A population health-oriented method for dynamic control of environmental risks and its storage medium

Abstract

This invention provides a population-oriented method and storage medium for dynamic control of environmental risks, relating to the field of public health. It systematically integrates macro-environmental monitoring with demographic data, multi-pollutant air quality health indices, and individual-level longitudinal cohort data. Through a constructed fusion module, the multi-pollutant air quality health index is used as prior knowledge to process individual-level causal models. The processed health risk function is coupled with the pollution control marginal cost function and refined health economic assessment parameters to construct a comprehensive optimization function. The uncertainty of all key parameters in the objective function is probabilistically characterized and propagated, ultimately solving for a control threshold range with confidence intervals. The optimized control threshold is then re-introduced into the multi-pollutant air quality health index calculation framework to assess its risk level changes under historical benchmarks, providing empirical support from historical data for the rationality and feasibility of the optimization results.

Description

Technical Field

[0001] This invention relates to the field of public health technology, and in particular to a method and storage medium for dynamic control of environmental risks oriented towards population health. Background Technology

[0002] Currently, the determination of air pollution control thresholds mainly relies on several independent research paradigms. These methods each have their own advantages, but they also have limitations. A complete solution that can systematically address the challenges of scientific validity, economic efficiency, and verifiability has not yet been formed.

[0003] In current technologies, the mainstream approach is epidemiological studies based on population health effects. These studies typically utilize large prospective cohort data and employ survival analysis techniques such as the Cox proportional hazards model to establish the exposure-response relationship between long-term pollutant exposure and health outcomes (such as all-cause mortality and the onset of specific diseases). This approach is scientifically rigorous and provides reliable causal evidence, forming the basis for setting health-oriented thresholds. However, this method is usually based on a single cohort, and its results become uncertain when extrapolated to other regions with different population structures, climatic conditions, or mixed pollution characteristics; model parameter estimates may be affected by unmeasured confounding factors, and robustness needs improvement; and its conclusions are often relative risks or theoretically optimal health thresholds, without fully considering the socioeconomic costs and feasibility of achieving these thresholds.

[0004] Some methods are based on macro-level correlation analysis between environmental monitoring and health statistics. These methods often use annual average pollutant concentrations at the regional or city level to perform correlation analysis with health statistics such as mortality and hospitalization rates, calculating health impact functions and attributable disease burden, which can then be further monetized into health economic losses. This method can quickly assess the health impacts of historical pollution and is easy to obtain data on. However, this method is based on ecological correlations, making it difficult to establish causal relationships and susceptible to ecological fallacies; it typically assesses single pollutants or simple summations, failing to accurately reflect the complex health effects of multiple pollutants working synergistically in the real world; and as a retrospective evaluation, its direct guiding role in setting prospective thresholds is limited.

[0005] Other approaches focus on developing comprehensive air quality health indices, such as cumulative risk indices. Air quality health indices aim to comprehensively reflect the combined acute or chronic health risks of multiple pollutants, offering advantages in health risk communication compared to traditional air quality indices. However, existing air quality health indices are primarily descriptive and warning tools. Their construction is largely based on acute effect coefficients obtained from time-series analysis, lacking direct calibration with long-term chronic health effects, and are not deeply integrated into optimization decision-making frameworks such as pollutant control thresholds.

[0006] Furthermore, although existing studies have attempted to incorporate health benefit assessments into the cost-benefit analysis of pollution control, their estimations of health benefits often rely on simplified exposure-response relationships provided by the second type of method mentioned above. This fails to effectively integrate with the most rigorous individual-level causal inference models, resulting in an insufficiently robust foundation for benefit estimation. Simultaneously, existing methods generally lack systematic quantification and transmission analysis of uncertainties in various parameters within the model during threshold optimization, and also lack closed-loop designs for retrospectively verifying optimization results against observable historical benchmarks. This renders the recommended thresholds insufficient in terms of scientific rigor and decision-making credibility.

[0007] Therefore, there is an urgent need for a new method that can connect micro-causal inference with macro-health assessment, integrate health risk assessment with economic cost analysis, and possess the ability to quantify uncertainty and verify historical data. Summary of the Invention

[0008] To achieve the above-mentioned objectives and other advantages of the present invention, a first objective of the present invention is to provide a population health-oriented method for dynamic regulation of environmental risks, comprising the following steps:

[0009] Acquire historical air pollution monitoring data, population health data, and longitudinal cohort data of the study subjects for the target area;

[0010] By using historical air pollution monitoring data and population health data, a multi-pollutant air quality health index applicable to the region is constructed to quantify the comprehensive impact coefficient of historical pollution levels on population health risks.

[0011] An individual-level causal risk prediction model for long-term air pollution exposure and health outcomes was established using the longitudinal cohort data.

[0012] The comprehensive impact coefficient of the multi-pollutant air quality health index is used as the prior knowledge of the fusion module to process the parameters of the individual-level causal risk prediction model, resulting in the processed health risk function.

[0013] A cost-benefit comprehensive objective function is established by coupling the pollution control marginal cost function with health economic assessment parameters based on the processed health risk function.

[0014] The uncertainty of the parameters in the objective function is quantified, and the concentration threshold and its confidence interval that maximize the expected social net benefit are solved as recommended control thresholds.

[0015] The recommended control threshold is substituted into the multi-pollutant air quality health index calculation framework to assess its risk level change relative to historical levels, and the output includes decision information including threshold recommendations, health benefit predictions, and historical verification comparisons.

[0016] Furthermore, the multi-pollutant air quality health index is an air quality health index constructed based on the cumulative risk index method, and its comprehensive influence coefficient is the exposure response coefficient vector obtained by the multi-pollutant generalized additive model.

[0017] Furthermore, the step of establishing an individual-level causal risk prediction model for long-term air pollution exposure and health outcomes using the longitudinal cohort data includes:

[0018] Based on the residential address history of each study subject in the longitudinal cohort data, and combined with the atmospheric pollutant concentration grid data, long-term pollutant exposure concentrations were assigned to each study subject year by year during the follow-up period.

