A method for assessing the risk of novel coronavirus infection based on GO Bayesian

Abstract

The present invention discloses a method for risk assessment of novel coronavirus infection based on GO Bayes, including risk zone division; definition and setting of risk factors; definition of unit structure of GO Bayesian network model, determination of internal attributes and conversion rules; construction of GO Bayesian network model according to the risk factors; acquisition of risk factor data of the system to be assessed, online risk assessment, and obtaining risk assessment results. The present invention can provide a method for quickly assessing the risk of infection of personnel by collecting basic behavior information of personnel and briefly judging clinical symptoms after personnel enter the risk zone, solving the problems of insufficient nucleic acid detection capacity and delayed notification of test results.

Description

Technical Field

[0001] The present invention relates to the field of epidemic prevention and control, and in particular to a method for assessing the risk of novel coronavirus infection based on GO Bayesian method. Background Art

[0002] Every year with the arrival of winter and the continuous drop in temperature, the virus is easily transmitted among people, especially in densely populated open communities. Due to the incubation period of the virus, confirmed cases are prone to enter and exit public places before being diagnosed, which makes people entering and exiting the same area at the same time period at risk of infection. This puts forward requirements for the assessment of the infection risk of close contacts of confirmed cases. The existing diagnostic method is to directly conduct nucleic acid testing and isolate close contacts. On the one hand, the nucleic acid testing process takes time. On the other hand, centralized isolation consumes a lot of material resources and increases the epidemic prevention burden of municipal departments. At the same time, urban communities are densely populated and highly mobile, and existing means are not efficient. Summary of the invention

[0003] Based on the technical problems existing in the background technology, the present invention proposes a method for assessing the risk of novel coronavirus infection based on GO Bayesian method.

[0004] The present invention proposes a method for assessing the infection risk of close contacts of COVID-19 based on a Bayesian network, and the specific operation steps are as follows:

[0005] 1) Risk area division: using Thiessen polygons or concentric circles according to the different distribution states of risk sources;

[0006] 2) The definition and setting of risk factors, including both personnel behavior and clinical manifestations;

[0007] 3) Definition of the unit structure, internal attributes and transformation rules of the GO Bayesian network model;

[0008] 4) constructing a GO method Bayesian network model according to the risk factors, obtaining risk factor data of the system to be evaluated, performing online risk evaluation, and obtaining risk evaluation results;

[0009] 5) Perform sensitivity analysis by fine-tuning model parameters.

[0010] Preferably, the risk zone is divided into two ways. When the number of risk sources in the region is large and dispersed, Thiessen polygons are used to divide the grid. When the number of risk sources is small or the geographical locations are dense, concentric circles of a certain radius are drawn with the risk source group as the center.

[0011] Preferably, the definition and setting of the risk factors are divided into two levels, including the behavior level of risk personnel and the clinical manifestation level. The preliminary set behavior level includes five factors: distance from the risk source, travel mode, closedness of entry and exit places, length of stay in the risk area, and wearing of masks. The preliminary set clinical manifestation level refers to the eighth edition of the "Diagnosis and Treatment Plan for New Coronavirus Pneumonia" issued by the National Health Commission. The main symptoms include fever, dry cough, and fatigue. The initial symptoms include loss of smell and taste. Severe symptoms include dyspnea and hypoxemia.

[0012] Preferably, the definition of the unit structure, determination of the internal attributes and the conversion rules of the GO method Bayesian network model are as follows:

[0013] Unit structure 1: Risk generation unit (signal generator). The risk generation unit represents the input of the system. The unit itself has no input and is an external event or signal independent of the system. It gives the status of whether the individual enters the risk zone. j There are two states: I1 (enter) or I2 (do not enter). The unit conversion rules are:

[0014] F R =I j The probability is P R (I j )=P j

[0015] Unit structure 2: Risk transfer unit, P(i) represents the probability of a person's infection state, Q(i) represents the probability of a person's health state, subscripts C, S, and R represent operators, input states, and output states, respectively, and θ represents the infection risk parameter. The two-state unit is a commonly used unit, which has two states, healthy or infected, and has input and output data. The conversion rule of this unit is:

[0016] P R (i) = P S (i)·P C (i)

[0017] Q R (i) = Q S (i)+P S (i) Q C (i)

[0018] θ R =θ S +θ C

[0019] Unit structure 3: Risk superposition unit, which indicates multiple risk data inputs (with 1, 2, ..., n added after the subscript), such as multiple clinical symptoms input, one risk probability data output, and the logical relationship that there is an output as long as one of the multiple inputs occurs. The conversion rule of this unit is:

[0020] Q R (i) = 1-[1-Q S1 (i)][1-Q S2 (i)]...[1-Q Sn (i)]

[0021] P S1 (i) = Q S1 (i) / {1-[1-Q S1 (i)][1-Q S2 (i)]...[1-Q Sn (i)]}

[0022] P S2 (i) = Q S2 (i) / {1-[1-Q S1 (i)][1-Q S2 (i)]...[1-Q Sn (i)]} ...

[0024] P Sn (i) = Q Sn (i) / {1-[1-Q S1 (i)][1-Q S2 (i)]...[1-Q Sn (i)]}

[0025] Unit structure 3 represents multiple risk data inputs (represented by 1, 2, ..., n after the subscripts), such as multiple clinical symptom inputs, one risk probability data output, and the logical relationship that as long as one of the multiple inputs occurs, there will be an output;

[0026] Unit structure 4: Risk selection unit, which means that a risk factor has multiple risk options. The output risk probability can choose one of n output paths, or all output paths can be closed. For example, there are multiple possible choices for the risk factor of transportation mode, and different paths will have different output results. The conversion rules of this unit are:

[0027]

[0028]

[0029]

[0030] Unit structure 5: risk feedback unit. Unit structure 5 is a closed-loop controllable structure. The state of some risk factors is not independently selectable, but continuous, and requires another control factor to control. As shown in the figure, S1 is the input data of the closed-loop controllable system, which serves as the main input of unit structure 5. S2 is the feedback signal of the closed-loop controllable system, which serves as the secondary input of unit structure 5. R is the output. The input data, feedback signal and output are equivalent to one component and simulated by unit structure 5. There is a feedback effect between the risk factor and the control factor, which can correct the output of the unit and reduce the impact of unreasonable input on the output of the unit. The difference between S1 and S2 is the deviation. The output exists only when the deviation is within a certain range. The conversion rule of this unit is:

[0031]

[0032]

[0033] Assume that the infection risk parameter of the main input data is θ S1 , the infection risk parameter of the second input data, i.e. the feedback signal, is θ S2 , the infection risk parameter of unit structure 5 itself is λ C , the infection risk parameter of the control factor unit is θ K .

[0034] Preferably, a GO method Bayesian network model is constructed according to the risk factors, risk factor data in the system to be evaluated is obtained, online risk evaluation is performed, and risk evaluation results are obtained.

[0035] Preferably, the sensitivity analysis is performed by fine-tuning the model parameters, adjusting the initial states of the risk factors at the behavioral and clinical levels, and the initial values ​​of the infection risk parameters, to derive the key factors that affect the risk of infection for personnel.

