Patient-Based Real-Time Quality Control Using ML-Predicted Test Results
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current internal quality control (IQC) methods in medical laboratories rely on quality control materials, which are not patient-based, leading to matrix effects, high costs, and insufficient sensitivity, particularly for metrics like white blood cell count and thyroid stimulating hormone, requiring a large number of samples for error detection.
Innovation Solution
A patient-based real-time quality control (PBRTQC) method using machine learning models trained on patient information such as disease, age, and gender to predict test results, allowing continuous monitoring and reducing the number of samples needed for error detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If quality control materials are used for internal quality control, then the analytical process can be monitored at specific time points, but the monitoring is not continuous and produces matrix effects that affect IQC results
Solution Approach 1:
The patent uses patient samples as copies of real test objects instead of artificial quality control materials. By analyzing actual patient samples continuously, the system eliminates matrix effects while maintaining monitoring reliability, as patient samples naturally reflect the true test conditions without artificial interference
Solution Approach 2:
The patent implements continuous quality control by continuously analyzing patient samples as they come through the testing instrument, rather than performing discrete IQC checks at scheduled intervals. This continuous monitoring eliminates gaps in detection and maintains constant oversight of analytical performance
2Measurement precision
If quality control materials are used for internal quality control, then statistical evaluation can be performed, but additional costs for quality control materials, reagents, and human resources are required
Solution Approach 1:
The patent makes the patient sample population self-service for quality control purposes. The continuous stream of patient samples automatically provides the data needed for statistical evaluation and performance monitoring, eliminating the need for separate quality control materials and associated costs while maintaining measurement precision
Solution Approach 2:
The patent makes patient samples serve dual functions: both clinical diagnosis and quality control monitoring. This multi-functionality eliminates the need for dedicated quality control materials, reagents, and personnel, as the same patient samples used for testing also provide continuous IQC data for statistical evaluation
3Reliability
If traditional IQC methods are used, then out-of-control factors can be detected, but a large number of samples are required for error detection in metrics like white blood cell count and thyroid stimulating hormone
Solution Approach 1:
The patent implements real-time feedback by continuously comparing actual patient sample results with expected values and immediately detecting deviations. This continuous feedback loop enables rapid error detection without accumulating large numbers of samples, as each new patient sample provides immediate quality control information
Solution Approach 2:
The patent uses continuous analysis of patient samples to provide uninterrupted quality control monitoring, allowing errors to be detected as soon as they occur rather than waiting for accumulated samples to show statistical significance. This continuous action reduces the number of samples needed for timely error detection
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Embodiments of the present application relate to a method and a system for patient-based real-time quality control, and a method for constructing a quality control model for patient-based real-time quality control. The method includes: while a testing instrument performs real-time tests on current patient samples for a target item, acquiring actual test results of the current patient samples for the target item obtained by the testing instrument and patient information of the current patient samples, wherein the patient information comprises at least one of patient's disease, patient's age, patient's gender or ordering department for the target item; quantifying the patient information of the current patient samples and inputting the quantified patient information into a pre-constructed machine learning model to obtain output results of the machine learning model as predicted test results for the target item, wherein the machine learning model is trained by using actual test results for the target item and patient information of a plurality of historical patient samples; acquiring monitoring indicators based on the actual test results and the predicted test results; and determining whether or not the testing instrument is under control, and/or determining whether or not to give an alarm prompt for indicating that the testing instrument is not under control, based on the monitoring indicators. The obtained quality control has high stability and sensitivity.