Health index assessment method based on combined exposure to multiple pollutants

Abstract

The present invention relates to the technical field of air quality health risk assessment, and discloses a health index assessment method based on composite exposure of multiple pollutants, including the following steps: S1: collecting air pollution, meteorology, hospitalization, and cause of death data, obtaining pollutant, meteorological factor, hospitalization, and death daily value data, and obtaining the 0-2 day moving average concentration of pollutants and the daily average value of meteorological factors; S2: selecting pollutants with statistical significance as pollutants for composite exposure analysis of multiple pollutants; S3: substituting the daily value data of hospitalization and death, the daily average value of meteorological factors, and the 0-2 day moving average concentration of selected pollutants into a multi-pollution model, a Bayesian weighted model, and a mixture analysis model to calculate the corresponding excess risk; S4: respectively calculating the pending CEHI under different air pollutant composite exposures corresponding to the excess risk under different models; S5: selecting a reasonable pending CEHI as the formal CEHI. The present invention can obtain a more accurate and comprehensive joint exposure health index.

Description

Technical Field

[0001] The present invention relates to the technical field of air quality health risk assessment, and in particular to a health index assessment method based on combined exposure to multiple pollutants. Background Art

[0002] AQHI (Air Quality Health Index) is an air quality evaluation index that comprehensively considers the acute health effects of various air pollutants on the population.

[0003] However, the existing methods for evaluating AQHI often simply calculate AQHI through a single model. It is impossible to verify whether the selected model has defects or biases, and thus it is impossible to know whether the calculated AQHI is accurate and reasonable. There is an urgent need for an evaluation method that can obtain a more accurate and comprehensive combined exposure health index. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a health index assessment method based on combined exposure to multiple pollutants, which can actively verify the obtained health index and obtain a more accurate and comprehensive combined exposure health index.

[0005] The technical solution adopted by the present invention is as follows: a health index assessment method based on multi-pollutant combined exposure includes the following steps:

[0006] S1: Collect data on air pollution, meteorology, hospital admissions, and causes of death, obtain daily values ​​of pollutants, meteorological factors, hospital admissions, and deaths, and obtain the 0-2 day moving average concentration of pollutants and the daily average of meteorological factors;

[0007] S2: Statistically significant pollutants were selected as pollutants for multi-pollutant composite exposure analysis through single-pollutant generalized additive models;

[0008] S3: Substitute the daily data of hospitalization and death, the daily average of meteorological factors, and the 0-2-day moving average concentration of selected pollutants into the multi-pollution model, Bayesian weighted model, and mixture analysis model to calculate the corresponding excess risk;

[0009] S4: Calculate the undetermined CEHI under different air pollutant composite exposures corresponding to the excess risk based on the multi-pollution model, Bayesian weighted model, and mixture analysis model;

[0010] S5: The K-fold cross-validation method and AQI fit analysis are used to verify the undetermined CEHIs calculated under different models, and a reasonable undetermined CEHI is selected as the official CEHI.

[0011] Explanation: CEHI stands for Combined Exposure to Multiple Chemicals Health Index, and AQI stands for Air Quality Index, which are mainly used to reflect the air quality status and degree of air pollution during the reporting period.

[0012] Compared with the prior art, the present invention has the following beneficial effects:

[0013] 1) The present invention is based on pollutants, meteorological factors, hospitalization, death daily data, 0-2 day moving average concentration of pollutants, daily average value data of meteorological factors, and three different models, calculates the pending CEHI of the corresponding model, and actively verifies the pending CEHI corresponding to different models through K-fold cross-validation method and AQI fit analysis, and finally selects a reasonable pending CEHI as the formal CEHI. Different models are based on different theoretical frameworks and algorithms. When processing the same data set, they will produce different prediction results. The present invention can make full use of the complementarity between models by combining the prediction results of multiple models, thereby improving the accuracy of the overall prediction, and can actively verify the obtained health index, and can obtain a more accurate and comprehensive joint exposure health index, thereby providing the public with more specific and targeted health advice and preventive measures;

[0014] 2) The present invention can provide a more accurate and comprehensive joint exposure health index, thereby providing more accurate suggestions for government policies and reducing the cost of government policy implementation (implementation costs such as reserving medical personnel, ambulances, and restricting economic activities);

[0015] 3) At present, when calculating AQHI, the health effects of various pollutants under the single pollutant model are simply added together. This method may not fully and truly reflect the impact of air pollution on human health due to the linear correlation between various air pollutants. A single model may be affected by factors such as data noise, unreasonable model assumptions or algorithm limitations, resulting in errors in the prediction results. Multiple models can interpret and predict data from different perspectives, thereby reducing the errors that may be caused by a single model.

