Method for constructing antibiotic use intensity parameter and storage medium
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INST OF URBAN ENVIRONMENT CHINESE ACAD OF SCI
- Filing Date
- 2026-04-17
- Publication Date
- 2026-08-07
AI Technical Summary
1）校准过程通常以数值最优为目标，缺乏明确的物理或结构解释；
[0011]本发明的有益效果在于：通过根据可提供绝对量级锚定的区域典型水平、可提供相对结构信息的PCU强度数据以及可补充区域内部差异性的生产规模，获取完整的国家-物种的先验强度，可体现各区域生产体系特征，尽量符合真实的生产用药强度；通过对先验强度和生产规模进行加权计算，保证国家整体强度水平与国家级平均强度一致；通过计算缩放系数并根据缩放系数对先验强度进行缩放，从而对先验强度进行校准。本发明生成的强度参数可同时满足准确性、稳定性与可解释性。
Smart Images
Figure CN122531795A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, and in particular to a method for constructing antibiotic use intensity parameters and a storage medium. Background Technology
[0002] When conducting veterinary antibiotic usage accounting at the global or national scale, the core technical foundation lies in constructing a parameter system for antibiotic use intensity at the national × species level. These parameters typically serve as key inputs to upstream production system models or downstream public health risk assessment models, and their accuracy and stability directly impact the overall calculation results. Due to differences in the level of monitoring system development, currently only a few countries (mainly concentrated in Europe and North America) are able to provide long-term, continuous direct observational data on national-level veterinary antibiotic use intensity. Many countries have long been in a state of data gaps or fragmented information at the national scale.
[0003] To address the aforementioned issue of incomplete data, existing research (such as that by van Boeckel et al., Schar et al., Tiseo et al., and Mulchandani et al.) has primarily developed the following technical implementation paths: 1. Directly use antibiotic use intensity data reported by national statistical departments, regulatory agencies, or literature as input parameters. The advantage is that the data source is clear, but due to incomplete coverage, only partial predictions can be made.
[0004] 2. Using regional-scale statistical results or model outputs as a basis, the regional average intensity value is extrapolated to each country within the region. This can achieve full numerical coverage, but it ignores differences in production structures and management among countries, and is prone to introducing systemic structural biases.
[0005] 3. By using regression analysis, global fitting, or optimization algorithms, the model derivation results are uniformly scaled to make them numerically close to the external reference strength. The advantage is that it can align data from different sources on a global scale and offers a degree of flexibility. However, it has the following disadvantages: 1) The calibration process usually aims at numerical optimization and lacks a clear physical or structural explanation; 2) The calibration process can easily overwrite the original structural relationships, destroying the internal consistency of the model; 3) When the number of samples in a region is small, overfitting or unstable adjustment results are very likely to occur.
[0006] In summary, given the reality of inconsistent sources of multi-source intensity information and extremely uneven distribution of evidence within a region, existing technologies generally suffer from the following problems: 1) It is impossible to perform strength calibration while maintaining the original production structure; 2) It is impossible to distinguish the differences in the reliability of intensity information from different countries or regions during the calibration process; 3) The calibration logic is mostly numerical alignment-oriented, lacking traceable and interpretable adjustment basis.
[0007] Therefore, existing technologies are unable to simultaneously meet the three key technical requirements for constructing global-scale strength parameters: accuracy, stability, and interpretability. Summary of the Invention
[0008] The technical problem to be solved by the present invention is to provide a method and storage medium for constructing antibiotic strength parameters, which can generate accurate, stable and interpretable national-level strength parameters.
[0009] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for constructing antibiotic efficacy parameters, comprising: Based on regional typical intensity data, intensity data based on population-corrected units, and the production scale of each species in each country, the a priori intensity of antibiotics for each species in each country is determined. Calculate the prior weighted strength of each country based on the prior strength and production scale of antibiotics for each species in each country. Based on the national average intensity observation values, determine the national average intensity of each country, and based on the national average intensity of each country, determine the target intensity of each country. Calculate the scaling factor for each country based on its prior weighted strength and target strength; The antibiotic use intensity for each species in each country is calculated based on the prior antibiotic strength for each country and the scaling factor for each country.
[0010] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.
