one, PBRTQC Research progress and types of operations

patient data AON Law in 1965 Year by year Hoffman and Waid propose, The technical principle is to select approximately normally distributed patient test results within the normal range after completing daily work and calculate the average value, Used for monitoring and analyzing the stability of the process, The limitation of this method is that it cannot monitor performance changes in real-time. Subsequently developed BULL law, MA law, EWMA law, movMed law, Moving percentile method, Multiple algorithms such as moving outlier method and moving standard deviation method, Each calculation method has its own characteristics and advantages. BULL The method is mainly applied to the average volume of red blood cells, Red blood cell mean hemoglobin and red blood cell mean hemoglobin concentration, etc 3 Quality control of a project, this 3 There is very little variation between different groups of people in a project, Not subject to blood dilution, concentrate, Pathological or technical factors affecting, Quality control of the method for calculating the mean of batch patient results based on this characteristic; Cembrowski and Westgard The algorithm was validated through computer simulation［11, 12］, This algorithm minimizes the impact of outliers, But this algorithm needs to complete batch detection processing before generating new ones MA value, Will delay the identification of system errors; MA Law is rarely applied in professional fields outside of the field of blood cells; movMed The method is more effective in detecting non normal distributions and extreme detection results; EWMA The law introduces weight coefficients (λ) , Through front and back 2 Allocation of weights for individual detection results, Calculate a new test result for each new one MA value, To achieve optimal and rapid bias detection, Represents a more realistic and continuous moving average, Its advantage is to detect small variations in the accuracy or precision of the analysis process. PBRTQC The key to sensitive identification of performance changes in detection systems through various computational methods is the selection of variables in the algorithm, as BULL law, MA Law and movMed The variable of the algorithm is the number of average values calculated from batch testing results, EWMA The variable of law isλ.

two, PBRTQC Clinical application value

Clinical laboratory analysis errors mainly come from reagents, calibrator, detection system, Personnel and environment, etc, quality control material QC Method is an important means commonly used to evaluate the imprecision and systematic error changes in detection systems, But there are certain limitations, Analysis errors are usually discovered when an out of control alarm is triggered; Westgard［13］Yu 2003 Propose to form the lowest cost in the year, Multi stage comprehensive quality control strategy for maximizing benefits (total quality control, TQC) , Including the best quality control products QC Law and patient data quality control method. Domestic and foreign research reports confirm that patient data quality control methods have the following advantages, Including no matrix effect, Regarding startup quality control products QC The stability of analytical performance of the sustainable monitoring and analysis system jointly used by France and France, Evaluate the comparability of clinical laboratory test items, The monitoring and analysis process originates from reagents, Analysis errors of instruments and calibration, Monitoring specimen collection, Pre analysis errors caused by improper transportation and handling processes, Improper selection of identifiable quality control products, Application of Continuous Monitoring and Testing System for Real Performance Changes Throughout the Period of Batch Number Replacement of Indoor Quality Control Products, Can compensate for quality control products in terms of error detection capability and quality control frequency QC Shortcomings, reduce costs, Can provide early warning of minor changes in the analysis performance of the detection system, And early warning, Avoiding the occurrence of potential quality risks, etc［14, 15, 16, 17］.

three, Related to domestic and international guidelines and industry standards PBRTQC suggestion

2011 The American Society for Clinical and Laboratory Standardization released EP23-A Approval Guidelines for Laboratory Quality Control Based on Risk Management［18］Suggest adding patient data quality control methods, In the case where the number of patients and the distribution of test results are relatively stable, This method not only eliminates matrix effects, It can also monitor the distribution trend of patient test results over time and the systematic error of the detection system; 2018 The health industry standards of the People's Republic of China issued in WS/T641-2018 Clinical Laboratory Quantitative Determination Indoor Quality Control Guidelines［19］We have also added quality control methods for applying patient data for quality control, including AON law, MA law, Differential inspection method and double difference quality control method for patient samples; IFCC PBRTQC The working groups are divided into 2019 Year and 2020 Separate documents were released annually, right PBRTQC The Information System, Program establishment, Provide guidance for performance validation and clinical implementation［1, 7］.

