Influenza epidemic intensity grading early warning method

Abstract

The invention discloses a grading early warning method for influenza epidemic intensity, and belongs to the field of infectious disease monitoring and early warning. The method comprises the following steps: collecting related historical monitoring data of influenza, calculating corresponding indexes, taking the influenza report morbidity as a target sequence, drawing a time sequence comparison diagram of the influenza report morbidity and other indexes according to the indexes, and screening indexes capable of being used for monitoring and early warning by performing time sequence diagram comparison with other indexes and constructing lagging correlation analysis; according to the screened indexes, adopting a comprehensive evaluation method to calculate an influenza comprehensive index; and constructing a grading early-warning model of a mobile epidemic interval method based on the influenza comprehensive index, and calculating a grading early-warning threshold value of the influenza epidemic intensity. According to the method, the limitation of a single data source and the influence of a subjective judgment standard on an early warning result are effectively avoided, the influenza epidemic situation can be reflected more objectively and comprehensively, and a reliable basis is provided for formulating a scientific prevention and control strategy.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of infectious disease monitoring and early warning, and particularly relates to a method for influenza epidemic intensity grading early warning. BACKGROUND

[0002] Influenza is an acute respiratory infectious disease. About 3-5 million severe cases are caused by influenza every year, and about 290,000-650,000 deaths are caused by respiratory diseases, which is an infectious disease that is the focus of global public health monitoring. Because influenza has the characteristics of strong antigen variability and universal susceptibility of the population, once an epidemic occurs, the situation is often difficult to control. Therefore, it is crucial to explore an influenza grading early warning method to achieve early warning and grading disposal for effective prevention and control of influenza epidemic.

[0003] However, different influenza monitoring data have certain differences in timeliness, and the early warning capabilities are also different, so it is not possible to simply rely on one early warning data for early warning. At the same time, the current early warning model mostly uses subjective criteria when determining the epidemic threshold, which is greatly affected by human subjective factors, and may result in insufficient objectivity and reliability of the early warning results. SUMMARY

[0004] The purpose of the present application is to provide a method for influenza epidemic intensity grading early warning, which integrates multiple sources of influenza monitoring data, fully utilizes the advantages of various data, and makes up for the defects of using single data in existing early warning methods. The method uses the moving epidemic method to automatically determine the start of the influenza epidemic period and calculate the grading early warning threshold, so as to realize more objective and comprehensive grading early warning of influenza epidemic, and provide stronger decision support for influenza prevention and control.

[0005] The technical solution adopted by the present application is as follows: a method for influenza epidemic intensity grading early warning, the method comprising the following steps: Collecting influenza-related historical monitoring data and calculating corresponding indicators; Taking the influenza reported incidence rate as the target sequence, drawing a time sequence comparison graph of the influenza reported incidence rate and the indicators according to the indicators, constructing a lag correlation analysis by comparing the time sequence graphs with other indicators, and screening indicators that can be used for monitoring and early warning; According to the screened indicators, the screened indicators are standardized and weighted by a comprehensive evaluation method, and an influenza comprehensive index is calculated; A grading early warning model is constructed to calculate the influenza grading early warning threshold and capture the early warning signal.

[0006] It should be noted that the indicators include influenza-like case percentage, influenza positive detection rate, general index, prevention index, symptom index, and treatment index, etc.

[0007] Further, the influenza-related historical monitoring data includes: Influenza network direct report data, including influenza report incidence rate; Influenza sentinel monitoring data, including influenza-like case percentage, influenza positive detection rate; Baidu search index data, including general index, prevention index, symptom index and treatment index; Among them, the Baidu search index data is calculated according to the total search volume of search keywords in four aspects of influenza general monitoring, prevention, symptoms and treatment every week.

[0008] Further, the time series comparison chart includes two-time series comparison charts of influenza report incidence rate and influenza-like case percentage, influenza positive detection rate, general index, prevention index, symptom index and treatment index. The curve trend and peak position of the report incidence rate and various data are compared to preliminarily judge the correlation and difference.

[0009] Further, the lag correlation analysis is constructed by R language software to calculate the correlation of different indicators at different lag weeks, and the lag correlation coefficient and the best lag week number are comprehensively judged whether the indicator can be used for monitoring and early warning.

[0010] Further, the lag correlation analysis includes the following: Calculate the best lag week number: Set a maximum lag week number, and take the influenza report incidence rate as the target sequence. Within the maximum lag week number, respectively, the influenza-like case percentage, the influenza positive detection rate, and the general index, the prevention index, the symptom index and the treatment index are moved to the left and right with a step of 1, and the correlation analysis is done Spearman The lag week number when the lag correlation coefficient is maximum is the best lag week number.