[0011] The beneficial effects of this invention are as follows: By obtaining complete national-species prior intensity based on typical regional levels that provide absolute magnitude anchoring, PCU intensity data that provides relative structural information, and production scale that supplements regional differences, it can reflect the characteristics of each region's production system and closely match the actual drug use intensity in production; by weighting the prior intensity and production scale, it ensures that the overall national intensity level is consistent with the national average intensity; by calculating a scaling factor and scaling the prior intensity accordingly, the prior intensity is calibrated. The intensity parameters generated by this invention simultaneously satisfy accuracy, stability, and interpretability. Attached Figure Description
[0012] Figure 1 This is a flowchart of the method for constructing antibiotic use intensity parameters according to Embodiment 1 of the present invention. Detailed Implementation
[0013] To explain in detail the technical content, objectives, and effects of the present invention, the following description is provided in conjunction with the embodiments and accompanying drawings.
[0014] In global agricultural production analysis, public health risk assessment, and resource use intensity accounting systems, it is usually necessary to construct use intensity parameters for different countries and different aquaculture categories under a unified calculation framework to support cross-country and cross-regional comparative analysis, scenario simulation, and model calculation.
[0015] In the actual operation of the aforementioned system, the construction of national-level intensity parameters is not a simple data aggregation problem, but a comprehensive computational problem facing multiple technical constraints within the computer system. These constraints include: highly inconsistent sources of national-level intensity data, with some countries possessing direct observational or documentary data, while others can only indirectly derive it through regional statistics or model results; a systematic deviation exists between the national yield structure output by the production system model and the actual biomass structure, which, if directly used for intensity calculation, can easily lead to structural distortion; significant differences in the number of available national-level observations across different regions, which, if a uniform regional calibration strategy is applied, can easily lead to over-adjustment of low-sample regions; and the superposition of multi-level regional calibration results may cause total offsets on a macro scale, affecting the overall consistency of the system.
[0016] Therefore, in this application scenario, there is an urgent need for a technical solution that can be automatically executed within a computer system without manual intervention on a country-by-country basis, and can generate stable, interpretable, and auditable national-level intensity parameters under conditions of incomplete multi-source data, in order to meet the reliability requirements of the global agricultural and public health assessment system.
[0017] To address at least the aforementioned technical problems, this disclosure provides a method for constructing antibiotic use strength parameters, comprising: Based on regional typical intensity data, intensity data based on population-corrected units, and the production scale of each species in each country, the a priori intensity of antibiotics for each species in each country is determined. Calculate the prior weighted strength of each country based on the prior strength and production scale of antibiotics for each species in each country. Based on the national average intensity observation values, determine the national average intensity of each country, and based on the national average intensity of each country, determine the target intensity of each country. Calculate the scaling factor for each country based on its prior weighted strength and target strength; The antibiotic use intensity for each species in each country is calculated based on the prior antibiotic strength for each country and the scaling factor for each country.
[0018] As can be seen from the above description, the beneficial effects of the present invention are that it can generate accurate, stable, and interpretable national-level strength parameters.
[0019] Furthermore, determining the prior antibiotic strength for each species in each country based on typical regional strength data, strength data based on population-corrected units, and the production scale of each species in each country includes: Based on the typical intensity levels of each species in each region of each country, determine the intensity center value for each species in each country. Based on intensity data with population-corrected units, the relative offset intensity for each species in each country is determined. Based on the production scale of each species in each country, perturbations corresponding to each species in each country are generated. The prior strength of antibiotics for each species in each country is determined based on the intensity center value, relative offset intensity, and perturbation for each species.
[0020] As described above, the regional typical level provides an absolute magnitude anchor, the PCU intensity data provides relative structural information, the production scale supplements the regional differences, and the complete country-species intensity prior distribution can be obtained through repeated sampling perturbation terms.
[0021] Furthermore, determining the relative offset intensity for each species in each country based on intensity data with population correction units includes: Based on the PCU intensity data of each country in each region corresponding to a species, the baseline drug use intensity for each species in each region is determined. The PCU intensity data is intensity data based on population correction units. The relative offset intensity of a country for a species is determined based on the PCU intensity data of a country for the species and the baseline drug use intensity of the region to which the country belongs for the species.
[0022] As described above, intensity data based on population-corrected units (PCUs) have good systematicity and cross-national comparability. Therefore, it can be used to extract cross-country comparable relative strength relationships, relative ranking relationships between countries, and structural differences of different species in the same country.
[0023] Furthermore, determining the relative offset intensity of each species for each country based on intensity data with population correction units also includes: Determine the PCU intensity data filling ratio for each species in each region; Based on the filling ratio of PCU intensity data for each species in each country's region, the reliability index of PCU intensity data for each species in each country is determined. Based on the reliability index of PCU intensity data for each species in each country and the preset first pruning threshold, extreme value pruning and screening are performed on the relative offset intensity for each species in each country to determine the final relative offset intensity for each species in each country.