[0016] 4) Multiple models can adapt to different data characteristics and pollutant concentration ranges, so that they can maintain good prediction performance under different environmental conditions. This makes the model calculation results reliable and stable in different regions and at different times. In practical applications, data may be affected by various interference factors, such as data missing. Multiple models can reduce the impact of these interference factors on the prediction results through mutual verification and correction, thereby improving the stability of the model;

[0017] 5) The three models selected in the present invention describe the actual situation from different perspectives and calculate the corresponding CEHI. For example, the Bayesian weighted model allows the quantification of uncertainty, and the mixture analysis model can more accurately deal with the collinearity problem between pollutants and evaluate the joint health effects of multiple pollutants.

[0018] As a preferred embodiment of the present invention, S1 comprises the following steps:

[0019] S11: Obtain PM through collected air pollution data 10 、PM 2.5 , NO2, O3, SO2, CO daily values, and PM 2.5 BC, OM, Daily data and calculate PM 10 、PM 2.5 , NO2, O3, SO2, CO, BC, OM, The daily moving average concentration is shown in formula (1):

[0020]

[0021] In formula (1), x pt0-2 Indicates that the pollutant p is at t 0-2 Daily moving average concentration, x pt0 represents the pollutant p concentration value on that day, x pt-1 represents the concentration value of pollutant p on the previous day, x pt-2 It represents the concentration value of pollutant p in the previous two days;

[0022] S12: Obtain daily value data of two meteorological factors, average temperature and average relative humidity, through the collected meteorological data, and calculate the average temperature and average relative humidity;

[0023] S13: Obtain hospitalization day data by hospital type, gender, age, and disease type through the collected hospital admission data;

[0024] S14: Obtain the daily death data of gender, age, and underlying cause of death through the collected cause of death data.

[0025] As a preferred embodiment of the present invention, S2 comprises the following steps:

[0026] S211: The exposure-response relationship coefficients between the 0-2 day moving average concentrations of different air pollutants and various health outcomes were calculated using the single-pollution generalized additive model, as shown in formula (2):

[0027] log[E(y jt )]=β p1 × pt0-2+s(time,k)+ns(te,df)+ns(h,df)+d+holidy (2),

[0028] In formula (2), y jt is the number of actual health outcomes j on day t, E(y jt ) is the expected number of health outcome j on day t, x pt0-2 is the pollutant p at t 0-2 Daily moving average concentration, β p1 is the exposure-response relationship coefficient of pollutant p, s is the cubic spline function, time is the time variable to control the long-term trend, the node k is 7, ns is the natural spline function, te is the average daily temperature, h is the average daily relative humidity, df is the degree of freedom, which is set to 3, d is the day of the week effect variable, and holidy is the variable to control the holiday effect;

[0029] S212: Obtain the exposure-response relationship coefficients of different air pollutants based on the single-pollution generalized additive model, and screen out pollutants corresponding to positive values ​​of the exposure-response relationship coefficients, and use the selected pollutants as pollutants for multi-pollutant composite exposure analysis.

[0030] As a preferred embodiment of the present invention, S3 includes the following steps:

[0031] S311: The exposure-response relationship coefficients between the 0-2 day moving average concentrations of selected different air pollutants and various health outcomes are calculated using a multi-pollutant generalized additive model, as shown in formula (3):

[0032] log[E(y jt )]=Σβ p2 × pt0-2 +s(time,k)+ns(te,df)+ns(h,df)+d+holidy (3),

[0033] In formula (3), y jt is the number of actual health outcomes j on day t, E(y jt ) is the expected number of health outcome j on day t, x pt0-2 is the pollutant p at t 0-2 Moving average concentration, β p2 is the exposure-response relationship coefficient of pollutant p, s is the cubic spline function, time is the time variable to control the long-term trend, the node k is 7, ns is the natural spline function, te is the average daily temperature, h is the average daily relative humidity, df is the degree of freedom, which is set to 3, d is the day of the week effect variable, and holidy is the variable to control the holiday effect;

[0034] S312: Calculate the first excess risk under the combined exposure of different air pollutants by calculating the exposure-response relationship coefficients of the selected different air pollutants in the 0-2 day moving average concentrations, as shown in formula (4):

[0035] ER jt =100×[exp(∑β p2 × pt0-2 )-1] (4),

[0036] In formula (4), ER jt is the cumulative excess risk of different air pollutants on health outcome j on day t, β p2 is the exposure-response relationship coefficient of pollutant p, x pt0-2 is the pollutant p at t 0-2 Moving average concentration.

[0037] As a preferred embodiment of the present invention, S3 includes the following steps:

[0038] S321: The Bayesian weighted model is used to calculate the overall exposure-response relationship coefficients between the 0-2 day moving average concentrations of different selected air pollutants and various health outcomes, as shown in formula (5).

