF 58 MBD PAPERS 38

T ABLE 2 5 S TATISTICAL TECHNIQUES AND COMPUTER SOFTWARE FOR SPATIAL ANALYSIS * 43

T ABLE 4 1 P RINCIPAL COMPONENT EIGENVALUES AND LOADING FOR EACH SEASON VARIABLES 83

T ABLE 4 2 S TATISTICAL CRITERIA FOR THE NUMBERS OF CLUSTERS 83

T ABLE 4 3 A NALYSIS OF VARIANCE IN CLUSTER ANALYSIS 84

T ABLE 5 1 S PEARMAN CORRELATION COEFFICIENTS BETWEEN MONTHLY

INCIDENCE OF RRV AND SOCIAL AND ENVIRONMENTAL VARIABLES IN

B RISBANE * 102

T ABLE 5 2 A DJUSTED RELATIVE RISKS FOR THE TRANSMISSION OF RRV IN

B RISBANE , A USTRALIA * 107

T ABLE 6 1 C HARACTERISTICS OF E XPLANATORY V ARIABLES 129

T ABLE 6 2 R EGRESSION COEFFICIENTS OF SARIMA ON THE MONTHLY INCIDENCE

OF RRV DISEASE IN B RISBANE , A USTRALIA , 1985 – 2000 137

T ABLE 7 1 S PEARMAN CORRELATION COEFFICIENTS BETWEEN MONTHLY

INCIDENCE OF RRV AND RAINFALL AND MOSQUITO DENSITY 160

T ABLE 7 2 PDL REGRESSION COEFFICIENTS OF RAINFALL AND MOSQUITO DENSITY

ON THE MONTHLY INCIDENCE OF RRV DISEASE IN B RISBANE , A USTRALIA * 163

T ABLE 8 1 C ROSS CORRELATION COEFFICIENTS BETWEEN MOSQUITO DENSITY AND INCIDENCE OF RRV 184

T ABLE 8 2 T IME SERIES P OISSON REGRESSION MODELS USED TO ADJUST FOR THE AUTOCORRELATION OF MONTHLY INCIDENCE RATES OF RRV AND

SEASONALITY * 185

LIST OF FIGURES

F IGURE 1 1 F LOWCHART OF 5 MANUSCRIPTS IN THESIS 28

F IGURE 2 1 T HE RESULTS OF SEARCH BASED ON GIS AND SPATIAL ANALYSIS IN

M EDLINE 31

F IGURE 2 2 T RENDS OF PUBLICATIONS ON GIS FOR GENERAL HEALTH DOMAINS 33

F IGURE 2 3 T RENDS AND DISTRIBUTION OF EMPIRICAL ARTICLES ON GIS AND

SPATIAL ANALYSIS FOR MBD 36

F IGURE 3 1 L OCATION OF THE STUDY AREA - B RISBANE 60

F IGURE 4 1 T HE ANNUAL INCIDENCE OF RRV INFECTIONS AND RAINFALL IN

B RISBANE , 1985 - 2001 79

F IGURE 4 2 H ISTOGRAM OF SEASONAL INCIDENCE OF RRV IN B RISBANE , 1985 –

2001 (X AXIS : SEASONAL INCIDENCE OF RRV, Y AXIS : FREQUENCY ( I E ,

NUMBERS FOR SLA S )) 79

F IGURE 4 3 S EASONAL INCIDENCE OF RRV DISEASE FOR SLA ACROSS B RISBANE

(F IGURE 4.3-A: S PRING ; F IGURE 4.3-B: S UMMER ; F IGURE 4.3-C: A UTUMN ;

F IGURE 4.3-D: W INTER ) 82

F IGURE 4 4 K- MEANS CLUSTERING ANALYSIS OF INCIDENCE RATE OF RRV IN

B RISBANE , A USTRALIA 84

F IGURE 5 1 L OCATION OF B RISBANE , A USTRALIA 97

F IGURE 5 2 T HE DISTRIBUTION OF RRV INFECTIONS IN 2001, B RISBANE (C ROSS REFERS TO MOSQUITO MONITOR STATIONS WHICH LOCATED IN HIGH RISK