[0011] Further, when the lag week number is positive, it means that the indicator sequence moves to the right relative to the target sequence, that is, it lags behind the influenza report incidence rate; when the lag week number is negative, it means that the indicator sequence leads the influenza report incidence rate; when the lag week number is 0, it means that the correlation coefficient is maximum when the indicator sequence does not move; Among them, the indicator sequence refers to the data sequence for lag correlation analysis with the influenza report incidence rate.

[0012] Further, the calculation of the influenza comprehensive index includes the following: According to the time series comparison and the lag correlation analysis, the selected indicators are determined by the expert consultation method to determine the weight coefficient of each indicator: ; Among them, and are the selected indicators, The actual values ​​of each selected indicator. The standard values ​​for each selected indicator, , These are the weighting coefficients for each indicator.

[0013] Furthermore, the standard values ​​for the indicators are calculated as follows: If the indicator data itself is normally distributed, then the influenza epidemic start threshold is used as the standard value of the indicator by using the moving epidemic interval method. If the indicator data itself is non-normally distributed, then the 75th percentile of the indicator data is taken as the standard value of the indicator.

[0014] Furthermore, a tiered early warning model is constructed. Using the historical influenza comprehensive index as the early warning baseline data, a moving epidemic interval method model is built to calculate the influenza epidemic start threshold and influenza tiered early warning threshold, namely, the medium, high and super high epidemic thresholds. Before each flu epidemic One-sided 95% of the arithmetic mean of the largest monitored value dataset CI The upper limit serves as the threshold for the onset of the influenza epidemic in the next influenza epidemic year, among which... , Number of years of influenza epidemic; During the flu season One-sided 40.0% of the geometric mean of the dataset with the largest monitored value. CI 90.0% CI and 97.5% CI Upper limits are set for three levels of prevalence: medium, high, and ultra-high.

[0015] It is important to note that CI The confidence interval is defined as a single side, indicating one direction, i.e., greater than.

[0016] Furthermore, the influenza composite index for this week will be compared with the influenza epidemic initiation threshold and the influenza classification warning threshold. If the influenza composite index reaches the influenza epidemic initiation threshold or the influenza classification warning threshold, the influenza composite index will be captured as a warning signal.

[0017] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: This invention compares time-series data from various influenza-related surveillance datasets, constructs lag correlation analysis to screen core indicators, and calculates a comprehensive influenza index based on these core indicators. Using this comprehensive index as a baseline, a tiered early warning model is established to achieve tiered early warning of influenza epidemic intensity. This method effectively avoids the limitations of single data sources and the influence of subjective judgment standards on early warning results, enabling a more objective and comprehensive reflection of the influenza epidemic situation and providing a reliable basis for formulating scientific prevention and control strategies. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a Baidu search keyword and its classification diagram for the present invention; Figure 3 This is a timing comparison diagram of the present invention. Detailed Implementation

[0019] The present invention will now be described in detail with reference to the accompanying drawings.

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0021] Currently, commonly used early warning data include the percentage of influenza-like illness cases and the influenza positivity rate. Among these, the influenza positivity rate is often used alone as an early warning indicator because it reflects the most reliable and specific influenza incidence level. However, the widespread use of Baidu search index in recent years has also provided new ideas for monitoring and early warning. These monitoring data have certain differences in timeliness and each has its own advantages in early warning capabilities. Therefore, using them in combination can improve the overall early warning effect.

[0022] Early warning models include the Autoregressive Moving Average (ARIMA) model, the Moving Percentile (MPM) method, the Cumulative Sum Control Chart (CUSUM) method, and artificial neural network models. Most of these models are based on time series methods and typically use subjective criteria to determine the epidemic threshold. The Moving Epidemic Interval (MEM) method determines the baseline of influenza activity based on historical influenza incidence data, uses automated methods to determine the start of the seasonal influenza epidemic, calculates the epidemic threshold, and assesses the intensity of influenza activity. It is one of the most widely used methods.

[0023] Therefore, this invention integrates various influenza-related monitoring data, selects suitable indicators, calculates an influenza comprehensive index through a comprehensive evaluation method, and establishes an influenza graded early warning model based on the influenza comprehensive index. This allows for a comprehensive early warning of the intensity of influenza epidemics, avoiding the limitations of single data and the influence of subjective standards on the early warning results. This enables a more objective and comprehensive implementation of influenza graded early warning, providing a basis for scientific prevention and control.