[0019] Using the occurrence time and status of health outcome events during the follow-up period as dependent variables, a statistical model was constructed with a Cox proportional hazards model as the framework and a nested distributed lag nonlinear model. The distributed lag nonlinear model, through cross basis functions, simultaneously characterizes the nonlinear relationship between atmospheric pollutant exposure concentration and health risk, as well as the exposure lag effect.

[0020] The individual covariates in the longitudinal cohort data are input into the statistical model as control variables, and the model is fitted using parameter estimation methods to obtain a complete set of model parameters, including the cross-basis function coefficients, lag effect coefficients, and covariate coefficients of pollutant exposure.

[0021] Based on the fitted model parameters, a health risk function is generated for prediction. This function is used to calculate the risk ratio or excess relative risk of a specific population for the target health outcome when the pollutant concentration changes from the reference level to the target level.

[0022] The health risk function and its parameters are encapsulated into a callable prediction module, which is then used as the output of the individual-level causal risk prediction model.

[0023] Furthermore, the allocation of long-term pollutant exposure concentration specifically involves: using a pollutant concentration reanalysis dataset with a spatial resolution better than a preset accuracy threshold, extracting the annual average concentration of the geographical grid where the study subject resides during the follow-up year based on the subject's residential address coordinates, and using this as the subject's individual annual exposure level;

[0024] The cross-basis function uses a penalized spline function to nonlinearly fit the pollutant exposure-response relationship, and uses a polynomial function or spline function to model the hysteresis effect structure.

[0025] The input interface of the callable prediction module includes at least: the change in the concentration of the target pollutant and the baseline covariate characteristic distribution of the target population, and its output interface includes at least: the predicted hazard ratio and the predicted change in the number of health outcomes.

[0026] Furthermore, the specific steps of the fusion module are as follows: taking the comprehensive impact coefficient corresponding to the multi-pollutant air quality health index as the center of the prior distribution, taking the parameters estimated by the individual-level causal risk prediction model as the likelihood function, and obtaining the posterior distribution of the parameters through the Markov chain Monte Carlo method, thereby realizing the correction and robustness of the estimation results of the individual-level causal risk prediction model.

[0027] Furthermore, the fusion module processes the parameters of the individual-level causal risk prediction model using a weighted integration method:

[0028] The health burden under different historical pollution scenarios is assessed using the multi-pollutant air quality health index, and compared with the same burden predicted by the individual-level causal risk prediction model to calculate the corrected weight of the model prediction.

[0029] The adjusted weights are applied to the prediction of future health benefits by the individual-level causal risk prediction model.

[0030] Furthermore, the disability-adjusted life years weight or disease cost in the health economic assessment parameters are differentiated based on the contribution ratio of different health endpoints in the historical assessment of the multi-pollutant air quality health index.

[0031] Furthermore, the established cost-benefit comprehensive objective function is specifically as follows:

[0032] The difference between the total health and economic benefits generated when the pollutant concentration changes to a preset value and the total pollution control cost required to achieve that concentration change is defined as the net social benefit function.

[0033] Wherein, the total health economic benefit is the sum of the product of the reduction in the number of type i health outcomes calculated by the processed health risk function and the localized health economic assessment parameter corresponding to type i health outcomes;

[0034] The total pollution control cost is obtained by integrating the pollution control marginal cost function from the current concentration to the target concentration;

[0035] The optimization objective of the cost-benefit comprehensive objective function is to maximize the net social benefit function;

[0036] The concentration threshold that maximizes the expected social net benefit is specifically achieved by maximizing the social net benefit function. The solution process is performed within the framework of Monte Carlo simulation to simultaneously handle the uncertainty of the parameters.

[0037] Furthermore, the historical verification comparison includes: comparing the predicted multi-pollutant air quality health index value and the predicted excess mortality reduction under the recommended threshold with the actual multi-pollutant air quality health index value and the actual excess mortality change in historical periods, and calculating the predicted improvement efficiency ratio as an auxiliary evaluation indicator for the feasibility of the threshold scheme.

[0038] A second objective of the present invention is to provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0039] Compared with the prior art, the beneficial effects of the present invention are:

[0040] Compared with the prior art, the present invention has achieved the following outstanding advantages:

[0041] This invention provides a population-oriented method for dynamic control of environmental risks and a storage medium that integrates macro-environmental monitoring and demographic data reflecting overall regional exposure and health status, multi-pollutant air quality health indices reflecting comprehensive health risks, and individual-level longitudinal cohort data revealing precise exposure-response causal relationships. Specifically, through a constructed fusion module, the macro-level multi-pollutant air quality health index is used as prior knowledge to perform Bayesian calibration or weighted integration on individual-level causal models. This effectively compensates for extrapolation biases that may arise from limited samples or unmeasured confounding factors in single-cohort data. Simultaneously, the causal rigor of individual models enhances the scientific basis of the multi-pollutant air quality health index, achieving a two-way enhancement of macro-level trend verification of micro-level causality and micro-level model calibration of macro-level parameters.

[0042] This invention couples a more robust health risk function, after fusion calibration, with a pollution control marginal cost function and refined health economic assessment parameters to construct a comprehensive optimization function aimed at maximizing net social benefits. This makes the recommended control thresholds not only more reliable in terms of health protection but also fully considers their socioeconomic feasibility and cost-effectiveness, providing decision-makers with a quantitative basis that combines public health value and economic rationality.

[0043] This invention introduces Monte Carlo simulation to probabilistically represent and propagate the uncertainty of all key parameters in the objective function. The final solution is a control threshold range with confidence intervals, which enables decision-makers to clearly define the risk level corresponding to different threshold selection schemes and supports them in making more robust and flexible decisions based on their risk preferences.

[0044] This invention re-introduces the optimized control thresholds into the multi-pollutant air quality health index calculation framework, assesses the changes in risk levels under historical benchmarks, and compares and verifies them with historically observed health improvement data. This not only provides intuitive risk communication indicators, but more importantly, by calculating quantitative indicators such as the predicted improvement efficiency ratio, it provides empirical support from historical data for the rationality and feasibility of the optimization results. This forms a complete evidence chain from historical learning to model optimization and backtracking verification, greatly enhancing the scientific persuasiveness and practical value of the final decision-making information.