four, PBRTQC Program establishment, Performance validation and implementation principles

PBRTQC Including multiple algorithms, Clinical laboratories can choose to use one or more of them based on different algorithm features. PBRTQC The biggest challenge in implementation is to obtain the best laboratory specific results PBRTQC program, Clinical laboratories should choose professional software tools, conduct PBRTQC Parameter Settings, Program establishment, performance verification, Optimization and implementation, To achieve the best identification of laboratory quality risks.

(one) PBRTQC Information system type

Independent statistical analysis software, Middleware and laboratory information systems can both support to varying degrees PBRTQC the establishment, Implementation and operation. Most blood cell analyzers typically support traditional methods XbarB law, Ruxi Sen Meikang, Mindray Hematology Analyzer, Moreover, Some automated assembly line middleware systems, such as Remisol Advance (beckman coulter ) support EWMA algorithm, Centralink data management system (Siemens) support MA algorithm, independent PBRTQC Intelligent professional software tools such as a real-time quality control intelligent monitoring platform for patient data based on artificial intelligence AI-MA (Senxu Medical) support BULL law, MA law, EWMA law, Cumulative and Control Chart Method (Cusum law) , The median method and Z Score, etc 6 Types of operations［20］.

(two) PBRTQC Basic characteristics of information systems

PBRTQC The design and implementation require real-time access to patient data, Establish appropriate rules, Operating procedures and selecting the best algorithm, IFCC PBRTQC The working group proposed, PBRTQC The information system should include the following basic functions: data collection, data storage, Data Extraction and Analysis, statistical analysis, Visual quality control chart, Test the performance verification capability of the algorithm, Real time operation function, Warning out of control correction and audit records, System log tracking, etc; Moreover, Propose advanced multidimensional visualization tools, Statistics of normal distribution of data, Data filtering and integration IQC Independent software or laboratory information systems with equivalent functional features are innovative, Non traditional differentiated functions［1］.

(three) PBRTQC Data information to be obtained, Data storage and security privacy requirements

PBRTQC The laboratory information management system required for program establishment (laboratory information management system, LIS) The data includes but is not limited to the following information, Including professional groups, Brand and model of detection system, Sample testing date and time (hour, minute, second) , Sample identification number, Sample Type, Department, Diagnostic information, Patient age, gender, Test project name, Test results, unit, etc; PBRTQC The instrument information obtained through program establishment and verification includes reagent batch numbers, Bottle No, Calibration Record, Indoor quality control, Instrument alarm information, Serum index information, Information on instrument malfunctions and maintenance. PBRTQC Program data storage needs to pay attention to security and privacy issues, It should comply with relevant data security and patient privacy regulatory requirements, Allow real-time extraction for saving. Transfer patient data to PBRTQC During the application process, Attention should be paid to removing the identification information of patient names in the dataset.

(four) PBRTQC Program establishment, Performance validation and optimal method selection

1. establish PBRTQC Key points of different computational programs: Including inclusion/Exclusion criteria, Calculation formula, batch size, weight factor, control limit, Quality Objective, Warning/Out of control rules, Control limit selection, Performance evaluation indicators and acceptable standards, etc.

2. PBRTQC Classification and proportion of program performance validation datasets: IFCC PBRTQC The performance validation recommendation document states that, In implementation PBRTQC Previously, its quality control effectiveness should be validated, To evaluate PBRTQC Actual performance in the operating environment; Extract at least from the historical database of patient test results 6 month, 1 Year or 1 Representative datasets of over years old, Divided into training dataset and validation dataset, Proportion from 50∶50 to 80∶20 not equal, for PBRTQC Parameter settings for, Performance validation and optimization, This can capture all situations more comprehensively, Including the environment, Changes in patient population, Reagent and calibration batch number replacement, etc［1］.

