Estimating the prevalence rate of a disease is crucial for controlling its spread, and for planning of healthcare services. Due to limited testing budgets and resources, prevalence estimation typically entails pooled, or group, testing where specimens (e.g., blood, urine, tissue swabs) from a number of subjects are combined into a testing pool, which is then tested via a single test. Testing outcomes from multiple pools are analyzed so as to assess the prevalence of the disease. The accuracy of prevalence estimation relies on the testing pool design, i.e., the number of pools to test and the pool sizes (the number of specimens to combine in a pool). Determining an optimal pool design for prevalence estimation can be challenging, as it requires prior information on the current status of the disease, which can be highly unreliable, or simply unavailable, especially for emerging and/or seasonal diseases. We develop and study frameworks for prevalence estimation, under highly unreliable prior information on the disease and limited testing budgets. Embedded into each estimation framework is an optimization model that determines the optimal testing pool design, considering the trade-off between testing cost and estimation accuracy. We establish important structural properties of optimal testing pool designs in various settings, and develop efficient and exact algorithms. Our numerous case studies, ranging from prevalence estimation of the human immunodeficiency virus (HIV) in various parts of Africa, to prevalence estimation of diseases in plants and insects, including the Tomato Spotted Wilt virus in thrips and West Nile virus in mosquitoes, indicate that the proposed estimation methods substantially outperform current approaches developed in the literature, and produce robust testing pool designs that can hedge against the uncertainty in model inputs.Our research findings indicate that the proposed prevalence estimation frameworks are capable of producing accurate prevalence estimates, and are highly desirable, especially for emerging and/or seasonal diseases under limited testing budgets. / Doctor of Philosophy / Accurately estimating the proportion of a population that has a disease, i.e., the disease prevalence rate, is crucial for controlling its spread, and for planning of healthcare services, such as disease prevention, screening, and treatment. Due to limited testing budgets and resources, prevalence estimation typically entails pooled, or group, testing where biological specimens (e.g., blood, urine, tissue swabs) from a number of subjects are combined into a testing pool, which is then tested via a single test. Testing results from the testing pools are analyzed so as to assess the prevalence of the disease. The accuracy of prevalence estimation relies on the testing pool design, i.e., the number of pools to test and the pool sizes (the number of specimens to combine in a pool). Determining an optimal pool design for prevalence estimation, e.g., the pool design that minimizes the estimation error, can be challenging, as it requires information on the current status of the disease prior to testing, which can be highly unreliable, or simply unavailable, especially for emerging and/or seasonal diseases. Examples of such diseases include, but are not limited to, Zika virus, West Nile virus, and Lyme disease. We develop and study frameworks for prevalence estimation, under highly unreliable prior information on the disease and limited testing budgets. Embedded into each estimation framework is an optimization model that determines the optimal testing pool design, considering the trade-off between testing cost and estimation accuracy. We establish important structural properties of optimal testing pool designs in various settings, and develop efficient and exact optimization algorithms. Our numerous case studies, ranging from prevalence estimation of the human immunodeficiency virus (HIV) in various parts of Africa, to prevalence estimation of diseases in plants and insects, including the Tomato Spotted Wilt virus in thrips and West Nile virus in mosquitoes, indicate that the proposed estimation methods substantially outperform current approaches developed in the literature, and produce robust testing pool designs that can hedge against the uncertainty in model input parameters. Our research findings indicate that the proposed prevalence estimation frameworks are capable of producing accurate prevalence estimates, and are highly desirable, especially for emerging and/or seasonal diseases under limited testing budgets.