AREAS BASED ON DISEASE MONITORING RECORDS ) 100

F IGURE 5 3 S PATIAL DISTRIBUTION MODEL USING INVERSE DISTANCE WEIGHTED INTERPOLATION 104

F IGURE 6 1 L OCATION OF B RISBANE , Q UEENSLAND , A USTRALIA ( INCLUDING LATITUDE AND LONGITUDE OF THE CITY ) 124

F IGURE 6 2 T HE RELATIONSHIPS BETWEEN MONTHLY INCIDENCE OF R OSS R IVER VIRUS AND CLIMATE VARIABLES IN B RISBANE BETWEEN 1985 AND 2001 ( USING SEASONALLY DIFFERENCED VARIABLES ) 134

F IGURE 6 3 C ROSS - CORRELATION FUNCTION BETWEEN R OSS R IVER VIRUS AND CLIMATE VARIABLES AFTER SEASONAL DIFFERENCING 136

F IGURE 6 4 A: A UTO - CORRELATION (ACF); B: P ARTIAL AUTO - CORRELATION OF RESIDUALS (PACF); AND C: S CATTERPLOT OF RESIDUALS 138

F IGURE 6 5 T HE VALIDATED SARIMA MODEL OF CLIMATE VARIATION IN

B RISBANE ( VALIDATION PERIOD : 1.2001 – 12 2001 IE , TO THE RIGHT OF THE VERTICAL DOTTED LINE ) 140

F IGURE 7 1 M OSQUITO DENSITY , RAINFALL AND R OSS R IVER VIRUS DISEASE IN

TO THE RIGHT OF THE VERTICAL DOTTED LINE ) 166

F IGURE 7 4 A UTO - CORRELATION , PARTIAL AUTO - CORRELATION OF RESIDUALS 167

F IGURE 8 1 L OCATION OF B RISBANE , A USTRALIA 179

F IGURE 8 2 10 MOSQUITO MONITOR STATIONS , B RISBANE , A USTRALIA 182

F IGURE 8 3 T HE DISTRIBUTION OF MOSQUITO SPECIES BY SEASON IN B RISBANE ,

A USTRALIA 183

F IGURE 8 4 P ROPORTION OF MOSQUITO SPECIES IN B RISBANE , A USTRALIA 183

F IGURE 8 5 C LASSIFICATION TREE FOR THE RELATIONSHIP BETWEEN

O CHLEROTATUS VIGILAX DENSITY AND RRV* 187

F IGURE 8 6 C LASSIFICATION TREE FOR THE RELATIONSHIP BETWEEN C ULEX

ANNULIROSTRIS DENSITY AND RRV* 187

F IGURE 9 1 F RAMEWORK OF RESEARCH RESULTS IN THIS THESIS 196

F IGURE 9 2 F RAMEWORK OF RESEARCH RECOMMENDATIONS IN THIS THESIS 210

DEFINITION OF TERMS

Classification and Regression Trees - builds classification and regression trees for

predicting continuous dependent variables (regression) and categorical predictor variables (classification)

Cluster Analysis – is one of data reduction methods that is used to group together

entities with similar properties

Eigenvalues - measure the amount of the variation explained by each principal

component (PC) and will be largest for the first PC and smaller for the subsequent PCs An eigenvalue greater than 1 indicates that PCs account for more variance than accounted by one of the original variables

El Niño/Southern Oscillation - is a systematic pattern of global climate variability

(Nicholls 1993) It affects most countries in the Pacific and Indian Oceans, bringing long droughts and extended wet periods every two to seven years

Generalised Linear Model - a model for linear and non-linear effects of continuous

and categorical predictor variables on a discrete or continuous but not necessarily normally distributed dependent (outcome) variable

Geographical Information System - can be seen as a system of hardware, software

and procedures (tools) designed to capture, manage, manipulate, analysis, modelling, and display spatial or geo-referenced data for solving complex planning and management problems

