The COVID-19 outbreak is spreading all over the world rapidly with a large number of deaths. The spreading nature of the COVID-19 is different from country to country. Therefore, it is required to predict the epidemic trend country-wise to determine the hospital capacity needed, quarantine measures needed, the number of required vaccines, etc. This study implemented an extended SIR model with an additional compartment for deaths to predict the COVID-19 incidences in the USA and South Korea. The model is helpful in estimating the primary reproduction number R_0, which estimates the average number of people an infected person infects. According to our estimates, the reproduction number ranges from 0-3 for different periods for an increasing trend of infections. Larger the R_0 the higher the probability of the pandemic, and it is harder to control the epidemic. We also estimated the number of vaccinations required to end the pandemic for the two countries which we used in our model. Further, the model accuracy was confirmed using the mean absolute percentage error. The fitted extended SIR models are recommended for the short-term predictions of the COVID-19 incidences.