[0045] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it according to the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Specific embodiments of the present invention are given in detail below with reference to the accompanying drawings. Attached Figure Description

[0046] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0047] Figure 1 Flowchart of a dynamic environmental risk control method oriented towards population health;

[0048] Figure 2 Flowchart for constructing an individual-level causal risk prediction model;

[0049] Figure 3 A flowchart for allocating long-term pollutant exposure concentrations;

[0050] Figure 4 A flowchart of the Bayesian calibration process for the fusion module;

[0051] Figure 5 A flowchart of the weighted integration process for the fusion module;

[0052] Figure 6 Flowchart for establishing and maximizing the social net benefit function;

[0053] Figure 7 Flowchart for historical verification and comparison;

[0054] Figure 8 This is a comparative analysis chart for historical verification;

[0055] Figure 9 A schematic diagram of computer equipment;

[0056] Figure 10 This is a schematic diagram of a computer-readable storage medium. Detailed Implementation

[0057] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be noted that, without conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0058] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0059] The drawing numbers in this application are only used to distinguish the steps in the scheme and are not used to limit the execution order of the steps. The specific execution order is as described in the specification.

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0061] Example 1

[0062] A population-oriented approach to dynamic control of environmental risks, such as Figure 1 As shown, it includes the following steps:

[0063] S100. Obtain historical air pollution monitoring data, population health data, and longitudinal cohort data of the research subjects for the target area.

[0064] This example uses PM2.5 in a specific urban area. 2.5 Taking the regulation of compound pollution with NO2 as an example, environmental data, health statistics, and longitudinal cohort data of the target city over the past ten years (e.g., 2015-2024) are collected.

[0065] Among them, environmental data refers to the daily average pollutant concentration data of national and provincial air quality monitoring stations, such as PM2.5. 2.5 Along with NO2, etc. Health statistics refer to the annual all-cause mortality rate and hospitalization rate by cause for the city and district levels during the same period, sourced from the "Health Statistics Yearbook" and the annual report of the Center for Disease Control and Prevention. Longitudinal cohort data refers to data from a chronic disease risk study cohort in the city that has been ongoing for more than ten years. This cohort includes long-term follow-up information of approximately tens of thousands of baseline healthy volunteers, including dynamically updated residential addresses, lifestyle questionnaires, and physical examination data. It also obtains daily-accurate information on deaths and major disease incidence events through links to the city's death registration and inpatient medical record homepage systems.

[0066] All data is cleaned and de-identified, and then aligned and matched according to a unified spatiotemporal benchmark, such as using streets as spatial units and years as time units.

[0067] S200. Construct a multi-pollutant air quality health index applicable to the region using the historical air pollution monitoring data and the population health data, and quantify the comprehensive impact coefficient of historical pollution levels on population health risks.

[0068] This embodiment employs a cumulative risk index-based method to construct a multi-pollutant air quality health index for the target city. Using daily all-cause mortality as the health endpoint, an index including PM2.5 is established. 2.5 A generalized additive model for multiple pollutants, including daily average concentrations of nitrogen, nitrogen, and phosphorus (NO2), was developed, controlling for long-term trends, meteorological factors (such as temperature and humidity), and day-of-week effects. Through model fitting, a set of exposure-response coefficients was obtained. , (,...). This coefficient vector is the comprehensive impact coefficient that quantifies the impact of historical compound pollution on the risk of acute death in the population. Using this coefficient, the daily cumulative risk value is calculated and standardized to an index scale of 1 to 10 or higher, forming a historical series of multi-pollutant air quality health indices for the target city.

[0069] Optionally, a sufficiently long historical period should be selected for the target city, such as from January 1, 2015 to December 31, 2024, to ensure sufficient coverage of pollution variations and health events. Daily average concentration data from all national and provincial air quality monitoring stations in the target city should be collected during this period. Core pollutants typically include PM2.5. 2.5 And NO2, etc. (daily maximum 8-hour average). Perform quality control on the data, handling missing values, such as using spatial-temporal interpolation. Collect the city's daily all-cause mortality data for the same period. This is the most stable and complete health endpoint indicator, usually obtained from the city's disease control and prevention center or mortality registry system. Collect meteorological data such as daily average temperature and daily relative humidity for the same period to control for confounding effects of meteorological factors in the model.

[0070] Establish a model that can simultaneously characterize the relationship between multiple pollutants and health outcomes:

[0071]

[0072] in, Let t be the number of all-cause deaths. Let be the expected number of deaths on day t. For the intercept term, For the i-th pollutant on day t (PM) 2.5 The concentration of (NO2, etc.) Let be the exposure-response coefficient for the i-th pollutant, which is the final vector of comprehensive impact coefficients to be obtained. It should be noted that... It is estimated under the condition that other pollutants are included in the model. It reflects the independent health effects of the pollutant after controlling for the effects of coexistence of other pollutants. A naturally smoothed spline function over calendar time is used to control for long-term trends and seasonal effects of mortality. (Degrees of freedom) are usually set according to the time span, such as 7 degrees of freedom per year. This is a smoothed spline function for temperature and humidity, used to control the nonlinear effects of meteorological factors. This is a dummy variable for the day-of-the-week effect, used to control for possible regular fluctuations in mortality throughout the week, such as the fact that mortality rates are usually higher on Mondays. This is the residual term.

[0073] The above model is fitted using the Generalized Additive Model (GAM) framework of statistical software such as the mgcv package in R. A quasi-Poisson distribution or a negative binomial distribution is employed to address the excessive dispersion of death counts. All parameters are estimated using the maximum likelihood method or the penalized likelihood method, with the most crucial step being obtaining... The estimated value and its 95% confidence interval.