Multicolinearity - in a multiple regression with more than one X variable, two or

more X variables are colinear if they show strong linear relationships This makes estimation of regression coefficients impossible It can also produce unexpectedly

Overdispersion - is the situation that occurs most frequently in Poisson and binomial

regression when variance is much higher than the mean (whereas it should be the same)

Poisson Regression - Analysis of the relationship between an observed count with a

Poisson distribution (i.e., outcome variable) and a set of explanatory variables

Polynomial - a sum of multiples of integer powers of a variable The highest power in

the expression is the degree of the polynomial

Principal Components Analysis - is a useful method of data interpretation which

assists in identifying and understanding data structure

Relative Risk – the ratio of the cumulative incidence rate in the exposed to the

cumulative incidence rate in the unexposed

Residuals - reflect the overall badness-of-fit of the model They are the differences

between the observed values of the outcome variable and the corresponding fitted values predicted by the regression line (the vertical distance between the observed values and the fitted line)

Southern Oscillation Index - defined as the normalized difference in atmospheric

pressure between Darwin (Australia) and Tahiti (French Polynesia) The SOI accounts for up to 40% of variation in temperature and rainfall in the Pacific

Statistical Local Areas - is a general purpose spatial unit It is the base spatial unit

used to collect and disseminate statistics other than those collected from the Population Censuses

ABBREVIATIONS

ABS Australian Bureau of Statistics

CARTs Classification and Regression Trees

EIP Extrinsic Incubation Period

ENSO EI Nino-Southern Oscillation

GIS Geographic Information System

GLM Generalised Linear Model

MBD Mosquito-Borne Disease

NNDSS National Notifiable Diseases Surveillance System PCA Principal Components Analysis

PDL Polynomial Distribution Lag

RRV Ross River Virus

SARIMA Seasonal AutoRegression Integrated Moving Average

SLA Statistical Local Areas

SOI Southern Oscillation Index

STATEMENT OF ORIGIANL AUTHORSHIP

The work contained in this thesis has not been previously submitted for a degree or diploma at any other higher education institution To the best of my knowledge and belief, the thesis contains no material previously published or written by another person except where due reference is made

Signed:

Date:

ACKNOWLEDGEMENTS

I reserve my greatest thanks and appreciation to my supervisory team, A/Prof Shilu Tong, Prof Kerrie Mengersen and Prof Brian Oldenburg, for their critical and thoughtful comments, and guidance, support, encouragement and advice through the course of my PhD study At all times throughout my candidature they have maintained diligence, interest and enthusiasm for my research I would like to thank A/Prof Shilu Tong, my principal supervisor, for his significant amount of time spent

on the professional guidance of my study and his generous financial support to assist

me to complete my thesis He has not only been an excellent mentor but also a constant source of inspiration and motivation It is difficult to imagine how I would have completed this thesis without his guidance I would like to thank Prof Kerrie Mengersen for her statistical advice and helpful comments on my project I would like

to thank Prof Brian Oldenburg in his capacity as an experienced researcher in looking over my project, and for his personal and professional guidance I would also like to heartfelt thank Prof Beth Newman and Dr John Aaskov for their invaluable advice

on my thesis It has been an honour for me to establish a strong personal and professional relationship with both of them

I am indebted to all the organisations involved in this project All of whom are acknowledged below:

The Queensland Department of Health for providing the health outcome data in Queensland

The Australian Bureau of Meteorology for providing the meteorological data

The Queensland Department of Transport for providing the high tide data

Brisbane City Council for providing the vegetation and mosquito density data

Australian Bureau of Statistics for providing the socio-demographic data

I would also like to acknowledge all my colleagues in the Centre for Health Research for their advice and assistance with research and personal friendship

Finally, I would like to specially thank my wife, Xiaodong, and my son Junqian, for their love, patience, encouragement and emotional support through this endeavour and for their suggestions and comments on my research

PUBLICATIONS BY THE CANDIDATE (2001 - 2004)

