Predicting the likelihood of falls among the elderly using likelihood basis pursuit technique.

This study reports on the application of the knowledge discovery in database process to generate models that can predict the likelihood of falls among the elderly who reside in long-term care facilities. This process was applied to data held in the Minimum Data Set, a comprehensive resident assessment instrument being used in all Medicare and Medicaid supported nursing homes in the United States. For this study, we incorporated a new data mining technique, Likelihood Basis Pursuit, into the process. Using this technique, we were able to correctly identify which of the variables in this data set were associated with falls and generate models that could make fall likelihood predictions based upon those variables. Because the model provides probabilities based upon the exact combination of variables present in a particular resident, models constructed using this new data mining technique have the potential to be more useful for assessing fall risk.