Using the diagnostic odds ratio to select multivariate sequential patterns in order to build an interpretable pattern-based classifier in a clinical domain (Preprint)

BACKGROUND It is important to exploit all available data on patients in settings such as Intensive Care Burn Units (ICBUs), where several variables are recorded over time. It is possible to take advantage of the multivariate patterns that model the evolution of patients in order to predict their survival. However, pattern discovery algorithms generate a large number of patterns, of which only some are relevant for classification. The interpretability of the model is, moreover, an essential property in the clinical domain. OBJECTIVE We propose to use the Diagnostic Odds Ratio (DOR) to select the multivariate sequential patterns used in the classification in a clinical domain, rather than employing frequency properties. This makes it possible to employ a terminology closer to the language of clinicians, in which a pattern is considered to be a risk factor or to have a protection factor. METHODS We employ data obtained from the ICBU at the University Hospital of Getafe, where six temporal variables for 465 patients were registered every day during 5 days, and to model the evolution of these clinical variables we use multivariate sequential patterns. We compare four ways in which to employ the DOR for pattern selection: 1) We use it as a threshold in order to select patterns with a minimum DOR; 2) We select patterns whose differential DORs are higher than a threshold as regards their extensions; 3) We select patterns whose DOR confidence intervals do not overlap; and 4) We propose the combination of threshold and non-overlapping confidence intervals in order to select the most discriminative patterns. As a baseline, we compare our proposals with Jumping Emerging Patterns (JEPs), one of the most frequently used techniques for pattern selection that utilize frequency properties. RESULTS We have compared the number and length of the patterns eventually selected, classification performance, and pattern and model interpretability. We show that discretization has a great impact on the accuracy of the classification model, but that a trade off must be found between classification accuracy and the physicians' capacity to interpret the patterns obtained. We have, therefore, opted to use expert discretization without losing too much accuracy. We have also identified that the experiments combining threshold and non-overlapping confidence intervals (Option 4) obtained the fewest number of patterns but also with the smallest size, thus implying the loss of an acceptable accuracy as regards clinician interpretation. CONCLUSIONS A method for the classification of patients’ survival can benefit from the use of sequential patterns, since these patterns consider knowledge about the temporal evolution of the variables in the case of ICBU. We have proved that the DOR can be used in several ways, and that it is a suitable measure with which to select discriminative and interpretable quality patterns.