Add variables/features to already trained tensorflow model when new variables come in. Possible?

Let’s assume I’ve trained a deep-learning tensorflow model with some data with n variables. At a later time, new data come in and there is an additional variable alongside the previous variables with a total of n+1 variables. If I load the previously trained model that is used to seeing n variables, is there a way to continue training and include the new variable in the model without completely retraining it using all of the data?

One way that I thought of, is to train the model on the whole dataset (assuming we know all the possible variables that may come in at a later time), with the missing values for the unseen variables/features being replaced by 0. However, this adds a lot of sparsity to the dataset, but it works.