Medical laboratory data quality control method based on maximum expectation

Abstract

The invention discloses a medical laboratory data quality control method based on maximum expectation. According to the method, three indexes exceeding a reference interval, a clinical decision limit and a critical value are used as positive standards, and the number of patients with the maximum expected increase unacceptable is calculated for quality control. The method comprises the following steps: firstly, selecting proper analytes, and selecting proper control values according to different analytes; then calculating the probability that the normal person detection result is not in a normal range; calculating the number of patients unacceptable to the maximum expected increase; and finally, obtaining the distribution condition of the number of unacceptable patients with maximum expected increase, and selecting proper quality control rules and quality control frequency in combination with the horizontal positioning of a laboratory.

Description

Technical Field

[0001] The present invention relates to a data quality control method, and more particularly to a medical laboratory data quality control method based on maximum expectation. Background Art

[0002] As continuous-production analyzers became more popular, traditional QC performance indicators such as false alarm rate (Pfr) and true detection rate (Ped) could not solve the problem of test frequency. Parvin proposed a new concept, the maximum expected increase in the number of unacceptable patients (MaxE(NUF)). After that, Martín Yago et al., James O Westgard et al., and Hassan Bayat further improved Parvin's model, simplifying and visualizing complex mathematical problems.

[0003] However, Martín Yago and others believe that the current model ignores the possibility that erroneous results will lead to decisions that harm patients. This is mainly because the model uses the allowable total error (TEa) as the judgment criterion. Different TEa standards will bring different results, and exceeding TEa obviously cannot directly represent the patient's risk. For example, the urine HCG of pregnant women can increase by thousands of times, and even if the error far exceeds TEa, it will not affect clinical judgment. For healthy people with high blood sugar, even if the test error is much smaller than TEa, they can easily be misdiagnosed as diabetics. Moreover, TEa is not an indicator that doctors and patients pay attention to. Doctors and patients often only pay attention to whether their test indicators are normal.

[0004] As for the indicators used for clinical classification or decision-making, it is easy to think of reference interval (RI), clinical decision limit (CDL) and critical value. Exceeding the reference interval may be considered abnormal, CDL is considered to have diagnostic significance for specific diseases, and critical value has a great impact on the health and even life of patients. Through these indicators, the number of false positives and false negatives increased due to undetected errors can be calculated, so as to judge whether the quality control rules and quality control frequency are reasonable. Summary of the invention

[0005] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a medical laboratory data quality control method based on maximum expectation, which calculates the number of false positives increased due to undetected errors through computer simulation, thereby exploring the feasibility of using RI, CDL and critical values ​​as evaluation criteria for quality control based on patient risk.

[0006] To achieve the above object, the present invention provides the following technical solution: a medical laboratory data quality control method based on maximum expectation, characterized in that it comprises the following steps:

[0007] Step 1, selecting a suitable analyte from the medical data, and selecting a corresponding indicator for the analyte; Step 2, calculating the probability that the test result of the analyte selected in step 1 is not within the normal range in a normal person;

[0008] Step 3: Calculate the maximum expected increase in the number of unacceptable patients MaxE(N) based on the probability obtained in step 2 UF );

[0009] Step 4: Based on the maximum expected increase in the number of unacceptable patients obtained in step 3, Max E(N UF ) Calculate the distribution of the maximum expected increase in the number of unacceptable patients.

[0010] As a further improvement of the present invention, the probability that the test result of a normal person is not within the normal range in step 2 is as follows:

[0011] Step 21: Use the reference interval to simulate the true value of normal people, so that the test result X conforms to the following distribution:

[0012] f X (x)~N(μ,σ 2 )

[0013] Among them, μ is the mean and σ is the standard deviation;

[0014] Step 22: Based on the test result y being normally distributed around the true value, the distribution is constructed using the deviation and coefficient of variation:

[0015] f Y|X (y|x)~N((1+bias)*x,(CV*x) 2 )

