Advances in parasitology global mapping of infectious diseases - part 1 docx

GRAHAM, TALA Research Group, Tinbergen Building, Depart-ment of Zoology, University of Oxford, South Parks Road, Oxford OX1 3PS, UK C.. GUERRA, TALA Research Group, Tinbergen Building, D

Series Editors

J R Baker, Royal Society of Tropical Medicine and Hygiene, London, UK

R Muller, London School of Hygiene and Tropical Medicine, London, UK

D Rollinson, The Natural History Museum, London, UK

Editorial Board

M Coluzzi, Director, Istituto de Parassitologia, Universita` Degli Studi di Roma ‘La Sapienza’, P le A Moro 5, 00185 Roma, Italy

C Combes, Laboratoire de Biologie Animale, Universite´ de Perpignan, Centre de Biologie et d’Ecologie Tropicale et Me´diterrane´enne, Avenue de Villeneuve, 66860 Perpignan Cedex, France

D.D Despommier, Division of Tropical Medicine and Environmental Sciences, Department of Microbiology, Columbia University, 630 West 168th Street, New York, NY 10032, USA

J.J Shaw, Instituto de Cieˆncias Biome´dicas, Universidade de Sa ˜ o Paulo,

av Prof Lineu Prestes 1374, 05508-900, Cidade Universita´ria, Sa˜o Paulo,

SP, Brazil

K Tanabe, Laboratory of Biology, Osaka Institute of Technology, 5-16-1 Ohmiya Asahi-Ku, Osaka 535, Japan

CONTRIBUTORS TO VOLUME 62

P M ATKINSON, School of Geography, University of Southampton, Highfield, Southampton SO17 1BJ, UK

D L BALK, CIESIN, Columbia University, Palisades, NY 10964, USA

School of Hygiene & Tropical Medicine, Keppel Street, London WC1E 7HT, UK

1818 H Street NW, Washington DC 20433, USA

A C A CLEMENTS, Department of Infectious and Tropical Diseases, London School of Hygiene & Tropical Medicine, Keppel Street, London WC1E 7HT, UK

1818 H Street NW, Washington, DC 20433, USA

02543-0296, USA

A J GRAHAM, TALA Research Group, Tinbergen Building, Depart-ment of Zoology, University of Oxford, South Parks Road, Oxford OX1 3PS, UK

C A GUERRA, TALA Research Group, Tinbergen Building, Depart-ment of Zoology, University of Oxford, South Parks Road, Oxford OX1 3PS, UK

ADVANCES IN PARASITOLOGY VOL 62

ISSN: 0065-308X $35.00

v

Series Editors’ Preface

Nearly six years ago, a special volume of Advances in Parasitology (vol 47) dealt with the uses of remote sensing and geographical in-formation systems in the study of disease epidemiology In a sense, this volume is a follow-on to that publication, dealing as it does with some practical applications of those techniques to the study of par-asitic and infectious diseases

We are once again fortunate in having Simon Hay, David Rogers and—a newcomer this time—Alastair Graham, of the University of Oxford in the United Kingdom, as guest editors They have assem-bled a formidable array of talented research workers from the UK and the USA as contributors to what we are sure will be a valuable source of both technical and epidemiological data in this rapidly growing field

We are sincerely grateful to the guest editors, authors and all those who have contributed to the production of this volume

John Baker Ralph Muller David Rollinson

ADVANCES IN PARASITOLOGY VOL 62

ISSN: 0065-308X $35.00

vii

Guest Editors’ Preface

It has been five years since an earlier special issue of Advances in Parasitology, Volume 47, outlined the advances that remote sensing (RS) and geographical information systems (GIS) could bring to epi-demiology During this interval a vast amount of work has been undertaken in this area and these RS data and GIS tools have moved from the novel to part of the mainstream of spatial epidemiology Data availability has continued to limit the engagement of many potential users, however This has been most obvious in continental and global scale public-health applications, and predictably these limitations have been particularly acute in regions with low band-width internet connections, often where the public health need is greatest The primary reason for compiling this new volume was to enable a wider range of epidemiologists to have access to the global environmental data (satellite and demographic), which we have been collectively working with for over a decade The second reason for devising this special issue was to demonstrate that RS and GIS do not simply create pretty maps, but biologically informative information and ultimately pragmatic control tools That being said, we also hope you like the front cover!