[0074] Based on the well-fitted model, calculate the daily cumulative risk index:

[0075]

[0076] in, Let be the cumulative risk index for day t, which represents the relative multiple of the risk of death from combined exposure to multiple pollutants on the current day, relative to the ideal situation where the concentrations of all pollutants are zero. For the i-th pollutant on day t (PM) 2.5 The actual concentrations of (NO2, etc.).

[0077] And calculate the overall relative risk of excess:

[0078]

[0079] in, This represents the overall excess risk percentage on day t. A higher value indicates a higher health risk.

[0080] To facilitate public understanding and risk communication, Standardize to a fixed exponential scale, such as 1 to 10 and above. The standardization formula is:

[0081]

[0082] in, Let be the air quality health index value for day t. Calculated within the study period (e.g., the past 10 years) The maximum value, or the 99th percentile, is used to avoid the influence of extreme values. This ensures that the exponent range is fixed.

[0083] Optionally, the risk levels are divided as follows: 1-3 represents low risk, 4-6 represents medium risk, 7-9 represents high risk, and 10 and above represents extremely high risk.

[0084] Output comprehensive influence coefficient vector ( , This is prior knowledge for the subsequent S400 steps, which quantifies the independent contribution weights of each pollutant to the health risks of the population in the target city's history.

[0085] S300. Establish an individual-level causal risk prediction model for long-term air pollution exposure and health outcomes using the longitudinal cohort data.

[0086] Specifically, such as Figure 2 As shown, the steps for establishing an individual-level causal risk prediction model for long-term air pollution exposure and health outcomes using the longitudinal cohort data include:

[0087] S310. Based on the residential address history of each research subject in the longitudinal cohort data, and combined with the atmospheric pollutant concentration grid data, long-term pollutant exposure concentrations are allocated to each research subject year by year during the follow-up period.

[0088] Among them, such as Figure 3 As shown, the allocation of long-term pollutant exposure concentrations specifically involves: S311, using a pollutant concentration reanalysis dataset with a spatial resolution better than a preset accuracy threshold; S312, extracting the annual average concentration of the geographical grid where the research subject resides in the follow-up year based on the residential address coordinates of the subject, as their individual annual exposure level.

[0089] For example, using high-precision pollutant reanalysis grid data with a spatial resolution of 1 km × 1 km, the PM2.5 concentration of each grid cell can be extracted based on the latitude and longitude of the residential addresses reported annually by cohort members. 2.5 The annual average concentration of NO2 was used as the individual's annual exposure level.

[0090] S320. Using the occurrence time and status of health outcome events during the follow-up period as dependent variables, a statistical model is constructed with a Cox proportional hazards model as the framework and a nested distributed lag nonlinear model. The distributed lag nonlinear model characterizes the nonlinear relationship between atmospheric pollutant exposure concentration and health risk, as well as the exposure lag effect, through cross basis functions.

[0091] S330. Input the individual covariates in the longitudinal cohort data as control variables into the statistical model, and use parameter estimation methods to fit the model to obtain a complete set of model parameters, including the cross-basis function coefficients, lag effect coefficients and covariate coefficients of pollutant exposure.

[0092] The cross-basis function uses a penalized spline function to nonlinearly fit the pollutant exposure-response relationship, and uses a polynomial function or spline function to model the hysteresis effect structure.

[0093] This embodiment uses the time from enrollment to death from cardiovascular disease or the end of follow-up as survival analysis data to construct a Bayesian stratified Cox proportional hazards model. The core variable is PM. 2.5 Long-term exposure levels to NO2 were characterized using a distributed lag nonlinear model: a penalized spline function was used to fit the nonlinear part of the exposure-response relationship; a natural cubic spline function was used to fit the lag effect structure spanning several years to capture chronic cumulative effects. Age, sex, smoking history, education level, and baseline BMI were included as covariates in the model to control for confounding factors. A Markov chain Monte Carlo algorithm was used for model fitting and parameter estimation to obtain the posterior probability distribution of the pollutant exposure hazard ratio.

[0094] Optionally, the longitudinal cohort data is converted to a classic person-period survival analysis format, where each row represents a record of a study subject within a specific risk period (in years). A Bayesian hierarchical Cox proportional hazards model is constructed using Bayesian statistical software (such as Stan, PyMC, or the rstanarm / brms package in R), and the model is fitted using a Stan sampler (such as NUTS). Typically, multiple independent Markov chains are run, each iterating 4000 times, with the first 2000 iterations discarded as a warm-up period. Convergence is then diagnosed to ensure that the MCMC sampling is effective. Extraction from the fitted data... , All MCMC samples, such as 4 chains * 2000 effective iterations = 8000 samples. These 8000 samples constitute... and The posterior probability distribution can be used to calculate the median as a point estimate and the 2.5% and 97.5% quantiles as 95% confidence intervals. Based on the median (or mean) parameter of the posterior distribution, a prediction function can be written as a callable prediction module.

[0095] Then, for each subject i in the cohort, at each follow-up year t, not only the exposure concentration for that year is needed, but also their exposure history needs to be constructed. Based on literature and research objectives, for example, L=10, indicating that exposure over the past 10 years is considered, for subject i in year t, their PM2.5 concentration... 2.5The exposure history is a vector of length L+1:

[0096]

[0097] in, This indicates that the research subjects were in Personal PM of the year 2.5 The same procedure was performed for NO2 based on the annual average exposure concentration, and will not be repeated here.

[0098] This is the core step completed through pre-computation outside the model using the dlnm R package. This embodiment will separately address PM. 2.5 Perform this operation with NO2.

[0099] The cross-basis functions for constructing the distributed lag nonlinear model are pre-computed outside the model using the dlnm R package. Specifically, penalized splines are chosen as basis functions to flexibly fit the nonlinear relationship; then natural cubic splines are selected to smooth the lag effect, assuming that the most recent year has the greatest impact, which decays over time. The two basis functions are then subjected to a tensor product operation to generate a two-dimensional basis function matrix that can simultaneously characterize the exposure of nonlinearity and lag structure. For example, the generated two-dimensional basis function matrix is ​​an n-row × (10*4=40) matrix, where each row corresponds to a person-year observation, and each column is a combination dimension of the cross-basis. These 40 columns will become a set of time-varying predictors in the Cox model.

