OBJECTIVES: To evaluate geostatistical description of influenza data from the Valencian Sentinel Network (VSN) in Spain using the kringing method and to assess the possibility of incorporating this method into routine surveillance. METHODS: We use influenza surveillance data on three influenza seasons (1997-1998, 1998-1999 and 1999-2000) from the VSN to construct a geocodified data matrix of the incidence of this disease. The geographic distribution was studied using the kringing method, which enables estimation of the incidence in a few strategically distributed points. Influenza isoincidence maps for each week were plotted. Cross validation was used to evaluate the method. RESULTS: In most of the weeks, the values of reduced standard deviation and reduced mean were close to the optimal values (0 and 1, respectively). Out of range reduced standard deviation values were found in 12 of 20 weeks in the last season only. The estimation of rates in all three seasons showed coherence in spatial distribution and temporal evolution. CONCLUSIONS: In most situations the results were acceptable. The method does not require extra computer resources or an excessive amount of time and requires only annual adaptation. Because it is easy to use, the technique is appropriate for routine use but the accuracy of estimations could be improved by increasing the complexity of the model.
OBJECTIVES: To evaluate geostatistical description of influenza data from the Valencian Sentinel Network (VSN) in Spain using the kringing method and to assess the possibility of incorporating this method into routine surveillance. METHODS: We use influenza surveillance data on three influenza seasons (1997-1998, 1998-1999 and 1999-2000) from the VSN to construct a geocodified data matrix of the incidence of this disease. The geographic distribution was studied using the kringing method, which enables estimation of the incidence in a few strategically distributed points. Influenza isoincidence maps for each week were plotted. Cross validation was used to evaluate the method. RESULTS: In most of the weeks, the values of reduced standard deviation and reduced mean were close to the optimal values (0 and 1, respectively). Out of range reduced standard deviation values were found in 12 of 20 weeks in the last season only. The estimation of rates in all three seasons showed coherence in spatial distribution and temporal evolution. CONCLUSIONS: In most situations the results were acceptable. The method does not require extra computer resources or an excessive amount of time and requires only annual adaptation. Because it is easy to use, the technique is appropriate for routine use but the accuracy of estimations could be improved by increasing the complexity of the model.
Authors: D Gomez-Barroso; M A Martinez-Beneito; V Flores; R Amorós; C Delgado; P Botella; O Zurriaga; A Larrauri Journal: Epidemiol Infect Date: 2014-01-27 Impact factor: 4.434