[0016] Step 2 and 3: In the absence of other errors, calculate the probability P1 that the normal person's test result exceeds the reference interval:

[0017]

[0018] Step 24: Calculate the probability P2 that the normal person's test result exceeds the clinical decision limit in the absence of other errors:

[0019]

[0020] Step 25: In the absence of other errors, the probability P3 that the normal person's test result exceeds the critical value (with both upper and lower limits) is calculated as:

[0021]

[0022] As a further improvement of the present invention, the maximum expected increase in the number of unacceptable patients Max E (N UF ) as follows:

[0023] Max E(N UF )=MAX[ΔP*N af ]

[0024] Where ΔP is the probability of the phase difference in the presence of a runaway error condition, N af is the number of people expected to be affected.

[0025] As a further improvement of the present invention, the deviation in step 22 is calculated by Passing-Bablok regression using fifteen results from laboratories participating in the inter-laboratory quality assessment, specifically by constructing a y=b+ax regression equation and then estimating the difference between the slope a and 1.

[0026] As a further improvement of the present invention, the coefficient of variation in step 22 is obtained through self-reporting by the laboratory.

[0027] The beneficial effects of the present invention are as follows: the present invention utilizes the reference interval to simulate the distribution of normal people, takes exceeding the reference interval, the clinical decision limit, and the critical value as the positive standard, and calculates the distribution of the maximum expected increase in the number of unacceptable patients. The laboratory can select appropriate quality control rules and quality control frequencies based on its own horizontal positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 The present invention is a flowchart of the medical laboratory data quality control method based on maximum expectation. DETAILED DESCRIPTION

[0029] The present invention will be further described below in detail with reference to the embodiments shown in the accompanying drawings.

[0030] Reference Figure 1 As shown, a medical laboratory data quality control method based on maximum expectation in this embodiment, the method steps are as follows:

[0031] Step 1: Select appropriate analytes and select appropriate indicators for different analytes;

[0032] Since it is impossible to obtain the actual distribution of patient results, the present invention uses reference intervals to simulate normal patient distribution; here, the normal distribution is used by default. In order to avoid the influence of pregnancy, race, age, gender, region, etc. on the reference interval, it is best to select analytes with small physiological changes, strictly regulated dynamic changes, and small differences among populations, such as sodium.

[0033] Theoretically, the conventional test results of healthy individuals all show Gaussian or nearly Gaussian distribution, while the skewed distribution is mainly caused by disease factors and physiological factors. In the present invention, the normal distribution is used by default, so analytes with severe skewed distribution, such as TSH, should not be selected.

[0034] For different analytes, it is necessary to select appropriate indicators. For analytes with only RI, only RI can be used. For analytes with both CDL and RI, it may be more appropriate to use CDL, and for analytes with critical values, critical values ​​can be considered.

[0035] In the present invention, the simulated monocyte distribution width (MDW, used to distinguish sepsis) is selected for the study of RI, diabetes is selected for the study of CDL, and sodium is selected for the study of critical values. The 95% reference interval of MDW is set to (16.22, 23.15). The 95% reference interval of fasting glucose in the routine biochemical mutual recognition reference interval of medical institutions in Beijing is (3.30, 6.11). The clinical decision limit for type 2 diabetes is 7.0, the 95% reference interval of sodium is set to (137, 147), the lower limit critical value is 120, and the upper limit critical value is 160.

[0036] Step 2: Calculate the probability that the test results of normal people are not within the normal range;

[0037] The reference interval is used to simulate the true value of normal people. At this time, the test result X conforms to the following distribution:

[0038] f X (x)~N(μ,σ 2 )

[0039] Where μ is the mean and σ is the standard deviation. The test result y will be normally distributed around the true value, and the distribution can be constructed using the bias and coefficient of variation (CV):

[0040] f Y|X (y|x)~N((1+bias)*x,(CV*x) 2 )

[0041] In the absence of other errors, the probability P1 that the test results of a normal person exceed the reference interval is:

[0042]