[0024] Example 1 like Figure 1 As shown, one embodiment of the present invention is a method for grading and issuing early warnings of influenza epidemic intensity, the method comprising the following steps: Step S100: Collect historical surveillance data related to influenza and calculate the corresponding indicators; Step S200: Draw a time-series comparison chart of influenza reported incidence rate and indicators based on the indicators; The influenza reported incidence rate was used as a benchmark value, and the influenza reported incidence rate was directly exported from the infectious disease surveillance and reporting system. Based on the curve trend and peak position of the influenza reported incidence rate and the indicator in the time series comparison chart, the correlation and differences between the influenza reported incidence rate and the indicator were determined. Step S300: Construct a lag correlation analysis, using the reported influenza incidence rate as the target sequence, calculate the lag correlation between the target sequence and the indicators, and screen indicators that can be used for monitoring and early warning. Step S400: Based on the selected indicators, standardize the selected indicators and set weights through a comprehensive evaluation method to calculate the influenza comprehensive index; Step S500: Construct a graded early warning model, calculate the influenza graded early warning threshold, and capture early warning signals.

[0025] This embodiment integrates historical influenza-related monitoring data and corresponding indicators, analyzes the lag correlation of the indicators, selects indicators for monitoring and early warning, and then calculates the influenza comprehensive index through a comprehensive evaluation method, thereby constructing a graded early warning model.

[0026] Example 2 Another embodiment of the present invention includes influenza-related historical surveillance data including: Influenza online reporting data, including reported influenza incidence rates; Influenza sentinel surveillance data, including the percentage of influenza-like illness cases and the influenza positivity rate; Baidu search index data includes general index, prevention index, symptom index, and treatment index; The Baidu search index data is calculated based on the total weekly search volume of keywords related to general influenza surveillance, prevention, symptoms, and treatment. Influenza online reporting data, influenza sentinel surveillance data, and Baidu search index data are all compiled on a weekly basis.

[0027] Taking historical influenza surveillance data and indicators from 2019 to 2024 in Suining City as an example: The influenza network reporting data comes from the "Infectious Disease Surveillance and Reporting System" subsystem of the "China Disease Prevention and Control Information System". The influenza reported incidence rate in Suining City is statistically analyzed on a weekly basis according to the statistical rules of "date of onset" and "current address". The influenza sentinel surveillance data comes from the "China Influenza Surveillance Information System," a subsystem of the "China Disease Prevention and Control Information System." Baidu search index data is calculated based on the total weekly search volume of keywords related to influenza general monitoring, prevention, symptoms, and treatment. Influenza-related Baidu search keywords and their classifications are as follows: Figure 2As shown, Figure 2 The data is derived by calculating the total weekly search volume.

[0028] Example 3 like Figure 3 As shown, another embodiment of the present invention is that the time-series comparison chart includes pairwise time-series comparison charts of influenza reported incidence rate and influenza-like illness percentage (ILI%), influenza positive detection rate (PR), general index, prevention index, symptom index, and treatment index. By comparing the curve trends and peak positions of the reported incidence rate with various data, the correlation and differences can be preliminarily determined.

[0029] By plotting time-series comparison charts, we can obtain the trends of various indicators and the reported incidence rate of influenza. By judging the correlation and differences between the reported incidence rate of influenza and the indicators, we can draw the following conclusions: The influenza epidemic in Suining City mainly shows a single peak in winter and spring. The overall trend and peak position of the influenza positive detection rate, symptom index, treatment index and reported incidence rate are similar.

[0030] Example 4 Another embodiment of the present invention involves constructing a lag correlation analysis using R 4.3.3 software. The correlation of different indicators under different lag periods is calculated, and the lag correlation coefficient and the optimal number of lag periods are used to comprehensively determine whether the indicator can be used for monitoring and early warning. The calculated optimal lag period and lag correlation coefficient are then incorporated into the early warning indicators. Furthermore, the lag correlation coefficient and the optimal number of lag periods are used to comprehensively determine whether the indicator can be used for monitoring and early warning. R is an open-source and free programming language and software environment.

[0031] Analysis results: The positive detection rate of influenza, symptom index, treatment index and the reported incidence rate of influenza showed a positive correlation that led the reported incidence rate for many years, which once again confirmed that these indicators can reflect early trend changes of influenza and are suitable for early warning of influenza epidemics.