[0100] The cross-basis is then integrated into a Bayesian hierarchical Cox model. Specifically, the cross-basis matrix is ​​used as data input, a set of coefficients is assigned to each cross-basis matrix, and a linear combination of the cross-basis is incorporated into the risk model.

[0101] Finally, MCMC sampling is run, and the cumulative exposure-response curve under a given hysteresis structure is calculated through post-processing. Furthermore, the complex model described above is encapsulated into a simplified interface.

[0102] S340. Based on the fitted model parameters, generate a health risk function for prediction. This function is used to calculate the risk ratio or excess relative risk of a specific population for the target health outcome when the pollutant concentration changes from the reference level to the target level.

[0103] Optionally, cross-basis feature vectors are generated for the two scenarios, and the posterior median of the model parameters (as a point estimate) is used for calculation. The hazard ratio and excess relative risk are calculated, and the number of attributable / avoidable cases is calculated. The above steps are encapsulated into a class for easy system calls.

[0104] S350. The health risk function and its parameters are encapsulated into a callable prediction module, which is used as the output of the individual-level causal risk prediction model.

[0105] The input interface of the callable prediction module includes at least: the change in the concentration of the target pollutant and the baseline covariate characteristic distribution of the target population, and its output interface includes at least: the predicted hazard ratio and the predicted change in the number of health outcomes.

[0106] This embodiment encapsulates the fitted model and its parameters into a standardized software function module. For example, the input interface of this module is: PM 2.5 The target concentration change (ΔC) of NO2 is output as follows: the relative change (e.g., hazard ratio) and absolute change (e.g., the number of deaths avoided) of the predicted cardiovascular disease mortality risk in the target population.

[0107] S400. The comprehensive impact coefficient of the multi-pollutant air quality health index is used as the prior knowledge of the fusion module to process the parameters of the individual-level causal risk prediction model to obtain the processed health risk function.

[0108] In some embodiments, such as Figure 4 As shown, the specific steps of the fusion module are as follows: S410, taking the comprehensive impact coefficient corresponding to the multi-pollutant air quality health index as the center of the prior distribution, and taking the parameters estimated by the individual-level causal risk prediction model as the likelihood function; S420, obtaining the posterior distribution of the parameters through the Markov chain Monte Carlo method, thereby realizing the correction and robustness of the estimation results of the individual-level causal risk prediction model.

[0109] This embodiment uses Bayesian calibration as the fusion method to integrate the PM estimated in step S200, after annualization and scale adjustment. 2.5 The NO2 exposure-response coefficient serves as the center of the prior distribution of the corresponding parameters of the individual Cox model in the S300 step.

[0110] The likelihood function of the individual Cox model parameter estimates is combined with the above prior, and the MCMC algorithm is used for updating and iteration to obtain the calibrated posterior distribution of the parameters.

[0111] Based on the calibrated posterior distribution, a new calibrated health risk prediction function is generated. This function incorporates the stability of the overall population risk pattern reflected by macro data while retaining the rigor of causal inference based on individual tracking, and its prediction uncertainty range is significantly narrowed compared to the original model.

[0112] In other embodiments, such as Figure 5 As shown, in application scenarios where resources are limited or rapid response is required, the fusion module uses a weighted integration method to process the parameters of the individual-level causal risk prediction model:

[0113] S430. The health burden under different historical pollution scenarios is assessed using the multi-pollutant air quality health index, and compared with the same burden predicted by the individual-level causal risk prediction model to calculate the corrected weight of the model prediction.

[0114] Optionally, several historical years with significant differences in pollution levels, such as a heavily polluted year and a lightly polluted year, are selected as baseline scenarios. Using the multi-pollutant air quality health index constructed in step S200, estimates of excess deaths attributable to air pollution are calculated for these two baseline scenarios. Actual pollution data from these two baseline years are input into the (i.e., uncalibrated) individual-level causal risk prediction model established in step S300 to obtain the model's predictions for excess deaths. The difference between the model's predictions and the multi-pollutant air quality health index assessments for each scenario is compared. The smaller the difference, the more reliable the model's performance in that scenario, and a higher weight W is assigned to it. The weight calculation formula can be: W = 1 / (1 + absolute relative deviation).

[0115] S440. Apply the adjusted weights to the prediction of future health benefits by the individual-level causal risk prediction model.

[0116] Optionally, when it is necessary to predict health benefits for a future pollution control scenario, the individual model first provides an initial prediction value, and then this value is multiplied by a weight W determined based on the similarity between the future scenario and the historical baseline scenario to obtain the final calibrated prediction value. This method uses historical macroeconomic data to intuitively and empirically correct the model predictions.

[0117] S500: Based on the processed health risk function, coupled with the pollution control marginal cost function and health economic assessment parameters, a cost-benefit comprehensive objective function is established.

[0118] The disability-adjusted life years weight or disease cost in the health economic assessment parameters are set differently based on the contribution ratio of different health endpoints in the historical assessment of the multi-pollutant air quality health index.

[0119] S600. Quantify the uncertainty of the parameters in the objective function, and solve for the concentration threshold and its confidence interval that maximizes the expected social net benefit, as a recommended control threshold.

[0120] Specifically, such as Figure 6 As shown, the specific cost-benefit comprehensive objective function is as follows:

[0121] S510. The difference between the total health and economic benefits generated when the pollutant concentration changes to a preset value and the total pollution control cost required to achieve that concentration change is defined as the net social benefit function.

[0122] Wherein, the total health economic benefit is the sum of the product of the reduction in the number of type i health outcomes calculated by the processed health risk function and the localized health economic assessment parameter corresponding to type i health outcomes;

[0123] The total pollution control cost is obtained by integrating the pollution control marginal cost function from the current concentration to the target concentration;

[0124] The optimization objective of the cost-benefit comprehensive objective function is to maximize the net social benefit function.

[0125] Optionally, a calibrated health risk function can be used to predict when PM2.5 levels rise. 2.5 The number of cardiovascular deaths that could be avoided when the concentration is reduced from the current level (e.g., 35 μg / m³) to different target values.