Made available in DSpace on 2011-05-04T12:36:24Z (GMT). No. of bitstreams: 0 Previous issue date: 2009 / Diversos fatores podem dificultar a caracterização acurada do perfil de umapopulação por amostragem. Se a característica que define a população é de difícil observação seja porque exige testes caros para detecção ou porque é uma característica de comportamento ilegal ou estigmatizado que dificulta a identificação, torna-se praticamente impossível aplicar os métodos clássicos de amostragem, pois não se pode definir uma base de amostragem (sampling frame). Populações desse tipo são conhecidas como populações ocultas, ou escondidas, e alguns exemplos comumente estudados são homens que fazem sexo com homens, trabalhadores do sexo e usuários de drogas. Essa dissertação discute a técnica de amostragem conhecida como Respondent-Driven Sampling (RDS), originalmente proposta por Heckathorn (1997), e que vem sendo amplamente utilizada na estimação de prevalências de doenças transmissíveis em populações ocultas. Esse método pertence à família de amostragens por bola-de-neve, na qual os elementos seguintes da amostra são recrutados a partir da rede de conhecidos dos elementos já presentes na amostra, formando as cadeias de referência. Com este método, além das informações individuais, é possível estudar também as relações entre os indivíduos. O recrutamento por bola de neve não gera uma amostra aleatória, e está sujeito às propriedades das redes sociais das populações em estudo, que deve mudar de lugar para lugar e potencialmente influenciar as medidas de prevalência geradas. As redes sociais são estruturas complexas, e compreender como que a amostragem RDS é influenciada por estas estruturas é um dos objetivos dessa dissertação. Além disso, se o interesse de um estudo epidemiológico é estimar a prevalência de uma doença transmissível, há de se considerar que muitas vezes a própria rede social pode estar correlacionada com as redes de transmissão, gerando potenciais dependências entre o processo de amostragem e a distribuição da variável desfecho. Essa dissertação teve por objetivo avaliar estimativas de prevalência geradas a partir de amostras obtidas com a utilização da metodologia RDS, considerando estruturas populacionais complexas, ou seja, populações com estruturas distintas de ligação entre os indivíduos e de disseminação de doenças. Para isso, foram realizados experimentos de simulação combinando quatro modelos geradores de redes sociais e quatro modelos de distribuição de casos infectados na população. Para cada uma, foram obtidas amostras utilizando RDS e as respectivas prevalências foram estimadas.Com os resultados encontrados, foi possível realizar uma avaliação tanto do RDS como forma de recrutamento, como o modelo proposto por Heckathorn (2002) para a ponderação e estimação de prevalências. Basicamente, três aspectos foram considerados nessa avaliação: 1. o tempo necessário para concluir a amostragem, 2. a precisão das estimativas obtidas, independente da ponderação, e 3. o método deponderação. De forma geral, o método apresentou bons resultados sob esses três aspectos, refletindo a possibilidade de sua utilização, ainda que exigindo cautela. Os achados apresentam-se limitados, pois são escassos os trabalhos que abordem essa metodologia e que permitam estabelecer comparações. Espera-se, no entanto,despertar o interesse para que outros trabalhos nessa linha sejam desenvolvidos. / Several factors may hamper the accurate characterization of a population. If the defining feature of the population is difficult to apply - either because it requires expensive tests for detection or because it is a stigmatized or illegal behavior that hinders the identification, it is virtually impossible to apply traditional methods for sampling, because sampling frame cannot be define. The latter are called “hidden populations”, and some examples are men who have sex with men, sexual workers and drug users. This dissertation focus on Respondent-Driven Sampling (RDS), a sampling method originally proposed by Heckathorn (1997), which has been widely used to estimate the prevalence of infectious diseases in hidden populations. RDS is a snowball sampling method, in which new elements for the sample are recruited from the network of the elements already present in the sample, forming reference chains. With this method, besides individual informations, it is also possible to study the relationships between individuals. Snowball sampling does not generate random samples, and its properties are likely to depend on the properties of the social networks underlying the recruitment process, which may change from place to place and potentially influence the measures of prevalence generated. Social networks are complex structures, and understanding how the different implementations of RDS sampling is influenced by these structures is one of the objectives of this dissertation. Moreover, if the interest of an epidemiological study is to estimate the prevalence of a disease, it is should be considered that very often, social network may be correlated with the transmission networks, generating potential dependencies between the process of sampling and distribution of outcome variable. The aim of this dissertation was to assess the behavior of prevalence estimators using RDS data in scenarios of populations organized in complex structures, i.e. Combinations of social networks structures and spreading patterns. To achieve that, theoretical experiments were performed using simulation models combining four generators of social networks and four models of distribution of infected cases in the population. For each one, samples were obtained using RDS and prevalence, estimated. Findings were used to evaluate RDS as a recruiting process itself, as well as Heckathorn’s (2002) model to estimate prevalences. Three aspects were considered in such analyses: 1. the time elapsed before obtaining the sample; 2. the accuracy of the estimates without taking in consideration the weighting strategies; and 3. the weighting strategy. Overall, RDS performed well in these three areas, showing it is a valid method to assess hidden populations, despite the fact its use should be made with the necessary caution. The interpretation of our findings was constrained by the scarcity of studies using the same methodology, what compromised the comparability of our findings. We hope, however, that our findings may foster the development of additional studies in this field.