[0032] Example 5 Another embodiment of the present invention is that the principle of lag correlation analysis is as follows: (1) Calculate the optimal lag period: Set a maximum lag period, take the reported influenza incidence rate as the target sequence, and within the maximum lag period, shift it left and right by a step size of 1 relative to the percentage of influenza-like cases, the positive detection rate of influenza, and the general index, prevention index, symptom index, and treatment index, and perform... Spearman Correlation analysis was performed, and the optimal lag number was determined by the lag number at which the lag correlation coefficient was maximized.

[0033] in, Spearman Correlation analysis uses monotonic equations to evaluate the correlation between two statistical variables.

[0034] (2) Spearman Correlation analysis is a nonparametric statistical method used to measure whether there is a correlation (statistic) between the trends of two variables. P <0.05), and judge the direction and strength of the correlation. The closer the absolute value of the correlation coefficient is to 1, the stronger the correlation. A positive number represents a positive correlation, and a negative number represents a negative correlation.

[0035] (3) A positive number of lag weeks indicates that the index sequence has shifted to the right relative to the target sequence, that is, it lags behind the reported incidence rate of influenza. Similarly, a negative number indicates that the sequence leads the reported incidence rate. 0 indicates that the correlation coefficient is the largest when there is no shift.

[0036] The indicator sequence is an interpretation of the optimal lag number sign in lag correlation analysis, referring to the data sequence that has undergone lag correlation analysis with the reported incidence rate of influenza.

[0037] Using R 4.3.3 software, a lag correlation analysis was constructed to calculate the correlation of different monitoring indicators under different lag periods. The optimal lag period and lag correlation coefficient were then incorporated into the early warning indicators.

[0038] Example 6 Another embodiment of the present invention is that the calculation of the influenza composite index includes the following: The weighting coefficients of each indicator were determined through expert consultation based on the indicators selected by time series plot comparison and lag correlation analysis. ; in, , and These represent the influenza positive detection rate, symptom index, and treatment index, respectively. for , and The actual value, for , and The standard value is defined as the 75th percentile (P75) of each indicator.

[0039] Example 7 Another embodiment of the present invention is to construct a graded early warning model, using the historical influenza comprehensive index as the early warning baseline data, constructing a mobile epidemic interval method MEM model, and calculating the influenza epidemic start threshold and influenza graded early warning threshold, namely, the medium, high and super high three-level epidemic thresholds. Before each flu epidemic One-sided 95% of the arithmetic mean of the largest monitored value dataset CI The upper limit serves as the threshold for the onset of the influenza epidemic in the next influenza epidemic year, among which... , Number of years of influenza epidemic; During the flu season One-sided 40.0% of the geometric mean of the dataset with the largest monitored value. CI 90.0% CI and 97.5% CI Upper limits are set for three levels of prevalence: medium, high, and ultra-high.

[0040] Based on the calculation of the weekly comprehensive index from 2019 to 2024, and considering that influenza in Suining City mainly exhibits a single peak in winter and spring, the period from the 26th week of each year to the 25th week of the following year is artificially divided into one epidemic year. Using the weekly comprehensive index of the five epidemic years from the 26th week of 2019 to the 25th week of 2024 as baseline data, a graded early warning model based on the moving epidemic interval method is constructed to calculate the influenza graded early warning threshold. (1) Sensitivity, specificity, positive predictive value, negative predictive value, positive likelihood ratio, negative likelihood ratio, Matthew correlation coefficient, and Youden index were calculated using the trial-and-error method to screen parameters. delta Select the optimal parameters.

[0041] (2) Use the “memgoodness” function to perform cross-validation on the baseline data, determine the stability of historical data, evaluate the model fit, remove outliers, and determine the data to be included. (3) After removing the poor-quality data, repeat the above parameters. delta The screening and fitting effect evaluation steps were carried out, and the accuracy of the MEM model was verified by cross-validation. (4) Use the “memmodel” function to calculate the influenza epidemic start threshold and the three epidemic levels of medium, high and super high.

[0042] Through calculation and verification using the above methods, the influenza epidemic initiation threshold and the three epidemic levels (medium, high, and super high) are obtained. The influenza comprehensive index is compared with the influenza epidemic initiation threshold and the influenza graded warning threshold. If the influenza comprehensive index reaches the influenza epidemic initiation threshold or the influenza graded warning threshold, the influenza comprehensive index is captured as a warning signal.

