Approcci basati sulle stime della dimensione della rete sociale nella sorveglianza epidemiologica di focolai

In the field of public health, determining the dimension of population it’s an important issue, especially of those who are hidden or hard to reach. Prevalence estimates are required in three key areas, which are resource allocation, monitoring of the target population and public health surveillance and epidemiology. Different techniques are available in literature to estimate these populations, such as enumeration method, capture recapture technique, multiplier method and the network scale up (NSUM). This last method has been developed to count the death after an earthquake in Mexico and was then widely applied in the field of hard to reach population. This thesis is developed in three stages. In the first part is presented a systematic review to retrieve studies that have used, both to implement or to only apply the NSU estimator. Considering some of the limitations relate to the bias that affects the estimator and the need to reduce the questions to pose, in the second part is proposed a new modified version of the Bayesian formulation of Maltiel. This modified version considers the network size as partially unknow since there is only one question to pose to estimate it. This modified version is applied to the dataset provided in the Maltiel study showing constant low level of error and a precision a little bit reduced compared to the original formulation. In the third part of the thesis are reported the full results of the study. The modified version is applied to define the size of the number of undocumented cases of COVID-19 during the pandemic. To evaluate the performance of the model in the same survey were asked also the number of COVID-19 cases positive. The number of transfers after lockdown restriction and the number of cohabitant of COVID-19 case were also considered as hidden populations. The new model, despite its loss in precision, has demonstrated to be sufficiently precise considering the increase in the easiness of use by respondents.