A framework for developing a hybrid regression model for short-term forecasting of COVID-19 cases has been proposed. It was demonstrated using three different models with varying predictor parameters.  Results show that the performance of forecasting models during training may not necessarily be the same when validation data is used. In addition, the performance of the model should be tested regularly to ensure whether its performance is still acceptable. In instances that the performance deteriorates, especially since forecast models tend to perform better for short-term prediction rather than long-term prediction, then decision-makers should consider modifying and recalibrating the model. By providing insights on the trend and expected number of COVID-19 cases a few days ahead, short-term forecasting models will be useful for planning resource utilization and allocation not just during this pandemic but other future crisis. Future work should consider other combinations of predictors and investigate the use of other indicators for testing model performance. In addition, these models can be calibrated using more localized regional data to provide better insights and guidance on the COVID-19 situation in a region.