[0043] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for graded early warning of influenza epidemic intensity, characterized in that, The methods include the following: Collect historical influenza-related surveillance data and calculate corresponding indicators; Using the reported influenza incidence rate as the target sequence, a time series comparison chart of the reported influenza incidence rate and the indicator was drawn. By comparing the time series charts with other indicators and constructing a lag correlation analysis, indicators that can be used for monitoring and early warning were screened. Based on the selected indicators, the selected indicators are standardized and weighted through a comprehensive evaluation method to calculate the influenza comprehensive index. A tiered early warning model was constructed to calculate the tiered early warning thresholds for influenza and to capture early warning signals.

2. The method for graded early warning of influenza epidemic intensity according to claim 1, characterized in that, Historical influenza surveillance data includes: Influenza online reporting data, including reported influenza incidence rates; Influenza sentinel surveillance data, including the percentage of influenza-like illness cases and the influenza positivity rate; Baidu search index data includes general index, prevention index, symptom index, and treatment index; The Baidu search index data is calculated based on the total weekly search volume of keywords related to influenza general monitoring, prevention, symptoms, and treatment.

3. The method for graded early warning of influenza epidemic intensity according to claim 2, characterized in that, The time-series comparison charts include pairwise time-series comparisons of the reported incidence rate of influenza and the percentage of influenza-like cases, the positive detection rate of influenza, the general index, the prevention index, the symptom index, and the treatment index. By comparing the curve trends and peak positions of the reported incidence rate with various data, the correlation and differences can be preliminarily judged.

4. The method for graded early warning of influenza epidemic intensity according to claim 3, characterized in that, The lag correlation analysis was constructed using R language software to calculate the correlation of different indicators under different lag periods. Based on the lag correlation coefficient and the optimal number of lag periods, a comprehensive judgment was made on whether the indicator could be used for monitoring and early warning.

5. The method for graded early warning of influenza epidemic intensity according to claim 4, characterized in that, Lag correlation analysis includes the following: Calculate the optimal lag period: Set a maximum lag period, using the reported influenza incidence rate as the target sequence, and within the maximum lag period, shift it left and right by a step size of 1 relative to the percentage of influenza-like cases, the influenza positive detection rate, and the general index, prevention index, symptom index, and treatment index. Spearman Correlation analysis was performed, and the optimal lag number was determined by the lag number at which the lag correlation coefficient was maximized.

6. The method for graded early warning of influenza epidemic intensity according to claim 5, characterized in that, A positive lag number indicates that the indicator sequence has shifted to the right relative to the target sequence, i.e., it lags behind the reported influenza incidence rate; a negative lag number indicates that the indicator sequence leads the reported influenza incidence rate. When the lag period is 0, it means that the correlation coefficient is the largest when the index series does not shift. Among them, the indicator sequence refers to the data sequence that has undergone a lag correlation analysis with the reported incidence rate of influenza.

7. The method for graded early warning of influenza epidemic intensity according to claim 6, characterized in that, The calculation of the influenza composite index includes the following: The weighting coefficients of each indicator were determined through expert consultation based on the indicators selected by time series plot comparison and lag correlation analysis. ； in, and These are the selected indicators. The actual values ​​of each selected indicator. The standard values ​​for each selected indicator, , These are the weighting coefficients for each indicator.

8. The method for graded early warning of influenza epidemic intensity according to claim 7, characterized in that, The standard values ​​for the indicators are calculated as follows: If the indicator data itself is normally distributed, then the influenza epidemic start threshold is used as the standard value of the indicator by using the moving epidemic interval method. If the indicator data itself is non-normally distributed, then the 75th percentile of the indicator data is taken as the standard value of the indicator.

9. The method for graded early warning of influenza epidemic intensity according to claim 8, characterized in that, A tiered early warning model was constructed, using historical influenza composite index as the early warning baseline data. A mobile epidemic interval method model was built to calculate the influenza epidemic initiation threshold and influenza tiered early warning threshold, namely, the medium, high and super high epidemic thresholds. Before each flu epidemic One-sided 95% of the arithmetic mean of the largest monitored value dataset CI The upper limit serves as the threshold for the onset of the influenza epidemic in the next influenza epidemic year, among which... , Number of years of influenza epidemic; During the flu season The one-sided geometric mean of the dataset with the largest monitored value is 40.0%. CI 90.0% CI and 97.5% CI Upper limits are set for three levels of prevalence: medium, high, and ultra-high.

10. The method for graded early warning of influenza epidemic intensity according to claim 9, characterized in that, The influenza composite index for this week is compared with the influenza epidemic initiation threshold and the influenza graded warning threshold. If the influenza composite index reaches the influenza epidemic initiation threshold or the influenza graded warning threshold, the influenza composite index is captured as a warning signal.

Images