Mathematical modeling and forecasting of COVID-19: experience in Santiago de Cuba province

Abstract

In the province of Santiago de Cuba, Cuba, the COVID-19 epidemic has a limited progression that shows an early small-number peak of infections. Most published mathematical models fit data with high numbers of confirmed cases. In contrast, small numbers of cases make it difficult to predict the course of the epidemic. We present two known models adapted to capture the noisy dynamics of COVID-19 in the Santiago de Cuba province. Parameters of both models were estimated using the approximate-Bayesian-computation framework with dedicated error laws. One parameter of each model was updated on key dates of travel restrictions. Both models approximately predicted the infection peak and the end of the COVID-19 epidemic in Santiago de Cuba. The first model predicted 57 reported cases and 16 unreported cases. Additionally, it estimated six initially exposed persons. The second model forecasted 51 confirmed cases at the end of the epidemic. In conclusion, an opportune epidemiological investigation, along with the low number of initially exposed individuals, might partly explain the favorable evolution of the COVID-19 epidemic in Santiago de Cuba. With the available data, the simplest model predicted the epidemic evolution with greater precision, and the more complex model helped to explain the epidemic phenomenology.

A new model of unreported COVID-19 cases outperforms three known epidemic-growth models in describing data from Cuba and Spain

“…The copyright holder for this preprint this version posted July 3, 2021. ; https://doi.org/10.1101/2021.06.29.21259707 doi: medRxiv preprint [8]. The Gompertz+U model estimates it at 60.3758 (95%CI : 51.5345 − 70.2727) (the estimate with all the daily credible intervals of the Gompert+U model with the data of the first wave in Santiago de Cuba can be visualized in Figure 1), the GRM+U model estimates it at 58.6924 (95%CI : 50.4297 − 63.5315) and the SIR+U model estimates it at 60.3758 (95%CI : 51.5345 − 70.2727).…”

“…The proposed model is able to provide reasonable estimation of the number of unreported infected cases. The total number of accumulated cases on the 50th day of the first wave in Santiago de Cuba was previously estimated at 73 [8]. The Gompertz+U model estimates it at 60.3758 (95%CI : 51.5345 − 70.2727) (the estimate with all the daily credible intervals of the Gompert+U model with the data of the first wave in Santiago de Cuba can be visualized in Figure 1), the GRM+U model estimates it at 58.6924 (95%CI : 50.4297 − 63.5315) and the SIR+U model estimates it at 60.3758 (95%CI : 51.5345 − 70.2727).…”

“…The Gompertz+U model estimates it at 60.3758 (95%CI : 51.5345 − 70.2727) (the estimate with all the daily credible intervals of the Gompert+U model with the data of the first wave in Santiago de Cuba can be visualized in Figure 1), the GRM+U model estimates it at 58.6924 (95%CI : 50.4297 − 63.5315) and the SIR+U model estimates it at 60.3758 (95%CI : 51.5345 − 70.2727). The approach of the present work has the advantage that it was not necessary to divide the wave into three parts for its analysis as in [8].…”

“…Also, we set 49 the lower limit of the improper uniform prior of K for the Gompertz+U and GRM+U models. Based on the expertise of epidemiologists and on estimates made in previous work [8], we set relatively small upper limits for the priors of C and U (20 and 19, respectively). This upper limit made it possible to reduce the search region of the posterior distributions of parameters and to reduce the risk of practical non-identifiability [29,30].…”

“…This was followed by a huge explosion of cases that has been widely studied and reported [2,6,7]. In the province of Santiago de Cuba, Cuba, there was a first wave of cases that had certain similarities with Wuhan, but with a much lower transmissibility [8]. The first infected case was reported on March 27, 2020, and the second case in the next six days.…”