Medical Quality Management System

Abstract

This invention discloses a medical quality management system that combines a medical information system with a generative training model. Through multi-level prompting engineering, retrieval-enhanced generation technology, and AI plugin technology, it automates risk assessment and real-time alerts to improve medical quality management. The invention first uses a risk assessment integration module to extract corresponding medical data based on the patient's user profile and generate a prompt instruction. Then, a computing platform analyzes and evaluates the prompt instruction and risk triggering factors to generate a risk assessment result. This result is then transmitted to an event notification module. When an alert condition is met, a risk event notification is sent to a user's terminal device for display, helping personnel to identify and monitor potential medical risks. This system is suitable for risk management and medical decision support in medical institutions, aiming to improve overall medical quality.

Claims

1. A medical quality management system, for a user terminal device to input a user characteristic file into a server, and after calculation and analysis through a computing platform, output a risk prediction result, comprising: the server having a central processing module, which is interconnected with a management platform, an event notification module, and a medical database, the medical database storing several medical data records; the management platform having a risk assessment integration module, a data retrieval module, and one or more medical record modules interconnected with each other, the risk assessment integration module having a pre-set prompt word library, the prompt word library storing multiple prompt words; The risk assessment integration module receives a risk assessment request, prompting the data retrieval module to send a query command to the medical database to access several medical data entries matching the user's profile. Each medical record module establishes a connection with each medical data entry by setting a module identification code, forming a structured medical record data set. The risk assessment integration module then selects multiple prompt words matching the user's profile from the prompt word library and integrates them according to a combination rule to generate a prompt command. The computing platform, connected to the server, has a computing module and a generative pre-trained model. The computing module drives the generative pre-trained model to receive the prompt command and perform evaluation calculations based on a risk trigger factor to output the risk prediction result. The risk prediction result is transmitted to an event notification module through the risk assessment integration module. After determining that an alarm condition is met, a risk event notification is generated and sent to the user's terminal device for message display.

2. The medical quality management system as described in claim 1, wherein, The prompt word library can be built from a relational data table or a JSON structured data format, and each prompt word can be a medical record format of subjective data, objective data, assessment, or plan.

3. The medical quality management system as described in claim 1, wherein, The risk trigger factor is any one of the following: an abnormal vital sign parameter, a marker of a test value deviating from the normal range, a marker of a specific disease or complication, a drug interaction risk indicator, or a composite risk score calculated according to clinical rules.

4. The medical quality management system as described in claim 1, wherein, The computing module performs word segmentation on the prompt instruction, generates a corresponding word vector through a word embedding table, and parses the prompt instruction into a basic layer, an intermediate layer and a high-level layer through a multi-level prompt engineering process. It also converts the risk trigger factor into a numerical vector or a risk weight vector and performs vector fusion with the parameters of the user profile to generate a final input vector.

5. The medical quality management system as described in claim 4, wherein, The computation module inputs the final input vector into the generative pre-trained model, extracts features through self-attention and a feed-forward network, and combines a retrieval-augmented generation (RAG) to call an external medical database or an AI plugin to generate a context, and then produces the risk prediction result after vector transformation.

6. The medical quality management system as described in claim 1, wherein, The risk prediction result contains multiple numerical fields, each containing at least one risk assessment value, which can be a probability value, a risk score, or a risk rating indicator.

7. The medical quality management system as described in claim 1, wherein, The warning condition can be any set of pre-defined judgment rules, such as a numerical threshold parameter, a logical judgment condition, or a weighted comparison rule.

8. The medical quality management system as described in claim 1, wherein, The server further includes a second medical database and a third medical database that are linked to the central processing module, each corresponding to a medical specialty or a medical field.

9. The medical quality management system as described in claim 1, wherein, The user terminal device can be replaced by a medical detection device, which can be any of a vital sign monitoring device integrating a sensing unit, a wearable medical device, and a bedside monitoring device, to capture a patient's physiological data in real time or periodically, and transmit the patient's physiological data to the server through a built-in communication module, and display the risk event notification on a display interface.

Images