Intelligent teletriage and personalized routing to manage patient access in a neurosurgery clinic

We consider the intake process of new low back pain (LBP) patients at a neurosurgery clinic to manage patient demand for improving access delays through personalized routing strategies rather than increasing care capacity. Using clinical notes from the first appointments with providers, we devise a decision-tree based intelligent teletriage tool that can be used by non-medically trained agents to predict the surgical class of a patient calling in to request an appointment. The intelligent teletriage tool is based on a classifier that uses surgical-nonsurgical labels that we have generated using a structured algorithm and features that are easy to obtain directly from the patient during the course of a phone conversation. We establish that the accuracy of the teletriage tool is in the order of 80% using 10-fold cross validation and out-of-sample testing on real-life data sets. We then present three priority-based routing strategies that are neutral with respect to care capacity, and show that when used in combination with the intelligent triage tool, these can result in 90% reduction in access delays for the higher priority surgical patients who should be seen urgently. We use detailed simulations of the appointment scheduling workflow to demonstrate our results. We comment on the managerial implications of our work and the potential for the use of needs-based personalized routing strategies with intelligent teletriage to reduce access delays, improve patient outcomes and provider satisfaction.