2020_04_23_ICS_benitez

Next Monday, April 27th, the Spanish Government will carry out a seroprevalence study with the goal to know the extent of Covid-19 in the country, designed by the high school de Salud Carlos III and the high school National Statistics Institute (INE). A total of 30,000 families throughout the country will be tested. "The work will provide clues to patterns of dispersion and speed of contagion, useful for pandemic management plans," said Edgar Benitez, an expert in methodology and statistics from the University of Navarra, research .

Dr. Benítez is a professor at the Master's Degree University in Big Data Science of the University of Navarra, coordinated this year by the Institute for Culture and Society (ICS) and, starting next year, by the Institute of Science of the data and Artificial Intelligence, attached to TECNUN School of Engineering.

The researcher reminds that this study will be repeated in two or three waves "because the problem is not static; it is necessary to evaluate the growth of the potentially immunized population".

Regarding the selection of more than 60,000 people, some 500-600 per province, the researcher assures that "the aim is to have a representative sample , so access to any of the households must be available. The INE has instructions of data that allows for easy identification".

He also points out two strategies that will facilitate the work. The first is multistage sampling: "Each stage makes it easier to reach each household with the least possible bias, that is, so that no groups are left without being evaluated". Another factor is stratification, "which makes it possible to take a minimum sample and one proportional to the population, so that those strata with a larger issue of people have a proportional size in the sample and are thus fairly represented".

Advantages of sampling versus census

One of the advantages of sampling, if it is carried out at agreement with the criteria mentioned at design of sample -he adds-, "is that it allows to have an instantaneous assessment of the status, a 'photograph'". He points out that a census, which consists of working on 100% of the population, "is such a complex process that it easily leads to errors that multiply with each new piece of data entered", which can result in "the result being worse in the end than if only a representative sample had been evaluated", in addition to the problems associated with logistics and the cost that the magnitude of a census would imply. 

With regard to the heterogeneity of the spread of the virus throughout Spain - not all provinces are affected in the same way - Edgar Benítez stresses that it influences two relevant elements of the size of the sample, variation and precision. 

"When assessing the presence or absence of infection, low values of disease prevalence are associated with high variability and the need to improve the level of precision, which results in the need to increase the size of sample". However, he indicates that in this strategy "the aim is to have minimum values of sample that can correct this bias" (about 500-600 per province).

Dr. Benitez alludes to the importance of this subject of programs of study."Everything that is measured is controlled and, therefore, improved. This is a phrase that is often repeated in project management. Otherwise, the tactics and strategies proposed would be randomly defined or, even worse, biased by interests or prejudices". 

"This subject project has a high level of complexity -he concludes-. What is presented to the public are only the final results of the work of a team of experts of great reputation in the area of sampling techniques and statistics, which ensures the quality of the methodology used and the results that can be obtained".