JOURNAL ARTICLES

In thesis

1 Hu W, Nicholls N, Lindsay M, Dale P, McMichael AJ, Machenzie J and Tong S

Development of a predictive model for Ross River virus disease in Brisbane, Australia American Journal of Tropical Medicine and Hygiene 2004;71:129-137

2 Hu W, Mengersen K, Oldenburg B and Tong S Spatial analysis of social and

environmental factors associated with Ross River virus in Brisbane, Australia Acta Tropica Under review

3 Hu W, Tong S, Mengersen K, Oldenburg B and Dale P Spatial and temporal

patterns of Ross River virus in Brisbane, Australia Arbovirus Research in Australia Under review

4 Hu W, Tong S, Mengersen K, Oldenburg B and Dale P Mosquito species and the

transmission of Ross River virus in Brisbane, Australia To be submitted

5 Hu W, Tong S, Mengersen K and Oldenburg B Rainfall, mosquito density and

the transmission of Ross River virus: a time series forecasting model Ecological Modelling Under review

6 Hu W, Zhang J, Oldenburg B and Tong S Applications of GIS and spatial

analysis in mosquito-borne disease research: a review of related literature International Journal of Health and Geographics Under review

Not included in thesis

7 Hu W, McMichael AJ and Tong S El Nino/Southern Oscillation and the

Transmission of Hepatitis A Virus in Australia Medical Journal of Australia 2004;180:488-489

8 Hu W, Tong S and Oldenburg B Applications of spatio-temporal analytical

methods in surveillance and control of communicable disease Australasian Epidemiologist.2004;11:6-12

9 Tong S, Hu W and McMichael AJ Climate variability and Ross River virus

transmission in Townsville region, Australia, 1985-1996 Tropical Medicine and International Health 2004;9:298-304

10.Tong S and Hu W Different responses of Ross River Virus to climate variability

between coastline and inland cities in Queensland, Australia Occupational and Environmental Medicine 2002;59:739-744

11.Tong S and Hu W Climate variables and incidence of Ross River virus in Cairns,

Australia: a time series analysis Environmental Health Perspectives 2001;109:1271-1273

12 Hu W, Tong S, Oldenburg B and Feng X Serum vitamin A concentration and

growth in children and adolescents in Gansu province, China Asia Pacific Journal of Clinical Nutrition 2001;10:63-66

13.Tong S and Hu W Effects of climate variation on the transmission of Ross River

virus in Queensland, Australia Environmental Health 2001;1:45-51

Published abstracts

1 Hu W, Mengersen K, Oldenburg and Tong S Spatial analysis of social and

environmental factors associated with Ross River virus in Brisbane, Australia

Epidemiology 2004;15:S98

2 Hu W, Tong S Ross River virus transmission and El Nino Southern-Southern

Oscillation in Australia Epidemiology 2003;14: S17

3 Hu W, Tong S Exploratory spatial analysis of Ross River virus in Brisbane,

Australia, 1987-2001 Australasian epidemiologist 2003;53

4 Hu W, Tong S Preliminary development of an epidemic forecasting model of

Ross River virus disease in relation to environmental variation Australasian epidemiologist 2003; 22

5 Hu W, Zhang J, Tong S, et al Application of geographic information systems

(GIS) and spatial analysis in epidemiological research Epidemiology 2003;14:S16

6 Hu W and Tong S Exploratory spatial analysis of Ross River virus in Brisbane,

Australia, 1987-2001 Australasian epidemiologist 2003;10:53

7 Tong S, Hu W Different responses of Ross River virus to climate variability

between coastline and inland cities in Queensland, Australia Epidemiology 2002;13:30

8 Hu W, Mengersen K, Tong S Spline regression and auto-regression models with

application to time-series data Epidemiology 2002;13:757

9 Tong S, Hu W Climate variability and Ross River virus transmission in

Townsville, Australia: A SARIMA model American Journal of Epidemiology 2002;1551:145

