Systems, methods, and devices for message control
Systems, methods, and devices include a message control platform for use in a hospital, clinic, or other urgent setting. The message control platform includes a neural network which generates a supervised machine learning model with a message prioritization system and/or a recommendation generation system. The message control platform provides improved message control for communication devices used to manage patient workflow. A message control system includes input messages, received at a first computing device from a second computing device. The message prioritization system generates an acuity indicator corresponding to the patient-related data of the input message. The acuity indicator is presented with the one or more messages at a first user interface of the first computing device. Additionally, a recommendation output message, generated by the recommendation generation system, is presented at the first user interface of the first computing device.
1 . A message control system comprising:
a message prioritization system comprising:
a processing unit comprising one or more processors;
one or more memory devices comprising computer program code configured, when executed by the one or more processors, to cause the one or more processors to:
receive training data corresponding to training messages;
receive assigned acuity levels for the training messages as part of a supervised learning procedure;
train a machine-learning architecture based on the training data and the assigned acuity levels for the training messages, wherein the machine-learning architecture comprises one or more decoders, wherein each decoder of the one or more decoders comprises a feed-forward neural network comprising an input layer of nodes, one or more intermediate layers, and an output layer comprising one or more output nodes, and wherein the machine-learning architecture does not include any encoders;
receive patient-related data corresponding to messages;
input the patient-related data corresponding to the messages into the machine-learning architecture at the input layer of nodes;
perform computations based on the patient-related data in the one or more intermediate layers;
generate, for the patient-related data for each patient, an acuity indicator corresponding to the patient-related data as an output of the one or more output nodes, wherein the acuity indicator is generated at least partially based on the computations; and
cause a message notification for each of the messages to be presented on a first user interface of a first computing device in a particular order based on the acuity indicator for each patient,
wherein the acuity indicator represents an urgency level associated with a condition of a patient.
2 . The message control system of claim 1 , wherein the machine-learning architecture uses autoregressive generation.
3 . The message control system of claim 1 ,
wherein the first computing device is configured to present the acuity indicator with the message notifications at the first user interface of the first computing device.
4 . The message control system of claim 1 , wherein the computer program code, upon being executed by the one or more processors, causes the one or more processors to:
generate a recommendation output message,
wherein the first computing device is configured to present the recommendation output message.
5 . The message control system of claim 1 , wherein the first computing device comprises computer readable code and one or more processors, the computer readable code of the first computing device comprising an acuity-based push notification integration with the first computing device, and, when the computer readable code of the first computing device is executed by the one or more processors of the first computing device, the acuity-based push notification integration causes a determination of a display order of the message notifications at a display screen of the first computing device.
6 . The message control system of claim 5 , wherein the computer readable code is configured, when executed by the one or more processors of the first computing device, to cause the one or more processors to:
receive a response metric; and
use the response metric as feedback to optimize the message prioritization system,
wherein the response metric is based on activity at the first computing device responsive to the messages.
7 . The message control system of claim 6 , wherein the response metric comprises one or more of a response content, a response time, an unaddressed message, or a subsequent action.
8 . The message control system of claim 1 , wherein the computer readable code is configured, when executed by the one or more processors, to cause the one or more processors to:
assign a high-acuity classification to a first message of the messages;
assign a medium-acuity classification or a low-acuity classification to a second message of the messages; and
cause the first computing device to present an acuity-based display arrangement in which a first message notification associated with the first message is displayed with a higher presentation prominence than a second message notification associated with the second message.
9 . The message control system of claim 8 , wherein the higher presentation prominence comprises at least one of a higher position on a list, a font change, a larger font size, a screen portion designation, or a presentation color.
10 . The message control system of claim 1 , wherein the acuity indicator is based on a color scheme which corresponds to a classification performed by the message prioritization system.
11 . The message control system of claim 1 , wherein the feed-forward neural network comprises a transformer integrated with a classifier.
12 . The message control system of claim 11 , wherein the transformer and the classifier are trained with training data including a plurality of training text messages including patient information.
13 . The message control system of claim 12 , wherein each decoder of the one or more decoders comprises an attention mechanism.
14 . The message control system of claim 13 , wherein the attention mechanism is a multi-headed attention mechanism.
15 . The message control system of claim 14 , wherein the machine-learning architecture comprises a human feedback unit configured to allow a human operator to make adjustments to a machine learning model.
16 . The message control system of claim 15 , wherein the machine-learning architecture is configured to be optimized by using one or more feedback loops.