Creating multiple prioritized clinical summaries using artificial intelligence

A method, comprising:

• receiving a message comprising a type of a user using a user device for displaying an aggregate summary, the type of the user associated with a hyper-parameter;
• receiving, at an artificial intelligence engine executing on at least one processor, patient data associated with a patient, the patient data including unstructured data and structured data collected from multiple computing devices associated with a medical facility;
• determining, using a plurality of natural language processing models in the artificial intelligence engine, active clinical issues from the unstructured data, wherein each natural language model is trained on historical data associated with a specific clinical issue to determine the specific clinical issue;
• determining, using a knowledge graph in the artificial intelligence engine, treatments associated with the active clinical issues;
• determining active diagnostic orders and active treatment orders from the structured data;
• prioritizing the active clinical issues, the treatments, the active diagnostic orders, and the active treatment orders into key concepts;
• summarizing, using a plurality of natural language generation models in the artificial intelligence engine, the key concepts into a plurality of summaries, wherein each natural language generation model is trained to summarize a corresponding clinical issue, treatment, diagnostic order or treatment order included in the key concepts;
• aggregating, using an aggregating natural language generation model, the plurality of summaries into the aggregate summary associated with the patient, wherein the aggregated summary is a paragraph and includes at least one hyperlink associated with at least one summary in the plurality of summaries embedded into the paragraph and wherein the at least one hyperlink comprises additional information associated with the at least one summary;
• smoothing the aggregated summary according to the hyper-parameter associated with the type of the user, wherein the smoothed summary having a first type of a user is different than the smoothed summary having a second type of a user;
• and
• displaying the aggregate summary including the at least one hyperlink on the user device, wherein the at least one hyperlink expands the aggregate summary on a display screen of the user device to display the additional information upon receiving input activating the at least one hyperlink.

A non-transitory computer-readable medium having instructions thereon, that when executed by a processor, cause the processor to perform operations for generating a patient summary, the operations comprising:

• receiving, over a network, patient data associated with a patient, the patient data including unstructured data and structured data collected from multiple medical devices connected to the network;
• determining, using at least one first natural language processing model, an active clinical issue from the unstructured data;
• determining, using a knowledge graph, a treatment associated with the active clinical issues;
• determining an active diagnostic order and an active treatment order from the structured data, the active diagnostic order and the active treatment order associated with the active clinical issue;
• generating, using a plurality of tasks in a plurality of natural language generation models, a plurality of summaries, one summary for one task, the plurality of summaries associated with the active clinical issue, the treatment, the active diagnostic order, and the active treatment order, and wherein each natural language model in the plurality of natural language models was trained on historical data to summarize a particular task;
• aggregating, using an aggregating natural language generation model, the plurality of summaries into the patient summary;
• receiving a message including a type of a user using the user device for displaying the aggregate summary;
• associating the type of the user in the message with a hyperparameter;
• smoothing, using the aggregating natural language generation model and the hyperparameter, the patient summary according to the type of the user included in the message, wherein a smoothed patient summary associated with a first user type is different from a smoothed patient summary associated with a second user type;
• and
• transmitting the smoothed patient summary to the user device.