10 Tong S, Hu W Effects of climate variation on the transmission of Ross River

virus in Australia American Journal of Epidemiology 2002;155:146

11 Hu W, Tong S Climate variation and incidence of Ross River virus in Cairns,

Australia: A time series analysis Epidemiology 2001;12:137

Book Chapter

Tong S, Bi P and Hu W Environmental Epidemiology In: Guo X et al, eds

Environmental Medicine Beijing, China: Beijing Medical University, 2002:15-30

CHAPTER 1: INTRODUCTION AND BACKGROUND

1.1 INTRODUCTION

1.1.1 The burden of Ross River virus disease in Australia

There are many vector-borne diseases (VBDs) in Australia, including Ross River virus (RRV) disease, Barmah forest virus, Australia encephalitis, dengue fever, Kunjin virus, etc RRV disease is the most prevalent vector-borne disease in Australia

and some Pacific island countries (Aaskov et al 1981a, Rosen et al 1981, Scrimgeous et al 1987, Mackenzie et al 1994) RRV causes a non-fatal, but

potentially debilitating, disease of humans known as epidemic polyarthritis or RRV disease (ICD-9: 663) The disease syndrome is characterized by headache, fever, rash, lethargy and muscle and joint pain The arthritic symptoms and lethargy may persist for many months and can be severe (Condon and Rouse 1995) Since 1991, several thousand cases of RRV disease throughout Australia have been reported each year to the National Notifiable Disease Surveillance System (NNDSS), and the majority of these cases are usually from Queensland (eg, approximately 82% of cases from Queensland in 2002) (Australian Department of Health and Aged Care 2004) The single largest reported outbreak occurred in the South Pacific islands in 1979-80,

during which more than 50,000 people were affected (Aaskov et al 1981a) RRV activity appears to have increased in Australia in the past decade (Harley et al 2001,

Australian Department of Health and Aged Care 2004), but the reasons for this remain

largely unknown (Harley et al 2001) It is estimated that the direct economic cost of RRV is approximately $2,500 per case (Hawkes et al 1985, Harley et al 2001), and

the economic impact of this disease is on the order of tens of millions of dollars

annually in direct and indirect health costs nationally (Hawkes et al 1985, Boughton

1994, Russell 1998b)

1.1.2 Transmission of RRV

Ross River virus circulates enzootically in reservoir populations of marsupials in Australia Infection is asymptomatic in host animals, but while they are viremic, host animals can infect mosquitoes that feed upon them After a variable period of time (the extrinsic incubation period), virus particles replicate to the point where the mosquito’s saliva is infective to the mosquito’s next non-immune vertebrate host If a human is bitten instead, clinical disease may result At least 20% of infected

individuals develop an acute disease (Weinstein 1997, Harley et al 2001, Russell

2002)

For the transmission of RRV, the virus and its reservoir, the vector, the human population, and environmental conditions are key factors The virus is dependent on the continuing presence of non-immune hosts in the reservoir population The distribution and abundance of the reservoir population will thus affect the availability

of viremic individuals to mosquitoes and a non-immune reservoir population leads to increased virus activity A number of vector-related factors also influence the level of RRV activity The mosquitoes are efficient vectors of the disease because of their susceptibility to the virus and the readiness with which they bite reservoir as well as human hosts The greater the abundance of mosquitoes, the greater the probability of being bitten (Weinstein 1997) The human population is susceptible to RRV infection

if individuals are non-immune and are exposed to the virus at the reservoir/mosquito/human interface Such exposure is enhanced by human intrusions

activities (Weinstein 1997, Harley et al 2001) Weather conditions directly affect the

breeding, survival, and abundance of mosquitoes and their extrinsic incubation period

In seasons with high temperatures and rainfall, the vegetation upon which kangaroos depend will flourish, and more non-immune reservoir hosts will be added to the

temporally and spatially expanding population (Weinstein 1997, Harley et al 2001,

Russell 2002)

1.1.3 Spatio-temporal modeling

In disease control programmes, there are several factors involved in the estimation of disease burden, monitoring of disease trend, identification of risk factors, planning and allocation of resources, etc; and a common thread involved in all these activities

is 'Geography' Geographic Information Systems (GIS) and spatio-temporal modelling potentially have great implications in public health research, and have already emerged as innovative and important tools for disease surveillance and assessments