[0024] Further, the step of performing extreme value pruning and filtering on the relative offset intensity of each country for each species based on the reliability index of PCU intensity data for each country and a preset first pruning threshold, to determine the final relative offset intensity of each country for each species, includes: Based on the PCU intensity data of the same species in each country within the same region, calculate the logarithmic standard deviation of the same species in the same region; Based on the logarithmic standard deviation of the same region corresponding to the same species, a first pruning threshold for the same region corresponding to the same species is determined; Based on the first pruning threshold corresponding to a species in a country, the relative offset intensity of the species in a country is pruned to extreme values. Then, based on the reliability index of the PCU intensity data of the species in a country and the pruned relative offset intensity, the final relative offset intensity of the species in a country is determined.
[0025] As described above, since PCU data is not entirely reliable (some countries are filled with mean values, and some species have low coverage), a reliability index is constructed. This means that PCU bias is only allowed to be included in the model when the national data is accurate and the overall species coverage is low. Extreme value pruning can avoid abnormal fluctuations.
[0026] Furthermore, the generation of perturbations for each species in each country based on their respective production scale includes: Within each country's territory, calculate the production scale quantile for each species in each country, based on the production scale of each country. Based on the PCU intensity data of the same species in each country within the same region, the logarithmic standard deviation of the same species in the same region is calculated, and based on the logarithmic standard deviation of the same species in each region, the maximum allowable fluctuation range and the minimum allowable fluctuation range of the same species are determined. Based on the production scale quantiles for each species in each country and the maximum and minimum permissible fluctuation ranges for each species, calculate the logarithmic standard deviation for each species in each country. Based on the reliability index of PCU intensity data for each species in each country, the indicator variables for each species in each country were determined. The perturbations for each species in each country are generated based on the logarithmic standard deviation, indicator variables, and a preset second pruning threshold.
[0027] As described above, by introducing heteroscedasticity based on production scale, excessive homogeneity of national strengths within a region is avoided when PCU information is insufficient. When PCU structural information is insufficient or unreliable, reasonable heterogeneity among countries is generated by introducing random perturbations. Indicator variables are set to prevent deviations in samples with reliable PCU data from being influenced by production scale. Prior strengths at the country-species level are generated by constructing conditional probability distributions.
[0028] In some embodiments, to ensure monotonicity and interpretability, the logarithmic standard deviation of each species for each country can be calculated using linear interpolation.
[0029] Furthermore, the step of generating perturbations for each species in each country based on the logarithmic standard deviation, indicator variable, and preset second pruning threshold includes: The second pruning threshold for a country corresponding to a species is determined based on the logarithmic standard deviation of a country to a species. Based on the logarithmic standard deviation of the species corresponding to the country, the indicator variable, and the second pruning threshold, a perturbation for the species corresponding to the country is generated.
[0030] As can be seen from the above description, abnormal fluctuation ranges can be avoided by setting a clipping threshold.
[0031] Furthermore, the process of determining the national average intensity of each country based on national average intensity observations, and determining the target intensity of each country based on its national average intensity, includes: If a country has a national average intensity observation value, then the national average intensity observation value of that country shall be used as the national average intensity and target intensity of that country. If a country does not have a national average intensity observation value, the median of the national average intensity of the countries within the region to which the country belongs that have national average intensity observation values is taken as the national average intensity of the country. The national average intensity of the country is then corrected according to a preset correction intensity parameter and a maximum correction range to obtain the target intensity of the country.
[0032] As described above, the principle of prioritizing national observations is adopted. If a national average intensity observation exists, it is directly used as the national average intensity and target intensity. If no national average intensity observation exists, the median of the established national average intensity within the region is used as a filler. Compared to the mean, which is easily affected by countries with extremely high intensity, the median is more robust when a long-tailed distribution exists and is more consistent with the statistical assumptions of subsequent logarithmic distribution and uncertainty analysis. By correcting the national average intensity filled by the median, the relative strength structure among countries can be restored.
[0033] Further, the step of correcting the national average intensity of a country based on preset correction intensity parameters and maximum correction magnitude to obtain the target intensity of the country includes: Obtain the log-median of the PCU national strength for each country, where the PCU national strength is the national average strength based on population-corrected units; The logarithmic shift strength of a country is determined based on the country strength of the PCU and the logarithmic median. Based on the preset correction intensity parameters, the maximum correction magnitude, and the logarithmic offset intensity of the country, the national average intensity of the country is corrected in logarithmic space to obtain the target intensity of the country.