[0036] In the present invention, the method for assessing the risk of infection with the new coronavirus based on the GO method Bayesian method gives full play to the advantages of the multi-state units in the GO method, simulates the multi-states of risk persons and risk factors, and can achieve independent risk assessment of risk groups based on the basic information of risk persons collected by non-clinical means, thereby reducing the time cost of epidemic prevention and control and reducing the pressure of nucleic acid testing. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a schematic diagram of the structure of five unit structures of a method for assessing the risk of infection with the novel coronavirus based on the GO method Bayesian approach proposed in the present invention;

[0038] Figure 2This is a schematic diagram of the GO Bayesian network model structure in an embodiment of a method for assessing the risk of infection with the new coronavirus based on the GO Bayesian approach proposed in the present invention. DETAILED DESCRIPTION

[0039] The technical solutions in the embodiments of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0040] A method for assessing the infection risk of close contacts of COVID-19 based on a Bayesian network, the specific steps of which are as follows:

[0041] 1) Risk area division: using Thiessen polygons or concentric circles according to the different distribution states of risk sources;

[0042] 2) The definition and setting of risk factors, including both personnel behavior and clinical manifestations;

[0043] 3) Definition of the unit structure, internal attributes and transformation rules of the GO Bayesian network model;

[0044] 4) constructing a GO method Bayesian network model according to the risk factors, obtaining risk factor data of the system to be evaluated, performing online risk evaluation, and obtaining risk evaluation results;

[0045] 5) Perform sensitivity analysis by fine-tuning model parameters.

[0046] In the present invention, the risk zone is divided into two ways. When the number of risk sources in the region is large and dispersed, Thiessen polygons are used to divide the grid. When the number of risk sources is small or the geographical locations are dense, concentric circles with a certain radius are drawn with the risk source group as the center.

[0047] In the present invention, the definition and setting of the risk factors are divided into two levels, including the risk personnel behavior level and the clinical manifestation level. The preliminary set behavior level includes five factors: distance from the risk source, travel mode, entry and exit tightness, stay time in the risk area, and wearing of masks. The preliminary set clinical manifestation level refers to the eighth edition of the "Diagnosis and Treatment Plan for New Coronavirus Pneumonia" issued by the National Health Commission. The main symptoms include fever, dry cough, fatigue, the first symptoms include loss of smell and taste, and severe symptoms include dyspnea and hypoxemia.

[0048] In the present invention, the definition of the unit structure, determination of internal attributes and conversion rules of the GO method Bayesian network model are as follows:

[0049] Unit structure 1: Risk generation unit (signal generator). The risk generation unit represents the input of the system. The unit itself has no input and is an external event or signal independent of the system. It gives the status of whether the individual enters the risk zone.j There are two states: I1 (enter) or I2 (do not enter). The unit conversion rules are:

[0050] F R =I j The probability is P R (I j )=P j

[0051] Unit structure 2: Risk transfer unit, P(i) represents the probability of a person's infection state, Q(i) represents the probability of a person's health state, subscripts C, S, and R represent operators, input states, and output states, respectively, and θ represents the infection risk parameter. The two-state unit is a commonly used unit, which has two states, healthy or infected, and has input and output data. The conversion rule of this unit is:

[0052] P R (i) = P S (i)·P C (i)

[0053] Q R (i) = Q S (i)+P S (i) Q C (i)

[0054] θ R =θ S +θ C

[0055] Unit structure 3: Risk superposition unit, which indicates multiple risk data inputs (with 1, 2, ..., n added after the subscript), such as multiple clinical symptoms input, one risk probability data output, and the logical relationship that there is an output as long as one of the multiple inputs occurs. The conversion rule of this unit is:

[0056] Q R (i) = 1-[1-Q S1 (i)][1-Q S2 (i)]...[1-Q Sn (i)]

[0057] P S1 (i) = Q S1 (i) / {1-[1-Q S1 (i)][1-Q S2 (i)]...[1-Q Sn (i)]}

[0058] P S2 (i) = Q S2 (i) / {1-[1-Q S1(i)][1-Q S2 (i)]...[1-Q Sn (i)]} ...

