Questions to the article: Predicting Unscheduled Emergency Department Return Visits Among Older Adults (Preprint)
UNSTRUCTURED The article published on 28 July 2021 is well-written and of interest, but remains several questions that are required for clarifications, such as (1) the Figure 1 is too complex to release the decision criteria for predicting unscheduled emergency department return visits (EDRVs); (2) the Table 1 with 11 rules is not succinct for readers to capture the core features of the influencing factors on the unscheduled EDRVs, and (3) the decision tree technique using Weka software did not demonstrate an online module that can be implemented in clinical settings. We suggested three ways to improve the study in methods and illustrated examples presented in previous studies using the decision tree technique. In addition, to solve the problem of class imbalance in data should be combined with an MP4 video(or a Multimedia Appendix) to make readers easily replicate similar research in the future. The patient characteristics and variables deposited in Multimedia Appendix 1 are insufficient. A small sample of data (e.g., one-tenth from the 10-fold cross-validation method to randomly partition the data set into ten subsets in the study) should be provided for readers to verify the decision tree that can yield appropriately 76.65% and 76.95% in sensitivity and specificity, respectively, as did in predicting the unscheduled EDRVs. Otherwise, the study results are doubtable.