[0043] In the absence of other errors, the probability P2 that the test results of normal people exceed the clinical decision limit (upper limit) is:

[0044]

[0045] In the absence of other errors, the probability P3 that the test results of a normal person exceed the critical value (with both upper and lower limits) is:

[0046]

[0047] Step 3: Calculate the maximum expected increase in the number of unacceptable patients Max E(N UF );

[0048] When a systematic error occurs, P1, P2, and P3 may change, and the systematic error can be regarded as an increase or decrease in the Lower bound and Upper bound. The probability of the difference between the previous and the next phases in the presence of an out-of-control error condition is defined as ΔP. ​​The maximum expected increase in the number of unacceptable patients MaxE(N UF ) can be defined as:

[0049] MaxE(N UF )=MAX[ΔP*N af ]

[0050] Where N af is the number of people expected to be affected.

[0051] Step 4: Calculate MaxE(N UF )’s distribution.

[0052] The probability that the test results of normal people exceed the reference interval will be affected by Bias and CV, thus affecting Max E(N UF ) is calculated. The inspection level of a laboratory is always subject to various conditions, and the Bias and CV of laboratories of different levels are often different.

[0053] To this end, in this embodiment, the inter-laboratory quality control data of different laboratories were collected to calculate the Max E (N UF ), and obtain the laboratory Max E(N UF ) is used as a reference for laboratories. Specifically, Bias (%) is calculated by Passing-Bablok regression using 15 results from laboratories participating in the external quality assessment. The calculation method is to construct a regression equation of y = b + ax, and Bias (%) is estimated by the difference between the slope a and 1. CV comes from the laboratory's self-reporting.

[0054] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A medical laboratory data quality control method based on maximum expectation, characterized by: The steps include: Step 1: Select an analyte in the medical data and select a corresponding indicator for the analyte; Step 2, calculating the probability that the test result of the analyte selected in step 1 in a normal person is not within the normal range; Step 3: Calculate the maximum expected increase in the number of unacceptable patients Max E(N) based on the probability obtained in step 2 UF ); Step 4: Based on the maximum expected increase in the number of unacceptable patients obtained in step 3, Max E(N UF ) Calculate the distribution of the maximum expected increase in the number of patients that would be unacceptable.

2. The medical laboratory data quality control method based on maximum expectation according to claim 1, characterized in that: The probability that the test result of a normal person in step 2 is not within the normal range is as follows: Step 21: Use the reference interval to simulate the true value of normal people, so that the test result X conforms to the following distribution: f X (x)~N(μ,σ 2 ) Among them, μ is the mean and σ is the standard deviation; Step 22: Based on the test result y being normally distributed around the true value, the distribution is constructed using the deviation and coefficient of variation: f Y|X (y|x)~N((1+bias)*x,(CV*x) 2 ) Step 2 and 3: In the absence of other errors, calculate the probability P1 that the normal person's test result exceeds the reference interval as: Step 24: In the absence of other errors, calculate the probability P2 that the normal person's test result exceeds the clinical decision limit as: Step 25: In the absence of other errors, the probability P3 of a normal person's test result exceeding the critical value (with both an upper limit and a lower limit) is calculated as:

3. The medical laboratory data quality control method based on maximum expectation according to claim 2, characterized in that: In step 3, the maximum expected increase in the number of unacceptable patients Max E (N UF ) is as follows: Max E(N UF )=MAX[ΔP*N af ] Where ΔP is the probability of the phase difference in the presence of a runaway error condition, N af is the number of people expected to be affected.

4. The medical laboratory data quality control method based on maximum expectation according to claim 3, characterized in that: The deviation in step 22 is calculated by Passing-Bablok regression using fifteen results from laboratories participating in the inter-laboratory quality assessment, specifically by constructing a y=b+ax regression equation and then estimating the difference between the slope a and 1.

5. The medical laboratory data quality control method based on maximum expectation according to claim 5, characterized in that: The coefficient of variation in step 22 is obtained through self-reporting by the laboratory.

Images