[0039]

[0040] In formula (5), y jt represents the number of actual health outcomes j on day t, E(y jt ) is the expected number of health outcome j on day t, β represents the overall exposure-response relationship coefficient caused by combined exposure to air pollutants, and w p is the weight of pollutant p, x pt0-2 The pollutant p is 0-2 Daily moving average concentration, s is the cubic spline function, time is the time variable to control the long-term trend, the node k is 7, ns is the natural spline function, te is the average daily temperature, h is the average daily relative humidity, df is the degree of freedom, which is set to 3, d is the day of the week effect variable, and holidy is the variable to control the holiday effect;

[0041] S322: Calculate the second excess risk based on the overall exposure-response relationship coefficient, as shown in formula (6),

[0042]

[0043] In formula (6), ER jt is the cumulative excess risk of different air pollutants on health outcome j on day t, β is the overall exposure-response relationship coefficient, and w p is the weight of pollutant p, x pt0-2 is the pollutant p at t0-2 Moving average concentration.

[0044] As a preferred embodiment of the present invention, S3 includes the following steps:

[0045] S331: Calculate the third excess risk according to the mixture analysis model, as shown in formula (7),

[0046]

[0047] ER ji =ψ1 (8)

[0048] In formulas (7) and (8), Y jt is the number of actual health outcomes j on day t, x q For multiple pollutants p at the t 0-2 Daily moving average concentration dataset, For a given x q The conditional expectation of the number of health outcomes j on day t under the condition of the variables, g is the link function in the generalized linear model, ψ0 is the model intercept, η is a set of model coefficients for the covariates, S q is the joint value of the exposure, Z is the covariate, ER jt is the cumulative excess risk of different air pollutants on health outcome j on day t.

[0049] As a preferred embodiment of the present invention, S4 includes the following steps:

[0050] S41: Substitute the daily average values ​​of pollutant concentrations recommended in the air quality guideline values ​​formulated by WHO into the multi-pollution model, Bayesian weighted model, and mixture analysis model to calculate the corresponding excess risk, and use the calculated excess risk as the upper limit of the excess risk of the first-level environmental health risk;

[0051] S42: Substitute the primary standard limit of the Ambient Air Quality Standard into the corresponding excess risk calculated based on the multi-pollution model, Bayesian weighted model, and mixture analysis model, and use the excess risk as the upper limit of the secondary environmental health risk excess risk;

[0052] S43: Substitute the secondary standard limit of the Ambient Air Quality Standard into the corresponding excess risk calculated based on the multi-pollution model, Bayesian weighted model, and mixture analysis model, and use it as the upper limit of the excess risk of the third-level environmental health risk;

[0053] S34: Calculate CEHI, as shown in formula (9),

[0054]

[0055] In formula (9), CEHI is the combined exposure health index, ER is the excess risk, and ER LO is the lower value of the excess risk range, ER HI CEHI is the high value of the excess risk range. LO It is the low value of the joint exposure health index corresponding to the ER interval.

[0056] As a preferred embodiment of the present invention, S5 comprises the following steps:

[0057] S51: Divide the collected data on air pollution, weather, hospital admissions, and causes of death into K equal parts according to the time series year, of which the K-1th part of the data is used as the training set, and the Kth part of the data is used as the test set;

[0058] S52: The K-1th data is used to calculate the mean of the exposure-response relationship coefficient under different models according to S1, S2, and S3;

[0059] S53: Based on the Kth data and the mean values ​​of the exposure-response relationship coefficients under different models, calculate the expected number of health outcomes of the test sets corresponding to different models according to S1, S2, and S3;

[0060] S54: Compare the mean absolute errors between the expected number of health outcomes in the test set calculated based on different models and the actual number of health outcomes in the test set;

[0061] S55: Calculate from the first to the Kth data in accordance with S51 to S54 to obtain K-1 mean absolute error values;

[0062] S56: Compare the fit of multiple continuous CEHIs calculated under different models and multiple continuous AQIs in corresponding time periods at the same time;

[0063] S57: The CEHI calculated by the corresponding model with a small mean absolute error and a goodness of fit close to 1 is selected as the official CEHI. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a logic block diagram of the health index assessment method based on multi-pollutant combined exposure of the present invention;

[0065] Figure 2 It is a curve graph corresponding to multiple CEHIs and multiple AQIs in the same period of time in the health index assessment method based on multiple pollutants combined exposure of the present invention;

[0066] Figure 3 It is a K-fold cross-validation diagram of the health index assessment method based on multi-pollutant combined exposure of the present invention;

[0067] Figure 4The present invention provides a public response guide for the combined exposure health index based on the health index assessment method under combined exposure to multiple pollutants. DETAILED DESCRIPTION

[0068] Typical embodiments that embody the features and advantages of the present invention will be described in detail in the following description. It should be understood that the present invention can have various changes in different embodiments without departing from the scope of the present invention, and the descriptions and illustrations therein are essentially used for illustration purposes rather than for limiting the present invention.