(Cressie 1991, Clarke et al 1996, Khan 1999, Brabyn and Skelly 2002, Hearnden et

al 2003) GIS are particularly well suited for the study of associations between

location, environment and disease due to their spatial analysis and modelling capabilities (Gesler 1986, Khan 1999) GIS are defined as ‘automated systems for the

capture, storage, retrieval, analysis, and display of spatial data’ (Clarke et al 1996)

Spatial modelling takes explicit and formal account of observations with a common spatial nature and leads to better statistical robustness and inferences (Cressie 1991)

In environmental epidemiological research, data are often correlated in space and time, and this correlation structure can be evaluated in its own right and also used to increase the accuracy of modelling and prediction efforts Recently, GIS and spatio-

temporal modelling have been used in studies of risk factors of VBDs (Hightower et

al 1998, Tong et al 2001, Tong and Hu 2001, Tong et al 2002, Tong and Hu 2002), water-borne disease (Clarke et al 1991, Hearnden et al 2003), sexually transmitted disease (Becker et al 1998), environmental health (Reeves et al 1994, Vine et al

1997, Ebi et al 2004), injury control and prevention (Braddock et al 1994) and the

analysis of disease control policy and planning (Gordon and Womersley 1997)

The transmission patterns of some VBDs are sensitive to ecological conditions

(Longley and Batty 1996, Kitron and Kazmierczak 1997, Weinstein 1997, Morrison et

al 1998) For example, mosquitoes can transmit many diseases (eg, malaria, dengue

and RRV) These mosquito-borne diseases usually have strong spatial and temporal patterns, because mosquito density and longevity depend on a number of environmental and ecological factors (eg, temperature, precipitation and mosquito-breeding habitats) It is generally agreed that GIS and spatio-temporal modelling are important tools to utilize These variables can be used in GIS and spatio-temporal modelling to predict the onset and severity of disease epidemics (Gill 1923,

Hightower et al 1998, Moore and Carpenter 1999) These techniques have been

increasingly employed in VBD surveillance and risk management

GIS and spatio-temporal modelling methods offer new and expanding opportunities for VBD research because they can display and model the spatial relationship between

cause and disease (Cressie 1991, Clarke et al 1996, Khan 1999) The applications of

GIS technology superimpose the temporal and spatial distributions of the ecological determinants of endemicity of RRV (eg, landscape ecology, climate, reservoir and vector populations, and human presence and activity) Spatio-temporal modelling can help us understand the distribution of RRV in space and time Improved surveillance

systems for RRV activity, such as the question of timing for control strategies can lead to an integrated management model for public health intervention based on a sound ecological understanding of the disease Endemic areas of RRV would expand

in both time (length of season) and space (geographic area) under environmental conditions (eg., optimal climatic, inadequate urban planning, increased tourists from non-endemic to endemic areas, ecosystem change etc) (Weinstein 1997) Visualisation demonstrates change or variation over space and time, and can illustrate where the transmission of diseases occurs However, caution is needed when interpreting the spatial pattern of RRV disease using GIS because the localities where cases occur sometimes differ from those where transmission occurs

socio-Display of these areas on a GIS-generated map has obvious benefits for the planning

of disease control strategies Therefore, there is a need to facilitate short-term epidemic forecasting and to improve scenario-based predictive modelling for the control and prevention of RRV It is anticipated that the analyses of spatio-temporal relationships between risk factors and disease transmission will improve our understanding of biological/ecological mechanisms of disease outbreaks, and will assist us to develop scientifically-sound, early-warning systems for this disease

1.2 AIMS AND HYPOTHESES

This study aims to examine the potential applications of GIS and spatio-temporal modelling in the surveillance and control of RRV disease

Develop a preliminary spatio-temporal epidemic forecasting model of RRV

1.2.2 Hypothesis

The central hypothesis to be tested is that the transmission of RRV is associated with

a range of socio-ecological factors and this association can be assessed using GIS and spatio-temporal modelling approaches As a result of this study, the applications of GIS and spatio-temporal modelling will assist the surveillance and control of RRV disease

