Prediction and Optimal Control of Triage Nurse Ordering: A Data-Driven Approach to Improve Patient Flow in Hospital Emergency Departments
DescriptionEmergency department overcrowding is a worldwide problem impairing the ability ofhospitals to provide emergency care within a reasonable time frame. After triage, apatient has to wait for a physician for consultation. We call this thefirst waiting, andit lasts from several minutes to several hours depending on the patient's clinical severity.After the consultation, the physician may order some diagnostic tests (lab test, x-ray,etc.), requiring the patient to wait for the tests to be completed and the results to bereturned. We call this thesecond waiting. Alternatively, if the required tests can bepredicted with the information available at triage (demographic information, vital signs,etc.), nurses can order the tests immediately after triage, in which case the patient goesdirectly to the test center to do tests while waiting for a physician. We call this practicetriage nurse ordering, or TNO. Once the test results are ready, the physician checks theresults and proposes a treatment plan. The main idea of TNO is to combine the twowaitings thus reduce the patient's length of stay. However, one should be aware thatTNO may over-request diagnostic tests that would not have been ordered by physicians.Hence, whether TNO should be implemented is not a trivial decision. In the proposedproject, we will develop a statistical model to predict which diagnostic tests should beordered using the patient's information at triage. We will use text mining to incorporatefree-text triage notes into our model to increase the prediction power. We also plan tobuild an Markov decision process formulation to study the optimal decisions on TNO. Tothe best of our knowledge, this will be the first analytical work on the decision makingof TNO.The current practice of TNO in hospitals relies on pre-set protocols that are created based on physicians' medical knowledge and experience. The protocols are usuallysymptom-specific and the over-requesting rates are high, all of which diminish the benefitsof TNO. In the era of Big Data, electronic health record data have been used intensivelyfor research designed to improve patient care. It is surprising that little work has beendone using electronic health record data to support TNO decision making. In addition, although healthcare administrators are aware of the risk of ordering excessive teststhrough TNO, little is known regarding when TNO should be implemented to balancethe benefits (time savings) with the potential drawbacks (unnecessary tests). Our projectaims to fill these gaps.In this proposal, we present our preliminary results and describe a set of tasks forachieving our research goals. TNO is a promising practice that could improve patientow and reduce emergency department overcrowding. Our team's training and researchexperiences in emergency medicine, statistics and operations research have prepared usto achieve our research goals.?
|Effective start/end date||1/01/18 → 26/11/21|