This special issue of Advances in Parasitology, Volume 62, ‘‘Global mapping of infectious diseases: methods, examples and emerging ap-plications’’ comprises 10 reviews and a DVD of global environmental and population data There are four introductory reviews: one on the various methods used to predict disease distributions (Rogers, this volume, pp 1–35); another on the global environmental datasets that can be used for disease mapping (Hey et al., this volume, pp 37–77);

a further one exploring the concepts of spatial resolution, accuracy and uncertainty measures in disease mapping based on remote

sens-ADVANCES IN PARASITOLOGY VOL 62

ISSN: 0065-308X $35.00

ix

C ONTRIBUTORS TO V OLUME 62 v

S ERIES E DITORS ’ P REFACE vii

G UEST EDITIORS ’ P REFACE ix

Models for Vectors and Vector-Borne Diseases D.J Rogers Abstract 1

1 A Brief History of Distribution Modelling 2

2 Families of Distribution Models 4

3 Predictor Variable Selection in Distribution Models 12

4 What to Do With Sparse Datasets? 14

5 Incorporating Spatial Information Into Models 19

6 Model Selection and Multi-Model Inference 21

7 Conclusion 30

Acknowledgements 33

References 33

Global Environmental Data for Mapping Infectious Disease Distribution S.I Hay, A.J Tatem, A.J Graham, S.J Goetz and D.J Rogers Abstract 38

1 Introduction 38

2 The AVHRR Sensor 39

3 Temporal Fourier Analysis (TFA) 49

4 Future Global Environmental Data 57

5 Conclusions 70

Acknowledgements 70

References 71

ADVANCES IN PARASITOLOGY VOL 62

ISSN: 0065-308X $35.00

xiii

Models for Vectors and Vector-Borne

Diseases

D.J Rogers

TALA Research Group, Tinbergen Building, Department of Zoology, University of Oxford, South Parks Road, Oxford 0X1 3PS, UK

Abstract 1

1 A Brief History of Distribution Modelling 2

2 Families of Distribution Models 4

2.1 Logistic Models: The Theory 4

2.2 Discriminant Analysis Models: The Theory 10

3 Predictor Variable Selection in Distribution Models 12

4 What to Do With Sparse Datasets? 14

4.1 Bootstrap Sampling 15

4.2 Environmental Envelope Expansion 17

5 Incorporating Spatial Information Into Models 19

6 Model Selection and Multi-Model Inference 21

6.1 Application to Vector and Disease Mapping 26

7 Conclusion 30

Acknowledgements 33

References 33

ABSTRACT

The development of models for species’ distributions is briefly re-viewed, concentrating on logistic regression and discriminant analyt-ical methods Improvements in each type of modelling approach have led to increasingly accurate model predictions This review addresses several key issues that now confront those wishing to choose the

‘‘right’’ sort of model for their own application One major issue is the number of predictor variables to retain in the final model

ADVANCES IN PARASITOLOGY VOL 62

ISSN: 0065-308X $35.00

Copyright r 2006 Elsevier Ltd.

All rights of reproduction in any form reserved

Logistic function, positive slope

a

0.00 0.20 0.40 0.60 0.80 1.00 1.20

Predictor variable, x2

Positive arg.

0.00 0.20 0.40 0.60 0.80 1.00 1.20

Predictor variable, x1

Negative arg.

Logistic functions

0 0.2 0.4 0.6 0.8 1 1.2

Predictor variables, x1, x2

Negative arg.

Positive arg.

c

Figure 1 Examples of logistic curves where the dependent variable (y in text Eq (3)) (a) increases and (b) decreases with the independent variable, x There can be only one ‘‘on’’–‘‘off ’’ or ‘‘off ’’–‘‘on’’ transition with any single variable when y is linearly related to x, so that two independent variables are required to define an ‘‘off ’’–‘‘on’’–‘‘off ’’ response (e.g c).

D J ROGERS 6

Table 1 Various accuracy metrics applicable to distribution modelling

Accuracy

metric

Range of

values

accuracy, all categories combined

Simple and easy to calculate

Presence and absence sites given equal weight Metric usually affected by prevalence False positives

(%)

0–100% % of total training

set sample wrongly predicted

as ‘presence’

Simple and easy to calculate

Should be considered with its

complement— false negatives False negatives

(%)

0–100% % of total training

set sample wrongly predicted

as ‘absence’

Simple and easy to calculate

Should be considered with its

complement— false positives

positives correctly

Derived from diagnostics.

Useful measure of positive test accuracy

Concentrates on positives only Should be considered with its complement— specificity

negatives correctly

Derived from diagnostics.

Useful measure of negative test accuracy

Concentrates on negatives only Should be considered with its complement— sensitivity Producer’s

accuracy

0–100% Ability to predict the

training set data correctly

A guide to the modeller to identify where current models are wrong

Not particularly useful to users

Consumer’s

accuracy

0–100% Accuracy of model

predictions

A guide to the user

to indicate the probability with which each model prediction is correct

An important metric for operational use, but not particularly useful

to the modeller in identifying model errors

for positive and negative samples combined

Adjusts for chance model agreement with training set data (for which

k ¼ 0) Applicable

to multiple categories of presence/absence

or abundance

Sensitive to overall prevalence at high and low prevalence levels

under the curve of

Effectively combines sensitivity and

Rather more time consuming to

D J ROGERS 8