[0060] P Sn (i) = Q Sn (i) / {1-[1-Q S1 (i)][1-Q S2 (i)]...[1-Q Sn (i)]}

[0061] Unit structure 3 represents multiple risk data inputs (represented by 1, 2, ..., n after the subscripts), such as multiple clinical symptom inputs, one risk probability data output, and the logical relationship that as long as one of the multiple inputs occurs, there will be an output;

[0062] Unit structure 4: Risk selection unit, which means that a risk factor has multiple risk options. The output risk probability can choose one of n output paths, or all output paths can be closed. For example, there are multiple options for the risk factor of transportation mode, and different paths will have different output results. The conversion rules of this unit are:

[0063]

[0064]

[0065]

[0066] Unit structure 5: risk feedback unit. Unit structure 5 is a closed-loop controllable structure. The state of some risk factors is not independently selectable, but continuous, and requires another control factor to control. As shown in the figure, S1 is the input data of the closed-loop controllable system, which serves as the main input of unit structure 5. S2 is the feedback signal of the closed-loop controllable system, which serves as the secondary input of unit structure 5. R is the output. The input data, feedback signal and output are equivalent to one component and simulated by unit structure 5. There is a feedback effect between the risk factor and the control factor, which can correct the output of the unit and reduce the impact of unreasonable input on the output of the unit. The difference between S1 and S2 is the deviation. The output exists only when the deviation is within a certain range. The conversion rule of this unit is:

[0067]

[0068]

[0069] Assume that the infection risk parameter of the main input data is θ S1 , the infection risk parameter of the second input data, i.e. the feedback signal, is θ S2, the infection risk parameter of unit structure 5 itself is λ C , the infection risk parameter of the control factor unit is θ K .

[0070] In the present invention, a GO method Bayesian network model is constructed according to risk factors, risk factor data in the system to be evaluated is obtained, online risk evaluation is performed, and risk evaluation results are obtained.

[0071] In the present invention, the sensitivity analysis is performed by fine-tuning the model parameters, adjusting the initial states of the risk factors at the behavioral level and the clinical level, and the initial values ​​of the infection risk parameters, to obtain the key factors affecting the risk of infection of personnel.

[0072] The present invention: divides the community with confirmed cases into polygonal grids or concentric circle areas of different levels according to rules, and the entire area is a risk area; after uninfected people enter the risk area, the basic behavior information of the people is collected and the clinical symptoms are briefly judged to obtain risk factor data, which is input into the GO method Bayesian network model, such as Figure 2 The specific meaning of each unit is as follows:

[0073]

[0074]

[0075] Unit 1-1 is a risk occurrence unit, which is an initial input, indicating that a person enters a risk area; at the behavioral level, a risk selection unit 4-2 is set to select the mode of transportation for the person, including three forms 2-3, 2-4, and 2-5; at the same time, a risk feedback unit 5-7 is set to control the distance between the person and the risk source and the length of stay in the risk area; unit 3-9 represents the superposition of these two risk factors at the behavioral level; unit 3-10 represents the state of the person wearing a mask; at the clinical manifestation level, unit 3-14 is a superposition unit of three main clinical symptoms 2-11, 2-12, and 2-13; unit 3-18 is a superposition unit of two serious clinical symptoms 2-17 and 2-16; at the same time, it is superimposed with the first clinical symptom unit 2-15 to the risk superposition unit 3-19; the risk factors at the behavioral level and the clinical manifestation level are finally superimposed through unit 3-20 to output the person's infection risk assessment result; for this result, the initial state of the risk factors at the behavioral level and the clinical level, as well as the initial value of the infection risk parameter, can also be adjusted to obtain the key factors affecting the risk of infection of the person.