[0069] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0070] Health index assessment methods based on multi-pollutant combined exposure, such as Figure 1 As shown, the following steps are included:

[0071] S1: Collect data on air pollution, meteorology, hospital admissions, and causes of death, obtain daily values ​​of pollutants, meteorological factors, hospital admissions, and deaths, and obtain the 0-2 day moving average concentration of pollutants and the daily average of meteorological factors;

[0072] S1 includes the following steps:

[0073] S11: Obtain PM through collected air pollution data 10 、PM 2.5 , NO2, O3, SO2, CO daily values, and PM 2.5 BC, OM, Daily data and calculate PM 10 、PM 2.5 , NO2, O3, SO2, CO, BC, OM, The daily moving average concentration is shown in formula (1):

[0074]

[0075] In formula (1), x pt0-2 Indicates that the pollutant p is at t 0-2 Daily moving average concentration, x pt0 represents the pollutant p concentration value on that day, x pt-1 represents the concentration value of pollutant p on the previous day, x pt-2 It represents the concentration value of pollutant p in the previous two days;

[0076] S12: Obtain daily value data of two meteorological factors, average temperature and average relative humidity, through the collected meteorological data, and calculate the average temperature and average relative humidity;

[0077] S13: Obtain hospitalization day data by hospital type, gender, age, and disease type through the collected hospital admission data;

[0078] S14: Obtain the daily death data of gender, age, and underlying cause of death through the collected cause of death data.

[0079] S15: The age in the obtained death day data is matched with the Chinese population life table standardized by the World Health Organization to calculate the daily YLL, as shown in formula (11):

[0080]

[0081] In formula (1), N i refers to the number of deaths in age group i, L i Refers to the number of years of life lost in the standard life table corresponding to the i-th age group.

[0082] In this embodiment, the 0-2 day average moving concentration refers to the moving average concentration accumulated over three days of the current day and the two days before. The existing AQHI evaluation method usually uses outpatient rate, hospitalization rate and mortality rate as indicators to measure health outcomes, but ignores age, which is an important information on the disease burden of the population. The present invention introduces YLL (years of life lost due to premature death) as an evaluation indicator, which combines information such as the number of deaths, age of death, and life expectancy, and can more comprehensively and truly reflect the impact of air pollution on human health.

[0083] S2: Statistically significant pollutants were selected as pollutants for multi-pollutant composite exposure analysis through single-pollutant generalized additive models;

[0084] S2 includes the following steps:

[0085] S211: The exposure-response relationship coefficients between the 0-2 day moving average concentrations of different air pollutants and various health outcomes were calculated using the single-pollution generalized additive model, as shown in formula (2):

[0086] log[E(y jt )]=β p1 × pt0-2 +s(time,k)+ns(te,df)+ns(h,df)+d+holidy (2),

[0087] In formula (2), y jt is the number of actual health outcomes j on day t, E(y jt ) is the expected number of health outcome j on day t, x pt0-2 is the pollutant p at t 0-2 Daily moving average concentration, β p1is the exposure-response relationship coefficient of pollutant p, s is the cubic spline function, time is the time variable to control the long-term trend, the node k is 7, ns is the natural spline function, te is the average daily temperature, h is the average daily relative humidity, df is the degree of freedom, which is set to 3, d is the day of the week effect variable, and holidy is the variable to control the holiday effect;

[0088] S212: Obtain the exposure-response relationship coefficients of different air pollutants based on the single-pollution generalized additive model, and screen out pollutants corresponding to positive values ​​of the exposure-response relationship coefficients, and use the selected pollutants as pollutants for multi-pollutant composite exposure analysis.

[0089] In this embodiment, the health outcome j refers to the daily hospitalization data, the daily death data, and the daily YLL data. That is to say, this scheme not only evaluates the excess risk of death, but also evaluates the excess risk of hospitalization caused by multiple pollution exposures, providing a more intuitive index to help policy makers, public health experts, and the public better understand the health risks of air pollution and make more effective risk management and decisions accordingly (such as the release of public response guidelines, the basis for emergency deployment of medical resources, the reference for urban activity planning and adjustment, and the assistance for traffic control strategy formulation, such as Figure 4 shown).