[0034] As described above, by introducing PCU-based national average intensity data as a ranking signal, the national average intensity filled with median is scaled by a limited proportion in the logarithmic space, thereby restoring the relative ranking differences between countries without changing the overall order of magnitude of the intensity.
[0035] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.
[0036] Example 1 Please refer to Figure 1 The first embodiment of the present invention is: a method for constructing antibiotic use intensity parameters, specifically a method for constructing and calibrating national and regional scale intensity parameters based on multi-source heterogeneous data fusion and consistency constraints, which can be applied to scenarios such as agricultural production system modeling, public health risk assessment, and multi-regional statistical parameter configuration.
[0037] This method models intensity as a structured probability distribution formed around the characteristics of the regional production system, such as... Figure 1 As shown, it includes the following steps: S1: Generate the prior strength of antibiotics for each species in each country.
[0038] To obtain the national-species-level prior strength of antibiotics and to reflect the characteristics of production systems in various regions and to conform as closely as possible to the actual production and drug use intensity, this step utilizes the following three types of structural information to form the research approach: regional typical levels provide absolute magnitude anchoring, PCU (population-corrected unit) intensity data provide relative structural information, production scale supplements the internal differences within regions, and through repeated sampling perturbation terms, the complete national-species-level prior distribution of strength can be obtained.
[0039] Specifically, this step includes the following steps: S101: Determine the intensity center value for each species in each country based on the typical intensity level of each species in each region.
[0040] Globally, animal production systems exhibit distinct regional characteristics (technological level, regulatory intensity, disease spectrum, production density, etc.). Even without national monitoring data, it is reasonable to assume that the intensity of a country's production system within the same region should fluctuate around a typical production system level. Therefore, by integrating international reports and monitoring networks (ESVAC, FDA, CIPARS, etc.), the typical intensity level μ of species k (r) in each region can be collected. r,k And regard it as the central value of all countries in the region: Center i,k :=logμ r(i),k Where, := is the operator used for variable assignment, logμ r(i),k The base is e, i.e., lnμ r(i),k .
[0041] S102: Determine the relative offset intensity for each species in each country based on intensity data with population correction units.
[0042] The Population Correction Unit (PCU) intensity data (hereinafter referred to as PCU intensity data) is a high-coverage antibiotic intensity database obtained by scholars in this field, such as van Boeckel, through linear extrapolation based on monitoring data from a few countries. Since the coverage of current national monitoring data is limited, and this dataset has good systematicity and cross-national comparability, it is possible to extract cross-national comparable relative strength relationships from it, such as which countries or species use antibiotics more intensively, the relative ranking relationships between countries, and the structural differences of different species within the same country.
[0043] Specifically, first determine the relative shifts in pesticide application intensity for each country and species. For different regions (r) and different species (k), define the median as the baseline central value of pesticide application intensity for the corresponding species in that region:
[0044] The PCU structure shift for species k corresponding to country i, which represents how much higher / lower the drug use intensity of that species in that country is relative to the median level in the region, i.e., the relative shift intensity, is:
[0045] The `median()` function returns the median of a given set of numbers. The median is the middle value in a set of numbers sorted by size; if the number of numbers is even, the median is the average of the two middle values. This represents the PCU intensity data for species k corresponding to country i. This represents the baseline drug use intensity center value for species k corresponding to region r to which country i belongs.
[0046] Furthermore, since PCU intensity data is not entirely reliable (some countries use mean-filled data, and some species have low coverage), a reliability index needs to be constructed. First, the fill ratio of the PCU intensity data is calculated:
[0047] Among them, F j,k =1 indicates that the PCU value of species k in country j is padded with the mean, N r,k Let r be the number of countries participating in the calculation for species k within region r (including countries with true values and those with filler values). Then, the reliability index of PCU intensity data can be defined as:
[0048] This means that PCU offsets are only allowed to be included in the model when the national data is accurate and the overall species fill rate is low.
[0049] To avoid abnormal fluctuations, a pruning threshold τ is set for species k corresponding to region r. r,k =2σ r,k , where σ r,k Let represent the logarithmic standard deviation of species k corresponding to region r, and its calculation formula is:
[0050] std() is used to calculate the standard deviation.
