Predicting Emergency Department Visits Based on Cancer Patient Types

Abstract

K MDI RE P OR T S Pr edi c ng Emer g enc y Depar t ment Vi s i t s Bas ed on Canc er Pa ent T ypes Aut hor s : Ma hs a Rouz ba hma n, Ma r k Chi g nel l &L i s a Ba r ber a AbstractThis study evaluated the predictive ability of patient types (clusters of similar patients) in identifying cancer patients at high risk for ED visits within one year (365 days) following their index date. A descriptive and retrospective cohort study of 254,552 unique cancer patients with only one primary cancer type was done using linked administrative sources of health care data. Three outcomes were investigated in this study. First, the time of ED visit following an index date was predicted using multiple linear regression. Second, those patients who visited an ED within seven days of their index date were detected using logistic regression. In addition to predicting emergency department visit, vital status of patients was also predicted using logistic regression. We implemented the linear/logistic regression once on unclustered raw data. Then cluster analysis was done before the prediction step and the results of these two analyses were compared to each other. Clustering was found to contribute to a modest improvement in prediction accuracy for all three outcome variables. In addition, for the first outcome variable, a privacy preserving analysis was carried out using summarized clustered data (mean, standard deviation and correlation). The results are discussed in terms of the ability of summarized patient type data to provide clinical support tools while also respecting the privacy of patients.