[0090] S3: Substitute the daily data of hospitalization and death, daily YLL data, daily mean values ​​of meteorological factors, and 0-2-day moving average concentrations of selected pollutants into the multi-pollution model (cumulative risk index CRI), Bayesian weighted model, and mixture analysis (QGComp) model to calculate the corresponding excess risk;

[0091] S3 includes the following steps:

[0092] S311: The exposure-response relationship coefficients between the 0-2 day moving average concentrations of selected different air pollutants and various health outcomes are calculated using a multi-pollutant generalized additive model, as shown in formula (3):

[0093] log[E(y jt )]=∑β p2 × pt0-2 +s(time,k)+ns(te,df)+ns(h,df)+d+holidy (3),

[0094] In formula (3), y jt is the number of actual health outcomes j on day t, E(y jt ) is the expected number of health outcome j on day t, x pt0-2 is the pollutant p at t 0-2 Moving average concentration, β p2is the exposure-response relationship coefficient of pollutant p, s is the cubic spline function, time is the time variable to control the long-term trend, the node k is 7, ns is the natural spline function, te is the average daily temperature, h is the average daily relative humidity, df is the degree of freedom, which is set to 3, d is the day of the week effect variable, and holidy is the variable to control the holiday effect;

[0095] S312: Calculate the first excess risk under the combined exposure of different air pollutants by calculating the exposure-response relationship coefficients of the selected different air pollutants in the 0-2 day moving average concentrations, as shown in formula (4):

[0096] ER jt =100×[exp(∑β p2 × pt0-2 )-1] (4),

[0097] In formula (4), ER jt is the cumulative excess risk of different air pollutants on health outcome j on day t, β p2 is the exposure-response relationship coefficient of pollutant p, x pt0-2 is the pollutant p at t 0-2 Moving average concentration.

[0098] In this embodiment, exp(∑β p2 × pt0-2 ) is equal to CRI jt , CRI jt is the cumulative risk index of different air pollutants on health outcome j on day t.

[0099] S321: The Bayesian weighted model is used to calculate the overall exposure-response relationship coefficients between the 0-2 day moving average concentrations of different selected air pollutants and various health outcomes, as shown in formula (5).

[0100]

[0101] In formula (5), y jt represents the number of actual health outcomes j on day t, E(y jt ) is the expected number of health outcome j on day t, β represents the overall exposure-response relationship coefficient caused by combined exposure to air pollutants, and w p is the weight of pollutant p, x pt0-2 The pollutant p is 0-2 Daily moving average concentration, s is the cubic spline function, time is the time variable to control the long-term trend, the node k is 7, ns is the natural spline function, te is the average daily temperature, h is the average daily relative humidity, df is the degree of freedom, which is set to 3, d is the day of the week effect variable, and holidy is the variable to control the holiday effect;

[0102] S322: Calculate the second excess risk based on the overall exposure-response relationship coefficient, as shown in formula (6),

[0103]

[0104] In formula (6), ER jt is the cumulative excess risk of different air pollutants on health outcome j on day t, β is the overall exposure-response relationship coefficient, and w p is the weight of pollutant p, x pt0-2 is the pollutant p at t 0-2 Moving average concentration.

[0105] In this embodiment, the prior distribution of the Bayesian weighted model parameters is set as follows:

[0106] β~Uniform(0,1)

[0107] W p ~Dirichlet(π p );π p ~Uniform(0,1)

[0108] Among them, by setting w p Obey Dirichlet distribution, π p Obeying a uniform distribution from 0 to 1, w can be limited p is a non-negative vector whose sum is 1, w p It can be regarded as the relative weight of each pollutant under the composite exposure of air pollutants. Four Markov chains were used, each with 5000 iterations, to estimate the posterior distribution of each parameter in the Bayesian weighted model. The convergence and efficiency diagnostics of the Markov chain were evaluated by R-hat and effective sample size (ESS), respectively. When R-hat < 1.05 and ESS of each chain > 100, the results were reliable.

[0109] S331: Calculate the third excess risk according to the mixture analysis model, as shown in formula (7),

[0110]

[0111] ER ji =ψ1 (8)

[0112] In formulas (7) and (8), Y jt is the number of actual health outcomes j on day t, x q For multiple pollutants p at the t 0-2 Daily moving average concentration dataset, For a given x qThe conditional expectation of the number of health outcomes j on day t under the condition of the variables, g is the link function in the generalized linear model, ψ0 is the model intercept, η is a set of model coefficients for the covariates, S q is the joint value of the exposure, Z is the covariate, ER jt is the cumulative excess risk of different air pollutants on health outcome j on day t.