[0051] In summary, the relative offset intensity of species k corresponding to country i, provided by PCU intensity data, can be defined as:
[0052] in, This represents the reliability index of PCU intensity data for species k corresponding to country i. The `clip()` function is used to restrict the elements in an array to between a specified minimum and maximum value; elements exceeding the minimum value are replaced with the minimum value, and elements exceeding the maximum value are replaced with the maximum value. In this embodiment, the PCU structure is offset. Limited to -τ r(i),k and τ r(i),k Between. λ k The PCU structure offset weight coefficient for species k represents the degree of trust in the relative PCU ordination signal for that species. These species-level weight parameters are manually set and adjustable, derived from prior judgments about the quality, coverage, and reliability of PCU data for each species.
[0053] The reliability of PCU intensity data varies significantly across different species. According to van Boeckel et al., the data reliability is: pigs > poultry > ruminants > fish. Therefore, the degree of trust in PCU ordination signals differs among species. In some embodiments, the PCU structure offset weighting coefficients for each species can be set as: λ 猪 =1, λ 家禽 =0.4, λ 反刍 =0.2, λ 鱼 =0.
[0054] S103: Generate perturbations for each species in each country based on the production scale of each species in each country.
[0055] To avoid excessive homogenization of national intensity within a region when PCU information is insufficient, this embodiment introduces a heteroscedasticity setting based on production scale. Countries with larger production scales tend to have more diverse internal production systems (e.g., different density levels, management models, and disease exposure conditions), and therefore their national average intensity may exhibit greater structural fluctuations. To reflect this reality, the logarithmic standard deviation of intensity perturbations is allowed to increase monotonically with the country's size quantile within the region, giving large producing countries a statistically wider range of possible values.
[0056] Specifically, this step includes the following steps: S1031: Calculate the production scale quantiles of each country within the region, that is, within the region to which each country belongs, calculate the production scale quantiles of each country for each species based on the production scale of each country.
[0057] Specifically, for country i and species k, the quantile of its production scale within the region is:
[0058] Among them, Q j,kThis represents the production scale of species k corresponding to country j. `rankpct()` represents the relative rank quantile function, used to calculate the relative position of the target element in a given sorted set; the closer the value is to 1, the higher the corresponding production scale within its region, and the closer the value is to 0, the lower the corresponding production scale within its region.
[0059] S1032: Define the logarithmic standard deviation, which is to calculate the logarithmic standard deviation for each species in each country.
[0060] When PCU structural information is insufficient or unreliable, random perturbations are introduced to generate reasonable heterogeneity among countries. However, if the same perturbation standard deviation is applied to all countries, the fluctuation amplitudes of countries within the region will be completely uniform, failing to reflect the structural complexity brought about by differences in production scale. Generally, the larger the country size, the greater the uncertainty and potential deviation of its national average strength should be. To characterize this phenomenon, the perturbation standard deviation of the logarithmic space is set as a function of the national production scale quantiles, so that larger countries have a wider prior distribution. If the standard deviation is directly amplified based solely on the size proportion, it may lead to perturbations of small-scale countries approaching zero (overdeterminacy) and perturbations of large-scale countries expanding infinitely (distribution divergence). Therefore, the minimum and maximum allowable fluctuation amplitudes of species k are set as follows:
[0061]
[0062] Where, σ r,k This represents the log-standard deviation of species k corresponding to region r; median r (σ r,k σ represents the median of the logarithmic standard deviation for species k across all regions, i.e., the logarithmic standard deviation σ for species k across all regions. r,k The median. Using the above method, a bounded range of permissible fluctuation for species k can be constructed based on the median fluctuation level of the corresponding species in each region. This avoids setting the fluctuation range too small, resulting in insufficient characterization of differences between countries, or setting the fluctuation range too large, resulting in instability in the intensity distribution.
[0063] To ensure monotonicity and interpretability, this embodiment uses linear interpolation for definition:
[0064] Where, σ i,k q represents the logarithmic standard deviation of species k corresponding to country i; i,k This represents the production scale quantile of species k corresponding to country i.
[0065] S1033: Perturbation term gating, which generates perturbations for each species corresponding to each country.
[0066] To avoid bias in samples with reliable PCU data due to production scale, complementary gates need to be set: when PCU is reliable, PCU is used to provide structural differences between country and species; when PCU is unreliable / missing, production scale is used instead to create (reasonable) inter-country heterogeneity, defining indicator variables:
[0067] in, This represents the reliability index of PCU intensity data for species k corresponding to country i.