[0076] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for assessing the infection risk of close contacts of COVID-19 based on GO Bayesian network, characterized in that: include: 1) Risk area division: using Thiessen polygons or concentric circles according to the different distribution states of risk sources; 2) The definition and setting of risk factors, including both personnel behavior and clinical manifestations; 3) Definition of the unit structure, internal attributes and transformation rules of the GO Bayesian network model; The definition of the unit structure of the GO method Bayesian network model, the determination of internal attributes and conversion rules, include: Unit structure 1: Risk occurrence unit. The risk occurrence unit is an initial input that gives the status of whether the individual enters the risk zone. j , there are two states, I1 or I2, I1 means entering, I2 means not entering; Unit structure 2: Risk transfer unit, P(i) represents the probability of a person's infection state, Q(i) represents the probability of a person's health state, subscripts C, S, and R represent operators, input states, and output states, respectively, and θ represents the infection risk parameter. There are both input and output data. The conversion rule of this unit is: P R (i)=P S (i)·P C (i) Q R (i)=Q S (i)+P S (i)·Q C (i) i R =θ S +θ C Unit structure 3: Risk superposition unit, which indicates multiple risk data inputs. The subscripts are followed by 1, 2, ..., n, which indicates that there is a logical relationship that there will be an output as long as one of the multiple inputs occurs. The conversion rules of this unit are: Q R (i)=1-[1-Q S1 (i)][1-Q S2 (i)]...[1-Q Sn (i)] P S1 (i)=Q S1 (i) / {1-[1-Q S1 (i)][1-Q S2 (i)]...[1-Q Sn (i)]} P S2 (i)=Q S2 (i) / {1-[1-Q S1 (i)][1-Q S2 (i)]...[1-Q Sn (i)]} … P Sn (i)=Q Sn (i) / {1-[1-Q S1 (i)][1-Q S2 (i)]...[1-Q Sn (i)]}; Unit structure 4: Risk selection unit, which means that there are multiple risk options for a risk factor. The output risk probability selects one output from n output paths or closes all output paths. The conversion rule of this unit is: Unit structure 5: risk feedback unit. Unit structure 5 is a closed-loop controllable structure. S1 is the input data of the closed-loop controllable structure, which serves as the main input of unit structure 5. S2 is the feedback signal of the closed-loop controllable structure, which serves as the secondary input of unit structure 5. R is the output. The input data, feedback signal and output are equivalent to one component, which is simulated by unit structure 5. There is a feedback effect between the risk factor and the control factor. The output of the unit can be corrected to reduce the impact of unreasonable input on the output of the unit. The difference between S1 and S2 is the deviation. The output exists only when the deviation is within a certain range. The conversion rule of this unit is: Assume that the infection risk parameter of the main input data is θ S1 , the infection risk parameter of the control factor unit is θ K ; 4) constructing a GO method Bayesian network model based on the risk factors, obtaining risk factor data, performing online risk assessment, and obtaining risk assessment results; 5) Perform sensitivity analysis by fine-tuning model parameters.

2. According to claim 1, a method for assessing the infection risk of close contacts of COVID-19 based on GO Bayesian network is characterized in that: There are two ways to divide the risk zone. When the number of risk sources in the region is large and scattered, Thiessen polygons are used to divide the grid. When the number of risk sources is small or the geographical locations are dense, concentric circles of a certain radius are drawn with the risk source group as the center.

3. According to claim 1, a method for assessing the infection risk of close contacts of COVID-19 based on GO method Bayesian network is characterized in that: The definition and setting of the risk factors are divided into two levels, including the behavior level of risk personnel and the clinical manifestation level. The preliminary set behavior level includes five factors: distance from the risk source, travel mode, closedness of entry and exit places, length of stay in the risk area, and wearing of masks. The preliminary set clinical manifestation level refers to the eighth edition of the "Diagnosis and Treatment Plan for New Coronavirus Pneumonia" issued by the National Health Commission. The main symptoms include fever, dry cough, and fatigue. The initial symptoms include loss of smell and taste. Severe symptoms include dyspnea and hypoxemia.

4. According to a method for assessing the infection risk of close contacts of COVID-19 based on GO Bayesian network according to claim 1, it is characterized in that: The sensitivity analysis is performed by fine-tuning the model parameters, adjusting the initial states of risk factors at the behavioral and clinical levels, and the initial values ​​of infection risk parameters, to derive the key factors that affect the risk of infection for personnel.

Images