Utilize este identificador para referenciar este registo: http://hdl.handle.net/10400.18/1255
Título: Nowcasting influenza epidemics using a sentinel network based influenza surveillance system
Autor: Nunes, Baltazar
Natário, Isabel
Carvalho, Maria Lucília
Palavras-chave: Estados de Saúde e de Doença
Hidden Markov Models
Nowcasting
Surveillance
Influenza
Data: 8-Set-2012
Resumo: Background: Timeliness of a public health surveillance system (SS) is one of its most important characteristics. In Europe the epidemiological surveillance of influenza is supported by general practitioners (GP) sentinel networks and, weekly, epidemiological bulletins are issued between Wednesday and Friday, reporting the previous week observed influenza values, representing a 2 to 4 days reporting delay. Some SS use web interfaces or computer routines that can provide up to date daily data streams accessing the current situation. The process of predicting the present week situation using the available incomplete information from the SS has received the term nowcasting and has high public health interest. Objective: Develop a statistical model to nowcast the influenza epidemic evolution in a weekly basis, by predicting two measures of interest: the current week influenza-like illness (ILI) rate (officially issued in the following week) and the probability that the influenza activity is epidemic. Methods: A two states (epidemic/non-epidemic) non-homogeneous hidden Markov model (HMM) is used, where the current week ILI rate is a function of an early observation of the ILI rate obtained by Friday of this week. The state-transition probabilities are modeled by a logistic function of the Friday ILI rate and of the number of ILI cases tested positive in the previous week. For comparison purposes a homogeneous HMM is also applied to the data. Bayesian inference is used to find estimates of the model parameters and of the nowcasted quantities. The models are applied to data provided by the Portuguese influenza SS for seasons 2008-09 up to 2010-11 to nowcast in real-time each week of the season 2010-11. Results: The non-homogenous HMM presents the best fit and is able to indentify four epidemic waves in the studied period, one in 2008-09 and 2010-11 and two in the pandemic season. Regarding the nowcast, the weekly ILI rates of season 2010-11 are predicted during the same week in a very satisfactory way, given that the estimates start to increase, reach the peak and decrease in synchrony with the observed rate (this reduces the reporting delay in 5 days). The non-homogenous model is able to detect the epidemic start two weeks before the homogeneous one. Conclusion: The present work shows the additional value of the non-homogeneous HMM to nowcast the ILI rate and the influenza activity state.
Peer review: no
URI: http://hdl.handle.net/10400.18/1255
Aparece nas colecções:DEP - Apresentações orais em encontros internacionais

Ficheiros deste registo:
Ficheiro Descrição TamanhoFormato 
nowcast_euroepi2012.pdf145,12 kBAdobe PDFVer/Abrir


FacebookTwitterDeliciousLinkedInDiggGoogle BookmarksMySpace
Formato BibTex MendeleyEndnote Degois 

Todos os registos no repositório estão protegidos por leis de copyright, com todos os direitos reservados.