[0068] To avoid abnormal fluctuations, set a pruning threshold κ. i,k =2σ i,k Generate disturbances:
[0069] S104: Determine the prior strength of antibiotics for each species in each country based on the intensity center value, relative offset intensity, and perturbation for each species in each country.
[0070] Specifically, the intensity prior at the national-species level is generated by constructing a conditional probability distribution:
[0071] Right now
[0072] Where, logμ r(i),k The intensity center value of species k corresponding to country i in region r obtained in step S101; The relative offset intensity of species k corresponding to country i obtained in step S102; ε i,k This is the perturbation of species k corresponding to country i obtained in step S103.
[0073] S2: Calculate the prior weighted strength of each country based on the prior strength and production scale of antibiotics for each species in each country.
[0074] The formula for calculating the prior weighting strength is:
[0075] in, Q represents the prior strength of antibiotics for species k corresponding to country i; i,k This represents the production scale of species k corresponding to country i.
[0076] S3: Based on the national average intensity observation values, determine the national average intensity of each country, and based on the national average intensity of each country, determine the target intensity of each country.
[0077] After obtaining the prior antibiotic strength for each species in each country and the weighted aggregation structure based on production scale, a core problem remains: the strength of all species can be raised or lowered as a whole while maintaining the relative structure, resulting in a systematic deviation between the national average strength and the target strength. To address this, a national average strength prior is introduced. This layer does not concern itself with specific species but rather imposes constraints on the weighted strength. This step primarily utilizes the national average strength data published by WOAH.
[0078] To construct national-level intensity anchors, this embodiment employs a structured construction method of "layered priority + robust fallback + soft correction based on sorting information" to generate national-level average intensity data. The WOAH has published national-level average intensities for some countries, as well as regional distribution data.
[0079] (1) Priority principle of national observations If country i has a national-level average intensity observation value Then it is directly used as the national average intensity Abar i ,Right now:
[0080] And the source is indicated: .
[0081] (2) Regional median supplement If country i has no observations, then the median of the national average intensity of countries within its region that have national observations (i.e., countries whose intensity has been determined) is used as the filler:
[0082] And the source is indicated:
[0083] The median is used instead of the mean because the mean is easily affected by countries with extremely high intensity, while the median is more robust when there is a long-tailed distribution and is more in line with the statistical assumptions of the subsequent logarithmic distribution and uncertainty analysis.
[0084] (3) Correction based on PCU data For countries lacking national observations and using regional medians as intensity anchors, directly using regional values would result in identical intensity across countries within the same region, artificially smoothing out the true differences between countries. To restore the relative strength structure among countries, this embodiment introduces national average intensity data based on the PCU from Thomas et al. (2017) as a sorting signal.
[0085] First, define the logarithmic offset of the PCU for each country:
[0086] in, This represents the logarithmic shift in intensity for country i, indicating whether the country's PCU intensity is high or low relative to the global median. PCU i Let represent the PCU national strength of country i, and median(logPCU) represent the log median of the PCU national strength of all countries globally.
[0087] For countries whose data source is a regional catch-all (i.e., countries without national observations), the target intensity is obtained by scaling in logarithmic space:
[0088] in, For the target intensity of country i; logAbar i Let be the national average intensity of country i; α is the intensity correction parameter, controlling the absorption ratio of the ranking signal. In this embodiment, α = 0.3. MAX_SHIFT is the maximum correction magnitude, used to limit the influence of extreme countries. This correction is equivalent to scaling the regional intensity by a limited ratio, thereby restoring the relative ranking differences between countries without changing the overall order of magnitude of the intensity.
[0089] This correction step does not apply to countries with national observations. In other words, for countries with national observations, their national average intensity observations can be directly used as the target intensity.
[0090] S4: Calculate the scaling factor for each country based on its prior weighted strength and target strength.
[0091] The formula for calculating the scaling factor is:
[0092] in, This represents the target intensity of country i. This represents the prior weighting strength of country i.
[0093] S5: Calculate the antibiotic use intensity for each species in each country based on the prior antibiotic strength for each species and the scaling factor for each country.
[0094] Specifically, the intensity of all species within the same country will be adjusted proportionally:
[0095] At the same time, the distribution parameters are updated as follows:
[0096]
[0097] This embodiment adopts a calibration method of "first generating the a priori intensity of the country-species, and then performing a unified scaling at the national level". The scaling factor only changes the common scale of the intensity of each species under the same country, without changing the relative proportional relationship between species determined by the a priori structure. Therefore, calibration can be completed without destroying the consistency of the original production structure and species structure.