[0126] Analysis of historical air quality health index data for multiple pollutants revealed that cardiovascular diseases account for as much as 60% of the pollution-related disease burden in the target city. Therefore, a localized statistical value of life (VW) method focusing on cardiovascular health outcome assessment was used for calculation, rather than applying national or international averages, to ensure that the benefit assessment more closely reflects local welfare.

[0127] Based on the pollution control project planning and budget data of the city's ecological and environmental protection department, PM2.5 was fitted. 2.5 The marginal cost curve for further reduction in concentration shows an exponential increase in cost, starting with easier reductions and progressing to more difficult ones.

[0128] Establish a net social benefit function: .in, For health economic benefits (number of deaths avoided × localized life value). To achieve concentration change The total governance cost required (obtained by integrating the marginal cost curve). The optimization objective is to find the cost that makes governance cost 100% higher. Maximize value.

[0129] The concentration threshold that maximizes the expected social net benefit is specifically achieved by S610, which involves maximizing the social net benefit function. The solution process is performed within the framework of Monte Carlo simulation to simultaneously handle the uncertainty of the parameters.

[0130] Specifically, probability distributions are set for the key uncertainty inputs in the objective function: the coefficients of the calibrated risk function are based on its posterior distribution, the localized statistical life value is set to a log-normal distribution, and the parameters of the marginal cost function are set to a triangular distribution to reflect the estimation range.

[0131] Then, multiple independent simulations were performed. In each simulation, a set of parameter values ​​was randomly selected from the above distributions, and these parameters were used to calculate the net social benefit NB as a function of PM. 2.5 The curve of target concentration change is used to find the PM value that maximizes the NB value in this simulation. 2.5 Concentration value.

[0132] After the simulation is completed, multiple optimal concentration values ​​are obtained. The average of this set is calculated as a point estimate of the recommended control threshold, for example, 28.5 μg / m³. The 2nd and 97th percentiles are then calculated as the 95% confidence interval for this threshold, for example, [26.8, 30.2] μg / m³. Risk indicators such as the probability of a positive net benefit can also be reported.

[0133] S700. Substitute the recommended control threshold into the multi-pollutant air quality health index calculation framework, assess its risk level change relative to historical levels, and output decision information including threshold recommendations, health benefit predictions, and historical verification comparisons.

[0134] Among them, such as Figure 7 As shown, the historical verification comparison includes: S710, comparing the predicted multi-pollutant air quality health index value and the predicted excess mortality reduction under the recommended threshold with the actual multi-pollutant air quality health index value and the actual excess mortality change in historical periods; S720, calculating the predicted improvement efficiency ratio as an auxiliary evaluation indicator for the feasibility of the threshold scheme.

[0135] Optionally, the recommended threshold, such as 28.5 μg / m³, can be substituted into the multi-pollutant air quality health index calculation model constructed in step S200 to simulate how much the number of high-health-risk days (such as the number of days with multi-pollutant air quality health index > 6) is expected to decrease if this standard is reached in a historically heavily polluted year (such as 2017).

[0136] The projected reduction was compared with the actual reduction in the number of high-risk days observed from 2017 to 2021 through actual policy actions.

[0137] Calculate the predicted improvement efficiency ratio (i.e., predicted reduction / actual reduction). If the ratio is close to 1, it indicates that the expected strength of the recommended threshold plan is comparable to that of historically effective measures, and its feasibility is high; if it is significantly greater than 1, it suggests that the plan may be too aggressive. Figure 8 By comparing the optimization results with historical data, the key quantitative indicator of the predicted improvement efficiency ratio is highlighted, and a visual output from data to decision is presented.

[0138] The system automatically integrates all the above results to generate a structured decision support report, which includes: recommended control thresholds and uncertainty ranges, expected health benefits and economic benefits, estimated governance costs and investment benefit analysis, and verification comparisons and predicted improvement efficiency ratios based on historical data. This report provides a quantitative, transparent, and evidence-based complete information chain for collaborative decision-making between environmental and public health departments.

[0139] For example, the output environmental health risk dynamic control decision support report includes a threshold recommendation module, a health benefit prediction module, a cost-benefit analysis module, a historical validation indicator module, and an implementation support information module. The threshold recommendation module outputs the technical parameters of pollutant control thresholds, including: pollutant type, recommended concentration threshold, confidence interval, and recommended time to achieve compliance, such as recommending PM2.5 concentration. 2.5 The annual average concentration threshold is 28.5 μg / m³, with a 95% confidence interval of [26.8, 30.2] μg / m³, and the recommended time to achieve the standard is before the end of 2030.

[0140] The health benefit prediction module outputs quantitative predictions of health improvement effects, including: the number of cases avoided for various health outcomes (such as death and hospitalization), the number of healthy life years gained, and the estimated value of monetized health benefits based on localized parameters, such as avoiding 620 premature deaths per year (95% CI: 520-720), reducing hospitalizations by 1850 cases per year, gaining 8750 healthy life years per year, and monetizing health benefits of RMB4.5 billion per year.

[0141] The cost-benefit analysis module outputs economic assessment results, including: total pollution control cost estimate (one-time investment and annual operation and maintenance cost), benefit-cost ratio (BCR), investment payback period, and key parameters of the marginal emission reduction cost curve, such as a total treatment cost estimate of RMB 1.5 billion (one-time + annual), an annual operation and maintenance cost of RMB 230 million / year, a benefit-cost ratio (BCR) of 3.0, and an investment payback period of <5 years.

[0142] The historical verification index module outputs feasibility verification indicators for the proposed solution, with the core being the predicted improvement efficiency ratio, which is the ratio of the expected unit cost health benefit of the recommended solution to the unit cost health benefit of historical actual measures. This ratio provides a quantitative assessment of the solution's realistic feasibility. For example, a predicted improvement efficiency ratio of 1.2 (range: 0.9-1.5) indicates a high feasibility rating and a medium uncertainty assessment.