[0113] S4: Calculate the undetermined CEHI under different air pollutant composite exposures corresponding to the excess risk based on the multi-pollution model, Bayesian weighted model, and mixture analysis model;

[0114] S4 includes the following steps:

[0115] S41: Substitute the daily average values ​​of pollutant concentrations recommended in the air quality guideline values ​​formulated by WHO into the multi-pollution model, Bayesian weighted model, and mixture analysis model to calculate the corresponding excess risk, and use the calculated excess risk as the upper limit of the excess risk of the first-level environmental health risk;

[0116] S42: Substitute the primary standard limit of the Ambient Air Quality Standard into the corresponding excess risk calculated based on the multi-pollution model, Bayesian weighted model, and mixture analysis model, and use the excess risk as the upper limit of the secondary environmental health risk excess risk;

[0117] S43: Substitute the secondary standard limit of the Ambient Air Quality Standard into the corresponding excess risk calculated based on the multi-pollution model, Bayesian weighted model, and mixture analysis model, and use it as the upper limit of the excess risk of the third-level environmental health risk;

[0118] S44: Calculate CEHI, as shown in formula (9),

[0119]

[0120] In formula (9), CEHI is the combined exposure health index, ER is the excess risk, and ER LO is the lower value of the excess risk range, ER HI CEHI is the high value of the excess risk range. LO It is the low value of the joint exposure health index corresponding to the ER interval.

[0121] In this embodiment, the daily average pollutant concentration recommended by the WHO in the air quality guideline value is PM 2.5 15 mg μg / m 3 , PM 10 45 μg / m 3 , O3100μg / m 3 , NO225μg / m 3, SO240μg / m 3 , CO4mg / m 3 ; The first level standard limit of "Ambient Air Quality Standard" is PM 2.5 35 μg / m 3 , PM 10 50 μg / m 3 , O3 100μg / m 3 , NO2 80μg / m 3 , SO2 50μg / m 3 , CO 4mg / m 3 ; The secondary standard limit of "Ambient Air Quality Standard" is PM 2.5 75 μg / m 3 , PM 10 150 μg / m 3 , O3 160μg / m 3 , NO2 80μg / m 3 , SO2 150μg / m 3 , CO 4mg / m 3 .

[0122] Joint exposure risk level CEHI interval Excess Risk (ER) Level 1 (0-3] <![CDATA[[0,ER HI1 ]]]> Level 2 (3-6] <![CDATA[(ER LO1 ,IS HI2 ]]> Level 3 (6-10] <![CDATA[(ER LO2 ,IS HI3 ]]> Level 4 >10 <![CDATA[>IS HI3 ]]>

[0123] Note: [a, b] and (a, b] in the table represent the interval, a is the lower value of the interval, and b is the higher value of the interval.

[0124] S5: The K-fold cross-validation method and AQI fit analysis are used to verify the undetermined CEHIs calculated under different models, and a reasonable undetermined CEHI is selected as the official CEHI.

[0125] like Figure 2 , 3 As shown, S5 includes the following steps:

[0126] S51: Divide the collected data on air pollution, weather, hospital admissions, and causes of death into K equal parts according to the time series year, with the K-1th part of the data used as the training set and the Kth part of the data used as the test set;

[0127] S52: The K-1th data is used to calculate the mean of the exposure-response relationship coefficient under different models according to S1, S2, and S3;

[0128] S53: Based on the Kth data and the mean values ​​of the exposure-response relationship coefficients under different models, calculate the expected number of health outcomes of the test sets corresponding to different models according to S1, S2, and S3;

[0129] S54: Compare the mean absolute errors between the expected number of health outcomes in the test set calculated based on different models and the actual number of health outcomes in the test set;

[0130] S55: Calculate from the first to the Kth data in accordance with S51 to S54 to obtain K-1 mean absolute error values;

[0131] S56: Compare the fit of multiple continuous CEHIs calculated under different models and multiple continuous AQIs in corresponding time periods at the same time;

[0132] S57: Select the CEHI calculated by the corresponding model with a small mean absolute error and a goodness of fit close to 1 as the official CEHI

[0133] In this embodiment, the fit between multiple continuous CEHIs and multiple continuous AQIs in the corresponding time period is calculated. If the fit is closer to 1, the CEHI predicted by the corresponding model is more accurate, as shown in formula (10),

[0134]

[0135] In formula (10), R-squared represents the goodness of fit, SSR is the sum of squares of differences between multiple CEHIs and the mean, in this embodiment, the mean difference refers to the difference between the means of multiple CEHIs, and SST is the sum of squares of differences between the AQI and the CEHI mean.

[0136] In this embodiment, the mean absolute error is small, indicating that the expected number of health outcomes of the test set calculated by the corresponding model is closer to the actual number of health outcomes of the test set and more in line with the actual situation. The degree of fit is close to 1, indicating that the fluctuation ranges of multiple consecutive CEHIs and multiple consecutive AQIs are consistent, the corresponding model has a high degree of fit, is more in line with the actual situation, and the undetermined CEHI obtained by the corresponding model is the most reasonable.

[0137] In this embodiment, the K-fold cross-validation method is used, that is, the collected data is divided into K equal parts. The first time, the first part of the data is used as the training set, and the second part of the data is used as the test set. The second time, the second part of the data is used as the training set, and the third part of the data is used as the test set, and so on. The mean absolute error corresponding to each model is calculated, and then the mean absolute errors corresponding to multiple models are compared.

