# US 2021/0082575 A1 · Computerized decision support tool for post-acute care patients > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2021/0082575 A1 - Document ID: US20210082575A1 - Status: Published application - Kind: A1 - Application number: 17/024,241 - Filed: 2020-09-17 - Published: 2021-03-18 - Earliest claimed: 2019-09-18 - Assignees: Cerner Innovation, Inc. - Applicants: Cerner Innovation, Inc. - Inventors: Leonardo Ji; Andrew Roberts; Sasanka Are - Web page: https://patentgrep.com/patent/US20210082575A1 ## CPC - G16H 40/20: for the management or administration of healthcare resources or facilities, e.g. managing hospital staff or surgery rooms - G06N 3/044: Recurrent networks, e.g. Hopfield networks - G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] - G06N 3/08: Learning methods - G06N 3/09: Supervised learning - G16H 20/00: ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance - G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems - G16H 50/30: for calculating health indices; for individual health risk assessment - G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract A decision support tool is provided for discharging a patient by predicting the probability of a patient's readmission based on information prior to discharge. The information used to make the prediction may include labs, vitals, diagnoses, and medications from prior encounters and from the current encounter. At least some of this information may be used to compute a readmission prediction. One or more recurrent neural networks may be utilized to generate the prediction, and one or more sequences may be generated from the patient information to feed into the one or more recurrent neural networks. Based on the prediction, one or more actions may be initiated to reduce the probability of readmission. ## Claims ### Claim 1 A computer storage media having computer-executable instructions embodied thereon that, when executed, provide a method for predicting readmission for a patient, the method comprising: receiving patient data recovered over time for the patient, the patient having received post-acute care services; generating one or more sequences based on the patient data recorded over time; based at least in part on the one or more sequences, determining, using a neural network, a readmission prediction that the patient will be readmitted into an acute care facility within a future time interval; and initiating an action based on the readmission prediction. ### Claim 2 (depends on 1) The computer storage media of claim 1, wherein the action comprises one or more of modifying discharge instructions for the patient or scheduling a follow-up appointment. ### Claim 3 (depends on 1) The computer storage media of claim 1, wherein generating the one or more sequences comprises arranging the patient data in reverse chronological. ### Claim 4 (depends on 1) The computer storage media of claim 1, wherein generating the one or more sequences comprises generating a visits sequence with feature values indicating a plurality of healthcare visits by the patient and generating an events sequence with feature values for one or more of a lab result, a diagnosis, a medication, and a procedure for the patient. ### Claim 5 (depends on 4) The computer storage media of claim 4, wherein the neural network is a recurrent neural network. ### Claim 6 (depends on 5) The computer storage media of claim 5, wherein the recurrent neural network comprises a visits layer into which the visits sequence is input and an events layer into which the events sequence is input. ### Claim 7 A computerized method for implementing a decision support tool for a patient receiving post-acute care services, the computerized method comprising: receiving patient data recovered over time for the patient, the patient having received post-acute care services; generating one or more sequences based on the patient data; based at least in part on the one or more sequences, determining, using a neural network, a readmission prediction that the patient will be readmitted into an acute care facility within a future time interval; and initiating an action based on the readmission prediction. ### Claim 8 (depends on 7) The computerized method of claim 7 further comprising converting the patient data that is in at least one nomenclature into a standard nomenclature using an ontology. ### Claim 9 (depends on 7) The computerized method of claim 7, wherein the action comprises displaying the readmission prediction and providing an alert when the readmission prediction is above a threshold. ### Claim 10 (depends on 7) The computerized method of claim 7, wherein the action comprises scheduling a visit for the patient with a healthcare provider. ### Claim 11 (depends on 10) The computerized method of claim 10, further comprising: determining a contribution score for at least some features within the patient data utilized to determine the readmission prediction for the patient; and providing the contribution score for display. ### Claim 12 (depends on 7) The computerized method of claim 7, wherein the one or more sequences comprises a visits sequence indicating a plurality of healthcare visits by the patient and an events sequence indicating a plurality of healthcare events for the patient, healthcare events comprising a diagnosis, a medication, or a procedure. ### Claim 13 (depends on 12) The computerized method of claim 12, wherein the neural network is a recurrent neural network comprising a visits layer into which the visits sequence is input and an events layer into which the events sequence is input. ### Claim 14 (depends on 7) The computerized method of claim 7, wherein the future time interval comprises 90 days and the readmission prediction is determined three days prior to a scheduled discharge of the patient from post-acute care services. ### Claim 15 A computerized system for predicting readmission for a patient who has received post-acute care services, the system comprising: one or more processors; and one or more computer storage media storing computer-useable instructions that, when used by the one or more processors, cause the one or more processors to perform a method, the method comprising: receiving patient data recovered over time for the patient; generating at least two sequences based on the patient data, wherein the at least two sequences comprise at least one sequence of feature values arranged in reverse chronological order; utilizing a recurrent neural network to determine, based on the at least two sequences, a likelihood that the patient will be readmitted to an acute care facility within a future time interval; and initiating an action based on the likelihood that the patient will be readmitted. ### Claim 16 (depends on 15) The computerized system of claim 15, wherein the at least one sequence of feature values is based on at least two patient visits, wherein at least one of the at least two patient visits includes services other than post-acute care services. ### Claim 17 (depends on 15) The computerized system of claim 15, wherein at least one of the at least two sequences comprises an events sequence in chronological order. ### Claim 18 (depends on 17) The computerized system of claim 17, wherein the at least two sequences comprises a visits sequence indicating a plurality of healthcare visits by the patient and an events sequence indicating a plurality of healthcare events for the patient, healthcare events comprising a diagnosis, a medication, or a procedure, wherein the visits sequence is arranged in reverse chronological order. ### Claim 19 (depends on 15) The computerized system of claim 15, wherein the future time interval comprises 90 days ### Claim 20 (depends on 15) The computerized system of claim 15, wherein the action comprises displaying the likelihood that the patient will be readmitted and providing an alert to a healthcare provider when the likelihood that the patient will be readmitted is above a threshold.