The COVID-19 pandemic has challenged authorities at different levels of government administration around the globe. When faced with diseases of this severity, it is useful for the authorities to have prediction tools to estimate in advance the impact on the health system as well as the human, material, and economic resources that will be necessary. In this paper, we construct an extended Susceptible-Exposed-Infected-Recovered model that incorporates the social structure of Mar del Plata, the 4°most inhabited city in Argentina and head of the Municipality of General Pueyrredón. Moreover, we consider detailed partitions of infected individuals according to the illness severity, as well as data of local health resources, to bring predictions closer to the local reality. Tuning the corresponding epidemic parameters for COVID-19, we study an alternating quarantine strategy: a part of the population can circulate without restrictions at any time, while the rest is equally divided into two groups and goes on successive periods of normal activity and lockdown, each one with a duration of days. We also implement a random testing strategy with a threshold over the population. We found that is a good choice for the quarantine strategy since it reduces the infected population and, conveniently, it suits a weekly schedule. Focusing on the health system, projecting from the situation as of September 30, we foresee a difficulty to avoid saturation of the available ICU, given the extremely low levels of mobility that would be required. In the worst case, our model estimates that four thousand deaths would occur, of which 30% could be avoided with proper medical attention. Nonetheless, we found that aggressive testing would allow an increase in the percentage of people that can circulate without restrictions, and the medical facilities to deal with the additional critical patients would be relatively low.

Disorder is a characteristic of real social networks generated by heterogeneity in person-to-person interactions. This property affects the way a disease spreads through a population, reaches a tipping point in the fraction of infected individuals, and becomes an epidemic. Disorder is usually associated with contact times between individuals, and normalized contact time values ω are taken from the distribution P (ω) = 1/(aω) that mimics "face-to-face" experiments [1,2]. To model more realistic systems, we study how heterogeneity in person-to-person interactions affects the spreading of diseases when two different kinds of disorder are present, each with a particular value of a. This allows two different types of interaction to emerge, such as close (family, coworkers) and distant (neighbors, strangers) contacts. We also develop a strategy for controlling distant contact times, which are easier to alter in practice, so as to reduce the total number of infected individuals.Finally, we use "face-to-face" experiments to generate a more accurate distribution of normalized contact times, and we repeat the analysis for this distribution.