3. PBRTQC Program performance verification and optimal method selection: Including trial and error method, Efficacy function analysis method［21, 22］, Simulation moving average verification model method with introduction of systematic error［16］Routine practice verification methods for real error data in clinical laboratories, etc［20］. PBRTQC Performance verification indicators include total allowable error (TEa) , Sensitivity specificity of error recognition, Number of false alarms (false positive rate) , False negative rate, The number of patient sample results required by the average method before error detection (ANPed) , The number of patient sample results required by the median method before error detection (MNPed) ［21, 22］and EWMA Probability of detecting legal errors (Ped) wait［20］, Clinical laboratories should determine acceptable quality management standards PBRTQC Performance evaluation indicators and objectives. IFCC PBRTQC Suggestions for performance verification are listed MA Performance verification method of law［1］, determine TEa and ANPed in PBRTQC Introducing positive values at different time points in the model validation dataset, Negative bias system error result, Through statistics ANPed Determine the optimal one MA program parameters; van Rossum and Kemperman［23］Optimize and validate using bias detection curves and moving average validation graphs MA Law or EWMA Legal procedure; Wen Dongmei and others［20］Extract real patient historical data, press 80∶20 The proportion is divided into training set and validation set, based on AI-MA PBRTQC Performance validation model, determine TEa, False positive rate and Ped Acceptance criteria, By changing the instrument reagent batch number, Calibration Record, Real quality event information such as instrument alarm information and EWMA Time nodes for error warning in graph analysis, Compliance statistics of positive and negative bias directions Ped, Determine the optimal solution based on this EWMA program. PBRTQC Performance verification and optimization of programs require attention to balance Ped Proportion of false positive rate, Ped smaller, Indicating a lower risk of issuing error reports when analyzing errors, The higher the false positive rate, It will require a significant amount of laboratory resources to deal with false alarms and their impact PBRTQC The credibility of. Choose the best option Ped And the false positive rate PBRTQC Quality control methods require continuous exploration and practice to accumulate professional experience.

five, Current clinical application status at home and abroad

Westgard［13］long ago 2003 It was pointed out in the year that clinical laboratories should choose to apply patient data quality control methods. With the development of patient risk based clinical laboratory indoor quality control technology and the improvement of information technology capabilities, Related to foreign countries PBRTQC Research reports from various sources PBRTQC The clinical application value of computational types, develop into PBRTQC Performance validation tools, methods, and key elements for establishing optimal programs. Internationally, optimization methods have shifted from simple statistical models to computer simulations, Bietenbeck wait［24］use R Programming language has been used to write simulation programs; Smith wait［25］Explained the distribution of patient results for each testing item PBRTQC The Importance of Program Establishment, And adopt the optimal method based on different distribution characteristics; Badrick［26］Suggest using biological variation PBRTQC Analyze the setting of quality objectives, It is more meaningful to use this monitoring and detection system to analyze performance changes; Zhou wait［27］Using the moving median method, Compare the probability of error detection using individual instruments and combined data from multiple instruments, The results show that data from different instruments can increase the data flow, Thereby improving the speed of error detection, But it will reduce the probability of error detection. PBRTQC The successful implementation in many complex laboratories has proven that this method is highly suitable for clinical laboratory applications with strict quality risk management awareness. Moreover, PBRTQC It is also applied to inter laboratory comparisons, Belgium Goossens wait［28, 29］Developed an independent online tool for median quality control of patient data "The Percentiler" conduct 124 Home Laboratory and 250 Comparison of Biochemical Test Results of Taiwan Analyzer, This method helps to discover the stability of analytical performance and inter batch differences of various brands of testing systems in the laboratory, To improve quality in this way. Quantitative project patient data inter laboratory comparison and monitoring platform developed by the Clinical Laboratory Center of the National Health Commission, Using the percentile method of patient data, By collecting data from across the country 181 Home Clinical Laboratory Clinical Biochemistry Project and National 105 Median patient data for new screening indicators in the home screening laboratory, Can vertically monitor the long-term stability of the laboratory itself and conduct inter laboratory comparisons［10］.

