Predicting diameter at breast height (DBH) of trees from stump information may be necessary to reconstruct silvicultural practices, to assess harvested timber and wood, or to estimate forest products’ losses caused by illegal cuttings or natural disasters (disturbances). A model to predict DBH of felled trees was developed by the first Italian National Forest Inventory in 1985 (IFNI85). The model distinguished between the two broad groups of conifers and broadleaves and used stump diameter as the sole quantitative variable. Using an original dataset containing data from about 1200 trees of sixteen species recorded throughout Italy, we assessed the performance of that model. To improve the prediction of the DBH of removed trees over large areas and for multiple species, we developed new models using the same dataset. Performance of the new models was tested through indices computed on cross-validated data obtained through the leave-one-out method. A new model that performs better than the old one was finally selected. Compared to the old NFI model, the selected model improved DBH prediction for fourteen species up to 31.28%. This study proved that species specification and stump height are variables needed to improve the models’ performance and suggested that data collection should be continued to get enhanced models, accurate for different ecological and stand conditions.