Using the electronic health record to identify children likely to suffer last-minute surgery cancellation

Cincinnati Childrens Hosp Med Ctr, Cincinnati OH

Investigators

Linked publications, trials & patents

Abstract

PROJECT SUMMARY/ABSTRACT Last-minute cancellation of surgery frequently leads to psychological stress and financial hardships that disproportionately affect individuals of low socioeconomic status. Moreover, cancellation leaves unutilized healthcare resources valued as high as $1 per second. Machine learning can uncover patterns in historical data to identify predictors, and captures relationships among many factors to allow assessment of risk associated with a particular set of conditions. There is a critical need to develop a model system through machine learning both to understand and to predict cancellation. The long-term goal is to develop evidence- based strategies for improved perioperative resource utilization and efficiency. The overall objective in this application is to develop and deploy an analytical model for predicting last-minute surgical cancellations. The central hypothesis is that a predictive model based on patient-specific and contextual factors will accurately determine the probability of cancellation. The hypothesis has been formulated on the basis of preliminary data showing that risk of cancellation varies substantially with respect to data extracted from the electronic health record (EHR). The rationale for the proposed research is that development of a model that can accurately predict the probability of last-minute cancellation of surgical procedures is likely to provide new opportunities for improving healthcare costs and efficiency. The hypothesis will be tested by pursuing three specific aims: 1) Develop computerized models for predicting surgery cancellation; 2) Identify key predictors of last-minute cancelation of surgery from the EHR and online data resources; and 3) Establish a scalable last-minute surgery cancellation prediction system. Under the first aim, using a pre-existing database and machine- learning techniques already established as feasible in the applicants' hands, patient-specific and contextual data from two pediatric surgical sites in a large Midwest conurbation will be mined. Under the second aim, these predictive models will be interrogated to generate actionable advice. Under the third aim, the optimal model will be identified and integrated into the clinical workflow to direct cancellation prevention and mitigation strategies. The approach is innovative, in the applicants' opinion, because it represents a substantial departure from the status quo by employing machine-learning techniques on large and detailed datasets drawing from the modern EHR and publicly available contextual data. The proposed research is significant because it will both have broad translational importance in perioperative management and also elucidate the etiology of cancellation. The positive impact is expected to be in facilitating quality improvement projects and operating room management strategies to increase utilization of expensive perioperative resources. Successful completion of the proposed research is also expected to lead to more timely surgeries at lower cost for hundreds of thousands of patients, which is of vital importance at a time when the cost of healthcare is ever increasing.