[0098] By constructing a PCU data reliability index and combining it with regional filling ratio, extreme value pruning, and gating mechanisms, the information quality of different countries, regions, and species is differentiated. This allows high-reliability information to participate in calibration first, while low-reliability information is suppressed or replaced, thereby improving the robustness of parameter construction.
[0099] In this embodiment, the regional typical intensity, PCU offset, production scale disturbance, national average intensity target, and scaling factor respectively serve different functions such as absolute anchoring, relative sorting, heterogeneity supplementation, total amount constraint, and consistency calibration. Each step has clear logic, clear source, and verifiable calculation process, thus having good interpretability and traceability.
[0100] Example 2 This embodiment is a specific application scenario of Embodiment 1.
[0101] This example illustrates the estimation of antibiotic use intensity in the pig farming industry of country A.
[0102] 1. Initialization of typical regional intensity: The typical intensity μ of Southeast Asian pigs was obtained from literature and monitoring data. r(A),pig =80mg / kg, the central value is defined in log space as Center=log(80)≈4.382, assuming this data is reliable ( If the value is 1, then the offset is entered into the model.
[0103] 2. PCU Structural Offset Calculation: Assuming that Thomas PCU data shows that the intensity of pigs in country A (i.e., the PCU intensity data of pigs corresponding to country A) is relatively high in the region, calculate the logarithmic offset. =0.35. Because This represents the relative offset in the natural logarithmic space, so after restoring it to the original scale, we have e 0.35 The value is approximately 1.42, indicating that the pig intensity in country A is about 1.42 times the typical level in the region, which is about 42% higher than the typical level in the region.
[0104] 3. Heteroscedasticity driven by production scale: Assume that the pig scale of country A is relatively high in the region (q). A,pig =0.80 indicates that country A is a major producer, according to the system's default settings. =0.15, =0.60, the logarithmic standard deviation σ is obtained by linear interpolation. A,pig=0.51, therefore This allows for greater volatility.
[0105] 4. Random perturbation sampling: Assume that the current sampling yields ε A,pig =0.08.
[0106] 5. Combined Logarithmic Strength:
[0107] Restore to intensity space:
[0108] That is, the prior antibiotic use intensity for pigs in country A is approximately 123 mg / kg.
[0109] 6. Calculate the uncalibrated weighted average intensity: Assume that country A has three species, and the prior intensities of each species are obtained according to the above steps as follows:
[0110] The weighted average strength (i.e., the prior weighted strength) is then:
[0111] 7. Assume that WOAH publishes the national average intensity for country A as 110 mg / kg, that is... Then the scaling factor S A =1.18, obtained by uniformly scaling all species:
[0112] Example 3 This embodiment is a computer-readable storage medium corresponding to the above embodiments, which stores a computer program. When the program is executed by a processor, it implements the various steps of the method for constructing an antibiotic use strength parameter as described in the above embodiments, and can achieve the same technical effect, which will not be repeated here.
[0113] In summary, the present invention provides a method and storage medium for constructing antibiotic strength parameters, which can perform strength calibration while maintaining the original production structure, distinguish the reliability differences of strength information from different countries or regions during the calibration process, and generate stable, interpretable and auditable national-level strength parameters under conditions of incomplete multi-source data, while simultaneously satisfying accuracy, stability and interpretability.
[0114] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.
Claims
1. A method for constructing antibiotic strength parameters, characterized in that, include: Based on regional typical intensity data, intensity data based on population-corrected units, and the production scale of each species in each country, the a priori intensity of antibiotics for each species in each country is determined. Calculate the prior weighted strength of each country based on the prior strength and production scale of antibiotics for each species in each country. Based on the national average intensity observation values, determine the national average intensity of each country, and based on the national average intensity of each country, determine the target intensity of each country. Calculate the scaling factor for each country based on its prior weighted strength and target strength; The antibiotic use intensity for each species in each country is calculated based on the prior antibiotic strength for each country and the scaling factor for each country.
2. The method for constructing antibiotic strength parameters according to claim 1, characterized in that, The determination of the prior antibiotic strength for each species in each country, based on typical regional strength data, strength data adjusted for population units, and the production scale of each species in each country, includes: Based on the typical intensity levels of each species in each region of each country, determine the intensity center value for each species in each country. Based on intensity data with population-corrected units, the relative offset intensity for each species in each country is determined. Based on the production scale of each species in each country, perturbations corresponding to each species in each country are generated. The prior strength of antibiotics for each species in each country is determined based on the intensity center value, relative offset intensity, and perturbation for each species.