[0138] The above-mentioned embodiments are only preferred embodiments of the present invention and cannot be used to limit the scope of protection of the present invention. Any non-substantial changes and substitutions made by technicians in this field on the basis of the present invention shall fall within the scope of protection required by the present invention.

Claims

1. A health index assessment method based on multiple pollutant combined exposure, characterized by: The following steps are involved: S1: Collect data on air pollution, meteorology, hospital admissions, and causes of death, obtain daily values ​​of pollutants, meteorological factors, hospital admissions, and deaths, and obtain the 0-2 day moving average concentration of pollutants and the daily average of meteorological factors; S2: Statistically significant pollutants were selected as pollutants for multi-pollutant composite exposure analysis through single-pollutant generalized additive models; S3: Substitute the daily data of hospitalization and death, the daily average of meteorological factors, and the 0-2-day moving average concentration of selected pollutants into the multi-pollution model, Bayesian weighted model, and mixture analysis model to calculate the corresponding excess risk; S4: Calculate the undetermined CEHI under different air pollutant composite exposures corresponding to the excess risk based on the multi-pollution model, Bayesian weighted model, and mixture analysis model; S5: The K-fold cross-validation method and AQI fit analysis are used to verify the undetermined CEHIs calculated under different models, and a reasonable undetermined CEHI is selected as the official CEHI.

2. The health index assessment method based on multiple pollutant combined exposure according to claim 1 is characterized by: S1 includes the following steps: S11: Obtain PM through the collected air pollution data 10 、PM 2.5 , NO2, O3, SO2, CO daily values, and PM 2.5 BC, OM, Daily data and calculate PM 10 、PM 2.5 , NO2, O3, SO2, CO, BC, OM, The daily moving average concentration is shown in formula (1): In formula (1), x pt0-2 Indicates that the pollutant p is 0-2 Daily moving average concentration, x pt0 represents the pollutant p concentration value on that day, x pt-1 represents the concentration value of pollutant p on the previous day, x pt-2 It represents the concentration value of pollutant p in the previous two days; S12: Obtain daily value data of two meteorological factors, average temperature and average relative humidity, through the collected meteorological data, and calculate the average temperature and average relative humidity; S13: Obtain hospitalization day data by hospital type, gender, age, and disease type through the collected hospital admission data; S14: Obtain the daily death data of gender, age, and underlying cause of death through the collected cause of death data.

3. The health index assessment method based on multiple pollutant combined exposure according to claim 1 is characterized in that: S2 includes the following steps: S211: The exposure-response relationship coefficients between the 0-2 day moving average concentrations of different air pollutants and various health outcomes were calculated using the single-pollution generalized additive model, as shown in formula (2): log[E(y jt )]=β p1 ×x pt0-2 +s(time,k)+ns(te,df)+ns(h,df)+d+holidy (2), In formula (2), y jt is the number of actual health outcomes j on day t, E(y jt ) is the expected number of health outcome j on day t, x pt0-2 is the pollutant p at t 0-2 Daily moving average concentration, β p1 is the exposure-response relationship coefficient of pollutant p, s is the cubic spline function, time is the time variable to control the long-term trend, the node k is 7, ns is the natural spline function, te is the average daily temperature, h is the average daily relative humidity, df is the degree of freedom, which is set to 3, d is the day of the week effect variable, and holidy is the variable to control the holiday effect; S212: Obtain the exposure-response relationship coefficients of different air pollutants based on the single-pollution generalized additive model, and screen out pollutants corresponding to positive exposure-response relationship coefficients, and use the selected pollutants as pollutants for multi-pollutant composite exposure analysis.

4. The health index assessment method based on multiple pollutant combined exposure according to claim 1 is characterized in that: S3 includes the following steps: S311: The exposure-response relationship coefficients between the 0-2 day moving average concentrations of different selected air pollutants and various health outcomes are calculated using a multi-pollutant generalized additive model, as shown in formula (3): log[E(y jt )]=∑β p2 ×x pt0-2 +s(time,k)+ns(te,df)+ns(h,df)+d+holidy (3), In formula (3), y jt is the number of actual health outcomes j on day t, E(y jt ) is the expected number of health outcome j on day t, x pt0-2 is the pollutant p at t 0-2 Moving average concentration, β p2 is the exposure-response relationship coefficient of pollutant p, s is the cubic spline function, time is the time variable to control the long-term trend, the node k is 7, ns is the natural spline function, te is the average daily temperature, h is the average daily relative humidity, df is the degree of freedom, which is set to 3, d is the day of the week effect variable, and holidy is the variable to control the holiday effect; S312: Calculate the first excess risk under the combined exposure of different air pollutants by calculating the exposure-response relationship coefficients of the selected different air pollutants in the 0-2 day moving average concentrations, as shown in formula (4): IS jt =100×[exp(Σβ p2 ×x pt0-2 )-1] (4), In formula (4), ER jt is the cumulative excess risk of different air pollutants on health outcome j on day t, β p2 is the exposure-response relationship coefficient of pollutant p, x pt0-2 is the pollutant p at t 0-2 Moving average concentration.