[0143] The implementation support information module outputs the technical support information required for the implementation of the plan, including: sensitivity analysis results of key parameters, identification of major risks and contingency plans, recommended monitoring indicator system and monitoring frequency suggestions. For example, a phased implementation roadmap: Phase 1 (2026-2027): In-depth treatment of industrial sources, PM2.5... 2.5Reduced to 35 μg / m³; Phase Two (2028-2029): Coordinated control of traffic and area source pollution, PM2.5 2.5 It decreased to 30 μg / m³.

[0144] This invention achieves deep integration and complementary advantages of multi-source heterogeneous data. Traditional methods often suffer from assessment bias due to single data sources, such as relying solely on macroeconomic statistical data or a single cohort. This invention systematically integrates three types of key data: macroeconomic environmental monitoring and demographic data reflecting overall regional exposure and health status; multi-pollutant air quality health indices reflecting comprehensive health risks; and individual-level longitudinal cohort data revealing precise exposure-response causal relationships. In particular, through the fusion module constructed using the S400 steps, the macroeconomic multi-pollutant air quality health index is used as prior knowledge to perform Bayesian calibration or weighted integration on individual-level causal models. This effectively compensates for extrapolation biases that may arise from single-cohort data due to sample limitations or unmeasured confounding factors. Simultaneously, the causal rigor of individual models enhances the scientific basis of multi-pollutant air quality health indices, achieving a two-way enhancement of macroeconomic trend verification of microeconomic causality and microeconomic model calibration of macroeconomic parameters.

[0145] This invention constructs an optimized decision-making framework that balances scientific rigor and economic efficiency. Existing threshold determination methods often separate health risk assessment from economic cost analysis or employ overly simplistic health benefit estimations. This invention, through the S500 step, couples a more robust health risk function, after fusion calibration, with a pollution control marginal cost function and refined health economic assessment parameters (such as disability-adjusted life years and disease costs), constructing a comprehensive optimization function aimed at maximizing net social benefits. This makes the recommended control thresholds not only more reliable in terms of health protection but also fully considers their socioeconomic feasibility and cost-effectiveness, providing decision-makers with a quantitative basis that combines public health value with economic rationality.

[0146] This invention quantifies uncertainty and provides risk-controlled decision-making solutions. Existing methods typically provide definite point estimation thresholds, ignoring the inherent uncertainties in model parameters, exposure-response relationships, and cost estimations, resulting in higher decision-making risks. This invention introduces Monte Carlo simulation in step S600 to probabilistically characterize and propagate the uncertainties of all key parameters in the objective function, ultimately solving for a concentration threshold range with confidence intervals. This allows decision-makers to clearly understand the risk levels corresponding to different threshold selection options, such as the probability of achieving the expected benefits, supporting more robust and flexible decisions based on their risk preferences.

[0147] This invention forms a complete decision-making closed loop of prediction, optimization, and verification, enhancing the credibility of the solution. Unlike most methods that only stay at the prediction or optimization stage, this invention, through step S700, re-introduces the optimized recommended thresholds into the multi-pollutant air quality health index calculation framework, assesses the changes in risk level under historical benchmarks, and compares and verifies them with historically observed health improvement data. This step not only provides intuitive risk communication indicators, but more importantly, by calculating quantitative indicators such as the predicted improvement efficiency ratio, it provides empirical support from historical data for the rationality and feasibility of the optimization results, forming a complete evidence chain from historical learning to model optimization and backtracking verification, greatly enhancing the scientific persuasiveness and practical value of the final decision information.

[0148] In summary, this invention solves the core problems of existing technologies, such as data fragmentation, single decision-making basis, unclear risks, and insufficient credibility, through systematic innovation in four dimensions: multi-source data fusion, health and economic coupling optimization, uncertainty quantitative assessment, and historical empirical verification. It provides a scientific, robust, verifiable, and economically reasonable dynamic control scheme for environmental risks.

[0149] Example 2

[0150] A computer device 800, such as Figure 9 As shown, the system includes a memory 810, a processor 820, and a computer program 830 stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of a population health-oriented dynamic environmental risk control method. For a detailed description of the method, please refer to the corresponding description in the above method embodiments; it will not be repeated here.

[0151] Example 3

[0152] A computer-readable storage medium, such as Figure 10 As shown, a computer program is stored thereon. When executed by a processor, the computer program implements the steps of a population health-oriented dynamic control method for environmental risks. For a detailed description of the method, please refer to the corresponding description in the above method embodiments, which will not be repeated here.

[0153] The number of devices and processing scale described herein are for the purpose of simplifying the description of the invention. Applications, modifications, and variations of the invention will be readily apparent to those skilled in the art.

[0154] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

[0155] The apparatus, computer device, and non-volatile computer storage medium and method provided in the embodiments of this specification are corresponding. Therefore, the apparatus, computer device, and non-volatile computer storage medium also have similar beneficial technical effects as the corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the corresponding apparatus, computer device, and non-volatile computer storage medium will not be repeated here.

[0156] Those skilled in the art will also know that, besides implementing the controller in the form of purely computer-readable program code, the same functions can be achieved by logically programming the method steps, making the controller take the form of logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers (PLCs), and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the devices included within it for implementing various functions can also be considered structures within that hardware component. Alternatively, the devices for implementing various functions can be considered as both software units implementing the method and structures within a hardware component.

[0157] The systems, apparatuses, or units described in the above embodiments can be implemented by computer chips or physical entities, or by products with certain functions. For ease of description, the above apparatuses are described separately as various units based on their functions. Of course, when implementing one or more embodiments of this specification, the functions of each unit can be implemented in one or more software and / or hardware.

[0158] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0159] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0160] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0161] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0162] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0163] This specification may be described in the general context of computer-executable instructions, such as program units, that are executed by a computer. Generally, program units include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification may also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program units may reside in local and remote computer storage media, including storage devices.

[0164] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0165] The above description is merely an embodiment of this specification and is not intended to limit the scope of one or more embodiments of this specification. Various modifications and variations can be made to one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of one or more embodiments of this specification.