3. The method for constructing antibiotic strength parameters according to claim 2, characterized in that, The determination of the relative offset intensity for each species in each country based on intensity data with population correction units includes: Based on the PCU intensity data of each country in each region corresponding to a species, the baseline drug use intensity for each species in each region is determined. The PCU intensity data is intensity data based on population correction units. The relative offset intensity of a country for a species is determined based on the PCU intensity data of a country for the species and the baseline drug use intensity of the region to which the country belongs for the species.
4. The method for constructing antibiotic strength parameters according to claim 3, characterized in that, The determination of the relative offset intensity for each species in each country based on intensity data with population correction units also includes: Determine the PCU intensity data filling ratio for each species in each region; Based on the filling ratio of PCU intensity data for each species in each country's region, the reliability index of PCU intensity data for each species in each country is determined. Based on the reliability index of PCU intensity data for each species in each country and the preset first pruning threshold, extreme value pruning and screening are performed on the relative offset intensity for each species in each country to determine the final relative offset intensity for each species in each country.
5. The method for constructing antibiotic strength parameters according to claim 4, characterized in that, The process involves cropping and filtering the relative offset intensity of each species in each country based on the reliability index of PCU intensity data for each species and a preset first pruning threshold, to determine the final relative offset intensity of each species in each country. This includes: Based on the PCU intensity data of the same species in each country within the same region, calculate the logarithmic standard deviation of the same species in the same region; Based on the logarithmic standard deviation of the same region corresponding to the same species, a first pruning threshold for the same region corresponding to the same species is determined; Based on the first pruning threshold corresponding to a species in a country, the relative offset intensity of the species in a country is pruned to extreme values. Then, based on the reliability index of the PCU intensity data of the species in a country and the pruned relative offset intensity, the final relative offset intensity of the species in a country is determined.
6. The method for constructing antibiotic strength parameters according to claim 2, characterized in that, The generation of perturbations for each species in each country, based on the production scale of each species in each country, includes: Within each country's territory, calculate the production scale quantile for each species in each country, based on the production scale of each country. Based on the PCU intensity data of the same species in each country within the same region, the logarithmic standard deviation of the same species in the same region is calculated, and based on the logarithmic standard deviation of the same species in each region, the maximum allowable fluctuation range and the minimum allowable fluctuation range of the same species are determined. Based on the production scale quantiles for each species in each country and the maximum and minimum permissible fluctuation ranges for each species, calculate the logarithmic standard deviation for each species in each country. Based on the reliability index of PCU intensity data for each species in each country, the indicator variables for each species in each country were determined. The perturbations for each species in each country are generated based on the logarithmic standard deviation, indicator variables, and a preset second pruning threshold.
7. The method for constructing antibiotic strength parameters according to claim 6, characterized in that, The process of generating perturbations for each species in each country based on the logarithmic standard deviation, indicator variable, and preset second pruning threshold includes: The second pruning threshold for a country corresponding to a species is determined based on the logarithmic standard deviation of a country to a species. Based on the logarithmic standard deviation of the species corresponding to the country, the indicator variable, and the second pruning threshold, a perturbation for the species corresponding to the country is generated.
8. The method for constructing antibiotic strength parameters according to claim 1, characterized in that, The process of determining the national average intensity of each country based on national average intensity observations, and then determining the target intensity of each country based on its national average intensity, includes: If a country has a national average intensity observation value, then the national average intensity observation value of that country shall be used as the national average intensity and target intensity of that country. If a country does not have a national average intensity observation value, the median of the national average intensity of the countries within the region to which the country belongs that have national average intensity observation values is taken as the national average intensity of the country. The national average intensity of the country is then corrected according to a preset correction intensity parameter and a maximum correction range to obtain the target intensity of the country.
9. The method for constructing antibiotic strength parameters according to claim 8, characterized in that, The step of correcting the national average intensity of a country based on preset correction intensity parameters and a maximum correction range to obtain the target intensity of the country includes: Obtain the log-median of the PCU national strength for each country, where the PCU national strength is the national average strength based on population-corrected units; The logarithmic shift strength of a country is determined based on the country strength of the PCU and the logarithmic median. Based on the preset correction intensity parameters, the maximum correction magnitude, and the logarithmic offset intensity of the country, the national average intensity of the country is corrected in logarithmic space to obtain the target intensity of the country.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-9.