5. The health index assessment method based on multiple pollutant combined exposure according to claim 1 is characterized in that: S3 includes the following steps: S321: The Bayesian weighted model is used to calculate the overall exposure-response relationship coefficients between the 0-2 day moving average concentrations of different selected air pollutants and various health outcomes, as shown in formula (5). In formula (5), y jt represents the number of actual health outcomes j on day t, E(y jt ) is the expected number of health outcome j on day t, β represents the overall exposure-response relationship coefficient caused by combined exposure to air pollutants, and w p is the weight of pollutant p, x pt0-2 The pollutant p is 0-2 Daily moving average concentration, s is the cubic spline function, time is the time variable to control the long-term trend, the node k is 7, ns is the natural spline function, te is the average daily temperature, h is the average daily relative humidity, df is the degree of freedom, which is set to 3, d is the day of the week effect variable, and holidy is the variable to control the holiday effect; S322: Calculate the second excess risk based on the overall exposure-response relationship coefficient, as shown in formula (6), In formula (6), ER jt is the cumulative excess risk of different air pollutants on health outcome j on day t, β is the overall exposure-response relationship coefficient, and w p is the weight of pollutant p, x pt0-2 is the pollutant p at t 0-2 Moving average concentration.

6. The health index assessment method based on multiple pollutant combined exposure according to claim 1 is characterized by: S3 includes the following steps: S331: Calculate the third excess risk according to the mixture analysis model, as shown in formula (7), IS ji =ψ1 (8) In formulas (7) and (8), Y jt is the number of actual health outcomes j on day t, x q For multiple pollutants p at the t 0-2 Daily moving average concentration dataset, For a given x q The conditional expectation of the number of health outcomes j on day t under the condition of the variables, g is the link function in the generalized linear model, ψ0 is the model intercept, η is a set of model coefficients for the covariates, S q is the joint value of the exposure, Z is the covariate, ER jt is the cumulative excess risk of different air pollutants on health outcome j on day t.

7. The health index assessment method based on multiple pollutant combined exposure according to claim 1 is characterized by: S4 includes the following steps: S41: Substitute the daily average values ​​of pollutant concentrations recommended in the air quality guideline values ​​formulated by WHO into the multi-pollution model, Bayesian weighted model, and mixture analysis model to calculate the corresponding excess risk, and use the calculated excess risk as the upper limit of the excess risk of the first-level environmental health risk; S42: Substitute the primary standard limit of the Ambient Air Quality Standard into the corresponding excess risk calculated based on the multi-pollution model, Bayesian weighted model, and mixture analysis model, and use the excess risk as the upper limit of the secondary environmental health risk excess risk; S43: Substitute the secondary standard limit of the Ambient Air Quality Standard into the corresponding excess risk calculated based on the multi-pollution model, Bayesian weighted model, and mixture analysis model, and use it as the upper limit of the excess risk of the third-level environmental health risk; S44: Calculate CEHI, as shown in formula (9), In formula (9), CEHI is the combined exposure health index, ER is the excess risk, and ER LO is the lower value of the excess risk range, ER HI CEHI is the high value of the excess risk range. LO It is the low value of the joint exposure health index corresponding to the ER interval.

8. The health index assessment method based on multiple pollutant combined exposure according to claim 1 is characterized by: S5 includes the following steps: S51: Divide the collected data on air pollution, weather, hospital admissions, and causes of death into K equal parts according to the time series year, of which the K-1th part of the data is used as the training set, and the Kth part of the data is used as the test set; S52: The K-1th data is used to calculate the mean of the exposure-response relationship coefficient under different models according to S1, S2, and S3; S53: Based on the Kth data and the mean values ​​of the exposure-response relationship coefficients under different models, calculate the expected number of health outcomes of the test sets corresponding to different models according to S1, S2, and S3; S54: Compare the mean absolute errors between the expected number of health outcomes in the test set calculated based on different models and the actual number of health outcomes in the test set; S55: Calculate from the first to the Kth data in accordance with S51 to S54 to obtain K-1 mean absolute error values; S56: Compare the fit of multiple continuous CEHIs calculated under different models and multiple continuous AQIs in corresponding time periods at the same time; S57: The CEHI calculated by the corresponding model with a small mean absolute error and a goodness of fit close to 1 is selected as the official CEHI.

Images