Specific hypotheses

(a) Spatio-temporal distribution of RRV can be assessed using GIS;

(b) The distribution of RRV disease is related to socio-ecological variability, and this relation can be determined by spatio-temporal modelling;

(c) Socio-ecological factors can be used to predict the occurrence of RRV by the combined use of GIS and spatio-temporal models

1.3 SIGNIFICANCE OF THE THESIS

This study assists in quantifying the relationships between socio-ecological factors (climate variables, mosquito density, vegetation and human population) and the epidemic potential of RRV infection in Brisbane, Queensland It contributes to the growing literature on the assessment of potential impacts of socio-environmental change upon the transmission of RRV infection Increased understanding of the relative importance of socio-ecological variables in the transmission cycles of RRV will aid public health planning and policy-making to develop effective strategies to control and prevent this wide-spread disease Epidemic forecasting models were developed and may be directly used for the decision-making process in the surveillance and control of RRV disease Additionally, the methods developed through this study may have a wider application to other public health problems

1.4 CONTENTS AND STRUCTURE OF THE THESIS

This thesis is presented in the publication style As such, it contains five manuscripts, each designed to stand on its own Chapter 2 critically reviews the literature relating

to applications of spatio-temporal model Chapter 3 provides the study design and methods

The five manuscripts are presented in Chapters 4 through 8 (Figure 1.1) Each manuscript was written in the conventional publication style for a particular journal Because each manuscript was designed to stand alone, there was an inevitable degree

of repetitiveness in their introduction, methods and discussion sections

The first manuscript aimed to visualize the spatio-temporal distributions of notified RRV infections in Statistical Local Areas (SLAs) of Brisbane and was submitted to

Arbovirus Research in Australia The second manuscript identified socio-economic

and environmental determinants of RRV disease transmission at an ecological level in

Brisbane and was submitted to Acta Tropica The third manuscript examined the

potential impact of climate variability on the transmission of RRV disease and explored the possibility of developing an epidemic forecasting system for RRV

disease using the multivariate SARIMA technique, which was published in American Journal of Tropical Medicine and Hygiene The fourth manuscript aimed to develop

an epidemic forecasting model using local mosquito density data to predict outbreaks

of RRV disease and was submitted to Ecological Modelling The fifth manuscript

aimed to identify major mosquito species of RRV disease and to explore the threshold

of mosquito density for transmission and is to be submitted to Journal of Medical Entomology

Chapter 9 summarizes the study findings across the five manuscripts, and discusses conclusions in relation to the overall aims of the study This chapter further discusses the study limitations, directions for future research, and public health implications of the research

Tables and figures are provided in the text to facilitate reading The references for each of the manuscripts are presented at the end of their corresponding chapters A complete reference list (including references cited in the manuscripts) is provided at the end of the thesis

Figure 1 1 Flowchart of manuscripts in thesis

Mosquito density

Chapter 8

Exploring the threshold of mosquito density

Manuscript 5

CHAPTER 2: APPLICATIONS OF GIS AND SPATIAL ANALYSIS IN MOSQUITO-BORNE DISEASE

RESEARCH: A REVIEW OF RELATED LITERATURE

Mosquito-borne diseases (MBDs) are prevalent and a significant cause of disease burden in more than 100 countries, infecting 700 million people and causing about 3 million deaths every year (Fradin and Day 2002) MBDs typically have strong spatial and temporal patterns, because mosquito density and longevity depend on a number of environmental and ecological factors (e.g., temperature, precipitation and mosquito-breeding habitats) GIS and spatio-temporal modelling methods offer new and expanding opportunities for MBD research because they can display and model the

spatial relationship between cause and disease (Cressie 1991, Clarke et al 1996, Khan

1999)

2.1 SYSTEMATIC REVIEW

Although there are some excellent reviews of GIS in public health (Clarke et al 1996,