Domestic related PBRTQC The research mainly focuses on the clinical application value of different algorithms. Liu Zhengmin and others［30］application EWMA Monitoring the stability of nucleic acid testing process using quality control charts, Establish an early warning mechanism; Xia Jun and others［31］Based on LIS Establish an indoor quality control program for clinical biochemical patient data mean and percentage method, Prove that it is an economically practical way, Can make up for the shortcomings of existing quality control; Song wait［32］Confirm the optimized analysis of armor function detection MA and AON Method for identifying and analyzing performance changes, Superior to quality control products in terms of trends and calibration events QC law.

Domestic related PBRTQC Professional software development, There are very few research reports on technology improvement, optimization, and performance verification, near 1 Annual increase. PBRTQC Selection of application programs, performance verification, Optimization and parameter optimization of dynamic models are the most challenging, Badrick wait［33］Pointing out that with PBRTQC Implementation of, In the future, it is necessary to shift the statistical process towards artificial intelligence methods, Wen Dongmei and others［20］Independently developed PBRTQC Online real-time intelligent monitoring platform AI-MA There have been significant improvements in both technology and functionality, Development based on medical big data mining and artificial intelligence technology, Including automatic statistical analysis of patient big data distribution characteristics, data filtering, 6 kind PBRTQC Automatic modeling and performance verification program for computational technology, Integrate quality control products QC, Real time dynamic operation and intelligent warning functions, Syncable tablet and visual large screen, The platform's artificial intelligence PBRTQC Multi parameter models are superior to traditional models Westgard rules PBRTQC model, compared to IFCC Recommended differentiation PBRTQC Software tools have certain characteristics and innovations; Yang Fan and others［17］based on PBRTQC Intelligent software tools for big data performance validation and real-time application of serum ion projects, Confirm the optimal EWMA The program can be sensitive, Accurately identify the source from the reagent, Quality risks caused by instrument malfunctions and calibration, Realize full time intelligent monitoring and early warning; Duan wait［34］Collect serum sodium from outpatient patients, chlorine, Alanine transaminase, Creatinine, etc 4 Detection results of conventional analytes with different distribution characteristics 906 552 copy, right 6 kind PBRTQC Compare and evaluate the performance of algorithms under different types of analysis errors, Indicate before selecting and running algorithms, Assessing potential sources of error in specific analytes and the importance of corresponding types of analytical errors. Duan Xincen and others［35］Summarized recent developments both domestically and internationally PBRTQC research results, Introduction PBRTQC Theoretical basis, Establishment of Quality Control Model, Existing problems and proposed future solutions PBRTQC research direction, domestically PBRTQC The establishment and application of have certain guiding significance. Jia Yin and others［36］The application research of artificial intelligence in laboratory medicine has become an important direction for the development of laboratory testing, But there are still many problems, PBRTQC Establish a program based on the results of patient specimen testing, Attention should be paid to privacy protection when applying, Desensitize or anonymize it.

The patient data quality control method has been proposed for half a century now, PBRTQC Clinical studies and successful implementation in laboratories of different scales have demonstrated that it is a valuable and promising new quality management tool, IFCC It is recommended to PBRTQC Widely used in clinical practice, but PBRTQC Limited by cognitive level, Lack of professional software tools, Constrained by various factors such as limited clinical practice experience, There is still a certain distance to be covered in domestic applications, How to advance PBRTQC The recognition and widespread application in the field of clinical laboratory testing in China is a new direction for the future industry, Believe that with the inspection of peers PBRTQC The increasing attention, The improvement of information technology, Promotion of academic platforms, Will accelerate PBRTQC The Development and Application Process of Clinical Laboratories in China.