Claims

1. A population-oriented method for dynamic control of environmental risks, characterized in that, Includes the following steps: Acquire historical air pollution monitoring data, population health data, and longitudinal cohort data of the study subjects for the target area; By using historical air pollution monitoring data and population health data, a multi-pollutant air quality health index applicable to the region is constructed to quantify the comprehensive impact coefficient of historical pollution levels on population health risks. An individual-level causal risk prediction model for long-term air pollution exposure and health outcomes was established using the longitudinal cohort data. The comprehensive impact coefficient of the multi-pollutant air quality health index is used as the prior knowledge of the fusion module to process the parameters of the individual-level causal risk prediction model, resulting in the processed health risk function. A cost-benefit comprehensive objective function is established by coupling the pollution control marginal cost function with health economic assessment parameters based on the processed health risk function. The uncertainty of the parameters in the objective function is quantified, and the concentration threshold and its confidence interval that maximize the expected social net benefit are solved as recommended control thresholds. Substitute the recommended control threshold into the multi-pollutant air quality health index calculation framework to assess its risk level change relative to historical levels, and output decision information including threshold recommendations, health benefit predictions, and historical verification comparisons. The steps for establishing an individual-level causal risk prediction model for long-term air pollution exposure and health outcomes using the longitudinal cohort data include: Based on the residential address history of each study subject in the longitudinal cohort data, and combined with the atmospheric pollutant concentration grid data, long-term pollutant exposure concentrations were assigned to each study subject year by year during the follow-up period. Using the occurrence time and status of health outcome events during the follow-up period as dependent variables, a statistical model was constructed with a Cox proportional hazards model as the framework and a nested distributed lag nonlinear model. The distributed lag nonlinear model, through cross basis functions, simultaneously characterizes the nonlinear relationship between atmospheric pollutant exposure concentration and health risk, as well as the exposure lag effect. The individual covariates in the longitudinal cohort data are input into the statistical model as control variables, and the model is fitted using parameter estimation methods to obtain a complete set of model parameters, including the cross-basis function coefficients, lag effect coefficients, and covariate coefficients of pollutant exposure. Based on the fitted model parameters, a health risk function is generated for prediction. This function is used to calculate the risk ratio or excess relative risk of a specific population for the target health outcome when the pollutant concentration changes from the reference level to the target level. The health risk function and its parameters are encapsulated into a callable prediction module, which is then used as the output of the individual-level causal risk prediction model.

2. The method for dynamic control of environmental risks oriented towards population health as described in claim 1, characterized in that, The multi-pollutant air quality health index is an air quality health index constructed based on the cumulative risk index method, and its comprehensive influence coefficient is the exposure response coefficient vector obtained by the multi-pollutant generalized additive model.

3. The method for dynamic control of environmental risks oriented towards population health as described in claim 1, characterized in that, The allocation of long-term pollutant exposure concentrations specifically involves: using a pollutant concentration reanalysis dataset with a spatial resolution better than a preset accuracy threshold, extracting the annual average concentration of the geographical grid where the study subjects reside during the follow-up year based on their residential address coordinates, and using this as their individual annual exposure level; The cross-basis function uses a penalized spline function to nonlinearly fit the pollutant exposure-response relationship, and uses a polynomial function or spline function to model the hysteresis effect structure. The input interface of the callable prediction module includes at least: the change in the concentration of the target pollutant and the baseline covariate characteristic distribution of the target population, and its output interface includes at least: the predicted hazard ratio and the predicted change in the number of health outcomes.

4. A population health-oriented dynamic control method for environmental risks as described in any one of claims 1 to 3, characterized in that, The specific steps of the fusion module are as follows: taking the comprehensive impact coefficient corresponding to the multi-pollutant air quality health index as the center of the prior distribution, taking the parameters estimated by the individual-level causal risk prediction model as the likelihood function, and obtaining the posterior distribution of the parameters through the Markov chain Monte Carlo method, thereby realizing the correction and robustness of the estimation results of the individual-level causal risk prediction model.

5. A population health-oriented dynamic control method for environmental risks as described in any one of claims 1 to 3, characterized in that, The fusion module processes the parameters of the individual-level causal risk prediction model using a weighted integration method: The health burden under different historical pollution scenarios is assessed using the multi-pollutant air quality health index, and compared with the same burden predicted by the individual-level causal risk prediction model to calculate the corrected weight of the model prediction. The adjusted weights are applied to the prediction of future health benefits by the individual-level causal risk prediction model.

6. The method for dynamic control of environmental risks oriented towards population health as described in claim 1, characterized in that, The disability-adjusted life years weight or disease cost in the health economic assessment parameters are set differently based on the contribution ratio of different health endpoints in the historical assessment of the multi-pollutant air quality health index.

7. The method for dynamic control of environmental risks oriented towards population health as described in claim 5, characterized in that, The specific cost-benefit comprehensive objective function is as follows: The difference between the total health and economic benefits generated when the pollutant concentration changes to a preset value and the total pollution control cost required to achieve that concentration change is defined as the net social benefit function. Wherein, the total health economic benefit is the sum of the product of the reduction in the number of type i health outcomes calculated by the processed health risk function and the localized health economic assessment parameter corresponding to type i health outcomes; The total pollution control cost is obtained by integrating the pollution control marginal cost function from the current concentration to the target concentration; The optimization objective of the cost-benefit comprehensive objective function is to maximize the net social benefit function; The concentration threshold that maximizes the expected social net benefit is specifically achieved by maximizing the social net benefit function. The solution process is performed within the framework of Monte Carlo simulation to simultaneously handle the uncertainty of the parameters.

8. The method for dynamic control of environmental risks oriented towards population health as described in claim 1, characterized in that, The historical verification comparison includes: comparing the predicted multi-pollutant air quality health index value and the predicted excess mortality reduction under the recommended threshold with the actual multi-pollutant air quality health index value and the actual excess mortality change in historical periods, and calculating the predicted improvement efficiency ratio as an auxiliary evaluation indicator for the feasibility of the threshold scheme.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 8.

Images