Moore and Carpenter 1999, Cromley 2003, Croner 2003, Ricketts 2003, Rushton 2003), there was still a need to examine systematically the applications of GIS and spatial analysis in MBD research This study aims to evaluate methodologies, strengths and limitations of GIS and spatial analysis tools, and to make recommendations for further applications of GIS and spatial analysis in MBD research

2.1.1 Design

The systematic review was based on empirical studies of MBD (e.g., malaria, dengue, lymphatic filariasis, West Nile virus, Japanese encephalitis, Rift Valley Fever and Ross River Virus diseases) that utilized GIS and spatial analysis These MBD were chosen because of their substantial health impact, causing about millions deaths worldwide every year (The Center for Disease Control and Prevention 2004)

2.1.2 Search methods

A comprehensive literature search was conducted using MedLine which contains bibliographic citations from more than 4,600 biomedical journals MedLine was selected as the main database because it covered over 95% of related articles in a pilot study The key words used in this study included “geograph* information system*” for general health domains and “(geograph* information system* or spa* analysis) and (malaria or dengue or lymphatic or Ross River virus or West Nile or Japanese encephalitis or Yellow fever or Rift valley fever)” for MBD (search methods were

defined by Medline EBSCOhost database) 815 articles (review articles: 10.7%;

empirical articles: 89.3%) that were published between 1986 and 2003 were reviewed for all health domains, as well as 58 empirical articles for MBD including malaria (43), dengue fever (7), lymphatic filariasis (4), West Nile virus (3) and Ross River virus (1) (Figure 2.1)

Figure 2 1 The results of search based on GIS and spatial analysis in Medline

2.1.3 Coding and analysis

A standardised coding system was developed for the study and codes were entered directly into a database All studies were coded on as many dimensions as possible, so that the characteristics of MBD studies could be quantified Categorizing of study design was established on the basis of data collection, GIS methods, spatial analysis methods, study purpose, study scale, exploratory factors and spatio-temporal model (Table 2.1) All 58 empirical articles in MBD were reviewed Cross-checking and double data entry were performed to ensure the quality of data All data processing

Medline

Keywords: Geograph* information system* or spa* analysis Keywords: Geograph* information system*

was performed using the Statistical Package for the Social Sciences (SPSS) program (Statistical Package for the Social Sciences 1997a)

Table 2 1 The coding categories for the literature review

Data collection Field survey

Disease surveillance system Remote Sensing and Global Positioning System

GIS methods Visualisation

Exploratory Modelling

Study scale Country

State City (Town) Suburb

Study purpose Identify disease risk factors

Improve disease prediction

Spatial analysis model Clustering

Dispersion (diffusion) Interpolation techniques

Exploratory factors Climate factors

Social-economic factors Ecological factors

Spatio-temporal model Time factors

Climate, social-economic and ecological factors

2.1.4 Results

A number of interesting trends have emerged from the analysis Figure 2.2 shows the distribution of relevant articles by year There has been a substantial increase in the use of GIS in the health research domain between 1986 and 2003

Table 2.2 shows the distribution of GIS/spatial analysis-related manuscripts by order

of number of papers in health science journals (first 50 journals) Environmental

Journal of Environmental Management, American Journal of Tropical Medicine and Hygiene and Social Science and Medicine were the most common vehicles of GIS- related articles Table 2.3 shows that the American Journal of Tropical Medicine and Hygiene, Southeast Asian Journal of Tropical Medicine and Public Health, Transactions of the Royal Society of Tropical Medicine and Hygiene, American Journal of Epidemiology, Annuals of Tropical Medicine and Parasitology, Bulletin of the World Health Organization, Computer Methods and Programs in Biomedicine, International Journal of Epidemiology and Tropical Medicine and International Health were the most common vehicles for empirical articles relating MBD and GIS

Figure 2.3 shows the percentage of the empirical papers on MBD by year (Figure 2.3a) and by disease (Figure 2.3b) Of all articles coded, 72.0% were related to malaria research, and others were related to dengue fever (12.0%), lymphatic filariasis (9.0%), West Nile (5.0%) and Ross River viruses (2.0%)

Figure 2 2 Trends of publications on GIS for general health domains